Professional Documents
Culture Documents
(Ramesh K. Shah, Dusan P. Sekulic) Fundamentals of PDF
(Ramesh K. Shah, Dusan P. Sekulic) Fundamentals of PDF
EXCHANGER DESIGN
FUNDAMENTALS OF
HEAT EXCHANGER
DESIGN
Ramesh K. Shah
Rochester Institute of Technology, Rochester, New York
Formerly at Delphi Harrison Thermal Systems, Lockport, New York
Dusan P. Sekulic
University of Kentucky, Lexington, Kentucky
1
This book is printed on acid-free paper. *
Copyright # 2003 by John Wiley & Sons, Inc. All rights reserved
Published by John Wiley & Sons, Inc., Hoboken, New Jersey
Published simultaneously in Canada
No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any
form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise,
except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either
the prior written permission of the Publisher, or authorization through payment of the appropriate
per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923,
(978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher
for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc.,
111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008,
e-mail: permcoordinator@wiley.com.
Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best
eorts in preparing this book, they make no representations or warranties with respect to the
accuracy or completeness of the contents of this book and specically disclaim any implied
warranties of merchantability or tness for a particular purpose. No warranty may be created or
extended by sales representatives or written sales materials. The advice and strategies contained
herein may not be suitable for your situation. You should consult with a professional where
appropriate. Neither the publisher nor author shall be liable for any loss of prot or any other
commercial damages, including but not limited to special, incidental, consequential, or other
damages.
For general information on our other products and services or for technical support, please contact
our Customer Care Department within the United States at (800) 762-2974, outside the United
States at (317) 572-3993 or fax (317) 572-4002.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print
may not be available in electronic books. For more information about Wiley products, visit our web
site at www.wiley.com
Library of Congress Cataloging-in-Publication Data:
Shah, R. K.
Fundamentals of heat exchanger design / Ramesh K. Shah, Dus an P. Sekulic.
p. cm.
Includes index.
ISBN 0-471-32171-0
1. Heat exchangersDesign and construction. I. Sekulic, Dus an P. II. Title.
TJ263 .S42 2003
621.402 0 5dc21
2002010161
Printed in the United States of America
10 9 8 7 6 5 4 3 2 1
Contents
Preface
xv
Nomenclature
xix
1
1
1.2
3
3
7
1.3
1.4
1.5
1.6
1.7
Gas-to-Fluid Exchangers
Liquid-to-Liquid and Phase-Change Exchangers
8
11
12
12
1.5.1
1.5.2
1.5.3
1.5.4
13
22
36
47
56
1.6.1
1.6.2
57
64
Single-Pass Exchangers
Multipass Exchangers
73
Summary
References
Review Questions
73
73
74
78
2.1
78
2.1.1
2.1.2
2.1.3
2.1.4
2.1.5
2.1.6
2.1.7
79
83
87
90
92
93
93
vi
CONTENTS
2.2
3.3
3.5
3.6
98
100
3.2.1
3.2.2
3.2.3
3.2.4
100
102
104
107
3.9
114
114
118
119
121
3.4.1
122
Single-Pass Exchangers
139
3.5.1
3.5.2
3.5.3
3.5.4
140
140
141
141
Temperature Eectiveness P
Number of Transfer Units, NTU
Heat Capacity Rate Ratio R
General PNTU Functional Relationship
PNTU Relationships
3.6.2
3.8
3.6.1
3.7
97
3.3.1
3.3.2
3.3.3
3.4
93
94
94
95
95
142
142
164
186
3.7.1
3.7.2
186
187
190
3.8.1
3.8.2
3.8.3
3.8.4
Counterow Exchanger
Parallelow Exchanger
Other Basic Flow Arrangements
Heat Exchanger Arrays and Multipassing
190
191
192
201
207
3.9.1
3.9.2
3.9.3
3.9.4
207
208
209
209
210
210
211
CONTENTS
212
213
213
213
214
214
215
216
216
219
219
220
227
232
4.1
232
244
4.2.1
4.2.2
4.2.3
248
249
251
4.2
4.3
4.4
vii
Temperature Eect
Length Eect
Combined Eect
236
236
239
239
239
239
258
4.3.1
4.3.2
4.3.3
4.3.4
259
272
288
289
291
291
296
297
298
298
299
302
308
5.1
308
5.1.1
5.1.2
5.1.3
308
310
312
viii
CONTENTS
5.2
5.3
316
316
320
320
321
326
The Method
337
5.3.1
5.3.2
5.3.3
341
344
345
5.4
348
5.5
355
5.5.1
355
5.6
5.7
Simplied Theory
360
360
366
371
372
373
376
378
6.1
Introduction
378
6.1.1
6.1.2
6.1.3
6.1.4
378
380
380
381
6.2
6.3
6.4
6.5
6.6
6.7
381
6.2.1
6.2.2
382
391
392
393
6.4.1
6.4.2
393
393
Tube Banks
Shell-and-Tube Exchangers
397
399
6.6.1
6.6.2
6.6.3
399
399
403
Pipe Losses
Sudden Expansion and Contraction Losses
Bend Losses
412
6.7.1
6.7.2
413
414
CONTENTS
6.8
ix
418
Summary
References
Review Questions
Problems
419
420
420
422
425
7.1
Basic Concepts
426
7.1.1
7.1.2
7.1.3
7.1.4
426
429
438
439
7.2
7.3
7.4
7.5
7.6
Dimensionless Groups
441
7.2.1
7.2.2
7.2.3
Fluid Flow
Heat Transfer
Dimensionless Surface Characteristics as a Function of the
Reynolds Number
443
446
450
7.3.1
7.3.2
7.3.3
7.3.4
451
460
467
471
7.8
449
473
7.4.1
7.4.2
7.4.3
7.4.4
7.4.5
7.4.6
475
499
502
507
508
510
511
7.5.1
7.5.2
7.5.3
7.5.4
7.5.5
512
514
515
519
523
Tube Bundles
Plate Heat Exchanger Surfaces
Plate-Fin Extended Surfaces
Tube-Fin Extended Surfaces
Regenerator Surfaces
7.7
Boundary Layers
Types of Flows
Free and Forced Convection
Basic Denitions
529
530
532
7.7.1
7.7.2
533
535
537
7.8.1
538
CONTENTS
563
8.1
563
8.1.1
8.1.2
563
566
8.2
8.3
Inline Arrangement
Staggered Arrangement
569
8.2.1
8.2.2
8.2.3
569
572
574
574
8.3.1
8.3.2
8.3.3
574
580
584
8.4
585
8.4.1
585
8.5
587
8.5.1
8.5.2
8.5.3
587
589
592
8.6
538
542
544
548
553
597
598
598
599
601
9.1
601
9.2
9.3
9.4
603
604
604
605
9.2.1
9.2.2
605
617
Rating Problem
Sizing Problem
631
9.3.1
9.3.2
9.3.3
9.3.4
631
631
632
632
Surface Geometries
Heat Transfer Calculations
Pressure Drop Calculations
Core Mass Velocity Equation
632
9.4.1
9.4.2
633
635
CONTENTS
9.4.3
9.4.4
9.5
9.6
Rating a PHE
Sizing a PHE
xi
637
645
646
9.5.1
9.5.2
9.5.3
9.5.4
646
650
658
663
664
Summary
References
Review Questions
Problems
667
667
668
669
Shell-and-Tube Exchangers
Plate Heat Exchangers
Extended-Surface Exchangers
Regenerator Surfaces
673
674
674
675
675
678
678
678
680
680
693
694
699
699
700
713
723
724
726
726
727
732
735
735
737
738
738
739
745
xii
CONTENTS
748
749
755
756
759
762
763
763
765
766
771
11.6 Energy, Exergy, and Cost Balances in the Analysis and Optimization
of Heat Exchangers
775
11.6.1
11.6.2
11.6.3
11.6.4
11.6.5
776
779
783
786
787
791
796
800
801
802
804
809
809
810
821
834
837
837
844
845
848
852
852
853
853
CONTENTS
Review Questions
Problems
13 Fouling and Corrosion
13.1 Fouling and its Eect on Exchanger Heat Transfer and Pressure Drop
13.2 Phenomenological Considerations of Fouling
13.2.1
13.2.2
13.2.3
13.2.4
13.2.5
13.2.6
Fouling Mechanisms
Single-Phase Liquid-Side Fouling
Single-Phase Gas-Side Fouling
Fouling in Compact Exchangers
Sequential Events in Fouling
Modeling of a Fouling Process
xiii
855
859
863
863
866
867
870
871
871
872
875
881
881
882
886
890
890
891
892
893
895
895
897
898
898
899
903
906
911
913
C.1
C.2
C.3
913
916
917
920
926
Index
931
Preface
Over the past quarter century, the importance of heat exchangers has increased immensely from the viewpoint of energy conservation, conversion, recovery, and successful
implementation of new energy sources. Its importance is also increasing from the standpoint of environmental concerns such as thermal pollution, air pollution, water pollution, and waste disposal. Heat exchangers are used in the process, power, transportation,
air-conditioning and refrigeration, cryogenic, heat recovery, alternate fuels, and
manufacturing industries, as well as being key components of many industrial products
available in the marketplace. From an educational point of view, heat exchangers
illustrate in one way or another most of the fundamental principles of the thermal
sciences, thus serving as an excellent vehicle for review and application, meeting the
guidelines for university studies in the United States and oversees. Signicant advances
have taken place in the development of heat exchanger manufacturing technology as well
as design theory. Many books have been published on the subject, as summarized in
the General References at the end of the book. However, our assessment is that none of
the books available seems to provide an in-depth coverage of the intricacies of heat
exchanger design and theory so as to fully support both a student and a practicing
engineer in the quest for creative mastering of both theory and design. Our book was
motivated by this consideration. Coverage includes the theory and design of exchangers
for many industries (not restricted to, say, the process industry) for a broader, in-depth
foundation.
The objective of this book is to provide in-depth thermal and hydraulic design theory
of two-uid single-phase heat exchangers for steady-state operation. Three important
goals were borne in mind during the preparation of this book:
1. To introduce and apply concepts learned in rst courses in heat transfer, uid
mechanics, thermodynamics, and calculus, to develop heat exchanger design
theory. Thus, the book will serve as a link between fundamental subjects mentioned and thermal engineering design practice in industry.
2. To introduce and apply basic heat exchanger design concepts to the solution of
industrial heat exchanger problems. Primary emphasis is placed on fundamental
concepts and applications. Also, more emphasis is placed on analysis and less on
empiricism.
3. The book is also intended for practicing engineers in addition to students.
Hence, at a number of places in the text, some redundancy is added to make the
concepts clearer, early theory is developed using constant and mean overall heat
transfer coefcients, and more data are added in the text and tables for industrial
use.
xvi
PREFACE
PREFACE
xvii
develop good understanding of the intricacies of heat exchanger design after going
through this material and prior to embarking on specialized work in their areas of
greatest interest.
For the thermal design of a heat exchanger for an application, considerable intellectual eort is needed in selecting heat exchanger type and determining the appropriate
value of the heat transfer coecients and friction factors; a relatively small eort is
needed for executing sizing and optimizing the exchanger because of the computerbased calculations. Thus, Chapters 7, 9, and 10 are very important, in addition to
Chapter 3, for basic understanding of theory, design, analysis, and selection of heat
exchangers.
Material presented in Chapters 11 through 13 is signicantly more interdisciplinary
than the rest of the book and is presented here in a modied methodological approach. In
Chapter 11 in particular, analytical modeling is used extensively. Readers will participate
actively through a set of examples and problems that extend the breadth and depth of the
material given in the main body of the text. A number of examples and problems in
Chapter 11 require analytical derivations and more elaborate analysis, instead of illustrating the topics with examples that favor only utilization of the formulas and computing numerical values for a problem. The complexity of topics requires a more diverse
approach to terminology, less routine treatment of established conventions, and a more
creative approach to some unresolved dilemmas.
Because of the breadth of the subject, the coverage includes various design aspects and
problems for indirect-contact two-uid heat exchangers with primarily single-phase
uids on each side. Heat exchangers with condensing and evaporating uids on one
side can also be analyzed using the design methods presented as long as the thermal
resistance on the condensing or evaporating side is small or the heat transfer coecient
on that side can be treated as a constant. Design theory for the following exchangers
is not covered in this book, due to their complexity and space limitations: two-phase
and multiphase heat exchangers (such as condensers and vaporizers), direct-contact
heat exchangers (such as humidiers, dehumidiers, cooling towers), and multiuid
and multistream heat exchangers. Coverage of mechanical design, exchanger fabrication
methods, and manufacturing techniques is also deemed beyond the scope of the
book.
Books by M. Jakob, D. Q. Kern, and W. M. Kays and A. L. London were considered
to be the best and most comprehensive texts on heat exchanger design and analysis
following World War II. In the last thirty or so years, a signicant number of books
have been published on heat exchangers. These are summarized in the General
References at the end of the book.
This text is an outgrowth of lecture notes prepared by the authors in teaching courses
on heat exchanger design, heat transfer, and design and optimization of thermal systems
to senior and graduate students. These courses were taught at the State University of
New York at Bualo and the University of Novi Sad, Yugoslavia. Over the past fteen
years or more, the notes of the rst author have been used for teaching purposes at a
number of institutions, including the University of Miami by Professor S. Kakac,
Rensselaer Polytechnic Institute by Professors A. E. Bergles and R. N. Smith,
Rochester Institute of Technology by Professor S. G. Kandlikar, Rice University by
Professor Y. Bayazitoglu, University of Tennessee Space Center by Dr. R. Schultz,
University of Texas at Arlington by Professor A. Haji-Sheikh, University of
Cincinnati by Professor R. M. Manglik, Northeastern University by Professor Yaman
Yener, North Carolina A&T State University by Professor Lonnie Sharpe, Auburn
xviii
PREFACE
University by Dr. Peter Jones, Southern Methodist University by Dr. Donald Price,
University of Tennessee by Professor Edward Keshock, and Gonzaga University by
Professor A. Aziz. In addition, these course notes have been used occasionally at a
number of other U.S. and foreign institutions. The notes of the second author have
also been used for a number of undergraduate and graduate courses at Marquette
University and the University of Kentucky.
The rst author would like to express his sincere appreciation to the management
of Harrison Thermal Systems, Delphi Corporation (formerly General Motors
Corporation), for their varied support activities over an extended period of time. The
second author acknowledges with appreciation many years of support by his colleagues
and friends on the faculty of the School of Engineering, University of Novi Sad, and
more recently at Marquette University and the University of Kentucky. We are also
thankful for the support provided by the College of Engineering, University of
Kentucky, for preparation of the rst ve and nal three chapters of the book. A special
word of appreciation is in order for the diligence and care exercised by Messrs. Dale Hall
and Mack Mosley in preparing the manuscript and drawings through Chapter 5.
The rst author is grateful to Professor A. L. London of Stanford University for
teaching him the ABCs of heat exchangers and for providing constant inspiration and
encouragement throughout his professional career and particularly during the course of
preparation of this book. The rst author would also like to thank Professors Sadik
Kakac of the University of Miami and Ralph Webb of the Pennsylvania State University
for their support, encouragement, and involvement in many professional activities
related to heat exchangers. The second author is grateful to his colleague and friend
Professor B. S. Baclic, University of Novi Sad, for many years of joint work and teaching
in the elds of heat exchanger design theory. Numerous discussions the second author
have had with Dr. R. Gregory of the University of Kentucky regarding not only what
one has to say about a technical topic, but in particular how to formulate it for a reader,
were of a great help in resolving some dilemmas. Also, the continuous support and
encouragement of Dr. Frederick Edeskuty of Los Alamos National Laboratory, and
Professor Richard Gaggioli of Marquette University were immensely important to the
second author in an eort to exercise his academic experience on both sides of the
Atlantic Ocean. We appreciate Professor P. V. Kadaba of the Georgia Institute of
Technology and James Seebald of ABB Alstom Air Preheater for reviewing the complete
manuscript and providing constructive suggestions, and Dr. M. S. Bhatti of Delphi
Harrison Thermal Systems for reviewing Chapters 1 through 6 and Dr. T. Skiepko of
Bialystok Technical University for reviewing Chapter 5 and providing constructive
suggestions. The constructive feedback over a period of time provided by many students
(too numerous to mention by name) merits a special word of appreciation.
Finally, we must acknowledge the roles played by our wives, Rekha and Gorana, and
our children, Nilay and Nirav Shah and Vis nja and Aleksandar Sekulic, during the
course of preparation of this book. Their loving care, emotional support, assistance,
and understanding provided continuing motivation to compete the book.
We welcome suggestions and comments from readers.
Ramesh K. Shah
Dus an P. Sekulic
NOMENCLATURE
xix
NOMENCLATURE
The dimensions for each symbol are represented in both the SI and English systems of
units, where applicable. Note that both the hour and second are commonly used as units
for time in the English system of units; hence a conversion factor of 3600 should be
employed at appropriate places in dimensionless groups.
A
Ac
Aeff
Af
Afr
Afr;t
Afr;w
Ah
Ak
Ak
A*k
Ao
Ao;bp
Ao;cr
Ao;sb
Ao;tb
Ao;w
Ap
Aw
a
a
{
total heat transfer surface area (both primary and secondary, if any) on one
side of a direct transfer type exchanger (recuperator), total heat transfer
surface area of all matrices of a regenerator,{ m2 , ft2
total heat transfer area (both primary and secondary, if any) on the cold side
of an exchanger, m2 , ft2
eective surface area on one side of an extended surface exchanger [dened by
Eq. (4.167)], m2 , ft2
n or extended surface area on one side of the exchanger, m2 , ft2
frontal or face area on one side of an exchanger, m2 , ft2
window area occupied by tubes, m2 , ft2
gross (total) window area, m2 , ft2
total heat transfer surface area (both primary and secondary, if any) on the
hot uid side of an exchanger, m2 , ft2
n cross-sectional area for heat conduction in Section 4.3 (Ak;o is Ak at the
n base), m2 , ft2
total wall cross-sectional area for longitudinal conduction [additional
subscripts c, h, and t, if present, denote cold side, hot side, and total (hot
cold) for a regenerator] in Section 5.4, m2 , ft2
ratio of Ak on the Cmin side to that on the Cmax side [see Eq. (5.117)],
dimensionless
minimum free-ow (or open) area on one uid side of an exchanger, heat
transfer surface area on tube outside in a tubular exchanger in Chapter 13
only, m2 , ft2
ow bypass area of one bae, m2 , ft2
ow area at or near the shell centerline for one crossow section in a shell-andtube exchanger, m2 , ft2
shell-to-bae leakage ow area, m2 , ft2
tube-to-bae leakage ow area, m2 , ft2
ow area through window zone, m2 , ft2
primary surface area on one side of an exchanger, m2 , ft2
total wall area for heat conduction from the hot uid to the cold uid, or total
wall area for transverse heat conduction (in the matrix wall thickness direction), m2 , ft2
short side (unless specied) of a rectangular cross section, m, ft
amplitude of chevron plate corrugation (see Fig. 7.28), m, ft
Unless clearly specied, a regenerator in the nomenclature means either a rotary or a xed-matrix regenerator.
xx
B
Bi
b
b
c
C
C
C
C
C*
C
CD
Cmax
Cmin
Cms
Cr
C*r
Cr
C*r
CUA
Cus
Cw
Cw
Cw*
CF
c
c
cp
cw
d
Dbaffle
Dctl
Dh
{
NOMENCLATURE
NOMENCLATURE
Dh;w
Dotl
Dp
Ds
d
dc
de
di
do
dw
d1
e_
E
E
E
Eu
e
e
F
f
fD
ftb
G
Gr
Gz
Gzx
g
gc
H
H
_
H
Hg
H
*
xxi
xxii
H1
*
H2
*
NOMENCLATURE
he
hg
I_irr
In
ij
Ji
correction factors for the shell-side heat transfer coecient for the Bell
Delaware method [see Eq. (9.50)]; i c for bae cut and spacing; i for
bae leakage eects, including both shell-to-bae and tube-to-bae leakage; i b for the bundle bypass ow (C and F streams); i s for variable
bae spacing in the inlet and outlet sections; i r for adverse temperature
gradient buildup in laminar ow, dimensionless
pressure loss coecient, p=u2m =2gc ; subscripts: b for a circular bend, s for
a miter bend, and v for a screwed valve in Chapter 6, and br for branches in
Chapter 12, dimensionless
K1
incremental pressure drop number for fully developed ow (see Table 7.2 for
the denition), dimensionless
Kc
Ke
Kn
kf
kw
Lf
Lh
Lp
L1
L2
L3
Lq
NOMENCLATURE
c
ef
s
*
c*
m
MA
Mw
m
m
m_
m_ n
N
N
Nb
Nc
Nf
Np
Np
Np0
Nr
Nr;c
Nr;cc
Nr;cw
Nt
Nt;b
xxiii
xxiv
Nt;c
Nt; p
Nt;w
Nt0
NTU
NTU1
NTUc
NTUh
NTUo
NTU*
Nu
n, np
nc
nf
nt
ntuc
ntu*cost
ntuh
P
P
P
}
Pc
Ph
Pr
Pt
Pe
Pr
p
{
NOMENCLATURE
NOMENCLATURE
xxv
p*
ratio of cold-uid inlet pressure to hot-uid inlet pressure, pc;i =ph;i , dimensionless
pd
pf
n pitch, 1=Nf , m, ft
pt
p
uid static pressure drop on one uid side of a heat exchanger core [see
Eq. (6.28)], Pa, psf (psi)
p*
pb
uid static pressure drop associated with a pipe bend, Pa, psf (psi)
pb;i
uid static pressure drop associated with an ideal crossow section between
two baes, Pa, psf (psi)
pc
uid static pressure drop associated with the tube bundle central section
(crossow zone) between bae tips, Pa, psf (psi)
pgain
ps
pw;i
uid static pressure drop associated with an ideal window section, Pa, psf (psi)
q
q*
q0
q 00
heat ux, heat transfer rate per unit surface area, q=A, W/m2, Btu/hr-ft2
qe
q0
qmax
heat capacity rate ratio [dened by Eqs. (3.105) and (3.106)], dimensionless
^
R
R*
ratio of thermal resistances on the Cmin to Cmax side, 1=o hA*; it is also the
same as the ratio of hot to cold reduced periods, h =c , Chapter 5, dimensionless
xxvi
R*
R~
^f
R
Ri
Ra
Re
Red
Redc
Reo
r
rc
rf
rh
ri
S
S*
S_irr
St
s
s
s
T
T
*
Tc;o
Th;o
T
Tm
NOMENCLATURE
total thermal resistance (wall, fouling, and convective) on the enhanced (or
plain with subscript p) outside surface side normalized with respect to the
thermal resistance 1=hAi; p of inside plain tube/surface (see Table 10.5
for explicit formulas), dimensionless
gas constant for a particular gas, R/m, J=kg K, 1bf-ft=1bm-8R
fouling factor or unit thermal resistance (fouling resistance), 1=hf ,
m2 K=W, hr-ft2-8F/Btu
pressure drop correction factor for the BellDelaware method, where i b for
bundle bypass ow eects (C stream), i for bae leakage eects (A and
E streams), i s for unequal inlet/outlet bae spacing eects, dimensionless
Rayleigh number [dened by Eq. (7.160)], dimensionless
Reynolds number based on the hydraulic diameter, GDh = , dimensionless
Reynolds number based on the tube outside diameter and mean velocity,
um do = , dimensionless
Reynolds number based on the collar diameter and mean velocity, um dc = ,
dimensionless
Reynolds number based on the tube outside diameter and free stream
(approach or core upstream) velocity, u1 do = , dimensionless
radial coordinate in the cylindrical coordinate system, m, ft
radius of curvature of a tube bend (see Fig. 6.5), m, ft
^ f 1=hf f =kf , m2 K=W, hr-ft2fouling factor or fouling resistance rf R
8F/Btu
hydraulic radius, Ao L=A or Dh =4, m, ft
tube inside radius, m, ft
entropy, J/K, Btu/8R
normalized entropy generation rate, S_ irr =C2 or S_irr =Cmax , dimensionless
entropy generation rate, W/K, Btu/hr-8R
Stanton number, h=Gcp , Sto U=Gcp , dimensionless
specic entropy in Chapter 11, J=kg K, Btu/lbm-8R
complex Laplace independent variable with Laplace transforms only in
Chapter 11, dimensionless
spacing between adjacent ns, pf , m, ft
uid static temperature to a specied arbitrary datum, except for Eqs. (7.157)
and (7.158) and in Chapter 11 where it is dened on an absolute temperature
scale, 8C, 8F
thermal boundary condition referring to constant wall temperature, both
axially and peripherally
ow area average cold-uid outlet temperature unless otherwise specied, 8C,
8F
ow area average hot-uid outlet temperature unless otherwise specied, 8C,
8F
temperature of the n tip, 8C, 8F
uid bulk mean temperature, 8C, 8F
NOMENCLATURE
Ts
Tw
T1
T*
Tc*
Th*
Tw*
T0
T
Tc
Th
Tlm
Tm
Tmax
U, Um
u, um
uc
ucr
uz , uw
u1
u*
V
V*
V_
Vm
Vp
Vv
v
W
wp
X*
xxvii
xxviii
Xd
Xd*
X
X*
Xt
Xt*
x
x
x*
xf
y
Z
z
w
*
f*
*
@,
b
bb
c
f
h
NOMENCLATURE
NOMENCLATURE
xxix
otl
s
leakage and bypass stream correction factor to the true mean temperature
dierence for the stream analysis method [dened by Eq. (4.170)], dimensionless
sb
tb
t
v
w
"
heat exchanger eectiveness [dened by Eq. (3.37) or (3.44) and Table 11.1];
represents an overall exchanger eectiveness for a multipass unit, dimensionless
"c
temperature eectiveness of the cold uid [dened by Eq. (3.52)], also as the
exchanger eectiveness of the cold uid in Appendix B, dimensionless
"cf
"h
temperature eectiveness of the hot uid [dened by Eq. (3.51)], also as the
exchange eectiveness of the hot uid in Appendix B, dimensionless
"h;o
"p
"r
"*
correction factors for shellside pressure drop terms for the BellDelaware
method [see Eq. (9.51)]; i for tube-to-bae and bae-to-shell leakage;
i b for bypass ow; i s for inlet and outlet sections, dimensionless
reduced time variable for a regenerator [dened by Eq. (5.69)] with subscripts
j c and h for cold- and hot-gas ow periods, dimensionless
f
o
extended surface eciency on one uid side of the extended surface heat
exchanger [see Eqs. (4.158) and (4.160) for the denition], dimensionless
o hA*
p
"
xxx
b
b
c
h
r
t
#
#
#*
T
m
*
,
m
d
d;min
s
w
*
c*, h*
NOMENCLATURE
angle between two radii intersected at the inside shell wall with the bae cut
(see Fig. 8.9), rad unless explicitly mentioned in degrees
bend deection angle (see Fig. 6.5), deg
disk sector angle for the cold-uid stream in a rotary regenerator, rad, deg
disk sector angle for the hot-uid stream in a rotary regenerator, rad, deg
disk sector angle covered by the radial seals in a rotary regenerator, rad, deg
h c r 2
3608, rad, deg
uid temperature for internal ow in Chapter 7, T Tw;m =Tm Tw;m ,
dimensionless
ratio of uid inlet temperatures, T1;i =T2;i in Chapter 11 where temperatures
are on absolute temperature scale, dimensionless
uid
temperature
for
external
ow,
T Tw =T1 Tw
or
T Tw =Te Tw , dimensionless
length eect correction factor for the overall heat transfer coecient [see Eqs.
(4.32) and (4.33)] dimensionless
isothermal compressibility, 1/Pa, ft2/lbf
reduced length for a regenerator [dened by Eqs. (5.84), (5.102), and (5.103)],
dimensionless
mean reduced length [dened by Eq. (5.91)], dimensionless
h =c , dimensionless
wavelength of chevron plate corrugation (see Fig. 7.28), m, ft
longitudinal wall conduction parameter based on the total conduction area,
kw Ak;t =Cmin L, c kw Ak;c =Cc Lc , h kw Aw;h =Ch Lh , dimensionless
uid dynamic viscosity, Pa s, 1bm/hr-ft
uid kinematic viscosity =, m2/s, ft2/sec
reduced length variable for regenerator [dened by Eq. (5.69], dimensionless
axial coordinate in Chapter 11, x=L, dimensionless
reduced period for a regenerator [dened by Eqs. (5.84), (5.104), and (5.105)],
dimensionless
harmonic mean reduced period [dened by Eq. 5.90)], dimensionless
uid density, kg/m3, 1bm/ft3
ratio of free ow area to frontal area, Ao =Afr , dimensionless
time, s, sec
delay period or induction period associated with initiation of fouling in
Chapter 13; dwell time, residence time, or transit time of a uid particle in a
heat exchanger in Chapter 5, s, sec
dwell time of the Cmin uid, s, sec
uid shear stress, Pa, psf
equivalent uid shear stress at wall, Pa, psf (psi)
time variable, =d;min , dimensionless
time variable for the cold and hot uids [dened by Eq. (5.26)], dimensionless
denotes a functional relationship
axial coordinate, x=L1 NTU, in Chapter 11 only, dimensionless
NOMENCLATURE
i
xxxi
Subscripts
A
unit (row, section) A
a
air side
B
unit (row, section) B
b
bend, tube bundle, or lateral branch
c
cold-uid side, clean surface in Chapter 13
cf
counterow
cp
constant properties
cr
crossow section in a segmental baed exchanger
cv
control volume
cu
cold utility
d
deposit
df
displaced uid
e
eective
f
fouling, uid in Section 7.3.3.2
g
gas side
H
constant axial wall heat ux boundary condition
h
hot-uid side
hu
hot utility
hex
heat exchanger
H1
thermal boundary condition referring to constant axial wall heat ux with
constant peripheral wall temperature
i
inlet to the exchanger
i
inside surface in Chapter 13
id
ideal
iso
isothermal
L
coupled liquid
leak
caused by a leak
lm
logarithmic mean
m
mean or bulk mean, manifold (in Chapter 12)
max
maximum
min
minimum
mixing
caused by mixing
ms
maldistributed uid
n
nominal or reference passage in Chapter 12
o
overall
o
outside surface in Chapter 13
xxxii
NOMENCLATURE
opt
optimal
otl
pf
parallelow
reentrainment
ref
std
tubeside, tube
tot
total
viscous
water
free stream
A variety of heat exchangers are used in industry and in their products. The objective of
this chapter is to describe most of these heat exchangers in some detail using classication
schemes. Starting with a denition, heat exchangers are classied according to transfer
processes, number of uids, degree of surface compactness, construction features, ow
arrangements, and heat transfer mechanisms. With a detailed classication in each category, the terminology associated with a variety of these exchangers is introduced and
practical applications are outlined. A brief mention is also made of the dierences in
design procedure for the various types of heat exchangers.
1.1
INTRODUCTION
A heat exchanger is a device that is used to transfer thermal energy (enthalpy) between
two or more uids, between a solid surface and a uid, or between solid particulates and
a uid, at dierent temperatures and in thermal contact. In heat exchangers, there are
usually no external heat and work interactions. Typical applications involve heating or
cooling of a uid stream of concern and evaporation or condensation of single- or
multicomponent uid streams. In other applications, the objective may be to recover
or reject heat, or sterilize, pasteurize, fractionate, distill, concentrate, crystallize, or control a process uid. In a few heat exchangers, the uids exchanging heat are in direct
contact. In most heat exchangers, heat transfer between uids takes place through a
separating wall or into and out of a wall in a transient manner. In many heat exchangers,
the uids are separated by a heat transfer surface, and ideally they do not mix or leak.
Such exchangers are referred to as direct transfer type, or simply recuperators. In contrast, exchangers in which there is intermittent heat exchange between the hot and cold
uidsvia thermal energy storage and release through the exchanger surface or matrix
are referred to as indirect transfer type, or simply regenerators. Such exchangers usually
have uid leakage from one uid stream to the other, due to pressure dierences and
matrix rotation/valve switching. Common examples of heat exchangers are shell-andtube exchangers, automobile radiators, condensers, evaporators, air preheaters, and
cooling towers. If no phase change occurs in any of the uids in the exchanger, it is
sometimes referred to as a sensible heat exchanger. There could be internal thermal
energy sources in the exchangers, such as in electric heaters and nuclear fuel elements.
Combustion and chemical reaction may take place within the exchanger, such as in
boilers, red heaters, and uidized-bed exchangers. Mechanical devices may be used in
some exchangers such as in scraped surface exchangers, agitated vessels, and stirred tank
reactors. Heat transfer in the separating wall of a recuperator generally takes place by
Fundamentals of Heat Exchanger Design. Ramesh K. Shah and Duan P. Sekulic
Copyright 2003 John Wiley & Sons, Inc.
conduction. However, in a heat pipe heat exchanger, the heat pipe not only acts as a
separating wall, but also facilitates the transfer of heat by condensation, evaporation,
and conduction of the working uid inside the heat pipe. In general, if the uids are
immiscible, the separating wall may be eliminated, and the interface between the uids
replaces a heat transfer surface, as in a direct-contact heat exchanger.
A heat exchanger consists of heat transfer elements such as a core or matrix containing
the heat transfer surface, and uid distribution elements such as headers, manifolds,
tanks, inlet and outlet nozzles or pipes, or seals. Usually, there are no moving parts in
a heat exchanger; however, there are exceptions, such as a rotary regenerative exchanger
(in which the matrix is mechanically driven to rotate at some design speed) or a scraped
surface heat exchanger.
The heat transfer surface is a surface of the exchanger core that is in direct contact
with uids and through which heat is transferred by conduction. That portion of the
surface that is in direct contact with both the hot and cold uids and transfers heat
between them is referred to as the primary or direct surface. To increase the heat transfer
area, appendages may be intimately connected to the primary surface to provide an
extended, secondary, or indirect surface. These extended surface elements are referred
to as ns. Thus, heat is conducted through the n and convected (and/or radiated) from
the n (through the surface area) to the surrounding uid, or vice versa, depending on
whether the n is being cooled or heated. As a result, the addition of ns to the primary
surface reduces the thermal resistance on that side and thereby increases the total heat
transfer from the surface for the same temperature dierence. Fins may form ow
passages for the individual uids but do not separate the two (or more) uids of the
exchanger. These secondary surfaces or ns may also be introduced primarily for structural strength purposes or to provide thorough mixing of a highly viscous liquid.
Not only are heat exchangers often used in the process, power, petroleum, transportation, air-conditioning, refrigeration, cryogenic, heat recovery, alternative fuel, and
manufacturing industries, they also serve as key components of many industrial products
available in the marketplace. These exchangers can be classied in many dierent ways.
We will classify them according to transfer processes, number of uids, and heat transfer
mechanisms. Conventional heat exchangers are further classied according to construction type and ow arrangements. Another arbitrary classication can be made, based on
the heat transfer surface area/volume ratio, into compact and noncompact heat exchangers. This classication is made because the type of equipment, elds of applications, and
design techniques generally dier. All these classications are summarized in Fig. 1.1 and
discussed further in this chapter. Heat exchangers can also be classied according to the
process function, as outlined in Fig. 1.2. However, they are not discussed here and the
reader may refer to Shah and Mueller (1988). Additional ways to classify heat exchangers
are by uid type (gasgas, gasliquid, liquidliquid, gas two-phase, liquid two-phase,
etc.), industry, and so on, but we do not cover such classications in this chapter.
1.2
Heat exchangers are classied according to transfer processes into indirect- and directcontact types.
1.2.1
In an indirect-contact heat exchanger, the uid streams remain separate and the heat
transfers continuously through an impervious dividing wall or into and out of a wall in a
transient manner. Thus, ideally, there is no direct contact between thermally interacting
uids. This type of heat exchanger, also referred to as a surface heat exchanger, can be
further classied into direct-transfer type, storage type, and uidized-bed exchangers.
FIGURE 1.2 (a) Classication according to process function; (b) classication of condensers;
(c) classication of liquid-to-vapor phase-change exchangers.
FIGURE 1.2 (d) classication of chemical evaporators according to (i) the type of construction,
and (ii) how energy is supplied (Shah and Mueller, 1988); (e) classication of reboilers.
and-tube and plate heat exchangers, although they are also considered as recuperators.
Recuperators are further subclassied as prime surface exchangers and extended-surface
exchangers. Prime surface exchangers do not employ ns or extended surfaces on any
uid side. Plain tubular exchangers, shell-and-tube exchangers with plain tubes, and
plate exchangers are good examples of prime surface exchangers. Recuperators constitute a vast majority of all heat exchangers.
1.2.1.2 Storage Type Exchangers. In a storage type exchanger, both uids ow alternatively through the same ow passages, and hence heat transfer is intermittent. The
heat transfer surface (or ow passages) is generally cellular in structure and is referred to
as a matrix (see Fig. 1.43), or it is a permeable (porous) solid material, referred to as a
packed bed. When hot gas ows over the heat transfer surface (through ow passages),
the thermal energy from the hot gas is stored in the matrix wall, and thus the hot gas is
being cooled during the matrix heating period. As cold gas ows through the same
passages later (i.e., during the matrix cooling period), the matrix wall gives up thermal
energy, which is absorbed by the cold uid. Thus, heat is not transferred continuously
through the wall as in a direct-transfer type exchanger (recuperator), but the corresponding thermal energy is alternately stored and released by the matrix wall. This
storage type heat exchanger is also referred to as a regenerative heat exchanger, or
simply as a regenerator.{ To operate continuously and within a desired temperature
range, the gases, headers, or matrices are switched periodically (i.e., rotated), so that
the same passage is occupied periodically by hot and cold gases, as described further in
Section 1.5.4. The actual time that hot gas takes to ow through a cold regenerator
matrix is called the hot period or hot blow, and the time that cold gas ows through the
hot regenerator matrix is called the cold period or cold blow. For successful operation, it
is not necessary to have hot- and cold-gas ow periods of equal duration. There is some
unavoidable carryover of a small fraction of the uid trapped in the passage to the other
uid stream just after switching of the uids; this is referred to as carryover leakage. In
addition, if the hot and cold uids are at dierent pressures, there will be leakage from
the high-pressure uid to the low-pressure uid past the radial, peripheral, and axial
seals, or across the valves. This leakage is referred to as pressure leakage. Since these
leaks are unavoidable, regenerators are used exclusively in gas-to-gas heat (and mass)
transfer applications with sensible heat transfer; in some applications, regenerators may
transfer moisture from humid air to dry air up to about 5%.
For heat transfer analysis of regenerators, the "-NTU method of recuperators needs
to be modied to take into account the thermal energy storage capacity of the matrix. We
discuss the design theory of regenerators in detail in Chapter 5.
1.2.1.3 Fluidized-Bed Heat Exchangers. In a uidized-bed heat exchanger, one side of
a two-uid exchanger is immersed in a bed of nely divided solid material, such as a
tube bundle immersed in a bed of sand or coal particles, as shown in Fig. 1.3. If the
upward uid velocity on the bed side is low, the solid particles will remain xed in
position in the bed and the uid will ow through the interstices of the bed. If the
upward uid velocity is high, the solid particles will be carried away with the uid. At a
proper value of the uid velocity, the upward drag force is slightly higher than the
weight of the bed particles. As a result, the solid particles will oat with an increase in
bed volume, and the bed behaves as a liquid. This characteristic of the bed is referred to
as a uidized condition. Under this condition, the uid pressure drop through the bed
remains almost constant, independent of the ow rate, and a strong mixing of the solid
particles occurs. This results in a uniform temperature for the total bed (gas and particles) with an apparent thermal conductivity of the solid particles as innity. Very high
heat transfer coecients are achieved on the uidized side compared to particle-free or
dilute-phase particle gas ows. Chemical reaction is common on the uidized side in
many process applications, and combustion takes place in coal combustion uidized
beds. The common applications of the uidized-bed heat exchanger are drying, mixing,
adsorption, reactor engineering, coal combustion, and waste heat recovery. Since the
{
Regenerators are also used for storing thermal energy for later use, as in the storage of thermal energy. Here the
objective is how to store the maximum fraction of the input energy and minimize heat leakage. However, we do not
concentrate on this application in this book.
In a direct-contact exchanger, two uid streams come into direct contact, exchange heat,
and are then separated. Common applications of a direct-contact exchanger involve mass
transfer in addition to heat transfer, such as in evaporative cooling and rectication;
applications involving only sensible heat transfer are rare. The enthalpy of phase change
in such an exchanger generally represents a signicant portion of the total energy transfer. The phase change generally enhances the heat transfer rate. Compared to indirectcontact recuperators and regenerators, in direct-contact heat exchangers, (1) very high
heat transfer rates are achievable, (2) the exchanger construction is relatively inexpensive,
and (3) the fouling problem is generally nonexistent, due to the absence of a heat transfer
surface (wall) between the two uids. However, the applications are limited to those cases
where a direct contact of two uid streams is permissible. The design theory for these
{
Th;i , inlet temperature of the hot uid to the uidized bed; Tf ;i , temperature of the uidized bed itself at the inlet.
exchangers is beyond the scope of this book and is not covered. These exchangers may be
further classied as follows.
1.2.2.1 Immiscible Fluid Exchangers. In this type, two immiscible uid streams are
brought into direct contact. These uids may be single-phase uids, or they may involve
condensation or vaporization. Condensation of organic vapors and oil vapors with
water or air are typical examples.
1.2.2.2 GasLiquid Exchangers. In this type, one uid is a gas (more commonly, air)
and the other a low-pressure liquid (more commonly, water) and are readily separable
after the energy exchange. In either cooling of liquid (water) or humidication of gas
(air) applications, liquid partially evaporates and the vapor is carried away with the gas.
In these exchangers, more than 90% of the energy transfer is by virtue of mass transfer
(due to the evaporation of the liquid), and convective heat transfer is a minor mechanism. A wet (water) cooling tower with forced- or natural-draft airow is the most
common application. Other applications are the air-conditioning spray chamber, spray
drier, spray tower, and spray pond.
1.2.2.3 LiquidVapor Exchangers. In this type, typically steam is partially or fully
condensed using cooling water, or water is heated with waste steam through direct
contact in the exchanger. Noncondensables and residual steam and hot water are the
outlet streams. Common examples are desuperheaters and open feedwater heaters (also
known as deaeraters) in power plants.
1.3
Most processes of heating, cooling, heat recovery, and heat rejection involve transfer of
heat between two uids. Hence, two-uid heat exchangers are the most common. Threeuid heat exchangers are widely used in cryogenics and some chemical processes (e.g., air
separation systems, a heliumair separation unit, purication and liquefaction of hydrogen, ammonia gas synthesis). Heat exchangers with as many as 12 uid streams have been
used in some chemical process applications. The design theory of three- and multiuid
heat exchangers is algebraically very complex and is not covered in this book.
Exclusively, only the design theory for two-uid exchangers and some associated
problems are presented in this book.
1.4
(213 ft2/ft3){ or a hydraulic diameter Dh 6 mm (14 in.) for operating in a gas stream and
400 m2 /m3 (122 ft2 /ft3 ) or higher for operating in a liquid or phase-change stream. A
laminar ow heat exchanger (also referred to as a meso heat exchanger) has a surface
area density greater than about 3000 m2 /m3 (914 ft2 /ft3 ) or 100 mm Dh 1 mm. The
term micro heat exchanger is used if the surface area density is greater than about
15,000 m2 /m3 (4570 ft2 /ft3 ) or 1 mm Dh 100 mm. A liquid/two-phase uid heat
exchanger is referred to as a compact heat exchanger if the surface area density on any
one uid side is greater than about 400 m2 /m3 . In contrast, a typical process industry shelland-tube exchanger has a surface area density of less than 100 m2 /m3 on one uid side with
plain tubes, and two to three times greater than that with high-n-density low-nned
tubing. A typical plate heat exchanger has about twice the average heat transfer coecient
h on one uid side or the average overall heat transfer coecient U than that for a shelland-tube exchanger for water/water applications. A compact heat exchanger is not necessarily of small bulk and mass. However, if it did not incorporate a surface of high-surfacearea density, it would be much more bulky and massive. Plate-n, tube-n, and rotary
regenerators are examples of compact heat exchangers for gas ow on one or both uid
sides, and gasketed, welded, brazed plate heat exchangers and printed-circuit heat exchangers are examples of compact heat exchangers for liquid ows. Basic ow arrangements of
two-uid compact heat exchangers are single-pass crossow, counterow, and multipass
cross-counterow (see Section 1.6 for details); for noncompact heat exchangers, many
other ow arrangements are also used. The aforementioned last two ow arrangements for
compact or noncompact heat exchangers can yield a very high exchanger eectiveness
value or a very small temperature approach (see Section 3.2.3 for the denition) between
uid streams.
A spectrum of surface area density of heat exchanger surfaces is shown in Fig. 1.4. On
the bottom of the gure, two scales are shown: the heat transfer surface area density
(m2 /m3 ) and the hydraulic diameter Dh ,{ (mm), which is the tube inside or outside
diameter D (mm) for a thin-walled circular tube. Dierent heat exchanger surfaces are
shown in the rectangles. When projected on the (or Dh ) scale, the short vertical sides of
a rectangle indicate the range of surface area density (or hydraulic diameter) for the
particular surface in question. What is referred to as in this gure is either 1 or 2 ,
dened as follows. For plate heat exchangers, plate-n exchangers, and regenerators,
1
Ah
Vh
or
Ac
Vc
1:1
Ah
Vtotal
or
Ac
Vtotal
1:2
The unit conversion throughout the book may not be exact; it depends on whether the number is exact or is an
engineering value.
{
The hydraulic diameter is dened as 4Ao =P, where Ao is the minimum free-ow area on one uid side of a heat
exchanger and P is the wetted perimeter of ow passages of that side. Note that the wetted perimeter can be
dierent for heat transfer and pressure drop calculations. For example, the hydraulic diameter for an annulus of a
double-pipe heat exchanger for q and p calculations is as follows.
Dh;q
Di
Di
Dh;p
4=4D2o D2i
Do Di
Do Di
where Do is the inside diameter of the outer pipe and Di is the outside diameter of the inside pipe of a double-pipe
exchanger. See also Eq. (3.65) for a more precise denition of the hydraulic diameter.
10
FIGURE 1.4 Heat transfer surface area density spectrum of exchanger surfaces (Shah, 1981).
Here A is the heat transfer surface area, V the exchanger volume, and the subscripts h and
c denote hot and cold uid sides, respectively. Vh and Vc are the volumes individually
occupied by the hot- and cold-uid-side heat transfer surfaces. From this point on in the
book, 1 is simply designated as and 2 is designated as :
1
2
1:3
Note that both and (with the denitions noted above) are used in dening the surface
area densities of a plate-n surface; however, only is used in dening the surface area
density of a tube-n surface since has no meaning in this case. The following specic
values are used in preparing Fig. 1.4:
. For a shell-and-tube exchanger, an inline arrangement{ is considered with
Xt*Xl* 1:88.
. For plate and plate-n exchangers, the porosity between plates is taken as 0.8333;
and for a regenerator, the porosity of matrix surface is taken as 0.8333. With these
values, (m2 /m3 ) and Dh (mm) are related as 3333=Dh .
The tube array is idealized as innitely large with thin-walled circular tubes. Xt* and Xl* are the transverse and
longitudinal tube pitches normalized with respect to the tube outside diameter. Refer to Table 8.1 for the denitions of tube pitches.
11
Note that some industries quote the total surface area (of hot- and cold-uid sides) in
their exchanger specications. However, in calculations of heat exchanger design, we
need individual uid-side heat transfer surface areas; and hence we use here the denitions of and as given above.
Based on the foregoing denition of a compact surface, a tube bundle having 5 mm
(0.2 in.) diameter tubes in a shell-and-tube exchanger comes close to qualifying as a
compact exchanger. As or varies inversely with the tube diameter, the 25.4 mm
(1 in.) diameter tubes used in a power plant condenser result in a noncompact exchanger.
In contrast, a 1990s automobile radiator [790 ns/m (20 ns/in.)] has a surface area
density on the order of 1870 m2 /m3 (570 ft2 /ft3 ) on the air side, which is equivalent
to 1.8 mm (0.07 in.) diameter tubes. The regenerators in some vehicular gas turbine
engines under development have matrices with an area density on the order of
6600 m2 /m3 (2000 ft2 /ft3 ), which is equivalent to 0.5 mm (0.02 in.) diameter tubes in a
bundle. Human lungs are one of the most compact heat-and-mass exchangers, having a
surface area density of about 17,500 m2 /m3 (5330 ft2 /ft3 ), which is equivalent to 0.19 mm
(0.0075 in.) diameter tubes. Some micro heat exchangers under development are as
compact as the human lung (Shah, 1991a) and also even more compact.
The motivation for using compact surfaces is to gain specied heat exchanger performance, q=Tm , within acceptably low mass and box volume constraints. The heat
exchanger performance may be expressed as
q
UA UV
Tm
1:4
where q is the heat transfer rate, Tm the true mean temperature dierence, and U the
overall heat transfer coecient. Clearly, a high value minimizes exchanger volume V
for specied q=Tm . As explained in Section 7.4.1.1, compact surfaces (having small Dh )
generally result in a higher heat transfer coecient and a higher overall heat transfer
coecient U, resulting in a smaller volume. As compact surfaces can achieve structural
stability and strength with thinner-gauge material, the gain in a lower exchanger mass is
even more pronounced than the gain in a smaller volume.
1.4.1
Gas-to-Fluid Exchangers
The heat transfer coecient h for gases is generally one or two orders of magnitude lower
than that for water, oil, and other liquids. Now, to minimize the size and weight of a gasto-liquid heat exchanger, the thermal conductances (hA products) on both sides of the
exchanger should be approximately the same. Hence, the heat transfer surface on the gas
side needs to have a much larger area and be more compact than can be realized practically with the circular tubes commonly used in shell-and-tube exchangers. Thus, for an
approximately balanced design (about the same hA values), a compact surface is
employed on the gas side of gas-to-gas, gas-to-liquid, and gas-to-phase change heat
exchangers.
The unique characteristics of compact extended-surface (plate-n and tube-n)
exchangers, compared to conventional shell-and-tube exchangers (see Fig. 1.6), are as
follows:
. Availability of numerous surfaces having dierent orders of magnitude of surface
area density
12
. Flexibility in distributing surface area on the hot and cold sides as warranted by
design considerations
. Generally, substantial cost, weight, or volume savings.
The important design and operating considerations for compact extended-surface
exchangers are as follows:
. Usually, at least one of the uids is a gas having a low h value.
. Fluids must be clean and relatively noncorrosive because of low-Dh ow passages
and no easy techniques for cleaning.
. The uid pumping power (and hence the pressure drop) is often as important as the
heat transfer rate.
. Operating pressures and temperatures are somewhat limited compared to shelland-tube exchangers, due to joining of the ns to plates or tubes by brazing,
mechanical expansion, and so on.
. With the use of highly compact surfaces, the resulting shape of the exchanger is one
having a large frontal area and a short ow length; the header design of a compact
heat exchanger is thus important for achieving uniform ow distribution among
very large numbers of small ow passages.
. The market potential must be large enough to warrant the sizable initial manufacturing tooling and equipment costs.
Fouling is a major potential problem in compact heat exchangers (except for plateand-frame heat exchangers), particularly those having a variety of n geometries or very
ne circular or noncircular ow passages that cannot be cleaned mechanically. Chemical
cleaning may be possible; thermal baking and subsequent rinsing are possible for small
units.{ Hence, extended-surface compact heat exchangers may not be used in heavy
fouling applications. Nonfouling uids are used where permissible, such as clean air or
gases, light hydrocarbons, and refrigerants.
1.4.2
1.5
Heat exchangers are frequently characterized by construction features. Four major construction types are tubular, plate-type, extended surface, and regenerative exchangers.
Heat exchangers with other constructions are also available, such as scraped surface
exchanger, tank heater, cooler cartridge exchanger, and others (Walker, 1990). Some
of these may be classied as tubular exchangers, but they have some unique features
compared to conventional tubular exchangers. Since the applications of these exchangers
{
Some additional techniques for cleaning and mitigation of fouling are summarized in Section 13.4.
13
are specialized, we concentrate here only on the four major construction types noted
above.
Although the "-NTU and MTD methods (see end of Section 3.2.2) are identical for
tubular, plate-type, and extended-surface exchangers, the inuence of the following
factors must be taken into account in exchanger design: corrections due to leakage
and bypass streams in a shell-and-tube exchanger, eects due to a few plates in a plate
exchanger, and n eciency in an extended-surface exchanger. Similarly, the "-NTU
method must be modied to take into account the heat capacity of the matrix in a
regenerator. Thus, the detailed design theory diers for each construction type and is
discussed in detail in Chapters 3 through 5. Let us rst discuss the construction features
of the four major types.
1.5.1
These exchangers are generally built of circular tubes, although elliptical, rectangular, or
round/at twisted tubes have also been used in some applications. There is considerable
exibility in the design because the core geometry can be varied easily by changing the
tube diameter, length, and arrangement. Tubular exchangers can be designed for high
pressures relative to the environment and high-pressure dierences between the uids.
Tubular exchangers are used primarily for liquid-to-liquid and liquid-to-phase change
(condensing or evaporating) heat transfer applications. They are used for gas-to-liquid
and gas-to-gas heat transfer applications primarily when the operating temperature and/
or pressure is very high or fouling is a severe problem on at least one uid side and no
other types of exchangers would work. These exchangers may be classied as shell-andtube, double-pipe, and spiral tube exchangers. They are all prime surface exchangers
except for exchangers having ns outside/inside tubes.
1.5.1.1 Shell-and-Tube Exchangers. This exchanger, shown in Fig. 1.5, is generally
built of a bundle of round tubes mounted in a cylindrical shell with the tube axis parallel
to that of the shell. One uid ows inside the tubes, the other ows across and along the
tubes. The major components of this exchanger are tubes (or tube bundle), shell, frontend head, rear-end head, baes, and tubesheets, and are described briey later in this
subsection. For further details, refer to Section 10.2.1.
A variety of dierent internal constructions are used in shell-and-tube exchangers,
depending on the desired heat transfer and pressure drop performance and the methods
employed to reduce thermal stresses, to prevent leakages, to provide for ease of cleaning,
to contain operating pressures and temperatures, to control corrosion, to accommodate
highly asymmetric ows, and so on. Shell-and-tube exchangers are classied and constructed in accordance with the widely used TEMA (Tubular Exchanger Manufacturers
Association) standards (TEMA, 1999), DIN and other standards in Europe and elsewhere, and ASME (American Society of Mechanical Engineers) boiler and pressure
vessel codes. TEMA has developed a notation system to designate major types of
shell-and-tube exchangers. In this system, each exchanger is designated by a three-letter
combination, the rst letter indicating the front-end head type, the second the shell
type, and the third the rear-end head type. These are identied in Fig. 1.6. Some common
shell-and-tube exchangers are AES, BEM, AEP, CFU, AKT, and AJW. It should be
emphasized that there are other special types of shell-and-tube exchangers commercially
available that have front- and rear-end heads dierent from those in Fig. 1.6. Those
exchangers may not be identiable by the TEMA letter designation.
14
FIGURE 1.5 (a) Shell-and-tube exchanger (BEM) with one shell pass and one tube pass; (b) shelland-tube exchanger (BEU) with one shell pass and two tube passes.
The three most common types of shell-and-tube exchangers are (1) xed tubesheet
design, (2) U-tube design, and (3) oating-head type. In all three types, the front-end
head is stationary while the rear-end head can be either stationary or oating (see Fig.
1.6), depending on the thermal stresses in the shell, tube, or tubesheet, due to temperature
dierences as a result of heat transfer.
The exchangers are built in accordance with three mechanical standards that specify
design, fabrication, and materials of unred shell-and-tube heat exchangers. Class R is
for the generally severe requirements of petroleum and related processing applications.
Class C is for generally moderate requirements for commercial and general process
applications. Class B is for chemical process service. The exchangers are built to comply
with the applicable ASME Boiler and Pressure Vessel Code, Section VIII (1998), and
other pertinent codes and/or standards. The TEMA standards supplement and dene the
ASME code for heat exchanger applications. In addition, state and local codes applicable
to the plant location must also be met.
The TEMA standards specify the manufacturing tolerances for various mechanical
classes, the range of tube sizes and pitches, baing and support plates, pressure
classication, tubesheet thickness formulas, and so on, and must be consulted for all
these details. In this book, we consider only the TEMA standards where appropriate, but
there are other standards, such as DIN 28 008.
Tubular exchangers are widely used in industry for the following reasons. They are
custom designed for virtually any capacity and operating conditions, such as from high
15
FIGURE 1.6 Standard shell types and front- and rear-end head types (From TEMA, 1999).
vacuum to ultrahigh pressure [over 100 MPa (15,000 psig)], from cryogenics to high
temperatures [about 11008C (20008F)] and any temperature and pressure dierences
between the uids, limited only by the materials of construction. They can be designed
for special operating conditions: vibration, heavy fouling, highly viscous uids, erosion,
corrosion, toxicity, radioactivity, multicomponent mixtures, and so on. They are the
most versatile exchangers, made from a variety of metal and nonmetal materials (such
as graphite, glass, and Teon) and range in size from small [0.1 m2 (1 ft2 )] to supergiant
[over 105 m2 (106 ft2 )] surface area. They are used extensively as process heat exchangers
16
Turbo - EHP
(a)
Turbo - CDI
(b)
FIGURE 1.7 Some enhanced tube geometries used in shell-and-tube exchangers: (a) internally
and externally enhanced evaporator tube; (b) internally and externally enhanced condenser tube.
(Courtesy of Wolverine Tube, Inc., Decatur, AL.)
{
A tube bundle is an assembly of tubes, baes, tubesheets and tie rods, and support plates and longitudinal baes,
if any.
17
FIGURE 1.9 Low-nned tubing. The plain end goes into the tubesheet.
reactive, or toxic uids and potable water, double-wall tubing is used. In most applications, tubes are bare, but when gas or low-heat-transfer coecient liquid is used on the
shell side, low-height ns (low ns) are used on the shell side. Also, special high-uxboiling surfaces employ modied low-n tubing. These are usually integral ns made
from a thick-walled tube, shown in Fig. 1.9. Tubes are drawn, extruded, or welded, and
they are made from metals, plastics, and ceramics, depending on the applications.
Shells. The shell is a container for the shell uid.{ Usually, it is cylindrical in shape with
a circular cross section, although shells of dierent shapes are used in specic applications and in nuclear heat exchangers to conform to the tube bundle shape. The shell is
made from a circular pipe if the shell diameter is less than about 0.6 m (2 ft) and is made
from a metal plate rolled and welded longitudinally for shell diameters greater than
0.6 m (2 ft). Seven types of shell congurations, standardized by TEMA (1999), are E,
F, G, H, J, K, and X, shown in Fig. 1.6. The E shell is the most common, due to its low
cost and simplicity, and has the highest log-mean temperature-dierence correction
factor F (see Section 3.7.2 for the denition). Although the tubes may have single or
multiple passes, there is one pass on the shell side. To increase the mean temperature
dierence and hence exchanger eectiveness, a pure counterow arrangement is desirable for a two-tube-pass exchanger. This is achieved by use of an F shell having a
longitudinal bae and resulting in two shell passes. Split- and divided-ow shells, such
as G, H, and J (see Fig. 1.6), are used for specic applications, such as thermosiphon
boiler, condenser, and shell-side low pressure drops. The K shell is a kettle reboiler used
for pool boiling applications. The X shell is a crossow exchanger and is used for low
pressure drop on the shell side and/or to eliminate the possibility of ow-induced
vibrations. A further description of the various types of shell congurations is provided
in Section 10.2.1.4.
Nozzles. The entrance and exit ports for the shell and tube uids, referred to as nozzles,
are pipes of constant cross section welded to the shell and channels. They are used to
distribute or collect the uid uniformly on the shell and tube sides. Note that they dier
from the nozzle used as a uid metering device or in jet engines, which has a variable
ow area along the ow length.
The uid owing in the tubes is referred to as the tube uid; the uid owing outside the tubes is referred to as the
shell uid.
18
Front- and Rear-End Heads. These are used for entrance and exit of the tube uid; in
many rear-end heads, a provision has been made to take care of tube thermal expansion. The front-end head is stationary, while the rear-end head could be either stationary (allowing for no tube thermal expansion) or oating, depending on the thermal
stresses between the tubes and shell. The major criteria for selection of the front-end
head are cost, maintenance and inspection, hazard due to mixing of shell and tube
uids, and leakage to ambient and operating pressures. The major criteria for selection
of the rear-end head are the allowance for thermal stresses, a provision to remove the
tube bundle for cleaning the shell side, prevention of mixing of tube and shell uids, and
sealing any leakage path for the shell uid to ambient. The design selection criteria for
the front- and rear-end heads of Fig. 1.6 are discussed in Sections 10.2.1.5 and 10.2.1.6.
Baes. Baes may be classied as transverse and longitudinal types. The purpose of
longitudinal baes is to control the overall ow direction of the shell uid such that a
desired overall ow arrangement of the two uid streams is achieved. For example, F,
G, and H shells have longitudinal baes (see Fig. 1.6). Transverse baes may be
classied as plate baes and grid (rod, strip, and other axial-ow) baes. Plate baes
are used to support the tubes during assembly and operation and to direct the uid in
the tube bundle approximately at right angles to the tubes to achieve higher heat
transfer coecients. Plate baes increase the turbulence of the shell uid and minimize
tube-to-tube temperature dierences and thermal stresses due to the crossow. Shown
in Fig. 1.10 are single- and multisegmental baes and disk and doughnut baes.
Single- and double-segmental baes are used most frequently due to their ability to
assist maximum heat transfer (due to a high-shell-side heat transfer coecient) for a
given pressure drop in a minimum amount of space. Triple and no-tubes-in-window
segmental baes are used for low-pressure-drop applications. The choice of bae type,
spacing, and cut is determined largely by ow rate, desired heat transfer rate, allowable
pressure drop, tube support, and ow-induced vibrations. Disk and doughnut baes/
support plates are used primarily in nuclear heat exchangers. These baes for nuclear
exchangers have small perforations between tube holes to allow a combination of
crossow and longitudinal ow for lower shell-side pressure drop. The combined
ow results in a slightly higher heat transfer coecient than that for pure longitudinal
ow and minimizes tube-to-tube temperature dierences. Rod (or bar) baes, the most
common type of grid bae, used to support the tubes and increase the turbulence of the
shell uid, are shown in Fig. 1.11. The ow in a rod bae heat exchanger is parallel to
the tubes, and ow-induced vibrations are virtually eliminated by the bae support of
the tubes. One alternative to a rod bae heat exchanger is the use of twisted tubes (after
attening the circular tubes, they are twisted), shown in Fig. 1.12. Twisted tubes provide rigidity and eliminate ow-induced tube vibrations, can be cleaned easily on the
shell side with hydrojets, and can be cleaned easily inside the tubes, but cannot be
retubed. Low-nned tubes are also available in a twisted-tube conguration. A
helical bae shell-and-tube exchanger with baes as shown in Fig. 1.13 also has the
following advantages: a lower shell-side pressure drop while maintaining the high heat
transfer coecient of a segmental exchanger, reduced leakage streams (see Section
4.4.1), and elimination of dead spots and recirculation zones (thus reducing fouling).
Every shell-and-tube exchanger has transverse baes except for X and K shells, which
have support plates because the sole purpose of these transverse baes is to support the
tubes. Bae types and their design guidelines are described further in Section 10.2.1.3.
19
Butterworth (1996) provides further descriptions of these designs, and they are compared in Table 1.1.
Tubesheets. These are used to hold tubes at the ends. A tubesheet is generally a round
metal plate with holes drilled through for the desired tube pattern, holes for the tie rods
(which are used to space and hold plate baes), grooves for the gaskets, and bolt holes
for anging to the shell and channel. To prevent leakage of the shell uid at the
20
FIGURE 1.11 (a) Four rod baes held by skid bars (no tubes shown); (b) tube in a rod bae
exchanger supported by four rods; (c) square layout of tubes with rods; (d) triangular layout of tubes
with rods (Shah, 1981).
FIGURE 1.12 Twisted tube bundle for a shell-and-tube exchanger. (Courtesy of Brown Fintube
Company, Houston, TX.)
21
FIGURE 1.13 Helical bae shell-and-tube exchanger: (a) single helix; (b) double helix. (Courtesy
of ABB Lumus Heat Transfer, Bloomeld, NJ.)
tubesheet through a clearance between the tube hole and tube, the tube-to-tubesheet
joints are made by many methods, such as expanding the tubes, rolling the tubes,
hydraulic expansion of tubes, explosive welding of tubes, stung of the joints, or
welding or brazing of tubes to the tubesheet. The leak-free tube-to-tubesheet joint
made by the conventional rolling process is shown in Fig. 1.14.
1.5.1.2 Double-Pipe Heat Exchangers. This exchanger usually consists of two concentric pipes with the inner pipe plain or nned, as shown in Fig. 1.15. One uid
ows in the inner pipe and the other uid ows in the annulus between pipes in a
counterow direction for the ideal highest performance for the given surface area.
However, if the application requires an almost constant wall temperature, the uids
may ow in a parallelow direction. This is perhaps the simplest heat exchanger. Flow
distribution is no problem, and cleaning is done very easily by disassembly. This conguration is also suitable where one or both of the uids is at very high pressure,
Segmental
Bae
Rod Bae
Twisted
Tube
No
Yes
Yes
Yes
Yes
No
No
Yes
With inserts
No
With inserts
Yes
Included
Yes
With inserts
No
No
Yes, with
square pitch
With special
designs
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes, with
square pitch
With double
helix
Yes
Helical
Bae
22
FIGURE 1.14 Details of a leak-free joint between the tube and tube hole of a tubesheet: (a) before
tube expansion; (b) after tube expansion.
because containment in the small-diameter pipe or tubing is less costly than containment
in a large-diameter shell. Double-pipe exchangers are generally used for small-capacity
applications where the total heat transfer surface area required is 50 m2 (500 ft2 ) or less
because it is expensive on a cost per unit surface area basis. Stacks of double-pipe or
multitube heat exchangers are also used in some process applications with radial or
longitudinal ns. The exchanger with a bundle of U tubes in a pipe (shell) of 150 mm
(6 in.) diameter and above uses segmental baes and is referred to variously as a hairpin
or jacketed U-tube exchanger.
1.5.1.3 Spiral Tube Heat Exchangers. These consist of one or more spirally wound
coils tted in a shell. Heat transfer rate associated with a spiral tube is higher than that
for a straight tube. In addition, a considerable amount of surface can be accommodated
in a given space by spiraling. Thermal expansion is no problem, but cleaning is almost
impossible.
1.5.2
Plate-type heat exchangers are usually built of thin plates (all prime surface). The plates
are either smooth or have some form of corrugation, and they are either at or wound in
an exchanger. Generally, these exchangers cannot accommodate very high pressures,
23
Basic Construction. The plate-and-frame or gasketed plate heat exchanger (PHE) consists of a number of thin rectangular metal plates sealed around the edges by gaskets
and held together in a frame as shown in Fig. 1.16. The frame usually has a xed end
cover (headpiece) tted with connecting ports and a movable end cover (pressure plate,
follower, or tailpiece). In the frame, the plates are suspended from an upper carrying
bar and guided by a bottom carrying bar to ensure proper alignment. For this purpose,
each plate is notched at the center of its top and bottom edges. The plate pack with
xed and movable end covers is clamped together by long bolts, thus compressing the
gaskets and forming a seal. For later discussion, we designate the resulting length of the
plate pack as Lpack . The carrying bars are longer than the compressed stack, so that
when the movable end cover is removed, plates may be slid along the support bars for
inspection and cleaning.
Each plate is made by stamping or embossing a corrugated (or wavy) surface pattern
on sheet metal. On one side of each plate, special grooves are provided along the periphery of the plate and around the ports for a gasket, as indicated by the dark lines in
Fig. 1.17. Typical plate geometries (corrugated patterns) are shown in Fig. 1.18, and over
60 dierent patterns have been developed worldwide. Alternate plates are assembled such
24
FIGURE 1.17 Plates showing gaskets around the ports (Shah and Focke, 1988).
that the corrugations on successive plates contact or cross each other to provide mechanical support to the plate pack through a large number of contact points. The resulting
ow passages are narrow, highly interrupted, and tortuous, and enhance the heat transfer
rate and decrease fouling resistance by increasing the shear stress, producing secondary
ow, and increasing the level of turbulence. The corrugations also improve the rigidity of
the plates and form the desired plate spacing. Plates are designated as hard or soft,
depending on whether they generate a high or low intensity of turbulence.
FIGURE 1.18 Plate patterns: (a) washboard; (b) zigzag; (c) chevron or herringbone;
(d) protrusions and depressions; (e) washboard with secondary corrugations; ( f ) oblique
washboard (Shah and Focke, 1988).
25
A channel is a ow passage bounded by two plates and is occupied by one of the uids. In contrast, a plate
separates the two uids and transfers heat from the hot uid to the cold uid.
Thus with the proper arrangement, gaskets also distribute the uids between the channels in addition to providing sealing to prevent leakage.
}
In a plate heat exchanger, a pass refers to a group of channels in which the ow is in the same direction for one full
length of the exchanger (from top to bottom of the pack; see Fig. 1.65). In an m pass n pass two-uid plate heat
exchanger, uid 1 ows through m passes and uid 2 through n passes.
{
26
Gasket Material
Natural rubber
Generic Name
cis-1,4polyisoprene
SBR (styrene
butadiene)
Neoprene
Maximum
Operating
Temperature
(8C)
70
80
trans-1,4polychloroprene
70
Nitrile
100140
Butyl
(resin cured)
120150
Ethylene propylene
(EDPM) rubber
Silicone rubber
Polydimethylsiloxane
140
Fluorinated rubber
175
Compressed
asbestos ber
140
200260
Applications
Comments
Oxygenated solvents,
acids, alcohols
General-purpose aqueous, Has poor fat
alkalies, acids, and
resistance
oxygenated solvents
Alcohols, alkalies, acids,
aliphatic hydrocarbon
solvents
Dairy, fruit juices,
Is resistant to
beverage, pharmaceutical
fatty materials;
and biochemical
particularly
applications, oil, gasoline, suitable
animal and vegetable oils, for cream
alkalies, aliphatic organic
solvents
Alkalies, acids, animal and Has poor fat
vegetable oils, aldehydes,
resistance;
ketones, phenols, and
suitable
some esters
for UHT
milk duties;
resists
inorganic
chemical
solutions up
to 1508C
Alkalies, oxygenated
Unsuitable for
solvents
fatty liquids
General low-temperature
use, alcohols, sodium
hypochlorite
High-temperature aqueous
solutions, mineral oils and
gasoline, organic solvents,
animal and vegetable oils
Organic solvents, highoperating-temperature
applications
common plate materials are stainless steel (AISI 304 or 316) and titanium. Plates made
from Incoloy 825, Inconel 625, and Hastelloy C-276 are also available. Nickel, cupronickel, and monel are rarely used. Carbon steel is not used, due to low corrosion resistance for thin plates. Graphite and polymer plates are used with corrosive uids. The
heat transfer surface area per unit volume for plate exchangers ranges from 120 to
660 m2 /m3 (37 to 200 ft2 /ft3 ).
27
FIGURE 1.19 A three-uid plate heat exchanger. (Courtesy of Alfa Laval Thermal, Inc., Lund,
Sweden.)
TABLE 1.3 Some Geometrical and Operating Condition Characteristics of Plate-and-Frame Heat
Exchangers
Unit
Maximum surface area
Number of plates
Port size
Operation
2
2500 m
3 to 700
Up to 400 mm
(for liquids)
Plates
Thickness
Size
Spacing
Width
Length
Hydraulic diameter
Surface area per plate
Pressure
Temperature
Maximum port velocity
Channel ow rates
Maximum unit ow rate
Performance
0.5 to 1.2 mm
0.03 to 3.6 m2
1.5 to 7 mm
70 to 1200 mm
0.4 to 5 m
2 to 10 mm
0.02 to 5 m2
Temperature approach
Heat exchanger eciency
Heat transfer coecients
for waterwater duties
As low as 18C
Up to 93%
3000 to 8000 W/m2 K
In the plate exchanger, the two outer plates serve as end plates and ideally do not participate in heat transfer
between the uids because of the large thermal resistance associated with thick end plates and air gaps between the
end plates and the header/follower. The remaining plates, known as thermal plates, transfer heat between the
uids.
28
the xed end cover, permitting easy disassembly and cleaning/repair of a PHE without
disconnecting any piping. In a multipass arrangement, the ports and uid connections
are located on both xed and movable end covers. A multipass arrangement is generally
used when the ow rates are considerably dierent or when one would like to use up
the available pressure drop by multipassing and hence getting a higher heat transfer
coecient.
Advantages and Limitations. Some advantages of plate heat exchangers are as follows.
They can easily be taken apart into their individual components for cleaning, inspection, and maintenance. The heat transfer surface area can readily be changed or rearranged for a dierent task or for anticipated changing loads, through the exibility of
plate size, corrugation patterns, and pass arrangements. High shear rates and shear
stresses, secondary ow, high turbulence, and mixing due to plate corrugation patterns
reduce fouling to about 10 to 25% of that of a shell-and-tube exchanger, and enhance
heat transfer. Very high heat transfer coecients are achieved due to the breakup and
reattachment of boundary layers, swirl or vortex ow generation, and small hydraulic
diameter ow passages. Because of high heat transfer coecients, reduced fouling, the
absence of bypass and leakage streams, and pure counterow arrangements, the surface
area required for a plate exchanger is one-half to one-third that of a shell-andtube exchanger for a given heat duty, thus reducing the cost, overall volume, and space
requirement for the exchanger. Also, the gross weight of a plate exchanger is about onesixth that of an equivalent shell-and-tube exchanger. Leakage from one uid to the
other cannot take place unless a plate develops a hole. Since the gasket is between the
plates, any leakage from the gaskets is to the outside of the exchanger. The residence
time (time to travel from the inlet to the outlet of the exchanger) for dierent uid
particles or ow paths on a given side is approximately the same. This parity is desirable for uniformity of heat treatment in applications such as sterilizing, pasteurizing,
and cooking. There are no signicant hot or cold spots in the exchanger that could lead
to the deterioration of heat-sensitive uids. The volume of uid held up in the exchanger is small; this feature is important with expensive uids, for faster transient response,
and for better process control. Finally, high thermal performance can be achieved in
plate exchangers. The high degree of counterow in PHEs makes temperature
approaches of up to 18C (28F) possible. The high thermal eectiveness (up to about
93%) facilitates economical low-grade heat recovery. The ow-induced vibrations,
noise, thermal stresses, and entry impingement problems of shell-and-tube exchangers
do not exist for plate heat exchangers.
Some inherent limitations of the plate heat exchangers are caused by plates and
gaskets as follows. The plate exchanger is capable of handling up to a maximum pressure
of about 3 MPa gauge (435 psig) but is usually operated below 1.0 MPa gauge (150 psig).
The gasket materials (except for the PTFE-coated type) restrict the use of PHEs in highly
corrosive applications; they also limit the maximum operating temperature to 2608C
(5008F) but are usually operated below 1508C (3008F) to avoid the use of expensive
gasket materials. Gasket life is sometimes limited. Frequent gasket replacement may
be needed in some applications. Pinhole leaks are hard to detect. For equivalent ow
velocities, pressure drop in a plate exchanger is very high compared to that of a shell-andtube exchanger. However, the ow velocities are usually low and plate lengths are
short, so the resulting pressure drops are generally acceptable. The normal symmetry
of PHEs may make phase-change applications{ more dicult, due to large dierences
in volumetric ows. For some cases, heat exchanger duties with widely dierent uid ow
29
rates and depending on the allowed pressure drops of the two uids, an arrangement of a
dierent number of passes for the two uids may make a PHE advantageous. However,
care must be exercised to take full advantage of available pressure drop while multipassing one or both uids.
Because of the long gasket periphery, PHEs are not suited for high-vacuum applications. PHEs are not suitable for erosive duties or for uids containing brous materials.
In certain cases, suspensions can be handled; but to avoid clogging, the largest suspended
particle should be at most one-third the size of the average channel gap. Viscous uids
can be handled, but extremely viscous uids lead to ow maldistribution problems,
especially on cooling. Plate exchangers should not be used for toxic uids, due to potential gasket leakage. Some of the largest units have a total surface area of about 2500 m2
(27,000 ft2 ) per frame. Some of the limitations of gasketed PHEs have been addressed by
the new designs of PHEs described in the next subsection.
Major Applications. Plate heat exchangers were introduced in 1923 for milk pasteurization applications and now nd major applications in liquidliquid (viscosities up to
10 Pa s) heat transfer duties. They are most common in the dairy, juice, beverage,
alcoholic drink, general food processing, and pharmaceutical industries, where their
ease of cleaning and the thermal control required for sterilization/pasteurization make
them ideal. They are also used in the synthetic rubber industry, paper mills, and in the
process heaters, coolers, and closed-circuit cooling systems of large petrochemical and
power plants. Here heat rejection to seawater or brackish water is common in many
applications, and titanium plates are then used.
Plate heat exchangers are not well suited for lower-density gas-to-gas applications.
They are used for condensation or evaporation of non-low-vapor densities. Lower vapor
densities limit evaporation to lower outlet vapor fractions. Specially designed plates are
now available for condensing as well as evaporation of high-density vapors such as
ammonia, propylene, and other common refrigerants, as well as for combined evaporation/condensation duties, also at fairly low vapor densities.
1.5.2.2 Welded and Other Plate Heat Exchangers. One of the limitations of the
gasketed plate heat exchanger is the presence of gaskets, which restricts their use to
compatible uids (noncorrosive uids) and which limits operating temperatures and
pressures. To overcome this limitation, a number of welded plate heat exchanger designs
have surfaced with welded pairs of plates on one or both uid sides. To reduce the
eective welding cost, the plate size for this exchanger is usually larger than that of the
gasketed PHE. The disadvantage of such a design is the loss of disassembling exibility
on the uid sides where the welding is done. Essentially, laser welding is done around
the complete circumference, where the gasket is normally placed. Welding on both sides
then results in higher limits on operating temperatures and pressures [3508C (6608F) and
4.0 MPa (580 psig)] and allows the use of corrosive uids compatible with the plate
material. Welded PHEs can accommodate multiple passes and more than two uid
streams. A Platular heat exchanger can accommodate four uid streams. Figure 1.20
shows a pack of plates for a conventional plate-and-frame exchanger, but welded on one
{
30
FIGURE 1.20 Section of a welded plate heat exchanger. (Courtesy of Alfa Laval Thermal, Inc.,
Richmond, VA.)
uid side. Materials used for welded PHEs are stainless steel, Hastelloy, nickel-based
alloys, and copper and titanium.
A Bavex welded-plate heat exchanger with welded headers is shown in Fig. 1.21. A
Stacked Plate Heat Exchanger is another welded plate heat exchanger design (from
Packinox), in which rectangular plates are stacked and welded at the edges. The physical
size limitations of PHEs [1.2 m wide 4 m long maximum (4 13 ft)] are considerably
extended to 1.5 m wide 20 m long (5 66 ft) in Packinox exchangers. A maximum
surface area of over 10,000 m2 (100,000 ft2 ) can be accommodated in one unit. The
potential maximum operating temperature is 8158C (15008F) with an operating pressure
of up to 20 MPa (3000 psig) when the stacked plate assembly is placed in a cylindrical
pressure vessel. For inlet pressures below 2 MPa (300 psig) and inlet temperatures below
2008C (4008F), the plate bundle is not placed in a pressure vessel but is bolted between
two heavy plates. Some applications of this exchanger are for catalytic reforming, hydrosulfurization, and crude distillation, and in a synthesis converter feed euent exchanger
for methanol and for a propane condenser.
A vacuum brazed plate heat exchanger is a compact PHE for high-temperature and
high-pressure duties, and it does not have gaskets, tightening bolts, frame, or carrying
and guide bars. It consists simply of stainless steel plates and two end plates, all generally
copper brazed, but nickel brazed for ammonia units. The plate size is generally limited to
0.3 m2 . Such a unit can be mounted directly on piping without brackets and foundations.
Since this exchanger cannot be opened, applications are limited to negligible fouling
cases. The applications include water-cooled evaporators and condensers in the refrigeration industry, and process water heating and heat recovery.
A number of other plate heat exchanger constructions have been developed to address
some of the limitations of the conventional PHEs. A double-wall PHE is used to avoid
mixing of the two uids. A wide-gap PHE is used for uids having a high ber content or
coarse particles/slurries. A graphite PHE is used for highly corrosive uids. A ow-ex
exchanger has plain ns on one side between plates and the other side has conventional
plate channels, and is used to handle asymmetric duties (a ow rate ratio of 2 : 1 and
higher). A PHE evaporator has an asymmetric plate design to handle mixed process ows
(liquid and vapors) and dierent ow ratios.
31
1.5.2.3 Spiral Plate Heat Exchangers. A spiral plate heat exchanger consists of two
relatively long strips of sheet metal, normally provided with welded studs for plate
spacing, wrapped helically around a split mandrel to form a pair of spiral channels
for two uids, as shown in Fig. 1.22. Alternate passage edges are closed. Thus, each uid
has a long single passage arranged in a compact package. To complete the exchanger,
covers are tted at each end. Any metal that can be cold-formed and welded can be used
for this exchanger. Common metals used are carbon steel and stainless steel. Other
metals include titanium, Hastelloy, Incoloy, and high-nickel alloys. The basic spiral
element is sealed either by welding at each side of the channel or by providing a gasket
(nonasbestos based) at each end cover to obtain the following alternative arrangements
of the two uids: (1) both uids in spiral counterow; (2) one uid in spiral ow, the
other in crossow across the spiral; or (3) one uid in spiral ow, the other in a
combination of crossow and spiral ow. The entire assembly is housed in a cylindrical
shell enclosed by two (or only one or no) circular end covers (depending on the ow
arrangements above), either at or conical. Carbon steel and stainless steel are common
materials. Other materials used include titanium, Hastelloy, and Incoloy.
A spiral plate exchanger has a relatively large diameter because of the spiral turns.
The largest exchanger has a maximum surface area of about 500 m2 (5400 ft2 ) for a
maximum shell diameter of 1.8 m (72 in.). The typical passage height is 5 to 25 mm
(0.20 to 1.00 in.) and the sheet metal thickness range is 1.8 to 4 mm (0.07 to 0.16 in.).
32
Fluid 2 in
Fluid 2 out
Fluid 1 in
Fluid 1 out
FIGURE 1.22 Spiral plate heat exchanger with both uids in spiral counterow.
The heat transfer coecients are not as high as in a plate exchanger if the plates are not
corrugated. However, the heat transfer coecient is higher than that for a shell-and-tube
exchanger because of the curved rectangular passages. Hence, the surface area requirement is about 20% lower than that for a shell-and-tube unit for the same heat duty.
The counterow spiral unit is used for liquidliquid, condensing, or gas cooling
applications. When there is a pressure drop constraint on one side, such as with gas
ows or with high liquid ows, crossow (straight ow) is used on that side. For condensation or vaporization applications, the unit is mounted vertically. Horizontal units
are used when high concentrations of solids exist in the uid.
The advantages of this exchanger are as follows: It can handle viscous, fouling liquids
and slurries more readily because of a single passage. If the passage starts fouling, the
localized velocity in the passage increases. The fouling rate then decreases with increased
uid velocity. The fouling rate is very low compared to that of a shell-and-tube unit. It is
more amenable to chemical, ush, and reversing uid cleaning techniques because of the
single passage. Mechanical cleaning is also possible with removal of the end covers. Thus,
maintenance is less than with a shell-and-tube unit. No insulation is used outside the
exchanger because of the cold uid owing in the outermost passage, resulting in negligible heat loss, if any, due to its inlet temperature closer to surrounding temperature.
The internal void volume is lower (less than 60%) than in a shell-and-tube exchanger,
and thus it is a relatively compact unit. By adjusting dierent channel heights, considerable dierences in volumetric ow rates of two streams can be accommodated.
The disadvantages of this exchanger are as follows: As noted above, the maximum
size is limited. The maximum operating pressure ranges from 0.6 to 2.5 MPa gauge (90 to
370 psig) for large units. The maximum operating temperature is limited to 5008C
(9308F) with compressed asbestos gaskets, but most are designed to operate at 2008C
(3928F). Field repair is dicult due to construction features.
This exchanger is well suited as a condenser or reboiler. It is used in the cellulose
industry for cleaning relief vapors in sulfate and sulte mills, and is also used as a
thermosiphon or kettle reboiler. It is preferred especially for applications having very
viscous liquids, dense slurries, digested sewage sludge, and contaminated industrial euents. A spiral version free of welded studs for plate spacing on one or both uid sides but
33
with reduced width is used for sludge and other heavily fouling uids. It is also used in the
treatment of bauxite suspensions and mash liquors in the alcohol industry.
1.5.2.4 Lamella Heat Exchangers. A lamella heat exchanger consists of an outer
tubular shell surrounding an inside bundle of heat transfer elements. These elements,
referred to as lamellas, are at tubes (pairs of thin dimpled plates, edge welded, resulting
in high-aspect-ratio rectangular channels), shown in Fig. 1.23. The inside opening of the
lamella ranges from 3 to 10 mm (0.1 to 0.4 in.) and the wall thickness from 1.5 to 2 mm
(0.06 to 0.08 in.). Lamellas are stacked close to each other to form narrow channels on
the shell side. Lamellas are inserted in the end ttings with gaskets to prevent the
leakage from shell to tube side, or vice versa. In a small exchanger, lamellas are of
increasing width from either end to the center of the shell to fully utilize the available
space, as shown in Fig. 1.23a. However, in a larger exchanger, lamellas consist of two
(see Fig. 1.23b) or more at tubes to contain operating pressures. There are no baes.
One end of the tube bundle is xed and the other is oating, to allow for thermal
FIGURE 1.23 (a) Lamella heat exchanger; (b) cross section of a lamella heat exchanger;
(c) lamellas. (Courtesy of Alfa Laval Thermal, Inc., Lund, Sweden.)
34
35
the same techniques as those employed for making printed circuit boards. A block
(stack) of chemically etched plates is then diusion bonded, and uid inlet/outlet headers are welded to make the exchanger. For the two uid streams, there are dierent
etching patterns, as desired to make a crossow, counterow, or multipass cross-counterow exchanger. Multiple passes and multiple uid streams can be made in a single
block. Several blocks are welded together for large heat duty applications. The channel
depth is 0.1 to 2 mm (0.004 to 0.08 in.). High surface area densities, 650 to 1300 m2 /m3
(200 to 400 ft2 /ft3 ), are achievable for operating pressures 50 to 10 MPa (7250 to 290
psi),{ and temperatures 150 to 8008C (300 to 15008F). A variety of materials, including
stainless steel, titanium, copper, nickel, and nickel alloys, can be used. It has been used
successfully with relatively clean gases, liquids, and phase-change uids in the chemical
processing, fuel processing, waste heat recovery, power and energy, refrigeration, and
air separation industries. They are used extensively in oshore oil platforms as compressor aftercoolers, gas coolers, cryogenic processes to remove inert gases, and so on.
Having a small channel size, the uid pressure drop can be a constraint for low-tomoderate pressure applications. However, the main advantage of this exchanger is high
pressure/strength, exibility in design, and high eectivenesses.
1.5.2.6 Panelcoil Heat Exchangers. The basic elements of this exchanger are called
panelcoils, platecoils, or embossed-panel coils, as shown in Fig. 1.25. The panelcoil
serves as a heat sink or a heat source, depending on whether the uid within the coil
is being cooled or heated. As a result, the shape and size of the panelcoil is made to t
the system, or at panelcoils are immersed in a tank or placed in the atmosphere for heat
transfer. Basically, three dierent methods have been employed to manufacture panelcoils: a die-stamping process, spot-weld process, and roll-bond process. In the diestamping process, ow channels are die-stamped on either one or two metal sheets.
When one sheet is embossed and joined to a at (unembossed sheet), it forms a onesided embossed panelcoil. When both sheets are stamped, it forms a double-sided
embossed panelcoil. The two plates are joined by electric resistance welding of the
metal sheets. Examples are shown in Fig. 1.25a and b.
In the spot-weld process, two at sheets are spot-welded in a desired pattern (no die
stamping), and then are inated by a uid under high pressure to form ow passages
interconnected by weld points. An example is shown in Fig. 1.25d.
In a roll-bond process, two sheets of metal (copper or aluminum) are bonded with a
true metallurgical bond, except in certain desired specied channels, where a special
stopweld material is applied. On one of the metal sheets, the stopweld material is applied
in the desired ow pattern. This sheet is then stacked with another plain sheet without
stopweld material on it. The sheets are then heated and immediately hot-rolled under
high pressure to provide a metallurgical bond. Subsequent cold rolling follows to provide
an appropriate increase in length. After annealing the panelcoil, a needle is inserted at the
edge, exposing stopweld material, and high-pressure air inates the desired ow passages
when the panelcoil is placed between two plates in a heavy hydraulic press. The roll-bond
process limits the panelcoils to a at in shape.
The most commonly used materials for panelcoils are carbon steel, stainless steel,
titanium, nickel and its alloys, and monel. The panelcoil sheet metal gauges range
between 1.5 and 3.0 mm (0.06 to 0.12 in.) depending on the materials used and whether
Note that operating pressures for 650- and 1300-m2 /m3 surface area densities are 50 and 10 MPa, respectively.
36
FIGURE 1.25 Die-stamped plate coils: (a) serpentine, (b) multizone, (c) a vessel; (d) spot-welded
Econocoil bank. (Courtesy of Tranter PHE, Inc., Wichita, TX.)
or not the panels are single or double embossed. The maximum operating pressure
ranges from 1.8 MPa (260 psig) for double-embossed and 1.2 MPa (175 psig) for singleembossed carbon steel, stainless steel, and monel panelcoils to 0.7 MPa (100 psig) for
double-embossed titanium panelcoils.
Panelcoil heat exchangers are relatively inexpensive and can be made into desired
shapes and thicknesses for heat sinks and heat sources under varied operating conditions.
Hence, they have been used in many industrial applications, such as cryogenics, chemicals, bers, food, paints, pharmaceuticals, and solar absorbers.
1.5.3
The tubular and plate-type exchangers described previously are all prime surface heat
exchangers, except for a shell-and-tube exchanger with low nned tubing. Their heat
exchanger eectiveness (see Section 3.3.1 for the denition) is usually 60% or below, and
the heat transfer surface area density is usually less than 700 m2 /m3 (213 ft2 /ft3 ). In some
applications, much higher (up to about 98%) exchanger eectiveness is essential, and the
box volume and mass are limited so that a much more compact surface is mandated.
Also, in a heat exchanger with gases or some liquids, the heat transfer coecient is quite
low on one or both uid sides. This results in a large heat transfer surface area requirement. One of the most common methods to increase the surface area and exchanger
37
FIGURE 1.26 Basic components of a plate-n heat exchanger (Shah and Webb, 1983).
compactness is to add the extended surface (ns) and use ns with the n density ( n
frequency, ns/m or ns/in.) as high as possible on one or both uid sides, depending on
the design requirement. Addition of ns can increase the surface area by 5 to 12 times the
primary surface area in general, depending on the design. The resulting exchanger is
referred to as an extended surface exchanger. Flow area is increased by the use of thingauge material and sizing the core properly. The heat transfer coecient on extended
surfaces may be higher or lower than that on unnned surfaces. For example, interrupted
(strip, louver, etc.) ns provide both an increased area and increased heat transfer coecient, while internal ns in a tube increase the tube-side surface area but may result in a
slight reduction in the heat transfer coecient, depending on the n spacing. Generally,
increasing the n density reduces the heat transfer coecient associated with ns. Flow
interruptions (as in oset strip ns, louvered ns, etc.) may increase the heat transfer
coecient two to four times that for the corresponding plain (uncut) n surface. Plate-n
and tube-n geometries are the two most common types of extended surface heat
exchangers.{
1.5.3.1 Plate-Fin Heat Exchangers. This type of exchanger has corrugated ns (most
commonly having triangular and rectangular cross sections) or spacers sandwiched
between parallel plates (referred to as plates or parting sheets), as shown in Fig. 1.26.
Sometimes ns are incorporated in a at tube with rounded corners (referred to as a
formed tube), thus eliminating the need for side bars. If liquid or phase-change uid
ows on the other side, the parting sheet is usually replaced by a at tube with or
without inserts or webs (Fig. 1.27). Other plate-n constructions include drawn-cup
If the heat transfer surface in a prime surface heat exchanger is rough (due either to the manufacturing process or
made articially) or small-scale ns (the n height is approximately 5% or less of the tube radius) are made integral
to the prime surface, the exchanger is sometimes referred to as a micron heat exchanger.
38
FIGURE 1.27 Flat webbed tube and multilouver n automotive condenser. (Courtesy of Delphi
Harrison Thermal Systems, Lockport, NY.)
(Fig. 1.28) and tube-and-center{ congurations. The plates or at tubes separate the two
uid streams, and the ns form the individual ow passages. Alternate uid passages are
connected in parallel by suitable headers to form the two or more uid sides of the
exchanger. Fins are die or roll formed and are attached to the plates by brazing,{
soldering, adhesive bonding, welding, mechanical t, or extrusion. Fins may be used
on both sides in gas-to-gas heat exchangers. In gas-to-liquid applications, ns are generally used only on the gas side; if employed on the liquid side, they are used primarily
for structural strength and ow-mixing purposes. Fins are also sometimes used for
pressure containment and rigidity. In Europe, a plate-n exchanger is also referred to
as a matrix heat exchanger.
Plate ns are categorized as (1) plain (i.e., uncut) and straight ns, such as plain
triangular and rectangular ns, (2) plain but wavy ns (wavy in the main uid ow
direction), and (3) interrupted ns, such as oset strip, louver, perforated, and pin
ns. Examples of commonly used ns are shown in Fig. 1.29. Louver form of the multilouver n is shown in Fig. 7.29, along with a sketch of its louver form at section AA in
{
In the automotive industry, corrugated ns in the plate-n unit are referred to as centers, to distinguish them
from the at ns outside the tubes in a tube-n exchanger. The latter are referred to simply as ns in the
automotive industry.
{
In the automotive industry, the most common brazing technique is controlled atmosphere brazing (CAB; brazing
at atmospheric pressure in a nitrogen environment and with noncorrosive ux; also known as a Nocolok process),
and sometimes vacuum brazing is used. In the cryogenics industry, only vacuum brazing is used.
39
FIGURE 1.28 U-channel ribbed plates and multilouver n automotive evaporator. (Courtesy of
Delphi Harrison Thermal Systems, Lockport, NY.)
(a)
(b)
(c)
(d)
(e)
(f)
FIGURE 1.29 Corrugated n geometries for plate-n heat exchangers: (a) plain triangular n;
(b) plain rectangular n; (c) wavy n; (d) oset strip n; (e) multilouver n; ( f ) perforated n.
(Courtesy of Delphi Harrison Thermal Systems, Lockport, NY.)
40
Fig. 7.29c. Strip ns are also referred to as oset ns, lance-oset ns, serrated ns, and
segmented ns. Many variations of interrupted ns are used in industry since they employ
the materials of construction more eciently than do plain ns and are therefore used
when allowed by the design constraints.
Plate-n exchangers are generally designed for moderate operating pressures [less
than about 700 kPa gauge (100 psig)], although plate-n exchangers are available
commercially for operating pressures up to about 8300 kPa gauge (1200 psig).
Recently, a condenser for an automotive air-conditioning system (see Fig. 1.27) using
carbon dioxide as the working uid has been developed for operating pressures of 14
MPa (2100 psia). A recently developed titanium plate-n exchanger (manufactured by
superelastic deformation and diusion bonding, shown in Fig. 1.30) can take 35 MPa
(5000 psig) and higher pressures. The temperature limitation for plate-n exchangers
depends on the method of bonding and the materials employed. Such exchangers have
been made from metals for temperatures up to about 8408C (15508F) and made from
ceramic materials for temperatures up to about 11508C (21008F) with a peak temperature
of 13708C (25008F). For ventilation applications (i.e., preheating or precooling of incoming air to a building/room), the plate-n exchanger is made using Japanese treated
(hygroscopic) paper and has the operating temperature limit of 508C (1228F). Thus,
FIGURE 1.30 Process of manufacturing of a super elastically deformed diusion bonded platen exchanger (From Reay, 1999).
41
plates and ns are made from a variety of materials, metals, ceramics, and papers. Platen exchangers have been built with a surface area density of up to 5900 m2 /m3 (1800 ft2 /
ft3 ). There is total freedom in selecting the n surface area on each uid side, as required
by the design, by varying the n height and n density. Although typical n densities are
120 to 700 ns/m (3 to 18 ns/in.), applications exist for as many as 2100 ns/m (53 ns/
in.). Common n thickness ranges between 0.05 and 0.25 mm (0.002 to 0.01 in.). Fin
heights may range from 2 to 25 mm (0.08 to 1.0 in.). A plate-n exchanger with 600 ns/m
(15.2 ns/in.) provides about 1300 m2 (400 ft2 /ft3 ) of heat transfer surface area per cubic
meter of volume occupied by the ns. Plate-n exchangers are manufactured in virtually
all shapes and sizes and are made from a variety of materials. A cryogenic plate-n
exchanger has about 10% of the volume of an equivalent shell-and-tube exchanger
(Reay, 1999).
Plate-n exchangers have been produced since the 1910s in the auto industry (copper
nbrass tubes), since the 1940s in the aerospace industry (using aluminum), and in gas
liquefaction applications since the 1950s using aluminum because of the better mechanical characteristics of aluminum at low temperatures. They are now used widely in
electric power plants (gas turbine, steam, nuclear, fuel cell, etc.), propulsive power plants
(automobile, truck, airplane, etc.), systems with thermodynamic cycles (heat pump,
refrigeration, etc.), and in electronic, cryogenic, gas-liquefaction, air-conditioning, and
waste heat recovery systems.
1.5.3.2 Tube-Fin Heat Exchangers. These exchangers may be classied as conventional and specialized tube-n exchangers. In a conventional tube-n exchanger, heat
transfer between the two uids takes place by conduction through the tube wall.
However, in a heat pipe exchanger (a specialized type of tube-n exchanger), tubes
with both ends closed act as a separating wall, and heat transfer between the two uids
takes place through this separating wall (heat pipe) by conduction, and evaporation
and condensation of the heat pipe uid. Let us rst describe conventional tube-n
exchangers and then heat pipe exchangers.
Conventional Tube-Fin Exchangers. In a gas-to-liquid exchanger, the heat transfer coefcient on the liquid side is generally one order of magnitude higher than that on the gas
side. Hence, to have balanced thermal conductances (approximately the same hA) on
both sides for a minimum-size heat exchanger, ns are used on the gas side to increase
surface area A. This is similar to the case of a condensing or evaporating uid stream
on one side and gas on the other. In addition, if the pressure is high for one uid, it is
generally economical to employ tubes.
In a tube-n exchanger, round and rectangular tubes are most common, although
elliptical tubes are also used. Fins are generally used on the outside, but they may be used
on the inside of the tubes in some applications. They are attached to the tubes by a tight
mechanical t, tension winding, adhesive bonding, soldering, brazing, welding, or
extrusion.
Depending on the n type, tube-n exchangers are categorized as follows: (1) an
individually nned tube exchanger or simply a nned tube exchanger, as shown in
Figs. 1.31a and 1.32, having normal ns on individual tubes; (2) a tube-n exchanger
having at (continuous) ns, as shown in Figs. 1.31b and 1.33; the ns can be plain, wavy,
or interrupted, and the array of tubes can have tubes of circular, oval, rectangular, or
other shapes; and (3) longitudinal ns on individual tubes, as shown in Fig. 1.34. A tuben exchanger with at ns has been referred to variously as a plate-n and tube, plate
42
FIGURE 1.31 (a) Individually nned tubes; (b) at (continuous) ns on an array of tubes. The at
ns are shown as plain ns, but they can be wavy, louvered, or interrupted.
nned tube, and tube in-plate n exchanger in the literature. To avoid confusion with a
plate-n exchanger dened in Section 1.5.3.1, we refer to it as a tube-n exchanger having
at (plain, wavy, or interrupted) ns. A tube-n exchanger of the aforementioned categories 1 and 2 is referred to as a coil in the air-conditioning and refrigeration industries
and has air outside and a refrigerant inside the tube. Individually nned tubes are
probably more rugged and practical in large tube-n exchangers. The exchanger with
at ns is usually less expensive on a unit heat transfer surface area basis because of its
simple and mass-production construction features. Longitudinal ns are generally used
in condensing applications and for viscous uids in double-pipe heat exchangers.
Shell-and-tube exchangers sometime employ low nned tubes to increase the surface
area on the shell side when the shell-side heat transfer coecient is low compared to the
tube-side coecient, such as with highly viscous liquids, gases, or condensing refrigerant
vapors. The low nned tubes are generally helical or annular ns on individual tubes; the
n outside diameter (see Fig. 1.9) is slightly smaller than the bae hole. Longitudinal ns
on individual tubes are also used in shell-and-tube exchangers. Fins on the inside of the
tubes are of two types: integral ns as in internally nned tubes, and attached ns.
Internally nned tubes are shown in Fig. 1.35.
Tube-n exchangers can withstand ultrahigh pressures on the tube side. The highest
temperature is again limited by the type of bonding, materials employed, and material
thickness. Tube-n exchangers usually are less compact than plate-n units. Tube-n
exchangers with an area density of about 3300 m2 /m3 (1000 ft2 /ft3 ) are available commercially. On the n side, the surface area desired can be achieved through the proper n
density and n geometry. Typical n densities for at ns vary from 250 to 800 ns/m
(6 to 20 ns/in.), n thicknesses vary from 0.08 to 0.25 mm (0.003 to 0.010 in.), and n
ow lengths vary from 25 to 250 mm (1 to 10 in.). A tube-n exchanger having at ns
with 400 ns/m (10 ns/in.) has a surface area density of about 720 m2 /m3 (220 ft2 /ft3 ).
43
Tube-n exchangers are employed when one uid stream is at a higher pressure and/
or has a signicantly higher heat transfer coecient than that of the other uid stream.
As a result, these exchangers are used extensively as condensers and evaporators in airconditioning and refrigeration applications, as condensers in electric power plants, as oil
coolers in propulsive power plants, and as air-cooled exchangers (also referred to as nfan exchangers) in process and power industries.
FIGURE 1.33 Flat ns on an array of round, at, or oval tubes: (a) wavy n; (b) multilouver n;
both ns with staggered round tubes; (c) multilouver n with inline elliptical tubes. (Courtesy of
Delphi Harrison Thermal Systems, Lockport, NY.)
44
FIGURE 1.34 Longitudinal ns on individual tubes: (a) continuous plain; (b) cut and twisted;
(c) perforated; (d) internal and external longitudinal ns. (Courtesy of Brown Fintube Company,
Houston, TX.)
FIGURE 1.35 Internally nned tubes. (Courtesy of Forged-Fin Division, Noranda Metal
Industries, Inc., Newtown, CT.)
45
heat pipe, and hot and cold gases ow continuously in separate parts of the exchanger,
as shown in Fig. 1.36. Heat is transferred from the hot gas to the evaporation section of
the heat pipe by convection; the thermal energy is then carried away by the vapor to
the condensation section of the heat pipe, where it transfers heat to the cold gas by
convection.
As shown in Fig. 1.37, a heat pipe is a closed tube or vessel that has been evacuated,
partially lled with a heat transfer uid (a working uid sucient to wet the entire wick),
46
and sealed permanently at both ends. The inner surfaces of a heat pipe are usually lined
with a capillary wick (a porous lining, screen, or internally grooved wall). The wick is
what makes the heat pipe unique; it forces condensate to return to the evaporator by the
action of capillary force. In a properly designed heat pipe, the wick is saturated with the
liquid phase of the working uid, while the remainder of the tube contains the vapor
phase. When heat is applied at the evaporator by an external source, the working uid in
the wick in that section vaporizes, the pressure increases, and vapor ows to the condenser section through the central portion of the tube. The vapor condenses in the
condenser section of the pipe, releasing the energy of phase change to a heat sink (to a
cold uid, owing outside the heat pipe; see Fig. 1.37). The heat applied at the evaporator
section tries to dry the wick surface through evaporation, but as the uid evaporates, the
liquidvapor interface recedes into the wick surface, causing a capillary pressure to be
developed. This pressure is responsible for transporting the condensed liquid back to
the evaporator section, thereby completing a cycle. Thus, a properly designed heat pipe
can transport the energy of phase change continuously from the evaporator to the condenser without drying out the wick. The condensed liquid may also be pumped back to
the evaporator section by the capillary force or by the force of gravity if the heat pipe is
inclined and the condensation section is above the evaporator section. If the gravity force
is sucient, no wick may be necessary. As long as there is a temperature dierence
between the hot and cold gases in a heat pipe heat exchanger, the closed-loop evaporationcondensation cycle will be continuous, and the heat pipe will continue functioning.
Generally, there is a small temperature dierence between the evaporator and condenser
section [about 58C (98F) or so], and hence the overall thermal resistance of a heat pipe in
a heat pipe exchanger is small. Although water is a common heat pipe uid, other uids
are also used, depending on the operating temperature range.
A heat pipe heat exchanger (HPHE), shown in Fig. 1.36 for a gas-to-gas application,
consists of a number of nned heat pipes (similar to an air-cooled condenser coil)
mounted in a frame and used in a duct assembly. Fins on the heat pipe increase the
surface area to compensate for low heat transfer coecients with gas ows. The ns can
be spirally wrapped around each pipe, or a number of pipes can be expanded into at
plain or augmented ns. The n density can be varied from side to side, or the pipe may
contain no ns at all (liquid applications). The tube bundle may be horizontal or vertical
with the evaporator sections below the condenser sections. The tube rows are normally
staggered with the number of tube rows typically between 4 and 10. In a gas-to-gas
HPHE, the evaporator section of the heat pipe spans the duct carrying the hot exhaust
gas, and the condenser section is located in the duct through which the air to be preheated
ows. The HPHE has a splitter plate that is used primarily to prevent mixing between the
two gas streams, eectively sealing them from one another. Since the splitter plate is thin,
a heat pipe in a HPHE does not have the usual adiabatic section that most heat pipes
have.
Unit size varies with airow. Small units have a face size of 0.6 m (length) by 0.3 m
(height), and the largest units may have a face size up to 5 m 3 m. In the case of gas-toliquid heat exchangers, the gas section remains the same, but because of the higher
external heat transfer coecient on the liquid side, it need not be nned externally or
can even be shorter in length.
The heat pipe performance is inuenced by the angle of orientation, since gravity
plays an important role in aiding or resisting the capillary ow of the condensate.
Because of this sensitivity, tilting the exchanger may control the pumping power and
ultimately the heat transfer. This feature can be used to regulate the performance of a
47
heat pipe heat exchanger. For further details on the design of a HPHE, refer to Shah and
Giovannelli (1988).
Heat pipe heat exchangers are generally used in gas-to-gas heat transfer applications.
They are used primarily in many industrial and consumer productoriented waste heat
recovery applications.
1.5.4
Regenerators
The regenerator is a storage-type heat exchanger, as described earlier. The heat transfer
surface or elements are usually referred to as a matrix in the regenerator. To have
continuous operation, either the matrix must be moved periodically into and out of
the xed streams of gases, as in a rotary regenerator (Figs. 1.38 through 1.40), or the
gas ows must be diverted through valves to and from the xed matrices as in a xedmatrix regenerator (Fig. 1.41). The latter is also sometimes referred to as a periodic-ow
regenerator,{ a swing regenerator, or a reversible heat accumulator. Thus, in a rotary
Heated air
to furnace
Guide
bearing
Axial
seal
Support
feet
Cold-end element
removal door
Cold air in
Modular rotor
section
FIGURE 1.38 Ljungstrom air preheater. (Courtesy of ABB Alstom Power Air Preheater, Inc.,
Wellsville, NY.)
{
Both rotary matrix and xed-matrix regenerators have been designated as periodic-ow heat exchangers by Kays
and London (1998), because from the viewpoint of an observer riding on the matrix, periodic conditions are
experienced in both types of regenerators.
48
FIGURE 1.39 Heat wheel or a rotary regenerator made from a polyester lm.
regenerator, the matrix (disk or rotor) rotates continuously with a constant fraction of
the core (having disk sector angle h ) in the hot-uid stream and the remaining fraction
(having the disk sector angle c ) in the cold-uid stream; the outlet uid temperatures
vary across the ow area and are independent of time. The two uids generally ow in the
opposite directions and are separated by some form of ductwork and rubbing seals on
49
Manhole
Dome
Insulation
Skinwall
Checkerwork
Mushroom
Gate
Elbow
Goggle
Butterfly
Plaster
(profile) wall
Combustion
chamber
Three-lever
Checkers
Checker shoes
Grids
Columns
Hot blast
main
Cold blast
air
Shell
Clean gas
Andco
Koppers
McKee
Typical Modern Checker Brick
FIGURE 1.41 Cowper stove. (Courtesy of Andco Industries, Inc., Bualo, NY.)
the matrix. In a xed-matrix regenerator, the hot and cold uids are ducted through the
use of valves to the dierent matrices (with a minimum of two identical matrices for
continuous operation) of the regenerator in alternate operating periods Ph and Pc ; the
outlet uid temperatures vary with time. Here again, the two uids alternately ow in
opposite directions in a given matrix.
A third type of regenerator has a xed matrix (in disk form) and xed streams of
gases, but the gases are ducted through rotating hoods (headers) to the matrix as shown
in Fig. 1.42. This Rothemuhle regenerator is used as an air preheater in some powergenerating plants. Since the basic thermal design theory of all types of regenerators is the
same, no specic attention will be given to the Rothemuhle regenerator for the thermal
design.
The desired material properties for the regenerator are high volumetric heat capacity
(high cp ) and low eective thermal conductivity in the longitudinal (gas ow) direction.
It should be noted that at very low temperatures, 20 K (368R) and below, the specic
heat of most metals decreases substantially, thus aecting the regenerator performance
signicantly.
The thermodynamically superior counterow arrangement is usually employed for
storage type heat exchangers by introducing gases successively at the opposite ends.
When the rotational speed or frequency of switching hot and cold uids through such
a regenerator is increased, its thermal performance ideally approaches that of a pure
counterow heat exchanger; but in reality, the carryover leakage may become signicant
with increased speed, thus reducing the regenerator performance. For some applications,
a parallelow arrangement (gases introduced successively at the same end) may be used,
but there is no counterpart of the single- or multipass crossow arrangements common
in recuperators. For a rotary regenerator, the design of seals to prevent leakages of hot
50
FIGURE 1.42 Rothemuhle regenerator. (Courtesy of Babcock and Wilcox, New Orleans, LA.)
to cold uids, and vice versa, becomes a dicult task, especially if the two uids are at
signicantly dierent pressures. Rotating drives also pose a challenging mechanical
design problem. For a xed-matrix regenerator operating at high temperatures, due to
thermal distortion of housing and valves, various large and small cracks occur in the
matrix housing and the valves do not seal the ow of gases perfectly, resulting in pressure
leakages.
Major advantages of the regenerators are the following. A much more compact surface may be employed than in a recuperator, thus providing a reduced exchanger volume
for given exchanger eectiveness and pressure drop and thereby making a regenerator
economical compared to an equivalent recuperator. The major reason for having a much
more compact surface for a regenerator is that the hot and cold gas streams are separated
by radial seals or valves, unlike in a recuperator, where the primary surface is used to
separate the uid streams. The cost of manufacturing such a compact regenerator
surface per unit of heat transfer area is usually substantially lower than that for the
equivalent recuperator. Similarly, material cost could be lower in a regenerator than in
a recuperator. Hence, a compact regenerator usually has a smaller volume and is lower
in weight than an equivalent recuperator. Eectively, many n congurations of plate-n
exchangers and any nely divided matrix material (high specic heat preferred) that
provides high surface area density may be used. However, the leakproof core required
in a recuperator is not essential in a regenerator, due to the mode of operation.
Regenerators have been made from metals, ceramics, nylon, plastics, and paper, depending on the application. Another important advantage of a counterow regenerator over a
counterow recuperator is that the design of inlet and outlet headers used to distribute
hot and cold gases in the matrix is simple. This is because both uids ow in dierent
sections (separated by radial seals) of a rotary regenerator, or one uid enters and leaves
51
It should be emphasized that in a regenerator matrix, however, the entire surface acts as a direct heat-absorbing
and heat-releasing surface (a primary surface); there is no secondary surface or ns, although the surface between
spacers is usually referred to as ns.
52
FIGURE 1.43 Continuous-passage matrices for a rotary regenerator: (a) notched plate;
(b) triangular passage.
to minimize the primary leakage from the high-pressure uid to the low-pressure
uid.
A number of seal congurations are used in rotary regenerators. Two common shapes
are shown in Fig. 1.44. For the annular sectorshaped seals shown in Fig. 1.44a, ow
passages at every radial location experience the same ow exposure and seal-coverage
histories. For the uniform-width seals in Fig. 1.44b, ow passages at dierent radial
locations experience dierent ow exposure and seal coverage. For regenerators with
seals of equal area but arbitrary shape, the regenerator eectiveness is highest for annular
sectorshaped seals (Beck and Wilson, 1996).
Rotary regenerators have been designed for surface area densities of up to about
6600 m2 /m3 (2000 ft2 /ft3 ). They can employ thinner stock material, resulting in the lowest
amount of material for a given eectiveness and pressure drop of any heat exchanger
known today. Metal rotary regenerators have been designed for continuous operating
inlet temperatures up to about 7908C (14508F). For higher-temperature applications,
ceramic matrices are used. Plastics, paper, and wool are used for regenerators operating
below 658C (1508F). Metal and ceramic regenerators cannot withstand large pressure
dierences [greater than about 400 kPa (60 psi)] between hot and cold gases, because the
design of seals (wear and tear, thermal distortion, and subsequent leakage) is the single
most dicult problem to resolve. Plastic, paper, and wool regenerators operate approxi-
FIGURE 1.44 Seals used in rotary regenerators: (a) annular sector shaped; (b) uniform width
shape (Beck and Wilson, 1996).
53
mately at atmospheric pressure. Seal leakage can reduce the regenerator eectiveness
signicantly. Rotary regenerators also require a power input to rotate the core from one
uid to the other at the rotational speed desired.
Typical power plant regenerators have a rotor diameter up to 10 m (33 ft) and
rotational speeds in the range 0.5 to 3 rpm (rev per min). Air-ventilating regenerators
have rotors with diameters of 0.25 to 3 m (0.8 to 9.8 ft) and rotational speeds up to 10
rpm. Vehicular regenerators have diameters up to 0.6 m (24 in.) and rotational speeds up
to 18 rpm.
Ljungstrom air preheaters for thermal power plants, commercial and residential oiland coal-red furnaces, and regenerators for the vehicular gas turbine power plants are
typical examples of metal rotary regenerators for preheating inlet air. Rotary regenerators are also used in chemical plants and in preheating combustion air in electricity
generation plants for waste heat utilization. Ceramic regenerators are used for hightemperature incinerators and the vehicular gas turbine power plant. In air-conditioning
and industrial process heat recovery applications, heat wheels are made from knitted
aluminum or stainless steel wire matrix, wound polyester lm, plastic lms, and honeycombs. Even paper, wettable nylon, and polypropylene are used in the enthalpy or
hygroscopic wheels used in heating and ventilating applications in which moisture is
transferred in addition to sensible heat.
54
FIGURE 1.45 (a) Three-stove regenerator with seriesparallel arrangement; (b) operating
schedule. H, hot-gas period; C, blast period (Shah, 1981).
FIGURE 1.46 (a) Four-stove regenerator with staggered parallel arrangement; (b) operating
schedule. H, hot-gas period; C, blast period (Shah, 1981).
55
sponding amount to maintain the hot blast temperature approximately constant. In the
staggered parallel arrangement of Fig. 1.46, two stoves on air are maintained out of phase
by one-half period. In this arrangement, cold blast is routed through a hot stove and a
cool stove (i.e., through which a cold blast has blown for one-half period) rather than
being bypassed. The amount of blast through the hot stove is constantly increased while
that through the cool stove is decreased by the same amount to maintain the hot blast air
temperature approximately constant. At the end of one-half period, the hot stoves inlet
valve is fully open and the cool stoves inlet valve is fully closed. At this point, the cool
stove is put on gas, the hot stove becomes the cool stove, and a new hot stove is
switched in.
The heat transfer surface used in the aforementioned high-temperature xed-matrix
regenerator is made of refractory bricks, referred to simply as checkerwork. The commonly used checker shapes surface area density range is 25 to 42 m2 /m3 (8 to 13 ft2 /ft3 ),
as shown in Fig. 1.47. The checker ow passage (referred to as a ue) size is relatively
large, primarily to accommodate the fouling problem associated with highly corrosive
hot exhaust gases coming to the regenerator. A typical heat transfer coecient in such a
passage is about 5 W/m2 K (1 Btu/hr-ft2 -8F).
The surface geometries used for a compact xed-matrix regenerator are similar to
those used for rotary regenerators. The surface geometries used for packed beds are
quartz pebbles, steel, copper, or lead shots, copper wool, packed bers, powders,
randomly packed woven screens, and crossed rods. Heat transfer surface area densities
of 82,000 m2 /m3 (25,000 ft2 /ft3 ) are achievable; the heat transfer coecient range is 50 to
200 W/m2 K (9 to 35 Btu/hr-ft2 -8F).
The design exibility of selecting dierent frontal areas is not possible for a xedmatrix regenerator having multiple matrices, but instead, dierent hot and cold ow
periods are selected. The pressure leakage in a xed-matrix regenerator is through the
imperfect valves after wear and tear and through the cracks of matrix walls. Fixedmatrix regenerators can be used for large ow rates and can have very large surface areas
and high-heat-capacity material, depending on the design requirements.
FIGURE 1.47 Checkers used for a blast furnace regenerator (Shah, 1981).
56
1.6
Common ow arrangements of the uids in a heat exchanger are classied in Fig. 1.1.
The choice of a particular ow arrangement is dependent on the required exchanger
eectiveness, available pressure drops, minimum and maximum velocities allowed,
uid ow paths, packaging envelope, allowable thermal stresses, temperature levels,
piping and plumbing considerations, and other design criteria. Let us rst discuss the
concept of multipassing, followed by some of the basic ideal ow arrangements for a twouid heat exchanger for single- and multipass heat exchangers.
Multipassing. The concept of multipassing applies separately to the uid and heat
exchanger. A uid is considered to have made one pass if it ows through a section
of the heat exchanger through its full length. After owing through one full length, if
the ow direction is reversed and uid ows through an equal- or dierent-sized
section, it is considered to have made a second pass of equal or dierent size. A heat
FIGURE 1.48 (a) Two-pass cross-counterow exchanger; (b) single-pass crossow exchanger;
(c, d) unfolded exchangers of (a) and (b), respectively.
57
exchanger is considered as a single-pass unit if both uids make one pass in the
exchanger or if it represents any of the single-pass ow arrangements when the multipass uid side is unfolded (note that the folding is used to control the envelope size). To
illustrate the concept, consider one exchanger with two dierent designs of inlet headers
for uid 2 as shown in Fig. 1.48a and b; uid 1 makes a single pass, and uid 2 makes
two passes in both exchangers. If the exchanger of Fig. 1.48b with uid 2 unmixed in
the headers is unfolded to the horizontal direction (the exchanger length for uid 2 will
be 2L1 ), as in Fig. 1.48d,{ the resulting geometry is a single-pass exchanger having the
same inlet temperatures as uids 1 and 2 of Fig. 1.48b. Hence, the exchanger of Fig.
1.48b is considered a single-pass exchanger from the exchanger analysis point of view. In
contrast, the temperature of uid 1 at the inlet to the rst and second pass of uid 2 is
dierent in Fig. 1.48a. Hence, when it is unfolded vertically as in Fig. 1.48c, the inlet
temperature of uid 1 to each half of the exchanger will be dierent, due to the presence
of two passes, each with one-half of the original ow length L2 . This does not correspond to a single-pass exchanger of the unfolded exchanger height. Therefore, the
exchanger of Fig. 1.48a is considered as a two-pass exchanger. An additional degree
of freedom is introduced by unfolding. This degree of freedom describes how to lead a
uid between the passes (see, e.g., the case considered in Fig. 1.48c, uid 1). Depending
on how the uid is routed from the exit of one pass to the inlet of the following pass,
several distinct ow arrangements can be identied (see Section 1.6.2.1 for further
details).
1.6.1
Single-Pass Exchangers
In unfolded exchangers of Fig. 1.48c and d, the U-bend lengths of the tubes are neglected for the present
discussion since they do not take an active part in heat transfer between two uids.
{
This ow arrangement can be rigorously identied as a countercurrent parallel stream. However, based on Kays
and Londons (1998) terminology, used widely in the literature, we use the term counterow for this ow arrangement throughout the book.
58
FIGURE 1.49 (a) Double-pipe heat exchanger with pure counterow; (bf ) plate-n exchangers
with counterow core and crossow headers (Shah, 1981).
59
FIGURE 1.51 (a) Hot-side solid and uid temperature excursion; (b) balanced (Ch Cc )
regenerator temperature distributions at the switching instant (Shah, 1991b).
exists at the inlet side, which may induce high thermal stresses in the exchanger wall at
the inlet. Although this ow arrangement is not used for applications requiring hightemperature eectiveness, it may be used in the following applications:
1. It often produces a more uniform longitudinal tube wall temperature distribution
and not as high or low tube wall temperature as in a counterow arrangement
at the same surface area (NTU),{ uid ow rates (uid heat capacity rates or C*),
and uid inlet temperatures (see Example 3.2). For this reason, a parallelow
exchanger is sometimes used with temperature-sensitive materials, highly viscous
liquids, and metal recuperators having inlet temperatures in excess of 11008C
(20008F).
2. The lowest wall temperature for the parallelow exchanger is higher than that for
the counterow or other ow arrangements for the same NTU, C*, and uid inlet
temperatures, although the exchanger effectiveness will also be lower. Thus, if acid
vapors are present in the exhaust gas, the parallelow arrangement minimizes or
60
3.
4.
5.
6.
avoids condensation of acid vapors and hence corrosion of the metal surface. The
parallelow exchanger may be preferred when there is a possibility that the temperature of the warmer uid may reach its freezing point.
The highest wall temperature for the parallelow exchanger is lower than that for
the counterow or other ow arrangements for the same NTU, C*, and inlet
temperatures. This may eliminate or minimize the problems of fouling, wall
material selections, and uid decomposition.
It provides early initiation of nucleate boiling for boiling applications.
A large change in NTU causes a relatively small change in " for NTU > 2, as
shown in Fig. 3.8. Thus a parallelow exchanger is preferred if the desired exchanger effectiveness is low and is to be maintained approximately constant over a large
ow rate range (e.g., for NTU 1 to 5 or higher at C* 1, see Fig. 3.8).
The application allows piping suited only to parallelow.
1.6.1.3 Crossow Exchanger. In this type of exchanger, as shown in Fig. 1.53, the two
uids ow in directions normal to each other. Typical uid temperature variations are
idealized as two-dimensional and are shown in Fig. 1.54 for the inlet and outlet sections
only. Thermodynamically, the eectiveness for the crossow exchanger falls in between
that for the counterow and parallelow arrangements. The largest structural temperature dierence exists at the corner of the entering hot and cold uids, such as point a
in Fig. 1.54. This is one of the most common ow arrangements used for extendedsurface heat exchangers, because it greatly simplies the header design at the entrance
and exit of each uid. If the desired heat exchanger eectiveness is high (such as greater
than 80%), the size penalty for the crossow exchanger may become excessive. In such a
case, a counterow unit is preferred. This ow arrangement is used in a TEMA X shell
(see Fig. 1.6) having a single tube pass. The length L3 (or the height in the x direction)
FIGURE 1.53 (a) Plate-n unmixedunmixed crossow heat exchanger; (b) serpentine (one tube
row) tube-n unmixedmixed crossow heat exchanger (Shah, 1981).
61
in Fig. 1.53a does not represent the ow length for either uid 1 or uid 2. Hence, it is
referred to as noow height or stack height since the ns are stacked in the L3 direction.
In a crossow arrangement, mixing of either uid stream may or may not occur,
depending on the design. A uid stream is considered unmixed when it passes through
individual ow channels or tubes with no uid mixing between adjacent ow channels. In
this case within the exchanger, temperature gradients in the uid exist in at least one
direction (in the transverse plane) normal to the main uid ow direction. A uid stream
is considered completely mixed when no temperature gradient exists in the transverse
plane, either within one tube or within the transverse tube row within the exchanger.
Ideally, the uid thermal conductivity transverse to the ow is treated as zero for the
unmixed-uid case and innity for the mixed-uid case. Fluids 1 and 2 in Fig. 1.53a are
unmixed. Fluid 1 in Fig. 1.53b is unmixed, while uid 2 is considered mixed because there
is only one ow channel. The temperature of an unmixed uid, such as uid 1 in Fig. 1.53,
is a function of two coordinates z and y within the exchanger, and it cannot be treated as
constant across a cross section (in the y direction) perpendicular to the main ow direction x. Typical temperature distributions of the unmixed uids at exchanger outlet
sections are shown in Fig. 1.54. The outlet temperature from the exchanger on the
unmixed side is dened as a mixed mean temperature that would be obtained after
complete mixing of the uid stream at the exit. For the cases of Fig. 1.53, it is idealized
that there is no variation of the temperature of either uid in the x direction. The
temperature of a mixed uid (uid 2 in Fig. 1.53b) is mainly dependent on the coordinate
y. The temperature change per pass (in the x direction) of uid 2 in Fig. 1.53b is small
compared to the total.
In a multiple-tube-row tubular crossow exchanger, the tube uid in any one tube is
considered mixed at any cross section. However, when split and distributed in dierent
tube rows, the incoming tube uid is considered unmixed between the tube rows.
Theoretically, it would require an innite number of tube rows to have a truly unmixed
uid on the tube side. In reality, if the number of tube rows is greater than about four, it
will practically be an unmixed side. For an exchanger with fewer than about four or ve
tube rows, the tube side is considered partially unmixed or partially mixed. Note that
when the number of tube rows is reduced to one, the tube uid is considered mixed.
62
Mixing thus implies that a thermal averaging process takes place at each cross section
across the full width of the ow passage. Even though the truly unmixed and truly mixed
cases are the extreme idealized conditions of a real situation in which some mixing exists,
the unmixed condition is nearly satised in many plate-n and tube-n (with at ns)
exchanger applications. As will be shown in Section 11.3 and in the discussion of Example
3.5, for the same surface area and uid ow rates, (1) the exchanger eectiveness generally
decreases with increasing mixing on any uid side, although counter examples can be
found in the multipass case; and (2) if the Cmax uid is placed on the unmixed uid side,
the exchanger eectiveness and performance will be higher than that for placing Cmax on
the mixed uid side.
Seven idealized combinations of ow arrangements for a single-pass crossow
exchanger are shown symbolically in Fig. 1.55. The ow arrangements are:
(a) Both uids unmixed. A crossow plate-n exchanger with plain ns on both sides
represents the both uids unmixed case.
(b) One uid unmixed, the other mixed. A crossow plate-n exchanger with ns on one
side and a plain gap on the other side would be treated as the unmixedmixed case.
63
(c) Both uids mixed. This case is practically less important, and represents a limiting
case of some multipass shell-and-tube exchangers (e.g., 11 TEMA E and J), as
presented later.
(d) One uid unmixed and coupled in identical order, the other partially mixed. Here
identical order refers to the fact that a uid coupled in such order leaves the rst
row at the point where the other uid enters (leaves) the rst row, and enters the
other row where the second uid enters (leaves) that row (see the stream AA in
Fig. 1.55d). A tube-n exchanger with at ns represents the case of tube uid
partially mixed, the n uid unmixed. When the number of tube rows is reduced
to one, this exchanger reduces to the case of out-of-tube (n) uid unmixed the
tube uid mixed (case b). When the number of tube rows approaches innity (in
reality greater than four), the exchanger reduces to the case of both uids
unmixed (case a).
(e) One uid partially unmixed, the other partially mixed. The case of one uid (uid
1) partially unmixed (i.e., mixed only between tube rows) and the other (uid 2)
partially mixed (see Fig. 1.55e) is of less practical importance for single-pass
crossow exchangers. However, as mentioned later (see the middle sketch of
Fig. 1.58b with the notation of uids 1 and 2 interchanged),{ it represents the
side-by-side multipass crossow arrangement. When the number of tube rows is
reduced to one, this exchanger is reduced to the case of out-of-tube uid unmixed,
the tube uid mixed. When the number of tube rows approaches innity,
the exchanger reduces to the case of out-of-tube uid mixed, the tube uid
unmixed.
(f) One uid unmixed and coupled in inverted order, the other partially mixed. Here,
the term inverted order refers to the fact that a uid coupled in such order leaves
the rst row at the point where the other uid enters (leaves) the rst row and
enters the other row where the second uid leaves (enters) that row (see the
stream AA in Fig. 1.55f ). This case is also of academic interest for single-pass
crossow exchangers.
(g) One uid mixed, the other partially mixed. This is the case realized in plain tubular
crossow exchangers with a few tube rows.
1.6.1.4 Split-Flow Exchanger, TEMA G Shell. In this exchanger, shown in Fig. 1.56a,
the shell uid stream enters at the center of the exchanger and divides into two streams.
These streams ow in longitudinal directions along the exchanger length over a longitudinal bae, make a 1808 turn at each end, ow longitudinally to the center of the
exchanger under the longitudinal bae, unite at the center, and leave from the central
nozzle. The other uid stream ows straight in the tubes. Typical temperature distributions for the two uids in this exchanger are shown in Fig. 1.56. This single-pass ow
arrangement is found in the TEMA G shell (see Fig. 1.6). Another variant is a doublesplit ow arrangement, as found in the TEMA H shell (see Fig. 1.6), again having a
single tube pass.
{
In Fig. 1.58b, this means uid 1 partially mixed (considering mixed in each individual passes) and uid 2 partially
unmixed (i.e., unmixed within a pass and mixed between passes).
64
FIGURE 1.56 (a) Single-pass split-ow (TEMA G) exchanger; (b) idealized shell uid and tube
uid temperature distributions.
1.6.2
Multipass Exchangers
When the design of a heat exchanger results in either an extreme length, signicantly low
uid velocities, or a low eectiveness (sometimes maybe other design criteria), a multipass heat exchanger or several single-pass exchangers in series, or a combination of both,
is employed. Heat exchangers in any of the ve basic ow arrangements of Section 1.6.1
can be put into series to make a multipass unit. In addition, there exist other multipass
ow arrangements that have no single-pass counterpart. One of the major advantages of
proper multipassing is to increase the exchanger overall eectiveness over the individual
pass eectivenesses, but with increased pressure drop on the multipass side. If the overall
direction of the two uids is chosen as counterow (see Figs. 1.58a left and 1.62), the
exchanger overall eectiveness approaches that of a pure counterow exchanger as
65
FIGURE 1.57 (a) Single-pass divided-ow (TEMA J) exchanger with shell uid mixed;
(b) idealized shell and tube uid temperature distributions.
the number of passes increases. The multipass arrangements are classied according to
the type of construction: for example, extended surface, shell-and-tube, or plate
exchangers (see Fig. 1.1).
1.6.2.1 Multipass Crossow Exchangers. This arrangement is the most common for
extended surface exchangers; two or more passes are put in series, with each pass usually
having crossow, although any one of the single-pass basic ow arrangements could be
employed. The ow arrangements could be categorized as (a) a series coupling of n
passes or over-and-under passes, (b) a parallel coupling of n passes or side-by-side passes,
and (c) a combination of both or a compound arrangement. These are shown in Fig.
1.58. Each module in Fig. 1.58 can be either an individual pass or an individual heat
exchanger. In the series coupling of n passes, each of the uid streams is in series;
whereas in the parallel coupling of n passes, one uid stream is in series, the other in
parallel. The parallel coupling (side-by-side) two-pass arrangement is also referred to as
the face-U ow arrangement. For the same surface area, uid ow rates and inlet
temperatures, a series-coupled overall counterow multipass exchanger yields higher
eectiveness and heat transfer rate than that for a parallel-coupled multipass exchanger,
as will be shown in Example 3.5. In a series-coupled multipass exchanger, usually the
ow direction is chosen such that an overall counterow is obtained, as shown in Fig.
1.58a, to obtain higher exchanger eectiveness. This arrangement is then referred to as
n-pass cross-counterow. If the direction of uid 2 in Fig. 1.58a is reversed, overall
66
FIGURE 1.58 Examples of multipass exchangers: (a) series coupling or over-and-under pass
arrangement; (b) parallel coupling or side-by-side pass arrangement; (c) compound coupling.
67
FIGURE 1.59 Two-pass cross-parallelow exchangers with both uids unmixed throughout. For
the tube uid in the pass return: (a) identical order; (b) inverted order. Cases (c) and (d) are symbolic
representations of cases (a) and (b), respectively. In both cases (a) and (b), air (out-of-tube uid) is in
inverted order.
with the air in pass 1, and the same uid in row 3 (the rst tube in pass 2) is rst in contact
with air in pass 2. The tube uid between passes for this arrangement is then dened to be
in identical order. In Fig. 1.59b, the tube uid stream in row 1 (the rst tube in pass 1) is
again rst in contact with air in pass 1. However, the dierent uid stream (or row
connected to the second tube in pass 1) makes the rst contact with air in pass 2. In
this arrangement, the tube uid between passes is said to be in inverted order. In either
Fig. 1.59a or b, the air-side uid is in inverted order between passes. This is because the
airstream that crossed the tube inlet end in the rst pass then crosses the tube exit end in
the second pass (see the rst vertical airstream, S1 in Fig. 1.59c and d). Figures 1.59a and
b are represented symbolically as Fig. 1.59c and d, respectively, since one does not have
always tubes in the exchanger, for example, in a plate-n multipass exchanger.
Multipassing of crossow exchangers retains the header and ducting advantages of a
simple crossow exchanger, while it is possible to approach the thermal performance of
a true counterow heat exchanger using an overall cross-counterow arrangement. The
maximum temperature dierences in the wall across the wall thickness direction (sometimes referred to as structural temperature dierences) are considerably reduced in a
multipass cross-counterow exchanger relative to a single-pass crossow design for the
same terminal temperatures. For high-temperature applications (approximately above
4508C or 8508F), the heat exchanger can be divided into two or more passes having
distinct operating temperature ranges. Special metals (such as stainless steel and superalloys) may be used in passes having high operating temperatures, and ordinary metals
68
FIGURE 1.60 (a) A 12 TEMA E heat exchanger (one shell pass and two tube passes);
(b) corresponding temperature distributions.
69
E exchanger. Similarly, we simply designate the 12n TEMA E exchanger as the 12n
exchanger.
As the tubes are rigidly mounted only at one end, thermal expansion is readily
accommodated. If the shell uid is idealized as well mixed, its temperature is constant
at any cross section but changes from a cross section to another cross section along the
shell length direction. In this case, reversing the tube uid ow direction will not change
the idealized temperature distribution of Fig. 1.60b and the exchanger eectiveness.
Increasing the even number of tube passes of a 12n exchanger from two to four, six,
and so on, decreases the exchanger eectiveness slightly, and in the limit when the
number of tube passes approaches innity with one shell pass, the exchanger eectiveness
approaches that for a single-pass crossow exchanger with both uids mixed. Common
tube-side multipass arrangements are shown in Fig. 1.61.{
The odd number of tube passes per shell has slightly better eectiveness when the shell
uid ows countercurrent to the tube uid for more than one half the tube passes.
However, this is an uncommon design and may result in structural and thermal problems
in manufacturing and design.
Each sketch in Fig. 1.61 represents a cross section of the shell and tube uid nozzles at the inlet and pass
partitions. The dashed lines are the pass partitions on the other end of the tube bundle. No tubes or baes are
shown for clarity. Also, no horizontal orientation of the nozzles is shown, although horizontal nozzles are
common for some applications.
70
FIGURE 1.62 (a) Two shell passfour tube pass exchanger; (b) three shell passsix tube pass
exchanger.
Since the 12n exchanger has a lower eectiveness than that of a counterow exchanger, multipassing of the basic 12 arrangement may be employed with multiple shells
(each shell as a 12 exchanger) to approach the counterow eectiveness. The heat
exchanger with the most general ow arrangement would have m shell passes and n
tube passes. Figure 1.62 represents two such exchangers.
Split-Flow Exchanger, TEMA G Shell. In this exchanger, there is one central inlet and
one central outlet nozzle with a longitudinal bae, as shown in Fig. 1.63a. Typical
temperature distribution is shown in Fig. 1.63b. This arrangement is used in the TEMA
FIGURE 1.63 (a) A 12 split ow (TEMA G) exchanger; (b) idealized shell uid and tube uid
temperature distributions.
71
FIGURE 1.64 (a) A 12 divided ow (TEMA J) exchanger with shell uid mixed; (b) idealized
shell uid and tube uid temperature distributions.
72
the two looped patterns, the U-arrangement (Fig. 1.65a) is usually preferred over the Zarrangement (Fig. 1.65b) since it allows all connections to be made on the same side of
the frame. This eliminates the need for disconnecting pipework for maintenance and
cleaning purposes.
A complex ow arrangement results by combining Z-arrangements in series with a
generally identical number of thermal plates in each pass. Although only three such ow
arrangements are shown in Fig. 1.65ce, many other combinations are possible (see, e.g.,
Table 3.6). Primarily, these arrangements are used when there is a signicant dierence in
the ow rates of the two uid streams and the corresponding available pressure drops.
Generally, the uid, having very low permissible pressure drop, goes through the single
pass; the other uid goes through multiple passes in order to utilize the available pressure
drop and pumping power. Also, if the ow rates are signicantly dierent, the uid
having the lower ow rate goes through n (> 1) passes such that in each pass the heat
capacity rates of both uid streams are about equal. This would produce approximately
equal heat transfer coecients on each uid side, resulting in a balanced exchanger (hA
values approximately the same). Multipass arrangements always have ports located on
xed and movable end plates.
In the series ow arrangement (Fig. 1.65 f ), each ow passage represents a pass. The
series arrangement is used for small uid ow rates that must undergo a large temperature dierence. It is used for very close temperature approaches. Because of many ow
reversals, a signicant portion of the available pressure drop is wasted in reversals (i.e.,
the pressure drop in the series ow arrangement is extremely high). The manifold-
FIGURE 1.65 Single- and multipass plate heat exchanger arrangements. Looped or single-pass
arrangements: (a) U arrangement; (b) Z arrangement. Multipass arrangements: (c) 2 pass 1 pass,
(d) 3 pass 1 pass, (e) 4 pass 2 pass, and ( f ) series ow.
73
The basic heat transfer mechanisms employed for transfer of thermal energy from the
uid on one side of the exchanger to the wall (separating the uid on the other side) are
single-phase convection (forced or free), two-phase convection (condensation or
evaporation, by forced or free convection), and combined convection and radiation
heat transfer. Any of these mechanisms individually or in combination could be active
on each uid side of the exchanger. Such a classication is provided in Fig. 1.1.
Some examples of each classication type are as follows. Single-phase convection
occurs on both sides of the following two-uid exchangers: automotive radiators and
passenger space heaters, regenerators, intercoolers, economizers, and so on. Single-phase
convection on one side and two-phase convection on the other side (with or without
desuperheating or superheating, and subcooling, and with or without noncondensables)
occur in the following two-uid exchangers: steam power plant condensers, automotive
and process/power plant air-cooled condensers, gas or liquid heated evaporators, steam
generators, humidiers, dehumidiers, and so on. Two-phase convection could occur on
each side of a two-uid heat exchanger, such as condensation on one side and evaporation on the other side, as in an air-conditioning evaporator. Multicomponent two-phase
convection occurs in condensation of mixed vapors in distillation of hydrocarbons.
Radiant heat transfer combined with convective heat transfer plays a role in liquid
metal heat exchangers and high-temperature waste heat recovery exchangers.
Radiation heat transfer is a primary mode in fossil-fuel power plant boilers, steam
generators, coal gasication plant exchangers, incinerators, and other red heat
exchangers.
SUMMARY
Heat exchangers have been classied according to transfer processes, number of uids,
degrees of surface compactness, construction features, ow arrangements, and heat
transfer mechanisms. A summary is provided in Fig. 1.1. The major emphasis in this
chapter is placed on introducing the terminology and concepts associated with a broad
spectrum of commonly used industrial heat exchangers (many specialized heat exchangers are not covered in this chapter). To acquaint the reader with specic examples, major
applications of most types of heat exchangers are mentioned. With a thorough understanding of this broad overview of dierent types of exchangers, readers will be able to
apply the theory and analyses presented in the succeeding chapters to their specic needs.
REFERENCES
ASME, 1998, ASME Boiler and Pressure Vessel Code, Sec. VIII, Div. 1, Rules for Construction of
Pressure Vessels, American Society of Mechanical Engineers, New York.
Beck, D. S., and D. G. Wilson, 1996, Gas Turbine Regenerators, Chapman & Hall, New York.
74
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your answers
briey.
1.1
REVIEW QUESTIONS
1.2
1.3
75
1.4
Louver ns (as compared to similar plain uncut ns) are used primarily to:
(a) increase heat transfer coecient h (b) increase surface area A
(c) increase both h and A
(d) increase neither h nor A
1.5
A nned double-pipe exchanger has ns on the outside of the inner tube(s) for the
following reasons:
(a) The tube outside heat transfer coecient is high.
(b) The tube inside heat transfer coecient is more than double for tube outside
with longitudinal ow.
(c) Fouling is expected on the tube side.
1.6
Which one of the following is not a function fullled by transverse plate baes in a
shell-and-tube exchanger?
(a) to provide counterow operation
(b) to support the tubes
(c) to direct the uid approximately at right angles to the tubes
(d) to increase the turbulence and mixing in the shell uid
(e) to minimize tube-to-tube temperature dierences and thermal stresses
1.7
Which of the following properties of plate heat exchangers, due to their specic
construction features, make them particularly suited for the food processing
industry?
(a) close temperature control
(b) easy disassembly for cleaning
(c) low probability of one uid to other uid contamination
(d) high corrosion resistance
1.8
1.9
1.10
76
1.11
A single-coolant-tube-row car radiator is a crossow heat exchanger with following uid streams:
(a) mixedmixed
(b) mixedunmixed
(c) unmixedunmixed
1.12
A truck radiator with six coolant-tube rows and multilouver air centers is a crossow heat exchanger with following uid streams:
(a) mixedmixed
(b) mixedunmixed
(c) unmixedunmixed
1.13
1.14
1.15
Which of the following are possible reasons for using a cross-parallelow instead
of a cross-counterow multipass exchanger?
(a) higher eectiveness
(b) less prone to core freeze-up near the
cold uid inlet
(c) reduced thermal stresses
(d) reduced size
(e) reduced higher axial temperature gradient in the wall
1.16
1.17
Name the specic exchanger construction types used in the following applications:
(a) milk pasteurizing: __________
(b) sulfuric acid cooling: _________
(c) automotive radiator: _________
(d) blast furnace air preheating: _______
(e) air-cooled condenser: _________
REVIEW QUESTIONS
77
1.18
1.19
For the identical average inlet and outlet uid temperatures, arrange the following
exchangers in terms of decreasing largest structural temperature dierences across
the wall thickness direction:
(a) parallelow
(b) counterow
(c) four-pass overall crosscounterow
(d) two-pass overall cross-counterow
(e) single-pass crossow
Now can you tell which exchanger will have the highest thermal stresses in the
dividing walls between two uids and which will have the least thermal stresses?
Why? Hint: Review the temperature distributions of the hot and cold uids and
of the wall.
1.20
1.21
Name ve heat exchangers that you are familiar with and classify them in proper
subcategories of six major schemes of Fig. 1.1.
2.1
At the initial stage, an engineer must specify requirements and dene the main goal of
the system design. This must be based on a good understanding of customer needs. If the
problem is clearly formulated, an engineer evaluates alternative concepts of the system
design and selects one or more workable design solutions. Based on this analysis, detailed
sizing, costing, and optimization have to be completed. This activity leads to a proposed
design solution. Simultaneously, project engineering (construction/manufacturing)
78
79
considerations should be taken into account. The issues related to startups, transients,
steady and erratic operations, and ultimately, the retirement, should be considered as
well. Through consideration of these steps, a design team reconsiders the conclusions
and, in the light of the constraints imposed, iterates one or more steps until all the
requirements are met within the tolerable limits. Within the framework of these activities,
a particular design methodology has to be developed.
A methodology for designing a new (single) heat exchanger is illustrated in Fig. 2.1; it
is based on experience and presented by Kays and London (1998), Taborek (1988), and
Shah (1982) for compact and shell-and-tube exchangers. This design procedure may be
characterized as a case study (one case at a time) method. Major design considerations
include:
.
.
.
.
.
These design considerations are usually not sequential; there could be strong interactions and feedback among the aforementioned considerations, as indicated by doublesided arrows in Fig. 2.1, and may require a number of iterations before the design is
nalized. The overall design methodology is quite complex because of the many qualitative judgments, in addition to quantitative calculations, that must be introduced. It
should be emphasized that depending on the specic application, some (but not necessarily all) of the foregoing considerations of heat exchanger designs are applied in various
levels of detail during the design process. In the following, these broad considerations are
discussed in some detail, accompanied by several examples showing the main features of
the design procedures and interactions among them. Refer to appropriate blocks and
boxes in Fig. 2.1 for identication of the following specic sections and subsections.
2.1.1
The process and problem specication (the top dashed block in Fig. 2.1) is one of the
most important steps in heat exchanger design. A heat exchanger design engineer can add
the most value by working together with a system design engineer to develop smart
specications for the heat exchanger that dene an optimum system. The smart specications need to be completed based on discussions with the customer, on industry and
customer standards, and on design engineers own experiences.
Process or design specications include all necessary information to design and
optimize an exchanger for a specic application. It includes problem specications for
operating conditions, exchanger type, ow arrangement, materials, and design/manufacturing/operation considerations. In addition, the heat exchanger design engineer provides necessary and missing information on the minimum input specications required.
2.1.1.1 Problem Specications. The rst and most important consideration is to select
the design basis (i.e., design conditions). Next comes an analysis of the performance at
the design point and o-design (turndown) conditions. The design basis would require
the specication of operating conditions and the environment in which the heat exchan-
80
FIGURE 2.1 Heat exchanger design methodology. (Modied from Shah, 1982; Taborek, 1988;
and Kays and London, 1998.)
81
ger is going to be operated. These include uid mass ow rates (including uid types
and their thermophysical properties), inlet temperatures and pressures of both uid
streams, required heat duty and maximum allowed pressure drops on both uid
sides, uctuations in inlet temperatures and pressures due to variations in the process
or environment parameters, corrosiveness and fouling characteristics of the uids, and
the operating environment (from safety, corrosion/erosion, temperature level, and
environmental impact points of view). In addition, information may be provided on
overall size, weight, and other design constraints, including cost, materials to be used,
and alternative heat exchanger types and ow arrangements. If too many constraints
are specied, there may not be a feasible design, and some compromises may be needed
for a solution. The heat exchanger designer and system design engineer should work
together at this stage to prepare the complete smart specications for the problem.
However, in some industries the heat exchanger designer is constrained by dumb
specications that he or she inherits and has little or no opportunity to provide input.
2.1.1.2 Exchanger Specications. Based on the problem specications and the design
engineers experience, the exchanger construction type and ow arrangement (see Fig.
1.1) are rst selected. Selection of the construction type depends on the uids (gas,
liquid, or condensing/evaporating) used on each side of a two-uid exchanger, operating pressures, temperatures, fouling and cleanability, uids and material compatibility,
corrosiveness of the uids, how much leakage is permissible from one uid to the other
uid, available heat exchanger manufacturing technology, and cost. The choice of a
particular ow arrangement is dependent on the required exchanger eectiveness,
exchanger construction type, upstream and downstream ducting, packaging envelope/
footprint, allowable thermal stresses, and other criteria and design constraints. The
orientation of the heat exchanger, the locations of the inlet and outlet pipes, and so
on, may be dictated by the system and/or available packaging/footprint space and
ducting. Some guidelines on selection of the exchanger type and ow arrangements
are provided in Section 10.2.
Next, the core or surface geometry and material are selected. The core geometry (such
as shell type, number of passes, bae geometry, etc.) is selected for a shell-and-tube
exchanger, while the surface geometry is chosen for a plate, extended surface, or regenerative heat exchanger. There are several quantitative and qualitative criteria for surface
selection. Some of the qualitative and quantitative criteria for compact heat exchanger
surfaces are discussed in Sections 10.2 and 10.3. The qualitative criteria for surface
selection are the operating temperature and pressure, the designers experience and judgment, fouling, corrosion, erosion, uid contamination, cost, availability of surfaces,
manufacturability, maintenance requirements, reliability, and safety. For shell-andtube exchangers, the criteria for selecting core geometry or conguration are the desired
heat transfer performance within specied pressure drops, operating pressures and temperatures, thermal/pressure stresses, the eect of potential leaks on the process, corrosion
characteristics of the uids, fouling, cleanability, maintenance, minimal operational
problems (vibrations, freeze-up, instability, etc.), and total installed cost. Some of
these are discussed in Section 10.2.
For shell-and-tube exchangers, the tube uid is usually selected as the one having
more fouling, high corrosiveness, high pressure, high temperature, increased hazard
probability, high cost per unit mass, and/or low viscosity. Maximum allowable pressure
drop will also dictate which uid will be selected for the tube side (high-pressure uid)
and which for the shell side.
82
For compact heat exchangers, such as for plate-n exchangers, one may have a choice
of considering oset strip n, louver n, or other n geometry. For each n geometry
selected, the thermal/hydraulic design and mechanical design of Fig. 2.1 are carried out.
This is what we mean by one or more cases in the box. One or more cases also include
dierent heat exchanger technologies.
Example 2.1 A hydrocarbon gas has to be cooled in a chemical plant. A stream of a
liquid hydrocarbon is available to be used as a coolant. The gas stream has to change its
temperature from 2558C to 308C. The liquid stream has the inlet temperature of 258C.
The required enthalpy change of the hot gas is smaller than 300 kW (with a small mass
ow rate of an order of magnitude 0.01 kg/s). Both uids are at relatively high pressures
(i.e., the respective pressures are of an order of magnitude 10 MPa). Is it possible, using
this incomplete set of process data, to oer an unambiguous selection of a feasible heat
exchanger type that will be capable of performing the task? Consider the following heat
exchanger types: shell-and-tube, double-pipe, welded plate, gasketed plate, lamella, and
spiral plate heat exchanger.
SOLUTION
Problem Data and Schematic: An incomplete database is provided to support a process
specication. Only information regarding some selected operating conditions is available
(a situation often encountered in practice). All the information regarding various heat
exchanger types under consideration (shell-and-tube, double-pipe, welded plate,
gasketed plate, lamella, and spiral plate heat exchanger) are available (see Sections
1.5.1.1, 1.5.1.2, 1.5.2.2, 1.5.2.1, 1.5.2.4, and 1.5.2.3, respectively).
Determine: Based on the available process specication data, select a feasible heat
exchanger type for further design considerations.
Assumptions: Design specications for the heat exchanger types listed are valid, as
discussed in Chapter 1.
Analysis and Discussion: One possible approach to a selection of the feasible heat exchanger type is rst to eliminate the types characterized with specications that conict with
the process conditions. The rst important fact to be considered is related to the operating temperature ranges and pressures. A study of various designs (see Chapter 1) leads to
the conclusion that lamella, welded, and gasketed plate heat exchangers cannot be used
because the allowable operating pressures and temperatures for them are both substantially lower than the process condition data imposed by the problem formulation. More
precisely, lamella heat exchangers usually operate at pressures lower than 3.45 MPa and
temperatures lower than 2008C (see Section 1.5.2.4). For welded and gasketed plate heat
exchangers, these parameters are 4.0 MPa and 3508C, and 3 MPa and 1508C, respectively
(see Sections 1.5.2.2 and 1.5.2.1). A spiral heat exchanger can operate at much higher
temperatures (up to 5008C; see Section 1.5.2.3), but the pressure limitation is 2 MPa. So,
only two remaining types should be considered. This means that only shell-and-tube and
double-pipe heat exchangers are feasible candidates.
Both shell-and-tube and double-pipe heat exchangers (see Sections 1.5.1.1 and
1.5.1.2) can easily sustain high pressures and temperatures. Consequently, other criteria
should be considered for the selection. These criteria include the required heat exchanger
83
eectiveness, the heat load, uid characteristics (such as fouling and corrosion ability),
cost, and others. For a relatively small heat load (i.e., smaller than 500 kW), a doublepipe heat exchanger would be a cost-eective solution. Also, for higher performance, a
multitube double-pipe heat exchanger with or without ns should be considered if cost
considerations support this decision. See Example 2.4 for the inclusion of cost considerations in a heat exchanger selection.
Finally, a decision should be made whether to use nned or plain tubes in the doublepipe multitube heat exchanger selected. Due to the fact that the heat exchanger should
accommodate both a gas and a liquid, the heat transfer conductance (hA) on the gas side
(with low gas mass ow rate) will be low. Hence, employing ns on the gas side will yield
a more compact unit with approximately balanced hA values on the gas and liquid sides.
Note also that the tube uid (liquid hydrocarbon) is more prone to fouling. So a doublepipe multitube heat exchanger with nned tubes on the hydrocarbon gas side and liquid
hydrocarbon on the tube side has to be suggested as a feasible workable design.
2.1.2
Heat exchanger thermal/hydraulic design procedures (the second block from the top in
Fig. 2.1) involve exchanger rating (quantitative heat transfer and pressure drop evaluation) and/or exchanger sizing. This block is the heart of this book, covered in Chapters 3
through 9. Only two important relationships constitute the entire thermal design
procedure. These are:
1. Enthalpy rate equations
q qj m_ j hj
2:1
2:2
84
rating problem). Note that the seven variables on the right-hand side of Eqs. (2.1) and
(2.2) are (m_ cp )j , Tj;i , Tj;o , with j 1 or 2 and UA.
Let us dene rst the heat exchanger thermal design problems. Subsequently, the
concept of a design method to solve the problem is introduced. Finally, the main inputs
to the design procedures are discussed.
2.1.2.1 Heat Exchanger Thermal Design Problems. From the quantitative analysis
point of view, there are a number of heat exchanger thermal design problems. Two
of the simplest (and most important) problems are referred to as the rating and sizing
problems.
Rating Problem. Determination of heat transfer and pressure drop performance of
either an existing exchanger or an already sized exchanger (to check vendors design)
is referred to as a rating problem. Inputs to the rating problem are the heat exchanger
construction, ow arrangement and overall dimensions, complete details on the materials and surface geometries on both sides, including their nondimensional heat transfer
and pressure drop characteristics ( j or Nu and f vs. Re),{ uid ow rates, inlet temperatures, and fouling factors. The uid outlet temperatures, total heat transfer rate,
and pressure drops on each side of the exchanger are then determined in the rating
problem. The rating problem is also sometimes referred to as the performance or
simulation problem.
Sizing Problem. In a broad sense, the design of a new heat exchanger means the
determination/selection of an exchanger construction type, ow arrangement, tube/
plate and n material, and the physical size of an exchanger to meet the specied
heat transfer and pressure drops within all specied constraints. However, in a sizing
problem for an extended surface exchanger, we will determine the physical size (length,
width, height, and surface areas on each side) of an exchanger; we will assume that
selection of the exchanger construction type, ow arrangement, and materials was
completed beforehand based on some of the selection guidelines presented in
Sections 10.1 and 10.2. For a shell-and-tube exchanger, a sizing problem in general
refers to the determination of shell type, diameter and length, tube diameter and number, tube layout, pass arrangement, and so on. For a plate exchanger, a sizing problem
means the selection of plate type and size, number of plates, pass arrangements, gasket
type, and so on. Inputs to the sizing problem are surface geometries (including their
dimensionless heat transfer and pressure drop characteristics), uid ow rates, inlet and
outlet uid temperatures, fouling factors, and pressure drops on each uid side. The
sizing problem is also referred to as the design problem. However, in the literature the
design problem is variously referred to as a rating or sizing problem. To avoid confusion
with the term design problem, we will distinctly refer to heat exchanger thermal design
problems as rating and sizing problems. The sizing problem is a subset of the comprehensive design process outlined in Fig. 2.1.
2.1.2.2 Basic Thermal and Hydraulic Design Methods. Based on the number of variables associated with the analysis of a heat exchanger, dependent and independent
dimensionless groups are formulated. The relationships between dimensionless groups
{
j, Colburn factor; Nu, Nusselt number; f, Fanning friction factor; Re, Reynolds number. They are dened in
Section 7.2.
85
86
87
UA
m_ cp 1
m_ cp 2
T1;i
T1;o
T2;i
T2;o
problem), and (2) only two variables can be considered as unknown since we have only
two equations [Eqs. (2.1) and (2.2)] at our disposal. Thus, the number of combinations
among the seven variables is six (i.e., in each case the two variables will be unknown and
the remaining ve must be known). This constitutes the list of six types of dierent sizing
problems given in Table E2.2.
Discussion and Comments: Among the six types of sizing problems, four have both heat
capacity rates [i.e., the products m_ cp j known]; and in addition to UA, the unknown is
one of the four terminal temperatures. The remaining two problem types presented in
Table E2.2 have one of the two heat capacity rates unknown and the other heat capacity
rate and all four temperatures known.
Exactly the same reasoning can be applied to devise a total of 15 rating problems
(in each of these problems, the product UA must stay known). A complete set of design
problems (including both the sizing and rating problems), devised in this manner, is
given in Table 3.11.
2.1.3
Mechanical Design
Mechanical design is essential to ensure the mechanical integrity of the exchanger under
steady-state, transient, startup, shutdown, upset, and part-load operating conditions
during its design life. As mentioned in the beginning of Chapter 1, the exchanger consists
of heat exchanging elements (core or matrix where heat transfer takes place) and uid
distribution elements (such as headers, manifolds, tanks, inlet/outlet nozzles, pipes, and
seals, where ideally, no heat transfer takes place). Mechanical/structural design should be
performed individually for these exchanger elements. Also, one needs to consider the
structural design of heat exchanger mounting. Refer to the third dashed-line block from
the top in Fig. 2.1 for a discussion of this section.
The heat exchanger core is designed for the desired structural strength based on the
operating pressures, temperatures, and corrosiveness or chemical reaction of uids with
materials. Pressure/thermal stress calculations are performed to determine the thicknesses of critical parts in the exchangers, such as the n, plate, tube, shell, and tubesheet.
A proper selection of the material and the method of bonding (such as brazing, soldering,
welding, or tension winding) ns to plates or tubes is made depending on the operating
temperatures, pressures, types of uids used, fouling and corrosion potential, design life,
and so on. Similarly, proper bonding techniques should be developed and used for tube-
88
to-header joints,{ tube-to-tubesheet joints, expansion joints, anges, and so on. These
bonding methods are usually decided upon before conducting the thermal-hydraulic
analysis. At this stage, attention should also be paid to operational problems. Thermal
stress and fatigue calculations are performed to ensure the durability and desired life of
the exchanger for expected startup and shutdown periods and for part-load operating
conditions. In addition, some of the less obvious operating problems should also be
addressed up front. A check should be made to eliminate or minimize ow-induced
vibrations, which may result in erosion, fatigue, and so on, of tubes leading to failure.
Flow velocities are checked to eliminate or minimize erosion, corrosion, and fouling.
Adequate provisions are also made for thermal expansion. At this stage, considerations
are also given to other operating problems, if they exist, such as dynamic instability
and freezing. Field experience, if any, becomes quite valuable at this step of design.
Fouling and corrosion are covered in Chapter 13.
In addition to the heat exchanger core, proper design of ow distribution devices
(headers, tanks, manifolds, nozzles, or inletoutlet pipes) is made to ensure that there is
uniform ow distribution through the exchanger ow passages, and that neither erosion
nor fatigue will be a problem during the design life of the exchanger. Header design and
ow maldistribution are covered in Chapter 12.
Either the exchanger is mounted on the oor/ground in a room or open environment
or in a system along with other components. The structural support for the heat exchanger needs to be designed properly with proper tabs, brackets, and other mounting features on the exchanger to ensure no failure of the supporting structure due to vibration,
impact loads, fatigue, and so on. In the mechanical design, consideration is also given to
maintenance requirements (such as cleaning, repair, serviceability, and general inspection) and shipping limitations, such as for overall size.
Every heat exchanger must comply with applicable local, state, national, and/or
international codes and standards (such as ASME pressure vessel codes, TEMA
standards, etc.), and should be designed accordingly for good mechanical design that
will also result in good thermal design. Particularly, the exchanger may need extensive
structural design to meet the codes and standards for one or more of the following
conditions: severe duty (extreme pressures and temperatures); considerable number of
pressure and temperature cycles during the design life; earthquake criteria; special application where destructive testing is not feasible, where reliability is critical, or where
repair/replacement is not easily eected; or when a customer species it. Structural design
would include thermal stresses, fatigue, and creep analyses in such a situation to dene
the life of the exchanger also.
Although some aspects of mechanical design are considered upfront before the
thermal design, a common practice in some exchangers is rst to design the exchanger
to meet the thermal/hydraulic requirements and then check that design from a structural
design viewpoint, conducting necessary iterations until the thermal/hydraulic requirements and structural design are both satised. Thus, a mechanical design of the exchanger is equally important and probably more dicult than the thermal design primarily
because it is not all analytical and is dierent for each specic case; one must rely on
experimentation, prior experience, and good design practice. Many mechanical design
criteria should be considered simultaneously or iteratively with thermal design.
{
The tubesheet in a tube-n exchanger is referred to as a header in the automotive industry. However, conical,
triangular, trapezoidal, or other-shaped ow distribution devices at the exchanger inlet and outlet with gas ows
are referred to as headers. See, for example, Section 12.4 for further details.
89
As shown in Fig. 2.1, several optimized optional solutions may be available when the
thermal and mechanical designs are completed. The designer addresses, in both parallel
and series, manufacturing considerations and cost estimating followed by trade-o factors to arrive at an optimum solution. For shell-and-tube exchangers, since TEMA
standards detail most of the mechanical design, the pricing of the exchanger is done
before nalizing the mechanical design; the nal drawings (solid models) are made
after a rm order is placed.
Example 2.3 An engineer has to perform a stress analysis within the scope of a mechanical design of a shell-and-tube heat exchanger for a given task. The following is the
available information: A TEMA E shell-and-tube heat exchanger has to be used for
cooling oil with water. All process conditions are known (uid types, thermophysical
properties, mass ow rates, temperature and pressure terminal conditions, fouling factors, etc.) The heat load is fully dened. Design specication data (ow arrangement,
tube inside and outside diameters, pitch, angle, tube material, shell inside diameter,
TEMA designation, etc.) are also known. It is specied that the heat exchanger has to
operate indoors and that the plant site is in a seismically inactive area. All data regarding
allowable stress limits and fatigue life requirements for the materials used are known. In
addition, the engineer has all the data needed to calculate thermal loads caused by
dierential thermal expansions between the shell and tubes. Enlist the important missing
data to perform the stress analysis.
SOLUTION
Problem Data and Schematic: All the process/design specication data, as well as the
information about allowable stresses, are known. Specic application of the exchanger is
known. The operating environment of the exchanger is specied. The schematic of the
TEMA E shell-and-tube heat exchanger (BEM) is given in Fig. 1.5a.
Determine: The missing set of data necessary for performing the stress analysis of the
heat exchanger described in the example formulation.
Assumptions: All usual assumptions for the heat exchanger type mentioned are satised.
Analysis: Inspection of the data indicates that most of the information needed for stress
analysis is available:
.
.
.
.
.
.
.
However, very important pieces of information are missing. These are mechanical,
superimposed, and operating transient loads. The loads include:
90
Manufacturing considerations and cost estimates (see the rst dashed-line major block
from the bottom in Fig. 2.1) are made for those optimized solutions obtained from
thermal and mechanical design considerations.
2.1.4.1 Manufacturing Considerations. Manufacturing considerations may be subdivided into manufacturing equipment considerations, processing considerations, and
other qualitative criteria. The equipment considerations that may inuence which
design should be selected include existing tooling versus new tooling; availability and
limitations of dies, tools, machines, furnaces, and manufacturing space; production
versus oine setup; and funding for capital investment. Processing considerations are
related to how individual parts and components of a heat exchanger are manufactured
and eventually assembled. This includes manufacturing of individual parts within specied tolerances; ow of parts; stacking of a heat exchanger core and eventual brazing,
soldering, welding, or mechanical expansion of tubes or heat transfer surfaces; leak-free
mounting ( joining) of headers, tanks, manifolds, or return hairpins on the heat exchanger core; mounting of pipes; washing/cleaning of the exchanger; leak testing, mounting
of the exchanger in the system; and structural supports. Not only the manufacturing
equipment but also the complete processing considerations are evaluated upfront nowadays when a new design of a heat exchanger is being considered, particularly for an
extended surface heat exchanger. Other evaluation criteria include the shop workload,
delivery dates, company policy, and estimate of the strength of the competition.
2.1.4.2 Costing. The overall total cost, also called lifetime costs, associated with a heat
exchanger may be categorized as the capital, installation, operating, and sometimes also
disposal costs. The capital (total installed) cost includes the costs associated with
design, materials, manufacturing (machinery, labor, and overhead), testing, shipping,
installation, and depreciation. Installation of the exchanger on the site can be as high as
the capital cost for some shell-and-tube and plate heat exchangers. The operating cost
consists of the costs associated with uid pumping power, warranty, insurance, main-
91
tenance, repair, cleaning, lost production/downtime due to failure, energy cost associated with the utility (steam, fuel, water) in conjunction with the exchanger in the
network, and decommissioning costs. Some of the cost estimates are dicult to obtain
and best estimates are made at the design stage.
Example 2.4 A heat exchanger designer needs to make a preliminary selection of a heat
exchanger type for known heat transfer performance represented by q=Tm [Eqs. (1.4)
and Eq. (2.2)]. The exchanger should operate with q=Tm 6:3 104 W/K. The criterion for selection at that point in the design procedure is the magnitude of the unit cost per
unit of q=Tm . From a preliminary analysis, the engineer has already selected four
possible workable design types as follows: (1) a shell-and-tube heat exchanger, (2) a
double-pipe heat exchanger, (3) a plate-and-frame heat exchanger, and (4) a welded
plate heat exchanger. From the empirical data available, the unit costs (in dollars per
unit of q=Tm ) for the two values of q=Tm are given in Table E2.4A. Idealize the
dependence of the unit cost vs. q=Tm as logarithmic. What is going to be the engineers
decision? Discuss how this decision changes with a change in the heat exchanger performance level.
SOLUTION
Problem Data and Schematic: From the available empirical data, the heat exchanger unit
cost per unit of heat exchanger performance level is known (see Table E2.4A).
Schematics of heat exchanger types selected are given in Figs. 1.5, 1.15, 1.16, and 1.20.
Determine: The heat exchanger type for a given performance. Formulate the decision
using the unit cost per unit of q=Tm as a criterion.
Assumptions: The cost of heat exchangers vs. q=Tm (W/K) is a logarithmic relationship.
All heat exchanger types selected represent workable designs (meet process/design
specications). The heat exchanger performance is dened by q=Tm as given by Eq.
(1.4).
Analysis: The analysis should be based on data provided in Table E2.4A. Because there
are no available data for the performance level required, an interpolation should be
performed. This interpolation must be logarithmic. In Table E2.4B, the interpolated
data are provided along with the data from Table E2.4A.
5 10 W/K
1 105 W/K
Shell-and-tube
Double pipe
Plate-and-frame
Welded plate
0.91
0.72
0.14
1.0
0.134
0.140
0.045
0.108
92
5 10 W/K
1 105 W/K
Shell-and-tube
Double pipe
Plate-and-frame
Welded plate
0.91
0.72
0.14
1.0
0.180
0.180
0.054
0.152
0.134
0.140
0.045
0.108
5 10 W/K
1 105 W/K
Double pipe
Plate-and-frame
Welded plate
1.26
6.5
0.91
1
3.33
1.18
0.96
2.98
1.08
Discussion and Comments: The following conclusions can be formulated. The doublepipe heat exchanger is more economical than the shell-and-tube type only for small
performance values. For higher performance levels, the unit cost may be exactly the
same for both shell-and-tube and double-pipe heat exchangers (for q=Tm
6:3 104 W/K, see the corresponding column in Table E2.4B), for larger q=Tm , a
shell-and-tube exchanger may be even cheaper (compare the numbers in the last column
of Table E2.4B). The most economical is the plate-and-frame heat exchanger, regardless of
the heat transfer performance level (see Table E2.4B, the last column). A welded plate heat
exchanger is the least economical for the small performance level (see the corresponding
values in the rst column of Table E2.4B). For large duties, the least desirable solution is a
double-pipe heat exchanger (it has a reduced unit cost value 0.96 compared to 2.98 and
1.08 for plate-and-frame and welded plate heat exchangers; Table E2.4C). Data presented
are based on an approximate costing method developed by Hewitt et al. (1982).
2.1.5
Trade-o Factors
Fairly accurate cost estimates are made as above to the various case studies, after detailed
evaluation of thermal, mechanical, and manufacturing design considerations. Now we
93
Optimum Design
The nal output of the quantitative and qualitative analyses is an optimum design, or
possibly, several such designs (depending on the number of surface or core geometries
selected) to submit to the customer.
2.1.7
Other Considerations
If the heat exchanger incorporates new design features, is a critical part of the system, or
is going to be mass produced, model and prototype heat exchangers are built and tested
in the laboratory to conrm any of the following: its heat transfer and pressure drop
performance both as a component and part of a system, fatigue characteristics due to
vibration, the quality and life of the n-to-tube/plate joint, pressure and temperature
cycling, corrosion and erosion characteristics, and burst pressure limit.
2.2
The design methodology presented in Fig. 2.1 could be considered a series solution if we
ignore the connecting line branches with double arrows. In this case, one would complete
the process specications and thermal and hydraulic design blocks, followed by the
mechanical design, and so on. But in reality, these design considerations are dependent
on each other and in many cases cannot be addressed individually without considering
the eects on each other. Let us review two examples to illustrate this point.
Consider a shell-and-tube exchanger design with heavy fouling. In the process specications module, geometry and material are selected properly to minimize the fouling,
minimize corrosion, and facilitate cleaning. The heat exchanger is architecturally placed
and oriented such that it could provide on-location cleaning or could be removed easily
{
The design life of an exchanger may vary from a few days/weeks operating in a very hostile environment to 30
years or more for nuclear power plant exchangers. For a conventional exchanger, the design life is based on the
eld experience of similar exchangers. For exchangers with new designs, new heat transfer surfaces or new
materials, the design life may be based on accelerated tests in the laboratory.
94
for external cleaning. During thermalhydraulic design, the fouling resistance and hence
the amount of increased surface area required should be taken into account properly,
depending on the cleaning cycle schedule. Also the selection of the core geometry or heat
transfer surface should be such that it either minimizes fouling or provides easy cleaning.
During mechanical design, the proper material gauge should be selected for the desired
life of the exchanger from the fouling and corrosion point of view. The tube gauge could
in turn aect both the heat transfer and pressure drop. The decision should be made up
front in terms of what type of cleaning technique and maintenance schedule should
be employed: online, oine, chemical cleaning, mechanical cleaning, or a throwaway
exchangersince this may aect the exchanger construction, material selection, and
thermal/hydraulic design. The material and gauge selection as well as cleaning cycle
will aect the xed and operating cost of the exchanger. For example, two aluminum
heat exchangers may be less expensive than one titanium exchanger for the same total
service time when using ocean water for cooling. This material choice will aect the
geometry, thermalhydraulic design, and mechanical design.
Consider a plate-n exchanger for high-temperature (8158C or 15008F) waste heat
recovery. The application will dictate the choice of material as stainless steel or more
exotic materials. High-temperature operation means that a special brazing technique will
be required, and it needs to be developed for each dierent n geometry selected. The
resultant contact resistance between the n and the plate/tube could be nite and not
negligible, depending on the brazing material and the technique employed, which in turn
will aect the thermal performance. The cost and thermal performance considerations
will dictate the selection of material with respect to the desired life of the exchanger. The
cost and allowed pressure drop will dictate the choice of whether or not to use ns and
which types. The operating thermal stresses and required waste heat recovery will dictate
the choice of construction and ow arrangement (crossow, counterow, or multipass).
So there are many interdependent factors that must be considered while designing and
optimizing this exchanger.
SUMMARY
The problem of heat exchanger design is multidisciplinary. Only a part of the total design
process consists of quantitative analytical evaluation. Because of a large number of
qualitative judgments, trade-os, and compromises, heat exchanger design is more of
an art than a science at this stage. In general, no two engineers will come up with the same
heat exchanger design for a given application. Most probably, an experienced engineer
will arrive at a better design since he or she has a better feel for the qualitative
considerations.
Heat exchanger design is a complex endeavor and involves not only a determination
of one or more feasible solution(s) but also the best possible or nearly optimal design
solution. In the chapters that follow, we address systematically the most important
quantitative aspects of design methodology outlined here.
REFERENCES
Hewitt, G. F., and S. J. Pugh, 1998, Approximate design and costing methods for heat exchangers,
Proc. Int. Conf. Heat Exchangers for Sustainable Development, June 1418, Lisbon, Portugal,
pp. 801820.
REVIEW QUESTIONS
95
Hewitt, G. F., A. R. Guy, and R. H. Marsland, 1982, Heat transfer equipment, in User Guide on
Process Integration for the Ecient Use of Energy, B. Linho et al., eds., Institution of Chemical
Engineers, Rugby, UK, pp. 129186.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger Publishing, Malabar, FL.
Shah, R. K., 1982, Advances in compact heat exchanger technology and design theory, Heat Transfer 1982, Proc. 7th Int. Heat Transfer Conf., Vol. 1, pp. 123142.
Taborek, J., 1988, Strategy of heat exchanger design, in Two-Phase Flow Heat Exchangers: ThermalHydraulic Fundamentals and Design, S. Kakac, A. E. Bergles, and E. O. Fernandes, eds., Kluwer
Academic Publishers, Dordrecht, The Netherlands, pp. 473493.
REVIEW QUESTIONS
2.1 Establish one-to-one correspondence between design considerations identied
within the scope of the heat exchanger design methodology and the process stages
within the life-cycle design process.
2.2 An engineer has concluded during a heat exchanger design procedure that three
important constraints limit his decision-making process. The rst is that the pressure drop of one of the two uids is larger than the value permitted. The second is
that the cost of one of the two uids used in the exchanger limits the mass ow rate
of that uid through the exchanger. Finally, the third is that the temperature
dierence between the two uids at the exit of the exchanger must be reduced
due to the need to increase the plants overall eciency. From a study of the design
methodology chart presented in Fig. 2.1, uncover in which of the suggested
segments of the design procedure these constraints are probably identied.
2.3 Two competing feasible heat exchanger types have been identied. List the criteria
(as many as you can) to be used to select the heat exchanger type that will suit the
imposed requirements (impose your own requirements).
2.4 For a given (known) heat load and inlet temperatures, the outlet temperatures of
both uids can be calculated with the thermophysical properties known. A design
engineer does not know either the magnitudes of the required mass ow rates of the
uids or the required size of the heat exchanger. Can he or she determine these three
variables using the data available?
2.5 How many distinct rating problems can one identify to be characterized by:
(a) one known and other unknown mass ow rate?
(b) one known and other unknown inlet temperature?
2.6 List all the information you need to perform the stress analysis of a given heat
exchanger.
PROBLEMS
2.1 Develop a owchart showing detailed steps for thermal and mechanical design,
laboratory, and other tests involved in developing an automotive radiator from
an initial concept to the nal mass production.
96
2.2 Develop a ow diagram of various departments involved in carrying out the tasks
involved in Problem 2.1.
2.3 Repeat Problems 2.1 and 2.2 for a one-of-a-kind shell-and-tube exchanger for a
petroleum renery.
2.4 Repeat Problems 2.1 and 2.2 for an exchanger either of your interest or for any
industry of your interest.
97
98
T
T
R
UA1
T Rq
E
R
E Ri
3:1
With this notion, the formal analogy between various parameters is presented in Table
3.1. It is important to note that the relationships between current, potential, resistance,
conductance, capacitance, and time constant terms are analogous for these dierent
physical processes. While the expressions for power and energy are analogous for heat
and current ow from the physics point of view, they are not analogous from the resistance circuit point of view as their formulas dier as shown in Table 3.1. Moreover, there
is no thermal analogy to electrical inductance or inertia in this analogy (not shown in
Table 3.1). Note that heat capacity or thermal capacitance energy storage terminology
used in heat transfer is used incorrectly as thermal inertia in the literature.
Since we know electrical circuit symbolism, we will nd it convenient to borrow the
symbols for the thermal circuits used to describe the exchanger heat transfer process.
These are summarized in Fig. 3.2.
We will also need an analogy between uid ow in a pipe (exchanger) and electric
current for the pressure drop analysis of the exchanger. The basic parameters of pressure
drop (head), uid ow rate, and ow losses are analogous to the voltage potential,
99
TABLE 3.1 Analogies and Nonalogies between Thermal and Electrical Parameters
Parameter
Electrical
Current
Potential
Resistance
Conductance
Capacitance
Time constant
i
E
R
G
C
RC
Power
iE
Analogies
ampere, A
volts, V
ohms, ; V/A
siemens, S, A/V
farads, F, A s/V
s
Thermal
q
T
R 1=UA
UA
C
RC
W, Btu/hr
8CK; 8F8R
8C=W; 8F-hr/Btu
W/8C, Btu/hr-8F
W s=8C; Btu/8F
s, hr
q
W, Btu/hr
Nonanalogies
Energy
W
iE d
J, W s
q d
J, Btu
current, and resistance.y Since the ow losses are measured in terms of the pressure loss
or head or uid column, which have the same units as the potential, this analogy is not as
well dened as the analogy between heat transfer and electric current. Again, the relationship between analogous parameters for uid ow is not linear for transition and
turbulent ows or developing laminar ows in pipes.
y
Pipe and duct design based on one-dimensional lumped parameter analysis typically denes the ow resistance or
the ow loss coecient K as equal to the number of velocity heads lost due to frictional eects [see Eq. (6. 53)].
100
3.2
In this section, starting with the assumptions built into heat exchanger design theory, the
basic problem for the exchanger heat transfer analysis is formulated. This includes the
dierential equations used for the analysis as well as a list of independent and dependent
variables associated with heat exchanger design and analysis problems. Next, the basic
denitions of important dimensional variables and important terminologies are introduced. Finally, the thermal circuit, and expressions for UA and wall temperatures are
presented.
3.2.1
To analyze the exchanger heat transfer problem, a set of assumptions are introduced so
that the resulting theoretical models are simple enough for the analysis. The following
assumptions and/or idealizations are made for the exchanger heat transfer problem
formulations: the energy balances, rate equations, boundary conditions, and subsequent
analysis [see, e.g., Eqs. (3.2) and (3.4) through (3.6) in dierential or integral form].y
1. The heat exchanger operates under steady-state conditions [i.e., constant ow
rates and uid temperatures (at the inlet and within the exchanger) independent
of time].
2. Heat losses to or from the surroundings are negligible (i.e. the heat exchanger
outside walls are adiabatic).
3. There are no thermal energy sources or sinks in the exchanger walls or uids, such
as electric heating, chemical reaction, or nuclear processes.
4. The temperature of each uid is uniform over every cross section in counterow
and parallelow exchangers (i.e., perfect transverse mixing and no temperature
gradient normal to the ow direction). Each uid is considered mixed or unmixed
from the temperature distribution viewpoint at every cross section in single-pass
crossow exchangers, depending on the specications. For a multipass exchanger, the foregoing statements apply to each pass depending on the basic ow
arrangement of the passes; the uid is considered mixed or unmixed between
passes as specied.
5. Wall thermal resistance is distributed uniformly in the entire exchanger.
6. Either there are no phase changes (condensation or vaporization) in the uid
streams owing through the exchanger or the phase change occurs under the
following condition. The phase change occurs at a constant temperature as for
a single-component uid at constant pressure; the effective specic heat cp;eff for
the phase-changing uid is innity in this case, and hence Cmax m_ cp;eff ! 1;
where m_ is the uid mass ow rate.
7. Longitudinal heat conduction in the uids and in the wall is negligible.
8. The individual and overall heat transfer coefcients are constant (independent of
temperature, time, and position) throughout the exchanger, including the case of
phase-changing uids in assumption 6.
y
The complete set of dierential equations and boundary conditions describing the mathematical models of heat
exchangers, based on these assumptions, is presented in Section 11.2.
101
9. The specic heat of each uid is constant throughout the exchanger, so that the
heat capacity rate on each side is treated as constant. Note that the other uid
properties are not involved directly in the energy balance and rate equations, but
are involved implicitly in NTU and are treated as constant.
10. For an extended surface exchanger, the overall extended surface efciency o is
considered uniform and constant.
11. The heat transfer surface area A is distributed uniformly on each uid side in a
single-pass or multipass exchanger. In a multipass unit, the heat transfer surface
area is distributed uniformly in each pass, although different passes can have
different surface areas.
12. For a plate-bafed 1n shell-and-tube exchanger, the temperature rise (or drop)
per bafe pass (or compartment) is small compared to the total temperature rise
(or drop) of the shell uid in the exchanger, so that the shell uid can be treated as
mixed at any cross section. This implies that the number of bafes is large in the
exchanger.
13. The velocity and temperature at the entrance of the heat exchanger on each uid
side are uniform over the ow cross section. There is no gross ow maldistribution
at the inlet.
14. The uid ow rate is uniformly distributed through the exchanger on each uid
side in each pass i.e., no passage-to-passage or viscosity-induced maldistribution
occurs in the exchanger core. Also, no ow stratication, ow bypassing, or ow
leakages occur in any stream. The ow condition is characterized by the bulk
(or mean) velocity at any cross section.
102
3.2.2
Problem Formulation
To perform the heat transfer analysis of an exchanger, our objective is to relate the heat
transfer rate q, heat transfer surface area A, heat capacity rate C of each uid, overall
heat transfer coecient U, and uid terminal temperatures. Two basic relationships are
used for this purpose: (1) energy balance based on the rst law of thermodynamics, and
(2) rate equations for heat transfer, as outlined by Eqs. (2.1) and (2.2).
Consider a two-uid exchanger (a counterow exchanger as an example) shown in Fig
3.3 to arrive at the variables relating to the thermal performance of a two-uid exchanger. Schematic of Fig. 3.3 and the balance equations for dierent exchanger ow arrangements may be dierent, but the basic concept of modeling remains the same. The analysis
that will follow is intended to introduce variables important for heat exchanger analysis.
Detailed approaches to a general thermodynamic problem formulation are presented in
Chapter 11.
Two dierential energy conservation (or balance) equations (based on the energy
balance implied by the rst law of thermodynamics) can be combined as follows for
control volumes associated with the dierential element of dA area for steady-state ow,
an overall adiabatic system, and negligible potential and kinetic energy changes:
dq q 00 dA Ch dTh Cc dTc
FIGURE 3.3 Nomenclature for heat exchanger variables (From Shah, 1983).
3:2
103
The negative signs in this equation are a result of Th and Tc decreasing with increasing
A (these temperatures decrease with increasing ow length x as shown in Fig. 1.50);y also,
dq is the heat transfer rate from the hot to cold uid, C m_ cp is the heat capacity rate of
the uid, m_ is the uid ow rate, cp is the uid specic heat at constant pressure, T is the
uid temperature, and the subscripts h and c denote hot and cold uids, respectively. The
heat capacity rate C J/s 8C (Btu/hr 8F) is the amount of heat in joules (Btu) that must
be added to or extracted from the uid stream per second (hour) to change its temperature by 18C 8F. The product m_ cp C appears in the energy balance [Eq. (3.2)] for
constant cp , and hence C is commonly used in the heat exchanger analysis instead of m_
and cp as two parameters separately.
In general, for any isobaric change of state, Eq. (3.2) should be replaced by
dq m_ h dhh m_ c dhc
3:3
where h is the uid specic enthalpy, J/kg (Btu/lbm). If the change of state is a phase
change, enthalpy dierences should be replaced by enthalpies of phase change (either
enthalpy of evaporation or enthalpy of condensation). However, cp can be assumed as
innity for condensing or evaporating single-component uid streams. Hence, the phasechanging stream can be treated as a single-phase uid having T q=C or
dT dq=C, with C being innity for a nite q or dq since the T or dT 0 for
isothermal condensing or evaporating uid streams (see Fig. 3.1). Note that here
T Th;i Th;o or Tc;o Tc;i as appropriate.
The overall heat transfer rate equation on a dierential base for the surface area dA of
Fig. 3.3 is
dq UTh Tc local dA U T dA
3:4
where U is the local overall heat transfer coecient to be dened in Eq. (3.20).z Thus for
this dierential element dA, the driving potential for heat transfer is the local temperature
dierence Th Tc T and the thermal conductance is U dA.
Integration of Eqs. (3.2) and (3.4) together over the entire heat exchanger surface for
specied inlet temperatures will result in an expression that will relate all important
operating variables and geometry parameters of the exchanger. The derivation of such
an expression for a counterow exchanger will be presented in Section 3.3 together with
nal results for many industrially important exchanger ow arrangements. The common
assumptions invoked for the derivation and integration of Eqs. (3.2) and (3.4) are
summarized in Section 3.2.1.
Two basic equations, energy conservation (balance) and rate equations, could also be
written on an overall basis for the entire exchanger as follows (under the conditions
implied by the above-mentioned idealizations):
y
The sign convention adopted in Eq. (3.2) leads to positive value of heat transfer rate change along each dx
element, and should be considered only as formal (i.e., not necessarily in agreement with thermodynamic convention for heat).
z
Note that although the overall heat transfer coecient U is idealized as constant (see assumption 8 in Section
3.2.1), it can vary signicantly in a heat exchanger. In that case, the mean overall heat transfer coecient Um is
obtained from local U data (see Section 4.2.3). Even though U Um constant throughout this chapter, we
distinguish between U and Um in Sections 3.2.3 and 3.2.4 to emphasize how the theory is developed using U and
Um .
104
3:5
U T dA Um A Tm
3:6
Here the subscripts i and o denote inlet and outlet, respectively; Th;o and Tc;o represent the
outlet temperatures; they are bulk temperatures dened by Eq. (7.10) for a nonuniform
temperature distribution across a cross section; and Um and Tm are the mean overall
heat transfer coecient and the exchanger mean temperature dierence, respectively,
and will be dened shortly.
From Eqs. (3.5) and (3.6) and Fig. 3.3, the steady-state overall-adiabatic heat exchanger behavior can thus be presented in terms of dependent uid outlet temperatures or
heat transfer rate as functions of four operating condition variables and three designer
controlled parameters:
Th;o ; Tc;o or q Th;i ; Tc;i ; Ch ; Cc U; A; flow arrangement
|{z}
|{z} |{z}
dependent
variables
3:7
parameters under
designers control
operating condition
variables
|{z}
independent variables and parameters
This equation represents a total of six independent and one or more dependent variables for a given heat exchanger ow arrangement. Of course, any one of the independent
variables/parameters in Eq. (3.7) can be made dependent (if unknown); in that case, one
of the three dependent variables in Eq. (3.7) becomes an independent variable/parameter.
Thus the most general heat exchanger design problem is to determine any two unknown
variables from this set when the rest of them are known.
For heat exchanger analysis, it is dicult to understand and work with such a large
number of variables and parameters as outlined in Eq. (3.7). From dimensional analysis,
three dimensionless groups are formulated from six independent and one or more dependent variables of Eq. (3.7). The reduced number of nondimensional variables and parameters simplies much of the analysis, provides a clear understanding of performance
behavior, and the results can be presented in more compact graphical and tabular forms.
The specic form of these groups is to some extent optional.
Five such options have been used, depending on which method of heat transfer
analysis has been used: the eectivenessnumber of heat transfer units ("-NTU or PNTU) method, the mean temperature dierence (MTD) method, the nondimensional
mean temperature dierencetemperature eectiveness ( -P) method, and the P1 -P2
method. These methods are discussed in Sections 3.3 through 3.10.
3.2.3
Basic Denitions
The denitions of the mean overall heat transfer coecient and mean temperature
dierence are introduced rst.
The rate equation (3.4), after rearrangement, is presented in integral form as
dq
T
U dA
A
3:8
105
Now we dene the mean temperature dierence and mean overall heat transfer
coecient as follows using the terms in Eq. (3.8):
1
1
Tm q
Um
dq
T
1
A
3:9
U dA
A
3:10
Substitution of Eqs. (3.9) and (3.10) into Eq. (3.8) results into the following equation
after rearrangement:
q Um A Tm
3:11
Here Um is the mean overall heat transfer coecient, and Tm is the true (or eective)
mean temperature dierence (MTD), also referred to as mean temperature driving potential or force for heat transfer.
Generally, the overall heat transfer coecient is treated as constant in the heat exchanger analysis. It is simply designated as U, without subscripts or overbars, throughout the
book except for Section 4.2, where various denitions of mean overall heat transfer
coecients are introduced. Thus the overall rate equation (3.6) is simply
q UA Tm
3:12
1
A
T dA
3:13
Other commonly used important entities for heat exchangers are the inlet temperature
dierence, temperature range, temperature approach, temperature pinch, temperature
gap, temperature meet, and temperature cross. They are discussed below and
summarized in Table 3.2.
The inlet temperature dierence (ITD) is the dierence between inlet temperatures of
the hot and cold uids and is designated as Tmax in this book. Tmax Th;i Tc;i
is also sometimes referred to as the temperature span or temperature head in a heat
exchanger.
The temperature range for a uid is referred to as its actual temperature rise or drop
T within the exchanger. The temperature ranges for hot and cold uids in the exchangers are then Th Th;i Th;o and Tc Tc;o Tc;i , respectively.
The temperature approach for exchangers with single-phase uids is dened as the
dierence between outlet uid temperatures Th;o Tc;o for all single-pass and multipass
ow arrangements except for the counterow exchanger. For the latter, it is dened as
the smaller of Th;i Tc;o and Th;o Tc;i . The temperature approach for multiphase
multicomponent streams is dened as the minimum local temperature dierence between
hot and cold uid streams. This could occur anywhere in the exchanger, depending on
the ow arrangement, heat duty, and so on. It can be shown that the temperature
approach for single-phase exchangers is related to the exchanger eectiveness " dened
by Eq. (3.44) later as follows:
106
TABLE 3.2 Expressions for Temperature Span, Range, Approach, Pinch, Gap, Meet, and Cross
Item
Expression
Th;i Tc;i
Temperature
Temperature
Temperature
Temperature
Temperature
gap
meet, counterow sigle-phase exchanger
meet, all other single-phase exchangers
cross, single-pass exchangers
cross, multipass exchangers
temperature
approach
Th;i Th;o
Tc;o Tc;i
min jTh;i Tc;o ; Th;o Tc;i j
Th;o Tc;o
Th Tc with Th Tc Th;i Tc;o
or Th;o Tc;i
Th;o Tc;o with Th;o > Tc;o
Th;i Tc;o or Th;o Tc;i
Th;o Tc;o
Tc;o Th;o with Tc;o > Th;o
Tc Th with Tc > Th in one of the passes
1 " Tmax
1 1 C*" Tmax Th;o Tc;o
where Tmax Th;i Tc;i and C* Cmin =Cmax . For some shell-and-tube, multipass,
and two-phase exchangers, it may not be either easy or possible to determine quantitatively the magnitude of the temperature approach. In that case, while the foregoing
denition is valid, it loses its usefulness.
A temperature pinch refers to a local temperature dierence within an exchanger (or
an array of exchangers) that is substantially less than either of two terminal temperature
dierences and is minimum in the exchanger. In the limit, it can approach zero, which is
referred to as temperature meet dened below. The temperature pinch usually occurs in
an exchanger with two-phase multicomponent uids. In reality, a temperature pinch
close to zero will require a very large (approaching innity) surface area for a singlepass exchanger. Hence, for a nite-size heat exchanger, the exchanger would cease to
function eciently beyond the temperature pinch point (i.e., resulting in a more signicant reduction in heat transfer than justied). However, for a multipass exchanger, the
temperature pinch could occur in one pass, and in that case, the performance of that pass
beyond the temperature pinch is reduced signicantly.
A temperature gap refers to the temperature dierence between hot and cold uid
outlet temperatures provided that Th;o > Tc;o :
A temperature meet refers to the case when the temperature pinch is zero or the hot
and cold uid temperatures are the same locally somewhere in the exchanger or at
outlets. This is an idealized condition and does not occur in a single-pass heat exchanger,
but may occur in one of the passes of a multipass exchanger.
A temperature cross refers to the case when the cold uid temperature becomes equal
or greater than the hot uid temperature within the exchanger. External temperature
cross refers to the case when the cold uid outlet temperature Tc;o is greater than the hot
uid outlet temperature Th;o . There is no physical (actual) crossing of hot and cold uid
temperature distributions within an exchanger. This is quite common in a counterow
exchanger (see Fig. 1.50a) and other single-pass and multipass exchangers having high
107
NTUs (see Section 3.3.3 for the denition of NTU). The magnitude of the external
temperature cross is Tc;o Th;o . Internal temperature cross refers to the case when locally
somewhere within the exchanger Tc becomes equal to Th (within a pass or in a segment of
the exchanger), and beyond that point along the ow length, Tc > Th in that pass or
segment; subsequently, reverse heat transfer takes place (original cold uid transferring
heat to the original hot uid). The external temperature cross can occur with or without
the internal temperature cross; similarly an internal temperature cross can occur with or
without an external temperature cross (see Section 11.4.1).
3.2.4
To understand the exchanger overall heat transfer rate equation [Eq. (3.12)], consider the
thermal circuit model of Fig. 3.4. Scale or fouling deposit layers are also shown on each
side of the wall.
In the steady state, heat is transferred from the hot uid to the cold uid by the
following processes: convection to the hot uid wall, conduction through the wall, and
subsequent convection from the wall to the cold uid. In many heat exchangers, a fouling
lm is formed as a result of accumulation of scale or rust formation, deposits from
the uid, chemical reaction products between the uid and the wall material, and/or
biological growth. This undesired fouling lm usually has a low thermal conductivity
and can increase the thermal resistance to heat ow from the hot uid to the cold uid.
This added thermal resistance on individual uid sides for heat conduction through the
fouling lm is taken into account by a fouling factory rf 1=hf , where the subscript f
denotes fouling (or scale); it is shown in Fig. 3.4. Thus, the heat transfer rate per unit
area at any section dx (having surface areas dAh , dAc , etc.) can be presented by the
appropriate convection and conduction rate equations as follows:
dq
dRh
dRh; f
dRw
dRc; f
dRc
3:15
Alternatively
dq
Th Tc
U dATh Tc
dRo
3:16
where the overall dierential thermal resistance dRo consists of component resistances in
series (similar to those shown in Fig. 3.4b for a heat exchanger):
1
dRo dRh dRh; f dRw dRc; f dRc
U dA
3:17
or
1
1
1
1
1
dRw
U dA o h dAh o hf dAh
o hf dAc o h dAc
3:18
y
We also refer to the fouling factor as the unit fouling thermal resistance or concisely as fouling resistance
^ f 1=hf f =kf where f is the thickness of fouling layer and kf is the thermal conductivity of the fouling
rf R
material. Refer to Section 13.1 for more details on the fouling resistance concept.
108
FIGURE 3.4 (a) Thermal resistances; (b) thermal circuit for a heat exchanger (From Shah, 1983).
Various symbols in this equation are dened after Eq. (3.24). If we idealize that the heat
transfer surface area is distributed uniformly on each uid side (see assumption 11 in
Section 3.2.1), the ratio of dierential area on each uid side to the total area on the
respective side remains the same; that is,
dA dAh dAc dAw
Ah
Ac
Aw
A
3:19
Replacing dierential areas of Eq. (3.18) by using corresponding terms of Eq. (3.19),
we get
1
1
1
1
1
Rw
UA o hAh o hf Ah
o hf Ac o hAc
3:20
It should be emphasized that U and all hs in this equation are assumed to be local. Using
the overall rate equation [Eq. (3.6)], the total heat transfer rate will be
q Um A Tm Um ATh;e Tc;e
1
T Tc;e
Ro h;e
3:21
109
Th;e Th; f Th; f Tw;h Tw;h Tw;c Tw;c Tc; f Tc; f Tc;e
Rh
Rh; f
Rw
Rc; f
Rc
3:22
where the subscript e denotes the eective value for the exchanger, or
Th;e Tc;e Tm . To be more precise, all individual temperatures in Eq. (3.22) should
also be mean or eective values for respective uid sides. However, this additional subscript is not included for simplicity. In Eq. (3.21), the overall thermal resistance Ro
consists of component resistances in series as shown in Fig. 3.4b.
1
Ro Rh Rh; f Rw Rc; f Rc
Um A
3:23
1
1
1
1
1
Rw
Um A o hm Ah o hm; f Ah
o hm; f Ac o hm Ac
3:24
For constant and uniform U and hs throughout the exchanger, Eqs. (3.24) and (3.20)
are identical. In that case, Um U and we will use U throughout the book except for
Section 4.2. In Eqs. (3.20) and (3.24), depending on the local or mean value, we dene
Rh hot-fluid-side convection resistance
Rh; f hot-fluid-side fouling resistance
1
1
or
o hAh
o hm Ah
1
1
or
o hf Ah
o hm; f Ah
1
1
or
o hf Ac
o hm; f Ac
1
1
or
o hAc
o hm Ac
In the foregoing denitions, h is the heat transfer coecient, discussed in detail in Section
7.1.4.3; hf is referred to as the fouling coecient (inverse of fouling factor), discussed in
Chapter 13; A represents the total of primary and secondary (nned) surface area; and o
is the extended surface eciency of an extended (n) surface dened in Section 4.3.4. In
^ f is referred to as unit fouling resistance. Note that no
the literature, 1=o hf Rf A R
ns are shown in the upper sketch of Fig. 3.4; however, o is included in the aforementioned various resistance terms in order to make them most general. For all prime surface
exchangers (i.e., having no ns or extended surface), o;h and o;c are unity.
The wall thermal resistance Rw for at walls is given by
8
w
>
>
>
>
< kw Aw
Rw
X w
>
>
>
>
:
kw Aw j
j
110
Rw
8
lndo =di
>
>
>
>
< 2kw LNt
>
1 X lndj1 =dj
>
>
>
: 2LN
kw; j
t j
where w is the wall plate thickness, Aw the total wall area of all at walls for heat
conduction, kw the thermal conductivity of the wall material, do and di the tube outside
and inside diameters, L the tube or exchanger length, and Nt the number of tubes. A at
(or plain) wall is generally associated with a plate-n or an all-prime-surface plate
exchanger. In this case,
Aw L1 L2 Np
3:27
Here L1 , L2 , and Np are the length, width, and total number of separating plates,
respectively. The wall thickness w is then the length for heat conduction.
If there is any contact or bond resistance present between the n and tube or plate on
the hot or cold uid side, it is included as an added thermal resistance on the right-hand
side of Eq. (3.23) or (3.24). For a heat pipe heat exchanger, additional thermal resistances
associated with the heat pipe need to be included on the right-hand side of Eq. (3.23) or
(3.24); these resistances are evaporation resistance at the evaporator section of the heat
pipe, viscous vapor ow resistance inside the heat pipe (very small), internal wick resistance at the condenser section of the heat pipe, and condensation resistance at the
condenser section.
If one of the resistances on the right-hand side of Eq. (3.23) or (3.24) is signicantly
higher than other resistances, it is referred to as the controlling resistance. It is considered
signicantly dominant when it represents more than 80% of the total resistance. For
example, if the cold side is gas (air) and the hot side is condensing steam, the thermal
resistance on the gas side will be very high (since h for air is very low compared to that for
the condensing steam) and will be referred to as the controlling resistance for that
exchanger. However, for a water-to-water heat exchanger, none of the thermal resistances may be dominant if the water ow rates are about the same.
The lowest overall thermal resistance in a heat exchanger can be obtained by making
the hot- and cold-side thermal resistances about equal (considering wall and fouling
resistances is negligible or low). Hence, a low h is often compensated by a high A to
make o hAh o hAc . This is the reason the surface area on the gas side is about 5 to
10 times higher than that on the liquid-side when the liquid side heat transfer coecient h
is 5 to 10 times higher than the h on the gas side. This would explain why ns are used on
the gas sides in a gastoliquid or gastophase change exchanger.
In Eq. (3.24) or (3.12), the overall heat transfer coecient U may be dened optionally
in terms of the surface area of either the hot surface, the cold surface, or the wall
conduction area. Thus,
UA Uh Ah Uc Ac Uw Aw
3:28
111
3:29
Rw Ah h c h c
o hf c o hc
Uh o hh o hf h
^h 1 R
^ w Ah 1 R
^ c Ah
^ R
^ Ah R
R
Aw o;c c; f Ac
Ac
o;h h; f
3:30a
1
A =A
A =A
1
1
c h c h Rw Ac
Uc o hh o hf h
o hf c o hc
^ Ac R
^ R
^ h Ac 1 R
^ w Ac 1 R
^c
R
Ah o;h h; f Ah
Aw o;c c; f
3:30b
where 1=Uh is the unit overall thermal resistance based on the hot-uid-side surface area.
^ j 1=o hj , j h or c are unit thermal resistances for
Similarly, 1=Uc is dened. Also R
^ w w =kw unit
^
hot or cold uids, Rf 1=hf unit thermal fouling resistance, and R
wall thermal resistance. For a plain tubular exchanger, o 1; then from Eq. (3.30), Uo
based on the tube outside surface area is given as follows after inserting the value of Rw
from the rst equation of Eq. (3.26):
1
1
1
d lndo =di
d
d
o
o o
hi; f di hi di
Uo ho ho; f
2kw
3:31a
1
d
di
d lndo =di
1
1
i
i
Ui ho do ho; f do
2kw
hi; f hi
3:31b
Here the subscripts o and i denote the tube outside and inside, respectively; 1=Uo and
1=Ui are the unit overall thermal resistances based on the tube outside and inside surface
area, respectively.
Knowledge of wall temperature in a heat exchanger is essential to determine
the localized hot spots, freeze points, thermal stresses, local fouling characteristics, or
boiling/condensing coecients. In this case, Tw;h and Tw;c can be computed from
Eq. (3.22) on a local basis as follows:
Th Tw;h Tw;c Tc
Rh Rh; f Rc Rc; f
3:32
112
Based on the thermal circuit of Fig. 3.4, when Rw is negligible, Tw;h Tw;c Tw , and Eq.
(3.32) reduces to
Tw
3:33
When there is no fouling on either uid side (Rh; f Rc; f 0, this reduces further to
Tw
o hAh o hAc
1=Rh 1=Rc
3:34
Equations (3.32) through (3.34) are also valid for the entire exchanger if all temperatures
are used as mean or eective values on respective uid sides.
Example 3.1 In a shell-and-tube feedwater heater, cold water at 158C owing at the rate
of 180 kg/h is preheated to 908C by ue gases from 1508C owing at the rate of 900 kg/h.
The water ows inside the copper tubes di 25 mm, do 32 mm) having thermal
conductivity kw 381 W=m K. The heat transfer coecients on gas and water sides
are 120 and 1200 W/m2 K, respectively. The fouling factor on the water side is
0.002 m2 K=W. Determine the ue gas outlet temperature, the overall heat transfer
coecient based on the outside tube diameter, and the true mean temperature dierence
for heat transfer. Consider specic heats cp for ue gases and water as 1.05 and
4:19 J=g K respectively, and the total tube outside surface area as 5 m2 . There are no
ns inside or outside the tubes, and there is no fouling on the gas side.
SOLUTION
Problem Data and Schematic: Fluid ow rates, inlet temperatures, and cold uid outlet
temperature are provided for a shell-and-tube exchanger of prescribed tube inner and
outer diameters (Fig. E3.1). Also, the thermal conductivity of the tube and the thermal
resistance on the cold uid side are given. There are no ns on either side of the tubes.
Determine: Hot uid outlet temperature Th;o , overall heat transfer coecient U, and true
mean temperature dierence Tm .
Assumptions: The assumptions invoked in Section 3.2.1 are valid. Hot-uid-side fouling
is negligible.
FIGURE E3.1
113
Analysis: The required heat transfer rate may be obtained from the overall energy
balance for the cold uid [Eq. (3.5)].
q Cc Tc;o Tc;i m_ cp c Tc;o Tc;i
180 kg=h
q
m_ cp h
Since
m_ h
900 kg=h
0:25 kg=s
3600 s=h
15,713 W
90:18C
0:25 kg=s 1050 J=kg 8C
Ans:
Since U is based on A Ah do LNt , Eq. (3.31) reduces to the following form after
substituting the hot-uid-side fouling factor (1=hf ) as zero, and replacing the subscripts o
and i of U and h with h and c, respectively.
^ 1 1 1 do lndo =di do do
R
hc; f di hc di
Uh hh hh; f
2kw
1
0:032 mln32 mm=25 mm 0:002 m2 K=W 0:032 m
2
2 381 W=m K
0:025 m
120 W=m K
0:032 m
120 W=m2 K 0:025 m
Ans:
Note that the controlling thermal resistance for this feedwater heater is 69.6% on the ue
gas side. Now the mean temperature dierence can be determined from Eq. (3.12) as
Tm
q
15,713 W
37:68C
Uh Ah 83:54 W=m2 K 5 m2
Ans:
114
Discussion and Comments: Since the heat transfer coecient on the cold side is greater
than that on the hot side and the hot- and cold-side surface areas are about the same, the
hot side becomes the controlling resistance side. This can be seen from the unit thermal
resistance distribution as 69.6% of the total unit thermal resistance on the hot-gas side.
The tube wall, made of copper, turned out to be a very good conductor with very small
thermal resistance. Notice that the fouling resistance on the water side contributes about
one-fth (21.4%) of the total unit thermal resistance and hence about 21% surface area
penalty. If there had been no fouling on the water side, we would have reduced the heat
transfer surface area requirement by about one-fth. Hence, if it is desired to make a
single important improvement to reduce the surface area requirement in this exchanger,
the best way would be to employ ns on the gas side (i.e., employing low nned tubing in
the exchanger).
3.3
In the "-NTU method, the heat transfer rate from the hot uid to the cold uid in the
exchanger is expressed as
q "Cmin Th;i Tc;i "Cmin Tmax
3:35
where " is the heat exchanger eectiveness,y sometimes referred to in the literature as the
thermal eciency, Cmin is the minimum of Ch and Cc ; Tmax Th;i Tc;i is the uid
inlet temperature dierence (ITD). The heat exchanger eectiveness " is nondimensional,
and it can be shown that in general it is dependent on the number of transfer units NTU,
the heat capacity rate ratio C*, and the ow arrangement for a direct-transfer type heat
exchanger:
" NTU; C*; flow arrangement
3:36
Here the functional relationship is dependent on the ow arrangement. The three
nondimensional groups, ", NTU, and C* are rst dened below. The relationship
among them is illustrated next.
3.3.1
115
Here it is idealized that there are no ow leakages from one uid to the other uid, and
vice versa. If there are ow leakages in the exchanger, q represents the total enthalpy gain
(or loss) of the Cmin uid corresponding to its actual ow rate in the outlet (and not inlet)
stream. How do we determine qmax ? It would be obtained in a perfect counterow heat
exchanger (recuperator) of innite surface area, zero longitudinal wall heat conduction,
and zero ow leakages from one uid to the other uid, operating with uid ow rates
and uid inlet temperatures the same as those of the actual heat exchanger; also, assumptions 8 to 11, 13, and 14 of Section 3.2.1 are invoked for this perfect counterow
exchanger. This perfect exchanger is the meterbar (or yardstick) used in measuring
the degree of perfection of actual exchanger performance. The value of " ranges from 0 to
1. Thus " is like an eciency factor and has thermodynamic signicance. As shown
below, in such a perfect heat exchanger, the exit temperature of the uid with the smaller
heat capacity will reach the entering temperature of the larger heat capacity uid.y
Consider a counterow heat exchanger having innite surface area. An overall energy
balance for the two uid streams is
q Ch Th;i Th;o Cc Tc;o Tc;i
3:38
Based on this equation, for Ch < Cc , Th;i Th;o > Tc;o Tc;i . The temperature drop
on the hot uid side will thus be higher, and over the innite ow length the hot uid
temperature will approach the inlet temperature of the cold uid as shown by the two
bottom curves in Fig. 3.5, resulting in Th;o Tc;i . Thus for an innite area counterow
exchanger with Ch < Cc , we get qmax as
qmax Ch Th;i Tc;i Ch Tmax
3:39
Similarly, for Ch Cc C;
qmax Ch Th;i Tc;i Cc Th;i Tc;i C Tmax
3:40
Based on Eq. (3.38), for Ch > Cc , (Tc;o Tc;i > Th;i Th;o . Hence, Tc;o will
approach Th;i over the innite length, and therefore
qmax Cc Th;i Tc;i Cc Tmax
3:41
3:42
where
(
Cmin
Cc
for Cc < Ch
Ch
for Ch < Cc
3:43
y
It should be mentioned here that the second law of thermodynamics is involved implicitly in the denition of the
exchanger eectiveness since the maximum possible heat transfer rate is limited by the second law. Further
discussion of this and related issues is presented in Section 11.2.2.
116
FIGURE 3.5 Temperature distributions in a counterow exchanger of innite surface area (From
Shah 1983).
Thus qmax is determined by Eq. (3.42) for dening the measure of the actual performance
of a heat exchanger having any ow arrangement. Notice that Tmax Th;i Tc;i in
every case and Cmin appears in the determination of qmax regardless of Ch > Cc or
Ch Cc .
Using the value of actual heat transfer rate q from the energy conservation equation
(3.5) and qmax from Eq. (3.42), the exchanger eectiveness " of Eq. (3.37) valid for all ow
arrangements of the two uids is given by
"
Ch Th;i Th;o
Cc Tc;o Tc;i
3:44
Thus " can be determined directly from the operating temperatures and heat capacity
rates. It should be emphasized here that Th;o and Tc;o are the bulk outlet temperatures
dened by Eq. (7.10). If there is ow and/or temperature maldistribution at the exchanger inlet, not only the uid outlet temperatures but also the uid inlet temperatures
should be computed as bulk values and used in Eq. (3.44).
An alternative expression of " using q from the rate equation (3.12) and qmax from Eq.
(3.42) is
"
UA Tm
Cmin Tmax
3:45
Now let us nondimensionalize Eq. (3.7). The mean uid outlet temperatures Th;o and
Tc;o , the dependent variables of Eq. (3.7), are represented in a nondimensional form by
the exchanger eectiveness " of Eq. (3.44). There are a number of dierent ways to arrive
117
"
UA Cmin Th;i Th;o Tc;o Tc;i Tm
;
;
;
;
; flow arrangement
Tmax Tmax
Cmin Cmax Tmax
|{z}
|
{z}
independent
3:46
dependent
Note that Tmax Th;i Tc;i in the last three groups of Eq. (3.46) is an independent
parameter. In Eq. (3.46), Cmax Cc for Cc > Ch and Cmax Ch for Ch > Cc , so that
Cmin
Cmax
8
Cc
>
>
>
<C
h
>
C
> h
>
:
Cc
for Cc < Ch
3:47
for Ch < Cc
In order to show that the third through fth groups on the right-hand side of Eq.
(3.46) are dependent, using Eqs. (3.5) and (3.44), we can show that the rst two of the
three groups are related as
Th;i Th;o Cc Tc;o Tc;i
"
Tmax
Ch Tmax
Ch =Cmin
"
"Cmin =Cmax
for Ch Cmin
for Ch Cmax
3:48
and using Eq. (3.45), we can show that the fth group of Eq. (3.46) is
Tm
"
Tmax UA=Cmin
3:49
Since the right-hand side of the last equality of Eqs. (3.48) and (3.49) have ", Cmin =Cmax ,
and UA=Cmin as the only nondimensional groups and they are already included in Eq.
(3.46), the dimensionless groups of the left-hand side of Eqs. (3.48) and (3.49) are dependent. Thus Eq. (3.46) can be written in terms of nondimensional groups, without a loss of
generality, as follows:
UA Cmin
;
; flow arrangement NTU; C*; flow arrangement
"
Cmin Cmax
3:50
118
Th;i Th;o
Th
3:51
Similarly, the temperature eectiveness of the cold uid is dened as a ratio of the
temperature rise of the cold uid to the uid inlet temperature dierence:
"c
Tc;o Tc;i
Tc
3:52
From an energy balance [Eq. (3.5)], and denitions of "h , and "c , it can be shown that
Ch "h Cc "c
3:53
A comparison of temperature eectivenesses with Eq. (3.44) for the exchanger (heat
transfer) eectiveness reveals that they are related by
"
Ch
"
Cmin h
"
Cc
"
Cmax c
"h
for Ch Cmin
"h =C*
for Ch Cmax
"c
for Cc Cmin
"c =C*
for Cc Cmax
3:54
3:55
3.3.2
C* is simply a ratio of the smaller to larger heat capacity rate for the two uid streams so
that C* 1. A heat exchanger is considered balanced when C* 1:
m_ cp min
C
C* min
Cmax m_ cp max
for Ch Cmin
for Cc Cmin
3:56
3:57
119
3:58
Let us reemphasize that for a condensing or evaporating uid at ideally constant temperature, the T range (rise or drop) is zero, and hence the heat capacity rate C
approaches innity for a nite q C T. Since C m_ cp , the eective specic heat of
condensing or evaporating uid is hence innity. The C* 0 case then represents Cmin as
nite and Cmax approaching innity in the sense just discussed. The C* 0 case with
Cmin m_ cp min 0 is not of practical importance, since m_ 0 in that case and hence
there is no ow on the Cmin side of the exchanger.
3.3.3
The number of transfer units NTU is dened as a ratio of the overall thermal conductance
to the smaller heat capacity rate:
UA
1
U dA
3:59
NTU
Cmin Cmin A
If U is not a constant, the denition of the second equality applies. NTU may also be
interpreted as the relative magnitude of the heat transfer rate compared to the rate of
enthalpy change of the smaller heat capacity rate uid. A substitution of the UA
magnitude from Eq. (3.24) into Eq. (3.59) for U as constant results in
NTU
1
1
Cmin 1=o hm Ah Rh; f Rw Rc; f 1=o hm Ac
3:60
NTU designates the nondimensional heat transfer size or thermal size of the exchanger,
and therefore it is a design parameter. NTU provides a compound measure of the heat
exchanger size through the product of heat transfer surface area A and the overall heat
transfer coecient U. Hence, in general, NTU does not necessarily indicate the physical
size of the exchanger. In contrast, the heat transfer surface area designates the physical
size of a heat exchanger. A large value of NTU does not necessarily mean that a heat
exchanger is large in physical size. As a matter of fact, the automotive gas turbine
regenerator at the idle operating point may have NTU 10 and core volume
V 0:01 m3 , whereas a chemical plant shell-and-tube exchanger may have NTU 1
and V 100 m3 . However, when comparing heat exchangers for a specic application,
U=Cmin approximately remains constant; and in this case, a higher NTU value means a
heat exchanger larger in physical size. Hence, NTU is sometimes also referred to as a heat
exchanger size factor. In general, higher NTU is obtained by increasing either U or A or
both or by decreasing Cmin . Whereas a change in Cmin aects NTU directly, a change in
Cmax (i.e., its ow rate) aects h on the Cmax side. This in turn inuences U and NTU.
Thus, a change in the value of C* may have direct or indirect eect on NTU.
NTU is also variously referred to as a performance factor or thermal length in
the plate heat exchangery literature, and as reduced thermal ux in the shell-and-tube
y
In a PHE with chevron plates, a plate with high chevron angle has a high heat transfer coecient and high
pressure drop in general and is referred to as a hard plate; in contrast, a soft plate has a low chevron angle and low
heat transfer coecient and low pressure drop in general.
120
exchanger literature. Other names are given in the following text with appropriate interpretations.
At low values of NTU, the exchanger eectiveness is low. With increasing values of
NTU, the exchanger eectiveness generally increases, and in the limit, it approaches a
thermodynamic asymptotic value. Note that the perfect exchanger requires that
NTU ! 1 (because A ! 1) for qmax Cmin Tmax . The following approximate values
of NTU will illustrate this point further.
Automobile radiator:
NTU
1 ! " 63%
1
d
d
d*
1=UACmin 1=UACmin Ro Cmin
3:61
Here Ro 1=UA is the overall thermal resistance; Cmin Mcp min Cmin d is the minimum-side uid heat capacitance (M fluid mass in the exchanger at an instant of time)
in the exchanger at any instant of time; and d is the dwell time, residence time, or transit
time of a uid particle passing through the exchanger. Thus, NTU can be interpreted as a
nondimensional residence time or a ratio of residence time to the time constant of the Cmin
uid in the exchanger at an instant. NTU might well be expected to play an important
part in the transient problem! And from the viewpoint of an observer riding with a
particle in the Cmin stream, he or she would indeed have a transient temperaturetime
experience.
Yet another interpretation of NTU is as follows. NTU is related to Tm from Eq.
(3.45) with the use of Eq. (3.44) as
NTU
3:62
NTU
8
Th Th;i Th;o
>
>
> T T
<
for Ch Cmin
>
Tc;o Tc;i
>
Tc
>
:
Tm
Tm
for Cc Cmin
3:63
Thus, NTU Tmax;i =Tm , where Tmax;i is a maximum of Th and Tc . When Tm
is equal to Th or Tc whichever is larger, NTU 1. Notice that Eq. (3.63) is
convenient for small values of NTU, in which case Tm Tm;h Tm;c is a good
approximation; there is no need to calculate Tlm or F (see Section 3.7). Here Tm;h
and Tm;c are the arithmetic mean values of the corresponding terminal temperatures.
121
NTU is also directly related to the overall (total) Stanton number Sto formulated with
U in place of h of Eq. (7.28) as
NTU Sto
4L
Dh
3:64
Thus, NTU can also be interpreted as a modied Stanton number. Note that here the
hydraulic diameter Dh is dened as follows depending on the type of heat exchanger
surface geometry involved.
8
4 flow area
4Ao 4Ao L
>
>
>
< wetted perimeter P A
Dh
> 4 core flow volume
4pV 4p 4p
>
>
:
fluid contact surface area
A
3:65
where p is the porosity, a ratio of void volume to total volume of concern. Here the rst
denition of Dh is for constant cross-sectional ow passages in a heat exchanger.
However, when ow area is expanding/contracting across ow cross sections along the
ow length as in three-dimensional ow passages (such as in a corrugated perforated n
geometry of Fig. 1.29f ), the more general second denition is applicable. In the second
denition, Dh 4p= for plate-n type and regenerative surfaces; for tube bundles and
tube-n surfaces, Dh 4p=. Note that heat transfer and pressure drop Dh magnitudes
will be dierent if the heated and ow friction perimeters are dierent, as illustrated in the
footnote of p. 9.
Equations (3.63) and (3.59) may also be interpreted as the number of transfer units
required by the heat duty NTU Tmax;i =Tm and the number of transfer units
achieved by the heat exchanger NTU UA=Cmin , respectively.
The foregoing denitions and interpretations are for the overall NTU for the exchanger. The number of heat transfer units individually on the hot and cold sides of the
exchanger may be dened as:
ntuh
o hAh
Ch
ntuc
o hAc
Cc
3:66
We use ntuh and ntuc in Chapter 9 when outlining the solution procedure for the sizing
problem for extended surface heat exchangers. The overall thermal resistance equation
(3.24), in the absence of fouling resistances, can then be presented in terms of overall and
individual number of transfer units as
1
1
1
Rw Cmin
3.4
3:67
In the preceding section we demonstrated that " is a function of NTU, C*, and ow
arrangement. We now derive this functional relationship for a single-pass counterow
122
FIGURE 3.6 Counterow heat exchanger with temperature distributions and dierential
elements for energy balance and rate equation development.
exchanger and then summarize similar functional relationships for single-pass and multipass ow arrangements.
3.4.1
Single-Pass Exchangers
3.4.1.1 Counterow Exchanger.y Consider a counterow heat exchanger with the temperature distributions for hot and cold uids as shown in Fig. 3.6. The uid temperatures on the left-hand end of the exchanger are denoted with the subscript I, and those
on the other end with the subscript II.
In the analysis, we consider the overall counterow exchanger as shown in Fig. 3.3
with only two passages. This is because the idealizations made in Section 3.2.1 (such as
uniform velocities and temperatures at inlet, uniform surface area distribution, uniform
U, etc.) are also invoked here. Thus, in fact, the hot-uid passage shown represents all
hot-uid ow passages, and the cold-ow passage shown consists of all cold-uid ow
y
We derive the heat exchanger eectiveness in this section, and show how to obtain temperature distributions in
Chapter 11.
123
passages. This is the reason that Ch and Cc are associated with the respective single-ow
passages shown in Fig. 3.3 and not dCh and dCc .
Based on an energy balance consideration on the dierential element dx,
dq Ch dTh Cc dTc
3:68
Here Th represents the bulk temperature of the hot uid in the dierential element dx of
Fig. 3.6. Tc is dened in a similar manner for the cold uid.
The rate equations applied individually to the dx length elements of the hot uid, wall,
and the cold uid yield
8
dx
>
>
o hAh Th Tw;h
>
>
L
>
>
>
< k A dx
w
w
dq
Tw;h Tw;c
>
w L
>
>
>
>
>
>
: o hAc Tw;c Tc dx
L
3:69
3:70
3:71
Note that considering a general case, we have included the extended surface eciency
o in Eqs. (3.69) and (3.71), although ns are not shown in Fig. 3.6. After getting the
expressions for individual temperature dierences from Eqs. (3.69)(3.71), adding them
up, and rearranging, we get
dq UATh Tc
dx
L
3:72
Here U represents the local overall heat transfer coecient for the element dA or dx.
However, we treat this local U the same as Um for the entire exchanger, and hence U will
be treated as a constant throughout the exchanger. Hence UA in Eq. (3.72) is given by
Eq. (3.20) or (3.24) without the fouling resistances on the hot and cold sides. If the fouling
or other resistances are present, UA in Eq. (3.72) will have those resistances included.
Elimination of dq from Eqs. (3.68) and (3.72) will result in two ordinary dierential
equations. We need two boundary conditions to solve them and they are
Th x 0 Th;I
Tc x L Tc;II
3:73
Now let us solve the set of Eqs. (3.68), (3.72), and (3.73) as follows to obtain a ratio of
terminal temperature dierences so that we can determine the exchanger eectiveness
directly. We derive the temperature distributions in Section 11.2.1. Here, eectively, we
will solve one dierential equation that is based on the energy balance and rate equations. Substituting the values of dTh and dTc in terms of dq, Eq. (3.68) can be rearranged
to
1
1
dq
3:74
dTh Tc
Cc Ch
Eliminating dq from Eqs. (3.72) and (3.74) and rearranging yields
dTh Tc
C
UA dx
1 h
Cc Ch L
Th Tc
3:75
124
Integrating this equation from the hot-uid inlet (section I) to the outlet section (section
II) yields
Th;II Tc;II
UA
C
exp
1 h
Th;I Tc;I
Cc
Ch
3:76
Th;I Tc;I
1 Tc;I Tc II =Th;I Tc;II
3:77
Now employ the denitions of the temperature eectivenesses from Eqs. (3.51) and
(3.52), using the nomenclature of Fig. 3.6:
"h
Th;I Th ;II
Th;I Tc;II
"c
Tc;I Tc;II
Th;I Tc;II
3:78
Th;I Tc;I
1 "c 1 Ch =Cc "h
3:79
where Eq. (3.53) is used to obtain the last equality. Substituting Eq. (3.79) into (3.76) and
rearranging, we get
"h
1 expUA=Ch 1 Ch =Cc
1 Ch =Cc expUA=Ch 1 Ch =Cc
3:80
Now the temperature eectiveness of the cold uid, "c , can be determined either
directly by employing its denition [Eq. (3.52)], or by substituting Eq. (3.80) into
(3.53). Using the second approach, gives us
"c
Ch
1 expUA=Cc 1 Cc =Ch
"
Cc h Cc =Ch expUA=Cc 1 Cc =Ch
3:81
Notice that the argument of the exponential terms has been rearranged. By multiplying
both numerator and denominator by exp fUA=Cc 1 Cc =Ch g and rearranging, we
get
"c
1 expUA=Cc 1 Cc =Ch
1 Cc =Ch expUA=Cc 1 Cc =Ch
3:82
A comparison of Eqs. (3.80) and (3.82) reveals that Eq. (3.82) can be obtained directly
from Eq. (3.80) by replacing the subscripts h with c and c with h.
To generalize the solution, let Cmin Cc , C* Cc =Ch and NTU
UA=Cmin UA=Cc . In this case, " "c from Eq. (3.52), and Eq. (3.82) reduces to
" "cf
1 expNTU1 C*
1 C* expNTU1 C*
3:83
125
3:84
Note that when C* 0, the temperature of the Cmax uid remains constant throughout
the exchanger, as shown in Fig. 3.1b and c. In this case, the Cmin uid can have any
arbitrary ow arrangement. Hence Eq. (3.84) is valid for all ow arrangements when
C* 0.
For the special case of C* 1, Eq. (3.83) reduces to the 0/0 form. Hence, using
lHospitals rule (taking the derivatives of the numerator and the denominator separately
with respect to C* and substituting C* 1 in the resultant equation), we get
"
NTU
1 NTU
3:85
For all 0 < C* < 1; the value of " falls in between those of Eqs. (3.84) and (3.85), as
shown in Fig. 3.7.
126
By inverting Eq. (3.83), NTU for a counterow exchanger can be expressed explicitly
as a function of " and C*:
NTU
1
1 C*"
ln
1 C*
1"
3:86
NTU ln
1
1"
or
NTU ln
Th;i Tc;i
Th;i Tc;o
For C* 1 :
NTU
3:88
The "-NTU results for the counterow exchanger are presented in Fig. 3.7. The
following important observations may be made by reviewing Fig. 3.7:
1. The heat exchanger effectiveness " increases monotonically with increasing values
of NTU for a specied C*. For all C*, " ! 1 as NTU ! 1. Note that this is true
for the counterow exchanger, and " may not necessarily approach unity for many
other exchanger ow arrangements, as can be found from Tables 3.3 and 3.6.
2. The exchanger effectiveness " increases with decreasing values of C* for a specied
NTU.
3. For " 9 40%, the heat capacity rate ratio C* does not have a signicant inuence
on the exchanger effectiveness ". As a matter of fact, it can be shown that when
NTU ! 0, the effectiveness for all ow arrangements reduces to
" NTU1 12 NTU1 C*
3:89
This formula will yield highly accurate results for decreasing value of NTU for
NTU 9 0.4.
4. Although not obvious from Fig. 3.7, but based on Eq. (3.84), the " vs. NTU curve
is identical for all exchanger ow arrangements, including those of Tables 3.3 and
3.6 for C* 0.
5. Heat exchanger effectiveness increases with increasing NTU as noted above in item
1, but at a diminishing rate. For example, increasing NTU from 0.5 to 1 at C* 1
increases " from 0.333 to 0.50; a 50% increase in " for a 100% increase in NTU (or
approximately 100% increase in surface area). Increasing NTU from 1 to 2 at
C* 1 increases " from 0.50 to 0.667, a 33% increase in " for a 100% increase
in NTU (or size). Increasing NTU from 2 to 4 at C* 1 increases " from 0.667 to
0.8, a 20% increase in " for a 100% increase in NTU (or size). This clearly shows a
diminishing rate of increase in " with NTU.
6. Because of the asymptotic nature of the "-NTU curves, a signicant increase in
NTU and hence in the exchanger size is required for a small increase in " at high
values of ". For example, for C* 1, " 90% at NTU 9, and " 92% at
NTU 11:5. Thus an increase of 2 percentage points in " requires an increase
127
of 28% in NTU and hence in the exchanger size for the same surface area and
ow rates. Alternatively, a larger increase in NTU (or the size of the exchanger)
is required to compensate for the same (or small) amount of heat loss to the
surroundings at high values of " in comparison to that for a lower " exchanger.
The counterow exchanger has the highest exchanger eectiveness " for specied
NTU and C* of that for all other exchanger ow arrangements. Thus, for a given
NTU and C*, maximum heat transfer performance is achieved for counterow; alternatively, the heat transfer surface area is utilized most eciently for counterow
compared to all other ow arrangements.
Should we design heat exchangers at high eectiveness for maximum heat transfer?
Let us interpret the results of Fig. 3.7 from the industrial point of view. It should be
emphasized that many industrial heat exchangers are not counterow, and the following
discussion in general is valid for other ow arrangements.
When the heat exchanger cost is an important consideration, most heat exchangers
are designed in the approximate linear range of "-NTU curves (NTU 2 or " 60%,
such as in Fig. 3.7) that will meet the required heat duty with appropriate values of Cmin
and Tmax . The reason for this is that an increase in the exchanger size (NTU) will
increase with " approximately linearly and hence will result in a good return on the
investment of added surface area. However, when the exchanger is a component in
the system, and an increase in the exchanger eectiveness has a signicant impact on
reducing the system operating cost compared to an increase in the exchanger cost,
the exchangers are designed for high eectivenesses. For example, a 1% increase in the
regenerator eectiveness approximately increases about 0.5% the thermal eciency of an
electricity generating gas turbine power plant. This increase in the power plant eciency
will translate into millions of dollars worth of additional electricity generated annually.
Hence, the cost of the regenerator represents only a small fraction of the annual savings
generated. Therefore, most gas turbine regenerators are designed at about " 90%. A
1% increase in the eectiveness of a cryogenic heat exchanger used in an air separation
plant will decrease the compressor power consumption by about 5%. This will translate
into an annual savings of more than $500,000 (in 1995 dollars) in operating costs for a
1000-ton oxygen plant. That is why most cryogenic heat exchangers are designed for
" 95%:
3.4.1.2 Exchangers with Other Flow Arrangements. The derivation of "-NTU
formulas for other heat exchanger ow arrangements becomes more dicult as the
ow arrangement becomes more complicated (Sekulic et al., 1999). The solutions
have been obtained in the recent past for many complicated ow arrangements, as in
Table 3.6 and Shah and Pignotti (1989). Temperature distributions for counterow,
parallelow, and unmixedunmixed crossow exchangers are derived in Sections 11.2.1
and 11.2.4.
"-NTU Results. Table 3.3 and Figs. 3.8 through 3.11 show the "-NTU formulas and
results for some heat exchanger ow arrangements. NTU can be expressed explicitly as
a function of " and C* only for some ow arrangements and those formulas are
presented in Table 3.4. For all other ow arrangements, NTU is an implicit function
of " and C* and can be determined iteratively or by using appropriate analytical
methods to solve the equation f (NTU 0. C* cannot be expressed explicitly as a
function of " and NTU for any exchanger ow arrangements.
128
12 shell-and-tube exchanger;
shell uid mixed; TEMA E shell
C*n Pn NTU
n
X
1
n 1 j nj
y
n 1! j1
j!
n1
2
1 C* 1 C*2 1=2 coth=2
"
"
1
1 expfC*1 expNTUg
C*
1
"
1
C*
1
1 expNTU 1 expNTU C* NTU
Pn y
exp1 C*NTU
1
1 C*
2
p
2 coth=2
p
where 2 NTU
"
1
2=1 expNTU 1=NTU
"
"
2
1 C* 1 C*2 1=2
1
1 C*
"
"
" 1 expNTU
" 12 1 expNTU
1 expNTU1 C*
1 C*
"
Parallelow
NTU
1 NTU
"
1 expNTU1 C*
1 C* expNTU1 C*
"-NTU Formulas
for C* 1
"
"-NTU Formulas
Counterow
Flow Arrangement
TABLE 3.3 eNTU Formulas and Limiting Values of e for C* 1 and NTU ! 1 for Various Exchanger Flow Arrangements
FIGURE 3.9 Unmixedunmixed crossow exchanger " as a function of NTU and C*.
129
130
FIGURE 3.10 Unmixedmixed crossow exchanger " as a function of NTU and C*.
FIGURE 3.11 Mixedmixed crossow exchanger " as a function of NTU and C*.
131
TABLE 3.4 NTU as an Explicit Function of e and C* for Known Heat Exchanger Flow
Arrangements
Flow Arrangement
Formula
Counterow
NTU
1
1 C*"
ln
1 C*
1"
"
NTU
1"
NTU
Parallelow
12 TEMA E Shell-and-Tube
C* < 1
C* 1
1
NTU ln 1
ln1 C*"
C*
NTU
NTU
1
ln1 C* ln1 "
C*
1 2 "1 C* D
ln
D 2 "1 C* D
From a review of Fig. 3.11, it is found for C* > 0 that the eectiveness of the mixed
mixed crossow exchanger increases with increasing NTU up to some value of NTU
beyond which increasing NTU results in a decrease in ". The reason for this unexpected
behavior is that due to the idealized complete mixing of both uids, the hot-uid temperature is lower than the cold-uid temperature in a part of the exchanger, resulting in a
temperature cross (discussed in Section 3.6.1.2) and existence of entropy generation
extrema (discussed in Section 11.4). Hence, heat transfer takes place in the reverse
direction, which in turn reduces the exchanger eectiveness.
3.4.1.3 Interpretation of e-NTU Results. As noted in Eq. (3.7), the heat exchanger
design problem in general has six independent and one or more dependent variables.
For analysis and concise presentation of the results, we arrived at a total of three
dimensionless groups (", NTU, and C*) for this method. However, from the industrial
viewpoint, the nal objective in general is to determine dimensional variables and
parameters of interest in any heat exchanger problem. The nondimensionalization is
the shortest road to obtaining the results of interest. In this regard, let us review Eq.
(3.35) again:
q "Cmin Th;i Tc;i "m_ cp min Th;i Tc;i
3:90
If we increase the ow rate on the Cc Cmin side (such as on the air side of a water-to-air
heat exchanger), it will increase C* Cmin =Cmax , reduce NTU UA=Cmin , and
132
hence will reduce " nonlinearly at a lower rate (see Figs. 3.7 to 3.11). Thus from Eq.
(3.90), q will increase linearly with increasing Cmin and q will decrease nonlinearly at a
lower rate with decreasing ". The net eect will be that q will increase. From Eq. (3.5),
q Cc Tc;o Tc;i Ch Th;i Th;o
3:91
Thus, overall, a lower increase in q and a linear increase in Cc will yield reduced
Tc;o Tc;i ; it means that Tc;o will be lower for a given Tc;i . Again from Eq. (3.91),
Th;o will also be lower with increased q for given Ch and Th;i . These results are shown
qualitatively by an arrow up or down in Table 3.5 depending on whether the particular
value increases or decreases. It is interesting to note that both Th;o and Tc;o in this table
can either increase, decrease, or one increases and the other decreases! A thorough
understanding of the qualitative trends of the variables of this table is necessary to
interpret the experimental/analytical results and anomalies or unexpected test results.
Reviewing Eq. (3.90), the desired heat transfer rate q in a heat exchanger can be
obtained by varying either " of Cmin since the ITD Th;i Tc;i is generally a xed
design value in many applications. The eectiveness " can be varied by changing the
surface area A (the size of the exchanger) and hence NTU, impacting the capital cost; "
increases with A or NTU nonlinearly and asymptotically. The minimum heat capacity
rate Cmin m_ cp can be varied by changing the mass ow rate through a uid pumping
device (fan, blower, or pump); the change in Cmin has a direct (linear) impact on q. Hence,
in general, specied percentage change in Cmin will have more impact on q than the
change in " due to the same percentage change in A or the size of the exchanger.
Current practice in many industries is to maintain the heat exchanger eectiveness
60% or lower to optimize the heat exchanger cost. The desired heat transfer rate is
then obtained by using a proper capacity uid pumping device, which in turn has a
major impact on the operating cost. Thus the heat exchanger and the uid pumping
device are selected individually. However, to get the desired heat transfer rate q in an
exchanger, the better cost eective approach would be to select optimally the heat
exchanger size (NTU) and the uid pumping power to get the appropriate ow rate
(and hence Cmin :
Also, a review of Eq. (3.90) indicates that if there is an insucient uid ow rate m_ ,
the exchanger will not provide the desired performance even if " ! 100% for a given
TABLE 3.5 Eect of Increasing One Independent Variable at a Time on the Remaining Heat
Exchanger Variables Based on the e-NTU Relationshipsa
Specic Variable
with
Increasing Value
Variables Aected
"
NTU
C*
Th;o
Tc;o
m_ h or Ch
m_ c or Cc
Th;i
"
"
"
"
"
"
"
"
#
"
#
"
Tc;i
"
"
hc or Ac
"
"
"
"
hh or Ah
"
"
"
"
133
Tmax . This is important when the uid ow is blocked partially or totally in a heat
exchanger, due to inadequate manufacturing or specic application/installation in a
system.
Equation (3.90) also implies that a perfectly designed exchanger may not operate
properly in a system if the heat duty of the exchanger is dictated by the system. For
example, if the engine heat rejection is reduced due to the use of a smaller engine in a
given automobile, correspondingly, Th;i Tc;i will reduce for the same fan, and the
passenger compartment will not warm up to a comfortable level with the same automotive heater (i.e., given "). To be more specic, consider the coolant temperature Th;i
from the engine lower than the desired value (such as 758C vs. desired value of 908C) with
a smaller engine. Hence, the ambient air will not be heated to the desired value (such as
508C vs. the desired value of 628C) even though the heater is designed properly (i.e.,
desired ") due to the reduced ITD.
3.4.1.4 Stream Symmetry. A two-uid heat exchanger is considered to be stream
symmetric if the exchanger eectiveness " (or P) remains invariant by switching the
two uids (such as the tube uid to the shell side, and the shell uid to the tube side).
Alternatively, the log-mean temperature dierence correction factor F (dened in
Section 3.7.2) remains the same for the original and stream-switched congurations.
Thus, for a stream symmetric exchanger, as shown in Eq. (3.36),
" NTU; C*; flow arrangement
3:92
regardless of whether the Cmin uid is on the tube or shell side in a shell-and-tube
exchanger, for example.
All geometrically symmetrical exchangers are stream symmetric such as single-pass
counterow, parallelow, and mixedmixed and unmixedunmixed crossow exchangers. Some geometrically asymmetric exchangers (such as 12 and 11 TEMA E, 11
TEMA G, and 11 TEMA J exchangers, all with shell uid mixed) are also stream
symmetric exchangers.
If an exchanger is stream asymmetric, the formula for the exchanger eectiveness " is
dierent for the two uid sides of the exchanger (such as the shell side and tube side). This
means that
(
"
3:93a
3:93b
134
U 500 W/m2 K and surface area of 10 m2 . Consider cp 1:88 and 4.19 J/g K for oil
and water, respectively. If the ratio of convection thermal resistances of oil to water is 1.2,
and the wall and fouling resistances are negligible, calculate the wall temperature at each
end of the counterow and parallelow exchangers.
SOLUTION
Problem Data and Schematic: Fluid ow rates, inlet temperatures, and overall heat
transfer coecient are provided for counterow and parallelow arrangements
(Fig. E3.2A). Also, total heat transfer area and the ratio of thermal resistances of the
two uid sides are given.
Determine: Wall and uid exit temperatures at each end x=L 0 and 1) of the counterow and parallelow arrangements.
Assumptions: The assumptions invoked in Section 3.2.1 are valid, wall and fouling resistances and negligible, the ratio of thermal resistances of oil to water is uniform throughout the exchanger.
Analysis: Knowing the inlet temperatures, we could use the denition of heat exchanger
eectiveness " to nd the outlet temperatures of various exchangers. To determine ", we
need to determine C* and NTU. For the determination of C*, the heat capacity rates are
Ch 3700 W=K
Cc m_ cp c 0:6 kg=s 4:19 J=g K 1000 g=kg 2514 W=K
where the subscripts h and c denote the hot (oil) and cold (water) sides, respectively.
Hence, Cmin 2514 W=K. Using the denitions of C* and NTU from Eqs. (3.56) and
(3.59), we have
C*
NTU
Cc 2514 W=K
0:679
Ch 3700 W=K
UA
500 W=m2 K10 m2
1:989
Cmin
2514 W=K
FIGURE E3.2A
135
(a) Counterow Exchanger: The exchanger eectiveness for the counterow exchanger,
from Eq. (3.83), is
"
1 expNTU1 C*
1 exp1:9891 0:679
0:736
1 C* expNTU1 C* 1 0:679 exp1:9891 0:679
Since water is the Cmin uid, according to the denition of the eectiveness, Eq. (3.44),
"
0:736
Th;i Tc;i 100 158C
Hence,
Tc;o 158C 0:736100 158C 77:68C
Ans:
Employing the energy balance equation (3.5), we could nd the heat transfer rate on
the water side as
q Cc Tc;o Tc;i 2514 W=K77:6 158C 157:4 103 W
The oil outlet temperature from the energy balance equation (3.5) is then
Th;o Th;i
q
157:4 103 W
1008C
57:58C
Ch
3700 W=K
Ans:
87:88C
1 Rh =Rc
1 1:2
Ans:
Tw;II
34:38C
1 Rh =Rc
1 1:2
Ans:
(b) Parallelow Exchanger: The heat exchanger eectiveness for a parallelow exchanger, from Table 3.3, is
"
0:574
1 C*
1 0:679
The water outlet is then calculated from the denition of the eectiveness as
"
0:574
Th;i Tc;i 100 158C
so that
Tc;o 63:88C
Ans:
136
q
122:7 103 W
1008C
66:88C
Ch
3700 W=K
Ans:
53:68C
1 Rh =Rc
1 1:2
Ans:
Tw;II
65:28C
1 Rh =Rc
1 1:2
Ans:
FIGURE E3.2B
137
It can be shown that these observations are true for comparing a parallelow exchanger to an exchanger with any other ow arrangement for specied NTU, C*, and inlet
temperatures.
As expected, the parallelow exchanger provides lower heat transfer rate
78% 122:7 kW 100=157:4 kW than that of a counterow exchanger. However, if
the exchanger is designed for the eectiveness lower than 40%, there is not a signicant
dierence in the exchanger eectiveness and heat transfer rate between parallelow and
counterow exchangers at equal NTU and inlet temperatures. This is an industrially
important conclusion for low eectiveness waste heat recovery from exhaust gases
having SO2 as one of the constituents. The sulfuric acid condensation in a heat exchanger
can be prevented at atmospheric pressure if the minimum wall temperature is maintained
above about 1508C. For this case, the parallelow exchanger becomes an attractive
solution since its lowest wall temperature is higher than that of any other exchanger
ow arrangement.
Example 3.3 One important design point for a radiator design is to cool the engine at
50 km/h on a 7% grade road. Your responsibility as a design engineer is to make sure that
the coolant (50% water50% glycol) at the radiator inlet (top tank) does not exceed
1208C temperature at 100 kPa gauge radiator cap pressure. Determine the radiator top
tank temperature for the following conditions: engine heat rejection rate q 35 kW,
airow rate 0.75 kg/s, air inlet temperature 538C, and waterglycol ow rate 1.4 kg/s.
For this radiator, UA 1180 W/K. The specic heats for the air and the waterglycol
mixture are 1009 and 3664 J/kg K respectively. What will be the outlet temperature of
the waterglycol mixture? Consider the radiator with both uids unmixed.
SOLUTION
Problem Data and Schematic: Fluid ow rates, inlet temperature of the cold uid, heat
transfer rate, and the total thermal conductance are given (see Fig. E3.3).
Determine: The inlet temperature of the hot uid (waterglycol mixture).
FIGURE E3.3
138
Assumptions: The uid properties and UA are constant, and the maximum inlet
temperature for the hot uid is 1208C at 100 kPa, beyond which it will boil in the engine.
Analysis: We could nd the NTU from the information given. But rst, we have to nd
C* and Cmin :
Cair Cc m_ cp air 0:75 kg=s 1009 J=kg K 756:75 W=K Cmin
Cliquid Ch m_ cp liquid 1:4 kg=s 3664 J=kg K 5129:6 W=K
C*
NTU
Cair
756:75 W=K
0:148
Cliquid 5129:6 W=K
UA
1180 W=K
1:559
Cmin 756:75 W=K
From Fig. 3.9 or the Table 3.3 formula for an unmixedunmixed crossow exchanger, we
get
" 0:769
Hence, Th;i from Eq. (3.35) is given by
Th;i Tc;i
q
35 kW 1000 W=kW
538C
113:18C
"Cmin
0:769 756:75 W=K
Ans:
Since this is less than 1208C, the design is safe. If we would have determined Th;i > 1208C,
we would have changed the radiator design (such as increasing A and hence UA and
NTU) so that Th;i 1208C.
Using the energy balance equation (3.5), we could nd the waterglycol mixture outlet
temperature as follows:
Th;o Th;i
q
35 kW 1000 W=kW
113:18C
106:38C
Ch
5129:6 W=K
Ans:
Discussion and Comments: As we discussed in Section 2.1.2.1, the two most important
heat exchanger design problems are the rating and sizing problems. However, based
on Eq. (3.7), the six independent variables of the problem for the specied ow arrangement yields a total of 21 dierent problems, summarized later in Table 3.11. The
problem above is one example beyond the simple sizing problems (numbers 2 and 4 in
Table 3.11).
In reality, one needs to design/size a radiator (i.e., UA or NTU to be determined) such
that the top tank temperature does not exceed the boiling temperature of the water
glycol mixture at the worst design point (7% grade with air-conditioning on at the
highest blower speed for airow in a desert summer condition); and at the same time,
it requires low fan power to reduce the total cost of the radiator and fan as well as to
reduce a negative impact on the fuel economy. However, to simplify the present problem,
the UA and airow rate were given as part of the input data.
3.5
139
In the P-NTU method, the heat transfer rate from the hot uid to the cold uid in the
exchanger is expressed as
q P1 C1 Tmax P2 C2 Tmax
3:94
Note that we use the general subscripts 1 and 2 in the P-NTU method for uids 1 and 2 of
the exchanger; and individual uids 1 and 2 can be hot or cold, or Cmin or Cmax uids.
Somewhat arbitrarily, we choose the uid 1 side as the shell uid side regardless of
whether it is the hot- or cold-uid side in our representation here. For exchangers
other than shell-and-tube types, one of the two uid sides is specically designated as
uid 1 side, as shown in the corresponding sketches of Table 3.6.
It may not be a problem at all in a computer program to keep track of which uid is
the Cmin before using the proper "-NTU formula for a stream asymmetric exchanger.
140
3.5.1
Temperature Eectiveness P
The temperature eectiveness P is referred to as thermal eectiveness in the shell-andtube heat exchanger literature. It is dierent for each uid of a two-uid exchanger. For
uid 1, it is dened as the ratio of the temperature range (rise or drop) of uid 1
(regardless of whether it is a cold uid or a hot uid) to the inlet temperature dierence
(Tmax or ITD) of the two uids:
P1
T1;o T1;i
T2;i T1;i
3:96
T2;i T2;o
T2;i T1;i
P1 P2 R2
3:97
P2 P1 R1
3:98
where R1 and R2 are dened later in Eqs. (3.105) and (3.106). Note that the temperature
eectivenesses dened by Eqs. (3.96) and (3.97) are identical to those dened by Eqs.
(3.51) and (3.52).
Comparing Eqs. (3.44) and (3.96), it is found that the temperature eectiveness P1 and
the exchanger eectiveness " are related as
C
P1 min "
C1
Similarly,
P2
Cmin
"
C2
"
for C1 Cmin
"C*
for C1 Cmax
"
"C*
for C2 Cmin
for C2 Cmax
3:99
3:100
Thus, the values of P1 and P2 will always be less than or equal to ".
3.5.2
UA
C1
NTU2
UA
C2
3:101
141
NTU2 NTU1 R1
3:102
NTU
for C1 Cmin
NTU C*
for C1 Cmax
NTU
for C2 Cmin
NTU C*
for C2 Cmax
3:103
3:104
C1 T2;i T2;o
C2 T1;o T1;i
3:105
R2
C2 T1;o T1;i
C1 T2;i T2;o
3:106
1
R2
3:107
and hence
R1
C1
C2
C
R2 2
C1
C*
for C1 C min
1=C*
for C1 Cmax
C*
for C2 Cmin
1=C*
for C2 Cmax
3:108
3:109
Thus R1 and R2 are always greater than or equal to C*. Individually, values of R1 and R2
range from 0 to 1, zero being for pure vapor condensation and innity being for pure
liquid vaporization.
3.5.4
3:110
142
3.6.1.1 P-NTU Relationships. This exchanger has one shell pass and an even (rarely,
odd) number of tube passes. The simplest arrangement is one shell pass and two tube
passes, simply referred to as a 12 exchanger. Such an exchanger, having two dierent
nozzle arrangements with respect to the tube return end, is shown in Fig. 3.12 with
idealized temperature distributions. Note that the tube uid in one pass is in the
counterow direction and the tube uid in the second pass is in the parallelow
direction to the shell uid ow direction.
Although this exchanger is not geometrically symmetric, it is highly stream symmetric
if the shell uid is considered perfectly mixed at each cross section (see Section 3.4.1.4).
Hence, the temperature eectiveness P is identical for the exchangers of Fig. 3.12a and b
for the same NTU and R. In other words, it does not make any dierence in the theoretical P as to which end the exchanger inlet nozzle is located on. Also, the ow direction of
either the tube or shell uid or both uids can be reversed, and P will still be the same for
specied values of NTU and R. Note that in all these cases, all four terminal temperatures and the tube uid temperature at the end of the rst tube pass, Tt; will be identical,
as shown in Fig. 3.12a. The explicit P-NTU formula for this ow arrangement is given in
Table 3.6, and the results are shown in Fig. 3.13. From this gure, it is clear that when
NTU1 ! 1 and R1 > 0, P1 is less than 1. Also note that P1 monotonically increases with
NTU1 to asymptotic values for all values of R1 .
P-NTU RELATIONSHIPS
143
The P1 -NTU1 relationship for the 1n exchanger (one shell pass, even-n tube passes
with shell uid mixed) is also presented in Table 3.6. Note that this 1n exchanger n 4
is stream asymmetric. The temperature eectiveness P1 of the 1n exchanger n > 2 and
even) is lower than that of the 12 exchanger for specied values of NTU1 and R1 . The
maximum reduction in P1 of the 1n exchanger over the 12 exchanger occurs at R1 1.
The thermal eectivenesses of the 1n exchangers at R1 1 are shown in Table 3.7. From
a review of this table, if the P1 -NTU1 expression of Table 3.6 for the 12 exchanger is
used for the 1n exchanger n 4 and even), the error introduced is negligible for low
NTU1 (NTU1 < 3), particularly for R1 6 1. However, if this error is not acceptable, it is
recommended to employ the P1 -NTU1 expression [see Eq. (III.4) in Table 3.6] of the 14
exchanger for the 1n exchanger for n 4 and even. When n ! 1, the 1n exchanger
eectiveness approaches that of a crossow exchanger with both uids mixed, and the
formula is given in Table 3.6 as Eq. (III.5).
The P-NTU results are presented in Figs. 3.14, 3.15, and 3.16 for 12 TEMA G, H,
and J exchangers for comparisons with those in Fig. 3.13 for the 12 TEMA E exchanger.
3.6.1.2 Temperature Cross.y As dened earlier, the temperature cross refers to the
situation where the cold-uid outlet temperature Tc;o is greater than the hot-uid outlet
temperature Th;o ; and in this case, the magnitude of the temperature cross is
Tc;o Th;o . It derives its name from ctitious (e.g., counterow exchanger; see Fig.
1.50) or actual crossing (e.g., 12 TEMA E exchanger as shown in Fig. 3.17b) of the
temperature distributions of the hot and cold uids in heat exchangers. Note that the
tube uid temperature in the second pass beyond point X in Fig. 3.17b is higher than
the shell uid temperature at the corresponding axial position in the exchanger.
To further understand the temperature cross, consider a 12 exchanger with two
possible shell uid directions with respect to the tube uid direction as shown in Fig.
3.17. Since the 12 exchanger is symmetric with respect to P or F, when the shell uid is
considered mixed, the same overall performance is obtained regardless of the direction of
the shell uid or whether the hot or cold uid is the shell uid. In Fig. 3.17 the solid line
temperature distributions correspond to a high-NTUs case; and Ps and F are identical in
both cases for the same NTUs and Rs . Similarly, the dashed line temperature distributions in Fig. 3.17 correspond to a low-NTUs case, and Ps and F will again be identical,
although dierent from those for the solid lines.
The temperature distributions of Fig. 3.17b reveal that there is a temperature cross of
the shell uid and the tube uid in the second pass for the high-NTUs case. In region J of
this case, the second tube pass transfers heat to the shell uid. This is contrary to the
original design, in which ideally, heat transfer should have taken place only in one
direction (from the shell uid to the tube uid in Fig. 3.17) throughout the two passes.
This temperature cross can be explained qualitatively as follows: Although an addition
of surface area (a high value of NTUs or a low value of F) is eective in raising the
temperature Tt; of the tube uid at the end of the rst tube pass, the temperature of the
tube uid rises in the second tube pass up to a point X. Beyond this point, the temperature of the shell uid is lower than that of the tube uid, since we have considered the
shell uid mixed at a cross section and it is cooled rapidly by the rst tube pass. Thus the
y In all the discussion of this subsection, we consider only the external temperature cross; refer to Section 11.4.1 for
external and internal temperature crosses. Also, throughout this subsection, the shell uid is considered hot, the
tube uid cold, and heat transfer takes place from the shell uid to the tube uid. However, all the discussion and
results presented in this subsection are valid if the shell uid is cold and the tube uid is hot.
144
Flow Arrangementa
1
1 R1 P1
ln
1 P1
1 R1
NTU1
I.2.2
1
1
ln
1 R1 1 P1 1 R1
P1 1 expNTU1
II.1
1
Rn
n 1! 1
n
X
n 1 j
NTUj nj
j!
j1
n1
1
X
exp1 R1 NTU1
I.2.3
R1 1 ln1 R1 P1 =1 P1
R1 1 ln1 P1 1 R1
P1
I.2.1
1 expNTU1 1 R1
1 R1
F 1
NTU1
I.1.2
I.1.3
P1
1 expNTU1 1 R1
1 R1 expNTU1 1 R1
General Formulab
I.1.1
Eq. No.
1
1 R1
for R1 1
for R1 1
for R1 1
0:78
NTU0:22
1
eR1 NTU1 1
R1
1
P1 !
R1
P1 ! 1
F!0
NTU1 ! 1
P1 !
F1
NTU1 ! 1
P1 ! 1=R1
P1 ! 1 for R1 1
P1 1 exp
2P1
P1 1 ln1 2P1
Same as Eq. (II.1)
with R1 1
1
1
NTU1 ln
2 1 2P1
P1 12 1 exp2NTU1
F 1
P1
1 P1
NTU1
1 NTU1
NTU1
P1
Value for R1 1c
TABLE 3.6 P1 NTU1 Formulas and Limiting Values of P1 for R1 1 and NTU1 ! 1 for Various Exchanger Flow Arrangements
145
1 expKR1
R1
1
1 1=R1 ln1 R1 P1
K
R1
1
R
1
1
K1 K2 NTU1
1
K2 1 expR1 NTU1
K1 1 expNTU1
P1
II.4
ln1 R1 P1 =1 P1
1 1=R1 ln1 R1 ln1 P1
II.3.3
II.3.2
P 1 exp
II.3.1
K 1 expR1 NTU1
1
1
ln
NTU1
R1
1 R1 ln1 P1
ln1 R1 P1 =1 P1
R1 1 ln1 1=R1 ln1 R1 P1
NTU1 ln
K 1 expNTU1
P1
II.2.3
II.2.2
II.2.1
P1
P1 1 ln1 ln1 P1
1
1 ln1 P1
1
1 ln1 P1
2
1
K1 NTU1
1
P1
P1 1 ln1 ln1 P1
P1
NTU1 ln
K 1 expNTU1
P 1 expK
NTU1 ln
K 1 expNTU1
P1 1 expK
P1 !
1
R1
(continued over)
1
1 R1
F !0
NTU1 ! 1
P1 ! 1 exp
F !0
1 expR1
R1
NTU1 ! 1
P1 !
146
12 TEMA E shell-and-tube
exchanger, shell uid divided into
two streams individually mixed
12 TEMA E shell-and-tube
exchanger, shell uid mixed, stream
symmetric
Flow Arrangementa
III.2
III.1.3
III.1.2
III.1.1
Eq. No.
2
1 R1 E cothE NTU1 =2
1 2 P1 1 R1 E
ln
E 2 P1 1 R1 E
E expNTU1
NTU1 R1
B exp
2
1
2 R1 2E R1 B
1
R1
2 R1 2E R1 =B
E ln1 R1 P1 =1 P1
2 P1 1 R1 E
1 R1 ln
2 P1 1 R1 E
P1
NTU1
E 1 R21 1=2
P1
General Formulab
for R1 2
1
1 E 2
1
21 NTU1
2
P1 =1 P1
ln2 P1 =2 3P1
P1
2 P1
2 3P1
1
p p
1 cothNTU1 = 2= 2
NTU1 ln
P1
Value for R1 1c
2
1 R1 E
2
2 R1
P1 !
1
R1
for R1 2
for R1 2
P1 !
F !0
NTU1 ! 1
P1 !
147
13 TEMA E shell-and-tube
exchanger, shell and tube uids
mixed, one parallelow and two
counterow passes
III.4
III.3
B X1 R1 2 X2 R1 1 X3
exp 1 NTU1 =3
;
2
i 1; 2; 3
R1 NTU1
4
B tanh
D 4 R21 1=2
D NTU1
4
A coth
P1 421 R1 DA R1 B1
3 R1
1=2
3
9
R1 R1 1
1
2
4
1=2
3
9
2
R1 R1 1
2
4
1 2
Xi
C X2 3R1 1 X1 3R1 2 X3
B tanh
NTU1
4
p
P1 44 5A B1
p
5 NTU1
A coth
4
NTU1 5
9
expNTU1 =3
2
X1 R1 1 R1 2
X3
2 1
A
X2 R1 2 R1 1
1
2 2
1 R1
with R1 1
1
C
1
2
R1
AC B
A
P1
P1 !
P1 !
for R1 1
for R1 1
(continued over)
4
21 2R1 D R1
1
R1
P1 ! 1
148
Flow Arrangementa
III.7
1D
1 R1 D
2R1 1 2
2 R1
B 2
A 2 R1 B
NTU1 2 R1
exp
2
NTU1 2 R1
exp
4
4 2 R1
B
2 R1
P1
NTU1 1 R1
D exp
2
P1 A B AB1 R1 R1 AB2
1
NTU1 1 R1
1 exp
A
1 R1
2
III.6
1 2NTU1 2
4 4NTU1 1 2
expNTU1
for R1 2
P1
for R1 1
NTU1
B
2 NTU1
Value for R1 1c
General Formulab
III.5
Eq. No.
for R1 1
2 R1
R21 R1 2
1
R1
P1 !
1
R1
for R1 2
for R1 2
P1 !
P1 !
P1 ! 1 for R1 1
149
1 exp2
4=R1 1
1 exp
4=R1 1
1 exp
4=R1 1
NTU1 4 R1
8
NTU1 4 R1
8
E
D
G 1 D2 D2 E 2 D2 1 E2
B 1 H1 E2
III.9 P1
E A B ABR1 =2=2
D expNTU1 1 R1 =2=2
NTU1
2 NTU
for R1 2
NTU1
2
for R1 4
H NTU1
1
f1 expNTU1 1 R1 =2=2g
1 R1 =2
B 1 D=1 R1 D=2
AB1 BR1 =2
1 AR1 =2 ABR1
III.8 P1 E1 1 BR1 =2
1
R1
P1 !
1
R1
1
(continued over)
for R1 4
for R1 4
4 R1 3
4 R1 R31 16
R1 2
P1 ! R1
P1 !
for R1 2
P1 !
150
Flow Arrangementa
III.11
III.10
Eq. No.
1
2 R1 2A R1 B
1
R1
2 R1 2A R1 =B
A expNTU1 )
R2 1=2
1 1
4
A 1=2
A 1
A1 =2
1 1 A
C
D1
A 1
A 1 1
1
R
1 1 B 2 CD
2
P1
A expNTU1
NTU1 R1
B exp
2
P1
General Formulab
1
1 A2
1
21 NTU1
2
for R1 2
P1
Value for R1 1c
2
2 R1
P1 !
P1 !
for R1 2
1
R
1 1
2
1
R1
for R1 2
P1 !
151
R2 1=2
1 1
16
1 P1;Ai
NTU1
i1
n
X
NTU1;Ai
n
X
1
1
R1
R
1;Ai
i1
i1
IV.1.2
n
Y
P1 1
IV.1.3
E exp
A expNTU1
A 1=2
A 1
A1 =2
1 1 A
C
D1
A 1
A 1
1
R 1 3E
1 1
B 2 CD
4 1E
P1
IV.1.1
III.13
III.12
i1
n
X
1
R1;Ai
Same as Eq. (IV.1.3)
1
3R
1 1
4
(continued over)
NTU1 ! 1
P1 !
152
Flow Arrangementa
i1
n
X
P1
R1 R1;Ai ;
IV.3.1.
IV.3.2
NTU1
i1
n
X
NTU1;Ai
i 1; . . . ; n
n
Y
1
1
1 1 R1 P1;Ai
1 R1
i1
IV.3.3
NTU1;Ai
i 1; . . . ; n
n
1 X
NTU1;Ai FAi
NTU1 i1
NTU1
R1 R1;Ai ;
Qn
Qn
i1 1 R1 P1;Ai
i1 1 P1;Ai
Q
P1 Qn
n
1
R
P
R
1 1;Ai
1
i1
i1 1 P1;Ai
General Formulab
IV.2.4
IV.2.3
IV.2.2
IV.2.1
Eq. No.
i 1; . . . ; n
n
Y
1
1 2P1;Ai
1
2
i1
1 R1;Ai
P1
i 1; . . . ; n
Pn
i1 P1;Ai =1 P1;Ai
P
1 ni1 P1;Ai =1 P1;Ai
1 R1;Ai ;
P1
Value for R1 1
NTU1 ! 1
Same as Eq:IV:2:1
counterflow
153
Pp 1; y
1
1y
Flow Arrangementa
1 expx1 y
1y
V.2
P1 A
V.1
B Pc NTU1 ; R1
P1 B
A Pp NTU1 ; R1
General Formulab
Eq. No.
Pp x; 1 12 1 exp2x
Pp x; y
Single-Pass Parallelow
1 expx1 y
1 y expx1 y
1 exp2NTU1
2
NTU1
P1
1 NTU1
P1
Pc x; 1
1
1 R1
8
< 1
P1
1
:
R1
P1
(continued over)
for R1 > 1
for R1 1
x
1x
1
for y < 1
Pc 1; y
1=y for y > 1
Pc x; y
Single-pass Counterow
In all formulas of plate heat exchangers with the number of thermal plates N ! 1 (equation numbers starting with V.), the single-pass parallel ow and
counterow temperature eectivenesses are presented in implicit forms. Their explicit forms are as follows with x and y representing the appropriate values of
the number of transfer units and heat capacity rate ratio, respectively.
154
Flow Arrangementa
V.5
V.4
P1 12 A B 12 ABR1
R
A Pp NTU1 ; 1
2
R
B Pc NTU1 ; 1
2
V.3
R
NTU1 ; 1
3
R
NTU1 ; 1
3
R
B Pc NTU1 ; 1
3
A Pp
1
R A
R B
2 1
P1 A B 1 1
3
3
3
R
B Pc NTU1 ; 1
3
A Pp
1
R B
R A
2 1
P1 B A 1 1
3
3
3
General Formulab
Eq. No.
for R1 3
NTU1
B
1 NTU1
for R1 3
NTU1
B
1 NTU1
for R1 2
NTU1
B
1 NTU1
>
1
>
>
:
R1
2
2 R1
8
9 R1
>
>
>
< 9 3R
1
P1
> 1
>
>
:
R1
R1 > 3
R1 3
R1 > 3
R1 3
for R1 > 2
for R1 2
8
9 R1
>
>
>
< 3 R 2
1
P1
>
1
>
>
:
R1
P1
8
>
>
>
<
155
V.8
V.7
V:6
P1 B2 B1 R1
NTU1
; R1
B Pc
2
R
B Pc NTU1 ; 1
4
R
A Pp NTU1 ; 1
4
AR1 2
BR1 2
1
Q 1
4
4
1Q
P1
R1
NTU1
1 NTU1
NTU1
2 NTU1
for R1 1
for R1 4
16
4 R1 2
R1 > 4
R1 4
(continued over)
8
>
< 1 R 1 R1 1
P 1 R1 1
>
R1 > 1
:
R21
P1
8
>
>
>
<
156
Flow Arrangementa
V.11
V.10
V.9
Eq. No.
NTU1
; R1
2
2 D
A2 B2
9 3
D1 DABA B
3
D2 A2 B2
9
5 4D
AB
9
3
P1 A B
2A A2 1 R1
1 R1 A2
A Pp
P1
General Formulab
3
2
NTU1
2 NTU1
for R1
P1
R1
3
2
>
4R2 2R1 3
3
>
>
: 12
R1 >
2
2R1 3 2R1
8
9 2R1
>
>
> 9 6R
<
P1
157
V.13
V:12
1
2R1 H=3
NTU1 2R1
;
2
3
NTU1 2R1
;
2
3
B Pc
NTU1 R1
;
2
2
NTU1 R1
;
2
2
A B ABR1 =2
2
A Pp
P1 2D 1 R1 D2
H Pp
1
;
2R1 G=3
G Pc
C
2
AE 1
1A
; C
F
E
B
D R1 E 2 C R1 E R1
2R1 EF 2 2EF F F 2
2
2R1 E F 2 E 2 F 2 2EF E F
A 0:5B 0:5C D
P1
R1
3
2
NTU1
2 NTU1
for R1 2
for R1
NTU1
G
2 NTU1
P1
4
2 R1 2
>
R1 1
>
>
:
R21
8
>
>
>
<
3
2
3
2
R1
R1 >
(continued over)
R1 > 2
R1 2
8
27 12R1 4R21
>
>
>
< 27 12R 4R2
1
1
P1
>
> 1
>
:
R1
158
2D 1 R1 D2
P1
1 D2 R1
V.14
B Pc
A Pp
NTU1 R1
;
2
2
NTU1 R1
;
2
2
A B ABR1 =2
D
2
General Formulab
Eq. No.
for R1 2
NTU1
B
2 NTU1
P1
4
4 R1 2
>
1
>
>
:
R1
8
>
>
>
<
R1 > 2
R1 2
For those ow arrangements where stream symmetric is not mentioned explicitly, they are asymmetric.
All the formulas in this table are based on the uid 1 side. They can be converted to the uid side 2 using the following relations: (1) for stream symmetric exchangers, change P1 ,
NTU1 , and R1 to P2 , NTU2 , and, R2 . (2) For stream asymmetric exchangers, convert P1 -NTU1 -R1 expressions to P2 -NTU2 -R2 expressions using the following relationships:
P1 P2 R2 , NTU1 NTU2 R2 , and R1 1=R2 .
c
Value for R1 1 unless specied dierently.
Flow Arrangementa
P-NTU RELATIONSHIPS
159
FIGURE 3.12 Idealized temperature distributions in a 12 TEMA E exchanger with shell uid
mixed for a low-NTU case.
FIGURE 3.13 P1 as a function of NTU1 and R1 for a 12 TEMA E exchanger, shell uid mixed;
constant-F factor lines are superimposed (From Shah and Pignotti, 1989).
160
12
14
16
18
110
112
11
1.0
2.0
3.0
4.0
5.0
6.0
0.463
0.557
0.579
0.584
0.585
0.586
0.463
0.553
0.569
0.568
0.564
0.560
0.463
0.553
0.567
0.564
0.558
0.552
0.463
0.552
0.566
0.562
0.555
0.549
0.463
0.552
0.566
0.561
0.554
0.547
0.463
0.552
0.566
0.561
0.553
0.546
0.462
0.552
0.563
0.559
0.551
0.544
addition of the surface area in the second tube pass left of the X is not useful from the
thermal design point of view. A review of Fig. 3.13 does reveal that Ps increases monotonically with NTUs for a specied Rs , although the rate of increase in Ps decreases with
increasing values of NTUs . Thus, increasing NTUs (or decreasing F) will result in the
higher Tt; of Fig. 3.17b and higher overall Ps , but with an increased temperature cross
(point X will move farther right with increasing NTUs ). Theoretically, the maximum
possible temperature cross for a constant U and innite surface area is
Tt;o;max Ts;o Ts;o Tt;i
3:111
P-NTU RELATIONSHIPS
161
FIGURE 3.16 P1 as a function of NTU1 and R1 for a 12 TEMA J (divided-ow) exchanger, shell
uid mixed; overall ow can be either counterow or parallelow (From Shah and Pignotti, 1989).
162
Shell fluid
Shell fluid
Tube fluid
Tube fluid
Rs < 1
Ts,i
Rs < 1
Ts,i
J
Tt,o
2
2
Ts,o
Tt,o
2
2
Tt,
Tt,
Tt,i
Tt,i
Tube pass 1
Tube pass 1
L
(a)
L
(b)
FIGURE 3.17 Idealized temperature distributions in a 12 TEMA E exchanger with shell uid
mixed for (a) high-NTU2 case with solid lines, and (b) low-NTU2 case with dashed lines (From Shah,
1983).
This equation is derived by applying the rate equations at the tube return end
(section BB) in Fig. 3.17b; the heat transfer rate from the tube uid of the second pass
to the shell uid is equal to the heat transfer rate from the shell uid to the tube uid of
the rst pass.
Now changing the nozzle orientation of the shell uid as in Fig. 3.17a, we nd no
apparent crossing of the temperature distributions, although the temperature cross
Tt;o Ts;o does exist due to the counterow direction of the shell uid and the tube
uid in the second pass. Note that in a counterow exchanger, the cold-uid outlet
temperature can be higher than the hot-uid outlet temperature (refer to Fig. 1.50).
These cases then have an external temperature cross. It must be emphasized that Tt;
can never exceed Ts;o since the shell uid and the uid in tube pass 1 represent a parallelow exchanger, as seen in Fig. 3.17a.
The temperature cross is undesirable, particularly for shell-and-tube exchangers,
because the tube surface area is not utilized cost-eectively. A good design avoids
the temperature cross in a shell-and-tube exchanger. Theoretically, the optimum design
would have the temperature cross (or temperature meet) point just at the end of the
second tube pass (Shah, 1983); that is,
Tt;o Ts;o
or
Tt;o Ts;o 0
3:112
Now
Tt;o Ts;o Tt;o Tt;i Tt;i Ts;i Ts;i Ts;o
Pt 1 Ps Pt 1 Pt Rt 0
Ts;i Tt;i
Ts;i Tt;i
3:113
P-NTU RELATIONSHIPS
163
where Pt and Ps are substituted from their denitions using Eqs. (3.96)(3.97), and the
last equality of 0 comes from Eq. (3.112). Equation (3.113) can be simplied [using Eq.
(3.98)] to the following after generalizing the subscripts t and s to 1 and 2:
P1 P2 1
P1
1
1 R1
P2
1
1 R2
3:114
Thus for a given R, Eq. (3.114) provides the limiting (maximum) value of P; correspondingly NTU computed from the NTU equation for the 12 TEMA E exchanger
in Table 3.6 provides the limiting (maximum) value of NTUs beyond which there will be
a temperature cross. These limiting values for the 12 exchanger are shown in Fig. 3.26
as an Fmin line. Implications of these limiting values, valid for all exchangers that have
temperature crosses, are discussed in detail in Section 11.4.
This concept of the temperature cross (or meet) at the exchanger outlet can readily be
utilized to determine whether or not one or more shells in series will be necessary to meet
the required heat duty without having a temperature cross in any individual shell.
Consider desired inlet and outlet temperatures for the hot uid to be 310 and 1858C,
and those for the cold uid to be 105 and 2758C. Considering a linear variation in specic
heats, the hot- and cold-uid temperature proles for an overall counterow exchanger
can be drawn as shown in Fig. 3.18 (the length of the horizontal scale is arbitrary). Now
draw a horizontal line from the cold-uid outlet temperature until it meets the hot-uid
temperature prole,y where you draw a vertical line. Continue this process until the
horizontal line meets the right-hand-side ordinate. The number of horizontal lines
350
300
Temperature (C)
Hot stream
250
200
Cold stream
150
100
50
Heat transfer, qx
FIGURE 3.18 Estimate of a required number of shells in series to avoid the temperature cross in a
12 TEMA E exchanger (From Bell, 1998).
y This horizontal line then assures the temperature cross at the outlet section of individual shell-and-tube
exchangers.
164
then indicates the number of shells required (three for Fig. 3.18) to avoid the temperature
cross in the exchanger. Thus with this procedure, there is no need to conduct a detailed
thermal analysis to nd out how many shells are required in series to avoid the temperature cross.
3.6.2
Multipass Exchangers
A multipass exchanger, dened at the beginning of Section 1.6, is a single exchanger having
multipassing of one or both uids in the exchanger. A heat exchanger array consists of a
number of individual heat exchangers connected to each other in a specied arrangement.
If the two uids are the same in a given exchanger array, the analysis presented below
applies. However, in petrochemical and rening applications, there are many heat exchangers interconnected with more than two uids (although any given exchanger may
have only two uid streams) in such a heat exchanger train. It is beyond the scope of
this book to provide heat transfer analysis of such trains. We will now derive the overall
eectiveness and related results for multipass extended surface, shell-and-tube and plate
exchangers, and many other construction types, including heat exchanger arrays. An
important additional assumption made in the following analyses is: An individual uid
stream is perfectly mixed between passes or between exchangers in an array.
3.6.2.1 Multipass Crossow Exchangers and Exchanger Arrays. These ow arrangements could be categorized as (1) a series coupling of n passes of a multipass crossow
exchanger or n individual exchangers of an exchanger array, either overall counterow
or parallelow, (2) a parallel coupling of n passes of a multipass crossow exchanger or
n individual exchangers, and (3) a combination of series and parallel couplings or other
compound multipass/array exchangers. The P-NTU expressions are now presented
separately for each category.
Series Coupling: Overall Counterow. Consider exchangers A and B in series coupling
as shown in Fig. 3.19a. The results presented below are equally valid if the two exchangers are considered as two passes of a multipass crossow exchanger. Figure 3.19b
shows their combined equivalent exchanger C. In this gure, the temperatures are
shown by two subscripts: the rst subscript refers to either uid 1 or uid 2, and
the second subscript denotes the mean terminal temperature at the location outside
the passes or exchangers. The temperature eectivenesses of uid 1 for exchangers A
and B from their denitions [see Eq. (3.96)] are given by
P1;A
T1;2 T1;1
T2;2 T1;1
P1;B
T1;3 T1;2
T2;3 T1;2
3:115
FIGURE 3.19 (a) Series coupled exchangers A and B in overall counterow arrangement;
(b) combined equivalent exchanger C.
165
P-NTU RELATIONSHIPS
T1;3 T1;1
T2;3 T1;1
3:116
Also, we need an expression for the heat capacity rate ratio R1 C1 =C2 expressed in
the form of temperatures of Fig. 3.19 as given by Eq. (3.105). Since R1 is the same for
exchangers A, B, and C of Fig. 3.19, we get
R1
Exchanger
B
3:117
Exchanger
C
To relate P1;A and P1;B of Eq. (3.115) with the overall P1 P1;C of Eq. (3.116), let us
rst compute the following expression in terms of temperatures for exchanger A by
algebraic manipulation using Eqs. (3.115) and (3.117):
1 P1;A
1 T1;2 T1;1 =T2;2 T1;1
T T1;1 T1;2 T1;1
2;2
T2;2 T2;1
T1;2 T1;1
1 R1 P1;A
T2;2 T1;1 T2;2 T2;1
1
T1;2 T1;1
T2;2 T1;1
3:118
Similarly, we can arrive at the following expression for exchanger B, using the terminology of the right-hand side of Eq. (3.118):
1 P1;B
T2;3 T1;3
3:119
1 R1 P1 T2;1 T1;1
3:120
X
1 R1 P1
1 R1 P1;A 1 R1 P1;B
3:121
where the right-hand side of the rst equality sign is designated as X. Then, from Eq.
(3.121),
P1
3:122
166
3:123
Equation (3.121) could have been obtained easily without the algebraic manipulations
of Eqs. (3.118)(3.120) by the operating lineequilibrium line approach used in chemical
engineering or by the matrix formalism method.
If we refer to exchangers A and B of Fig. 3.19 as A1 and A2 , Eq. (3.122) can be
generalized to n exchangers (or passes) Ai in series in overall counterow as
Qn
Qn
i1 1 R1 P1;Ai
i1 1 P1;Ai
Qn
P1 Qn
i1 1 R1 P1;Ai R1
i1 1 P1;Ai
3:124
Pn
i1 P1;Ai =1 P1;Ai
P
1 ni1 P1;Ai =1 P1;Ai
3:125
These formulas are reported in Table 3.6 as Eq. (IV.2.1). For this case, the overall NTU1
and R1 are related to the corresponding individual exchanger/pass quantities by the
following relationships:
NTU1
n
X
NTU1;Ai
3:126
i1
R1 R1;Ai
i 1; 2; . . . ; n
3:127
Now let us summarize the additional assumptions made to derive the overall P1 from
the individual P1;Ai [i.e., the relationships of Eqs. (3.124) and (3.125)].
1. Both uids are considered perfectly mixed between exchangers or passes. This is
the reason we considered the outlet temperatures from individual exchangers, T1;2
and T2;2 in Fig. 3.19a as mixed mean temperatures.
2. Fluid properties and ow rates are idealized as constant, so that R1 is the same for
each exchanger or each pass.
Note that in the derivation of Eq. (3.124), we did not impose any constraints on the
NTUs, ow arrangement, and stream symmetry or asymmetry of individual exchangers.
Hence, Eq. (3.124) is valid for:
1. Any values of individual NTU1;Ai may be specied (they need not be the same).
2. Individual exchangers can have any ow arrangements, such as counterow,
12 TEMA E, parallelow, and crossow, as shown in a hypothetical array of
exchangers in Fig. 3.20.
3. Individual exchangers can be stream symmetric or asymmetric.
P-NTU RELATIONSHIPS
cf
pf
12
167
Crossflow
1 P1 =1 R1 P1 1
T1;3 T1;1 T2;3 T2;1
3:128
Here the terms after the rst and second equality signs are based on the energy balances
on exchangers A and C; the term after the third equality sign can be derived using Eqs.
(3.118) and (3.120). If there are n exchangers in series, the temperatures of uids 1 and 2
after the jth exchanger/pass are T1; j1 and T2; j1 , given by
T1;1 T1; j1 T2;1 T2; j1
zj
T1;1 T1;n1 T2;1 T2;n1
Qj
i1
1 P1;Ai =1 R1 P1;Ai 1
1 P1 =1 R1 P1 1
3:129
j
X
i1
1
1 P1;Ai
3:130
If all n exchangers (or passes) have identical ow arrangements and identical individual NTU (i.e., NTUp ), Eqs. (3.124) and (3.129) are further simplied as follows:
P1
1 R1 P1; p =1 P1; p n 1
1 R1 P1; p =1 P1; p n R1
3:131
zj
1 P1; p =1 R1 P1; p j 1
1 P1 =1 R1 P1 1
3:132
zj
1 P1 =1 R1 P1 j=n 1
1 P1 =1 R1 P1 1
3:133
168
Here P1; p is the temperature eectiveness of each pass (or individual identical
exchangers). Equation (3.131) reduces to the following forms when R1 1 and 0:
P1
8
>
<
>
:
nP1; p
1 n 1P1; p
for R1 1
3:134
1 1 P1; p n
for R1 0
3:135
The temperature eectiveness per pass P1; p can be expressed in terms of P1 , R1 , and n
by inverting Eq. (3.131) as follows:
P1; p
1 R1 P1 =1 P1 1=n 1
1 R1 P1 =1 P1 1=n R1
3:136
P1; p
8
>
<
>
:
P1
n n 1P1
for R1 1
3:137
1 1 P1 1=n
for R1 0
3:138
P 1 P1; p
ln1 R1 P1 =1 P1
! 1
ln 1 R1 P1; p = 1 P1; p R1 1 P1; p 1 P1
3:139
n
Y
1
1
1 1 R1 P1;Ai
1 R1
i1
3:140
P-NTU RELATIONSHIPS
169
The temperatures of uids 1 and 2 after the jth exchanger (pass) are given by
zj
Qj
P1;Ai
P1
i1
3:141
For this case of series coupling with overall parallelow, individual NTU1;Ai and R1;Ai
are related to the overall NTU1 and R1 by Eqs. (3.126) and (3.127).
If all n exchangers (or passes) have identical ow arrangements and identical individual NTUs (i.e., NTUp ), Eqs. (3.140) and (3.141) simplify to
P1
zj
1
1 1 1 R1 P1; p n
1 R1
P j1; p
P1
3:142
j=n
P1
1j=n
P1
P1
3:143
The temperature eectiveness per pass, P1; p , can be expressed as follows from Eq.
(3.142):
P1; p
1
f1 1 1 R1 P1 1=n g
1 R1
3:144
1
1 R1
3:145
Hence,
1 R1 P1 < 1 for NTU1 < 1
3:146
The condition of Eq. (3.146) applies to Eq. (3.144) for its use.
Finally, the number of passes can be expressed from Eq. (3.142) as
n
ln1 1 R1 P1
ln1 1 R1 P1; p
3:147
170
When individual passes of the multipass overall parallelow exchanger are crossow
exchangers, the resultant arrangement is referred to as the multipass cross-parallelow. There are many possible combinations of this ow arrangement: (1) each uid is
mixed or unmixed between passes (in the preceding section we considered the uids
mixed between passes; (2) each uid is mixed or unmixed within each pass; (3) the
uid that is unmixed between passes has an identical or inverted order (see Fig. 1.55);
(4) uid 1 is the Cmin or Cmax uid; and (5) the number of passes. Baclic (1990) has
provided closed-form formulas for 36 such two-pass cross-parallelow exchangers. The
simplied relationship of Eq. (3.142) is adequate for the design and analysis of most twoand three-pass industrial exchangers.
Example 3.4 The heat exchanger array shown in Fig. E3.4a has hot and cold uids with
equal heat capacity rates and inlet temperatures of 3008C and 1008C, respectively. Heat
exchangers A, B, and C are unmixedunmixed crossow, counterow, and parallelow
exchangers, respectively. The corresponding NTUs for each exchanger are 2, 3, and 3.
Determine:
(a) the overall exchanger array effectiveness and outlet temperatures at the exit of
exchanger C
(b) the outlet temperatures at the exits of exchangers A and B.
SOLUTION
Problem Data and Schematic: Fluid inlet temperatures and each exchangers NTU are
given as shown in Fig. E3.4a. Also given are the ow arrangements of each exchanger.
Determine: The overall eectiveness and the outlet temperatures at the exit of each
exchanger.
Assumptions: The assumptions invoked in Section 3.2.1 are valid, the hot and cold uids
have equal heat capacity rates, and both uids are mixed between passes.
Analysis: (a) The overall exchanger array eectiveness will be evaluated rst by considering exchangers A and B in the overall counterow direction. Since the heat capacity rates
are equal, R 1. Knowing R, NTU, and the ow arrangement of each exchanger, we
could obtain the eectiveness of each exchanger from the formulas in Table 3.6.
PA 0:6142
PB 0:7500
PC 0:4988
Note that P, NTU, and R are the same for uids 1 and 2 since R 1. Hence, the subscripts A, B, and C used here for P and NTU designate those values for exchangers A, B,
and C for either uid.
P-NTU RELATIONSHIPS
Hot fluid
300C
NTU1 = 2
NTU2 = 3
NTU3 = 3
171
Cold fluid
100C
(a)
Th,3
300C
Th,4
PAB = 0.8212
Pc = 0.4988
Tc,4
100C
Tc,3
(b)
Tc,2 = 207.3C
Th,2 = 243.1C
Th,3 = 135.8C
Th,4 = 199.8C
Th,1 = 300C
Tc,4 = 200.2C
X
Tc,3 = 264.2C
Tc,1 = 100C
(c)
FIGURE E3.4
0:6142=0:3858 0:7500=0:2500
0:8212
1 0:6142=0:3858 0:7500=0:2500
So the problem now looks as shown in Fig. E3.4b. These exchangers are in overall
parallelow arrangement. Hence, the overall eectiveness from Eq. (3.140) for R 1 is
P
1
f1 1 1 RPAB 1 1 RPB g
1R
Ans:
The denition of the overall P in terms of the temperatures for R 1 are as follows:
P
Th;i Tc;i
Th;1 Tc;1
172
Rearranging the equations above, we could solve for Th;4 and Tc;4 :
Th;4 Th;1 PTh;1 Tc;1 3008C 0:5008300 1008C 199:88C
Ans:
Ans:
Ans:
Ans:
Thus, we know all temperatures except for the hot-uid inlet temperature to exchanger B
and the cold-uid inlet temperature to exchanger A. Applying the denition of the
temperature eectiveness to exchanger A yields
Tc;3 Tc;2 PA 3008C Tc;2
With known Tc;3 264:28C, we get
Tc;2 207:38C
Ans:
Using the denition of the temperature eectiveness again for exchanger A, we get the
following equation:
Th;2 Th;1 PA Th;1 Tc;2 3008C 0:6142300 207:38C 243:18C
Ans:
P-NTU RELATIONSHIPS
173
FIGURE 3.22 (a) Parallel coupled exchangers A and B; (b) combined equivalent exchanger C.
the second subscript denotes the mean terminal temperature at the location outside the
passes or exchangers. The temperature eectiveness of uid 1 of exchangers A, B, and C
from their denitions [see Eq. (3.96)] is given by
P1;A
T1;3 T1;2
T2;1 T1;2
P1;B
T1;2 T1;1
T2;1 T1;1
P1;C
T 1;3 T1;1
T2;1 T1;1
3:148
To relate P1;A and P1;B with P1;C , let us rst compute the following expressions from
their denitions of Eq. (3.148):
1 P1;A 1
3:149
1 P1;B 1
3:150
1 P1;C 1
3:151
3:152
3:153
For this parallel coupling, the heat capacity rate of uid 2 is divided into two
components:
C2 C2;A C2;B
3:154
174
R1 R1;A R1;B
3:155
If we refer to exchangers A and B of Fig. 3.22 as A1 and A2 , Eqs. (3.153) and (3.155)
can be generalized to n exchangers (or passes) Ai in parallel coupling as
P1 1
n
Y
1 P1;Ai
3:156
i1
n
X
1
1
R1
R
1;Ai
i1
3:157
and
NTU1
n
X
NTU1;Ai
3:158
i1
In the foregoing derivations, we idealized both uids perfectly mixed between exchangers or passes as well as uid properties and ow rates as constant. Here again, we did not
impose any constraints on the magnitude of individual exchanger NTUs, ow
arrangement, and stream symmetry or asymmetry condition.
If all passes (exchangers) have the same NTU1;Ai and R1;Ai , they will have the same
temperature eectiveness P1; p , and Eq. (3.156) simplies to
P1 1 1 P1; p n
3:159
P1; p 1 1 P1 1=n
3:160
or
and
n
ln1 P1
ln1 P1; p
3:161
Equation (3.159) can be presented in terms of P2 (of uid stream 2 of Fig. 3.22) as
follows since P1 P2 R2 from Eq. (3.98) and hence P1;Ai P2;Ai R2;Ai :
P2
n
1
1
1 1 P2; p R2; p
1 1 P2; p R2; p
R2
nR2; p
3:162
where R2 nR2; p from Eq. (3.157) since R1 1=R2 and R1;Ai 1=R2;Ai .
Exchangers shown in Fig. 1.55e and g are parallel-coupled exchangers, and their
temperature eectiveness P1 is calculated from Eq. (3.159) or Eq. (IV.1.1) of Table 3.6.
P-NTU RELATIONSHIPS
175
Determine: The total surface area required to do the job for each case.
Assumptions: The assumptions invoked in Section 3.2.1 are valid, there are no ns on
either side of tubes, and the two 12 exchangers are identical (i.e., have the same NTUp .
Analysis: (a) Two exchangers in series in overall counterow (Fig. E3.5a): Since this is an
exchanger array problem, we cannot calculate total NTU directly. First, we need to
calculate the temperature eectiveness of one of the uids for one exchanger and then
determine NTUp to come up with total NTU and A. We need to evaluate R1 and P1 to
proceed. Here we designate the shell uid (water) as uid 1 and the tube uid (oil) as
uid 2.
Cwater Cs C1 m_ cp s 1:000 kg=s 4187 J=kg K 4187:0 W=K
Coil Ct C2 m_ cp t 1:666 kg=s 1675 J=kg K 2790:6 W=K
R1
C1 4187:0 W=K
1:50 R1; p
C2 2790:6 W=K
To calculate the shell uid temperature eectiveness P1 , we calculate the water outlet
temperature based on the energy balance, using Eq. (3.5).
2790:6 W=K60 158C 4187:0 W=K95 Th;o 8C
Hence,
Th;o Ts;o 65:08C
176
Water
ms = 1.00 kg/s
60C
Oil
Oil
Tt,i = 15C
mt = 1.666 kg/s
cp,t = 1675 J/kg K
(a)
(b)
U = 540 W/m2 K
Water in
Water out
Oil out
Shell fluid
in series
Oil in
(c)
Water in
Oil out
Tube fluid
in series
Oil in
Water out
(d)
FIGURE E3.5
Ts;i Ts;o 95
C 65
C
0:375
Ts;i Tt;i 95
C 15
C
Now use Eq. (3.136) to compute the temperature eectiveness of the same uid per pass:
P1; p
1 R1 P1 =1 P1 1=n 1
1 R1 P1 =1 P1 1=n R1
0:2462
177
P-NTU RELATIONSHIPS
Using Eq. (III.1.2) of Table 3.6 or Fig. 3.14, we get NTU1; p NTUs; p :
E 1 R21 1=2 1 1:502 1=2 1:8028
NTU1; p
2 P1; p 1 R1; p E
1
1
2 0:24621 1:5 1:8028
ln
ln
E
2 P1; p 1 R1; p E 1:8028 2 0:24621 1:5 1:8028
0:3686
Hence, using Eq. (3.126) gives us
NTU1
2
X
i1
and
A
5:72 m2
U
540 W=m2 K
Ans:
(b) Two exchangers in series in overall parallelow (Fig. E3.5b). In this case, R1 and P1
are specied as in part (a).
R1 1:50
P1 0:375
The temperature eectiveness per pass, P1; p , is computed from Eq. (3.144) as
P1; p
1
f1 1 1 R1 P1 1=n g
1 R1
1
f1 1 1 1:50 0:3751=2 g 0:300
1 1:50
Using Eq. (III.1.2) of Table 3.6 or Fig. 3.14, we get NTU1; p NTUs; p as follows with
E 1:8028 as before:
NTU1; p
2 P1; p 1 R1; p E
1
1
2 0:300 1 1:5 1:8028
ln
ln
E
2 P1; p 1 R1; p E 1:8028 2 0:300 1 1:5 1:8028
0:5138
Hence, using Eq. (3.126) yields
NTU1 2NTU1; p 2 0:5138 1:0277
and
A
U
540 W=m2 K
Ans:
178
(c) Parallel coupling with the shell uid in series (Fig. E3.5c). In this case, for the
equivalent combined exchanger, R1 and P1 are 1.50 and 0.375. The individual pass
temperature eectiveness of uid 1 is computed from Eq. (3.160) as
P1;p 1 1 P1 1=2 1 1 0:3751=2 0:2094
The heat capacity rate ratio for the individual pass will change for parallel coupling and
can be determined from Eq. (3.157) as
R1; p 2R1 3:00
Thus NTU1; p is computed from Eq. (III.1.2) of Table 3.6 with E 1 R21 1=2
1 3:002 1=2 3:1623 as
NTU1; p
2 P1; p 1 R1; p E
1
1
2 0:20941 3:00 3:1623
ln
ln
2 P1; p 1 R1; p E 3:1623 2 0:20941 3:00 3:1623
E
0:4092
Hence, using Eq. (3.158) gives us
NTU1
2
X
i1
and
A
U
540 W=m2 K
Ans:
(d) Parallel coupling with the tube uid in series (Fig. E3.5d). For the parallel
coupling, consider the temperature eectiveness of the tube uid in series. Hence, in
this case, for the equivalent combined exchanger, redene R1 and P1 on the tube side
and compute the values as
R1 Rt
Coil
2790:6 W=K
0:666
Cwater 4187:0 W=K
P1 Pt
Tt;o Tt;i 60 15
C
0:5625
Ts;i Tt;i 95 15
C
The individual pass temperature eectiveness of uid 1 is computed from Eq. (3.160) as
P1; p 1 1 P1 1=2 1 1 0:56251=2 0:3386
The heat capacity rate ratio for the individual pass will change for parallel coupling as
in the previous case and can be determined from Eq. (3.157) as
R1; p Rt; p 2R1 1:332
P-NTU RELATIONSHIPS
179
1 2 P1; p 1 R1; p E
1
2 0:33861 1:332 1:6656
ln
ln
E 2 P1; p 1 R1; p E 1:6656 2 0:33861 1:332 1:6656
0:6062 NTUt; p
Hence, using Eq. (3.158), the total tube side NTU is
NTUt 2NTU1; p 2 0:6062 1:2124
UA NTUt Ct 1:2124 2790:6 W=K 3383:3 W=K
Finally, the shell-side total surface area is
A
UA
3383:3 W=K
6:27 m2
U
540 W=m2 K
Ans:
To compare results with all three previous cases, we redene all important parameters
for this last case based on the shell side.
Ps Pt Rt 0:5625 0:666 0:375
Rs
Rs; p
1
1
1:50
Rt 0:666
1
1
0:75
Rt; p 1:332
Shell-side
Parameter
Series
Coupling,
Overall
Counterow
[Case (a)]
Ps
Rs
Ps; p
Rs; p
NTUs; p
NTUs
As m2
0.375
1.50
0.2462
1.50
0.3686
0.7372
5.72
Series
Coupling,
Overall
Parallelow
[Case (b)]
0.375
1.50
0.300
1.50
0.5138
1.0277
7.97
Parallel
Coupling,
Shell Fluid
Cmax in Series
[Case (c)]
Parallel
Coupling,
Tube Fluid
Cmin in Series
[Case (d)]
0.375
1.50
0.2094
3.00
0.4092
0.8184
6.35
0.375
1.50
0.4510
0.75
0.8075
0.8075
6.27
180
Discussion and Comments: The objective of this example is threefold: compare (1) series
coupling arrangements, (2) parallel coupling arrangements, and (3) series vs. parallel
coupling arrangements.
. A comparison of cases (a) and (b) of series coupling indicates as expected that the
overall counterow arrangement is superior to the overall parallelow exchanger
requiring 5.72 m2 vs. 7.97 m2 surface area (about 28% less) for the same desired
performance. As a matter of fact, the desired overall temperature eectiveness of
uid 1 over 40% cannot be achieved in the overall parallelow arrangement; NTU1
or A1 of innity should have resulted in P1 40%, as per Eq. (3.145).
. Comparison of cases (c) and (d) indicate that these arrangements feature a minimal
dierence in overall performance. Far reaching conclusions regarding the performance level of these arrangements vs. (a) and (b) cannot be provided in all details
due to specied low exchanger eectiveness. However, what can be stated is that
splitting the streams and mixing them again causes a deterioration of the overall
performance (for a further discussion see Section 11.5) when compared with a case
that does not feature ow splitting and has the same overall ow direction (like in
case a). Of course, cases (c) and (d) still have better performance than the overall
parallelow of case (b).
. Finally, a comparison of the best cases of series and parallel coupling, cases (a) and
(d), indicate clearly, as expected, that series coupling with overall counterow
arrangement yields smallest area requirement than does parallel coupling with
the Cmin uid in series. This is due to the fact that overall counterow leads to
smaller temperature dierences than in parallelow of two separated streams (in
parallel coupling); this, in turn, results in the smallest surface area requirement for
the same temperature eectiveness for the overall counterow arrangement.
In addition to the considerations above, several other points need to be considered
when designing a heat exchanger. From the heat transfer point of view, whenever the
uid velocity is reduced in the exchanger, the corresponding heat transfer coecient will
be lower in turbulent and transition ow regimes than that for the nominal uid velocity.
This, in turn, may reduce the overall heat transfer coecient and the heat transfer rate in
the exchanger. However, the lower velocity will also reduce the pressure drop on that
side. Now compare the parallel coupling vs. series coupling. The uid, which splits into
two streams [oil for case (c) and water for case (d)], will have lower heat transfer coecient and will result in lower overall U and higher required exchanger surface area than
those calculated in the example above. However, if the pressure drop specied is severely
constrained, parallel coupling may be a choice for the uid with the limited pressure
drop, to split it into two streams in the parallel coupling. Of course, this solution will
result in a twofold penalty in surface area requirement for the same heat transfer: (1) an
increase in the surface area for parallel coupling vs. that for series coupling with overall
counterow as in the example above, and (2) a reduction in the heat transfer coecient
and U for parallel coupling resulting in increased surface area. Example 7.6 also provides
some insight into performance of single-pass vs. two-pass series or parallel coupled
exchanges. In addition, there are other considerations, such as fouling, laminar vs.
turbulent ow on the shell side, pressure containment, and thermal stresses, that should
be considered before the specic choice is made for series vs. parallel coupling for a given
application.
P-NTU RELATIONSHIPS
181
Compound Coupling. Many examples are found in industrial applications having compound multipass exchangers and exchanger arrays: (1) a mixed coupling (a series
coupling of exchangers with a combination of overall counterow and parallelow;
Fig. 3.23), (2) a combination of series and parallel couplings (Fig. 3.24), and (3) a
compound coupling that cannot be reduced to any simple conguration (Fig. 3.25).
Let us illustrate how we can analyze such compound assemblies by three examples.
Series Coupling of a TEMA E Exchanger. Consider a 12 TEMA E shell-and-tube
exchanger with four transverse (plate) baes, as shown in Fig. 3.23a. The sections
between two baes with pairs of idealized single-pass unmixedunmixed crossow
exchangers are assumed to have individual uids mixed between passes, as shown in
Fig. 3.23b. Using the series coupling of Fig. 3.19, we coalesce exchangers B1 and B2 into
a single unit B. Performing an analogous operation with two pairs C1 and C2, D1 and
D2, and E1 and E2, we arrive at the conguration shown in Fig. 3.23c. The seriescoupled exchangers A and B of Fig. 3.23c can be combined into one which then can
be reduced with C, and so on, to end up eventually with one exchanger. Hence, the
FIGURE 3.23 (a) 12 TEMA E exchanger with four baes; (b) decomposition into coupled
unmixedunmixed crossow units mixed between passes; and (c) rst reduction of the model in
(b) (From Shah and Pignotti, 1989).
182
C1 =2 1 C1 R1
R1;B R1;C R1;D
2
C2
2 C2
3:163
C1;A
2
C1 =2
2 C1
3:164
Here A in NTU1 is the total surface area on uid 1 side of the 2 pass4 pass exchanger.
Exchangers A and B in Fig. 3.24c are in parallel coupling with uid 1 divided equally in
these exchangers. Note that uids 1 and 2 in this subassembly correspond to uids 2 and 1
in Fig. 3.22. Hence the combined temperature eectiveness P2;E of this subassembly
(referred to as E in Fig. 3.24d) is given by Eq. (3.153) with the subscript 1 replaced by 2:
P2E 1 1 P2;A 1 P2;B
3:165
3:166
FIGURE 3.24 (a) 2 pass4 pass plate heat exchanger; (b) decomposition into parallelow and
counterow units; (c) rst reduction of the model in (b); (d) next reduction of the model (c).
P-NTU RELATIONSHIPS
183
3:167
where
R1;E R1
P1;A Pcf
P1;B Ppf
3:168
3:169
The resulting exchangers E and F are in series coupling with overall counterow, similar
to Fig. 3.19a. Their combined eectiveness P1 , using Eq. (3.123), is given by
P1
3:170
2PI 1 R1 P2I
1 R1 P2I
3:171
184
FIGURE 3.25 (a) 14 TEMA E exchanger with two baes; (b) decomposition into coupled
unmixedunmixed crossow units mixed between passes; (c) irreducible conguration obtained
from (b) after several reduction steps (From Shah and Pignotti, 1989).
across the longitudinal bae, this 22 F shell exchanger is a true single-pass counterow exchanger. Hence, its P-NTU formula is the same as Eq. (I.1.1) in Table 3.6.
All of the exchangers in the preceding paragraph have only one shell pass (except for
the F shell) and one or more tube passes. A careful comparison at the same P1 and R1
indicates that NTU1 and hence the surface area requirement for all these congurations
is higher than that for the counterow exchanger. This fact will be obvious when we
compare the log-mean temperature correction factor F, which will be lower than unity
P-NTU RELATIONSHIPS
185
for these congurations. An increase in surface area (and hence NTU1 ) will increase the
temperature eectiveness and hence overall thermal performance of the exchanger for
most congurations (except for the J shell; see Fig. 3.16 for P1 -NTU1 results). However,
the gain in performance will be small compared to a large amount of added surface area
in the asymptotic region of the P-NTU curve.
A remedy for this situation (a lower gain or reduction in P1 with increasing NTU1 ) is
to employ multiple shells in series, with one or more tube passes in each shell. If the
overall ow direction of the two uids is chosen as counterow (that is generally the
case), the temperature eectiveness of a given uid of such an exchanger will approach
that of a pure counterow exchanger as the number of shells is increased. Up to six shells
in series have been used in applications; generally, the number of shells in series is limited
due to the pressure drop constraint on one of the two uid streams. With an increased
number of shells in series, the temperature change for each stream will represent only a
fraction of the total temperature change, and hence the ow arrangement within each
shell then has less importance. Note that if the individual exchangers are 12 TEMA E
exchangers, the number of multiple shells is determined by the procedure outlined in
Section 3.6.1.2 to avoid the temperature cross in the exchanger.
Multiple shells in series (with individual identical shell-and-tube exchangers) are used
in some applications to increase the overall temperature eectiveness, to obtain hightemperature eectiveness where it is essential, mainly for part-load operation where
multiple shells (individual shells of smaller size) are economical, and for shipping/handling considerations. For multiple E, F, and G shells in series, both shell and tube uids
are considered as mixed outside individual shells. In this case, obtain the overall temperature eectiveness as follows:
. For series coupling of exchangers in overall counterow
. Use Eq. (3.131) if all individual exchangers are identical and their eectivenesses
is known.
. Use Eq. (3.136) if the overall temperature eectiveness of one uid is specied
and one needs to determine individual exchanger temperature eectiveness of the
same uid for sizing. In this case, all individual exchangers are identical.
. If the exchangers connected in series are not identical, use Eq. (3.124) to compute
the overall temperature eectiveness of one uid of the exchangers in series.
. For series coupling of exchangers in overall parallelow
. Use Eq. (3.142) to compute P1 if all exchangers are identical and P1; p and R1 are
known.
. Use Eq. (3.144) to determine P1; p if all exchangers are identical and P1 and R1
are known.
. Use Eq. (3.140) to calculate P1 if all exchangers are not identical.
. For parallel coupling of exchangers
. Use Eq. (3.159) to compute P1 if all exchangers are identical and P1; p is known.
. Use Eq. (3.160) to determine P1; p if all exchangers are identical and P1 is known.
. Use Eq. (3.156) to determine P1 if all exchangers are not identical.
3.6.2.3 Multipass Plate Exchangers. In a plate exchanger, although a single-pass
counterow arrangement is most common, a large number of multipass arrangements
have been used in industry, depending on design criteria. Some of them are shown in
186
Fig. 1.65. A multipass plate exchanger is designated by the number of passes that each
stream makes in the exchanger. For example, in a 2 pass1 pass plate exchanger, uid 1
makes two passes and uid 2 makes one pass in the exchanger. In each pass, there can
be any equal or unequal number of thermal plates. Essentially, these are combinations
of counterow and parallelow arrangements, with heat transfer taking place in adjacent channels. These arrangements can be obtained simply by proper gasketing around
the ports in the plates. The single-pass arrangement is used for large ow rates but
relatively small T on each uid side. The n passn pass arrangement (with n being a
large number) is used for low ow rates and relatively large T on each uid side. The
other ow arrangements are used for intermediate ow rates and T ranges as well as
unbalanced ow rates. P-NTU results are presented for 14 ow arrangements for plate
exchangers in Table 3.6. For additional 10 ow arrangements (up to 4 pass4 pass
geometry), the P-NTU results are provided by Kandlikar and Shah (1989).
3.7
In this section we introduce the concepts of the log-mean temperature dierence LMTD,
mean temperature dierence MTD, and log-mean temperature dierence correction
factor F, then implicit and explicit functional relationships among three nondimensional
groups of this method.
3.7.1
TI TII
lnTI =TII
3:172
Here TI and TII are temperature dierences between two uids at each end of a
counterow or parallelow exchanger. For a counterow exchanger, from Fig. 1.50,
TI Th;i Tc;o
3:173
3:174
187
Some limiting values of Tlm dened by Eqs. (3.172) and (3.173) are
Tlm
8
TI TII
>
>
>
<
2
TI TII
>
>
>
:
0
It can be shown that when 1 T1 =TII 2:2, the error introduced by considering
the arithmetic mean instead of the log-mean temperature dierence is within 5%, i.e.,
Tam =Tlm < 1:05 where
Tam TI TII =2 Th;i Tc;o =2 Th;o Tc;i =2
Th;i Th;o =2 Tc;i Tc;o =2
3:175a
3:176
This relationship is obtained directly from the denitions of Tlm , P1 , P2 , ", and C*, and
hence is valid for all ow arrangements. Following are two limiting forms of Eq. (3.176):
8
Tlm
>
>
1"
>
< T
max
Tlm
for C* ! 1
3:177
for " ! 1
3:178
Equation (3.177) is the same as Eq. (3.175) for TI TII . Equations (3.177) and
(3.178) clearly show that Tlm ! 0 as " ! 1. Thus a decreasing LMTD means increasing exchanger eectiveness for a given exchanger. An alternative way of interpretation is
that Tlm decreases with increasing NTU and hence increasing A.
3.7.2
As shown in Eq. (3.12), the heat transfer rate in the exchanger is represented by
q UA Tm
3:179
Here UA is the exchanger overall thermal conductance dened by Eq. (3.24), and Tm is
the true (or eective) mean temperature dierence (TMTD), simply referred to as the
mean temperature dierence (MTD). The value of Tm is dierent for dierent exchanger
ow arrangements at the same inlet and outlet uid temperatures. In contrast, the
LMTD is the same for all exchanger arrangements as given by Eq. (3.172). From Eq.
188
(3.62) and denitions of " and Tmax , the mean temperature dierence Tm can be
presented in terms of ", NTU, and temperature drop (or rise) on the hot (or cold) side as
Tm
UA=Ch
UA=Cc
NTU
3:180
NTU1
NTU1
UA=C1
UA=C2
jTt;o Tt;i j jTs;i Ts;o j
UA=Ct
UA=Cs
3:181
3:182
is obtained by eliminating dq from the energy balance equation (3.2) and the rate
equation (3.4), and then integrating the resulting expression. For all other ow arrangements, integration of these dierential energy and rate equations yields a complicated
explicit or implicit expression for Tm . Hence, for these ow arrangements it is customary to dene a correction factor F as a ratio of the true mean temperature dierence to
the log-mean temperature dierence, or a ratio of the actual heat transfer rate in a given
exchanger to that in a counterow exchanger having the same UA and uid terminal
temperatures, as shown by the following two equalities:
F
Tm
q
Tlm UA Tlm
3:183
Thus,
q UAF Tlm
3:184
F is referred to as the log-mean temperature dierence correction factor, MTD correction factor, or exchanger conguration correction factor. It is dimensionless. It can be
shown that in general it is dependent on the temperature eectiveness P, the heat capacity
rate ratio R, and the ow arrangement.
F
1 P1 ; R1 1 P2 ; R2
1 P1 ; R1 2 P2 ; R2
3:185
3:186
As an example, the explicit relationships of Eqs. (3.185) and (3.186) are shown in
Table 3.8 for the stream symmetric 12 TEMA E exchanger and the stream asymmetric
crossow exchanger with one uid unmixed and the other mixed.
F is unity for a true counterow exchanger, and thus the maximum temperature
potential (driving force) Tm for any heat exchanger may approach the log-mean temperature dierence Tlm (computed considering two uids hypothetically arranged in
counterow). F is generally less than unity for all other ow arrangements provided that
189
TABLE 3.8 F as an Explicit Function of P1 and R1 only for the Specic Heat Exchanger Flow
Arrangements Listed Here.
Flow Arrangement
Formula
Counterow
Parallelow
Crossow (single-pass)
uid 1 unmixed, uid 2 mixed, stream
asymmetric
F 1
F 1
ln1 R2 P2 =1 P2
1 1=R2 ln1 R2 ln1 P2
ln1 R1 P1 =1 P1
1 1=R1 ln1 R1 ln1 P1
ln1 R1 P1 =1 P1
R1 1 ln1 1=R1 ln1 R1 P1
ln1 R2 P2 =1 P2
R2 1 ln1 1=R2 ln1 R2 P2
D1 ln1 R1 P1 =1 P1
2 P1 1 R1 D1
1 R1 ln
2 P1 1 R1 D1
D2 ln1 R2 P2 =1 P2
2 P2 1 R2 D2
1 R2 ln
2 P2 1 R2 D2
D2 1 R22 1=2
F 1
R1 , and P1 are both dierent from zero. F can be explicitly presented as a function of P1 ,
R1 and NTU1 by Eq. (3.202). For these ow arrangements, a limited meaning of F should
be clearly understood. It does not represent the eectiveness of a heat exchanger. But it
represents a degree of departure for the true mean temperature dierence from the
counterow log-mean temperature dierence. Alternatively, F is a gauge of the actual
exchanger performance (in terms of Tm or NTU) in comparison with the counterow
exchanger performance. An F value close to unity does not necessarily mean a highly
ecient heat exchanger; it means a close approach to the counterow performance for
the comparable operating conditions of ow rates and uid inlet temperatures.
It should be emphasized that all idealizations mentioned in Section 3.2.1 for heat
exchanger analysis are invoked for the derivation of F and Tm and are not built into the
concept of the log-mean temperature dierence Tlm as sometimes referred to in the
literature. The denition of Tlm is given by Eq. (3.172) for all exchangers, with TI and
TII dened by Eq. (3.173) for all exchangers except for the parallelow exchanger and
by Eq. (3.174) for the parallelow exchanger.
190
The MTD method is generally used for designing shell-and-tube exchangers because it
provides a feel to the designer regarding the size if the reduction in eective T (given
by Tm against the best possible T (given by Tlm for the counterow exchanger.
A large reduction in Tm over Tlm means a low value of F or a large value of NTU
[see Eq. (3.200) or (3.204)] for the same counterow eectiveness or NTUcf , or a reduction in the temperature eectiveness for the same NTU [see Eq. (3.203) or (3.204)]. In
such a case, the exchanger operates in the asymptotic region of the "-NTU or P-NTU
curve; and a large increase in surface area provides only a small increase in heat transfer.
This is true in general regardless of whether or not there is a temperature cross for a
specied exchanger ow arrangement. Because the capital cost for a shell-and-tube
exchanger is an important design consideration, generally it is designed in the steep
region of the "-NTU or P-NTU curve " < 60%; and as a rule of thumb, the F value
selected is 0.80 and higher. However, a better guideline for Fmin is provided by Eq. (3.114)
when the temperature meet is at the end of the second tube pass in a 12 TEMA E
exchanger. Further detailed comparisons between the MTD and "-NTU methods are
provided in Section 3.9.
When the temperatures of the hot and cold uids remain constant (as in a phasechange condition or C* 0) in a heat exchanger, Eqs. (3.182) and (3.183) become
Tm Tlm Th;i Tc;i ITD Tmax
and
F 1
3:187
This is also a good approximation when condensation takes place on one uid side and
evaporation on the other uid side (with each uid side having a single component or an
azeotropic uid), or one of the uids can be water or other liquids with a high heat
capacity rate and having a high ow rate. In this case, Eq. (3.179) becomes
q UA Tm UA Th;i Tc;i UAmod Th;i Tc;i
3:188
where is dened by Eq. (3.212). Here (UA)mod UA when the hot and cold uid
temperatures are truly constant. However, when they are not constant, sometimes in
the literature UAmod UA is used to take into account the nonconstancy of the mean
temperature dierence. Since the MTD method is more commonly used than the -P
method (discussed in Section 3.10), it is suggested Eq. (3.184) be used in all cases.
3.8 F FACTORS FOR VARIOUS FLOW ARRANGEMENTS
3.8.1
Counterow Exchanger
We derive an expression for Tm and hence for F for the counterow exchanger of Fig.
3.6. Following the same algebraic details starting from Eq. (3.68), and integrating Eq.
(3.75) from the hot-uid inlet (section I) to outlet (section II), we get
TII
1
1
UA
3:189
ln
TI
Cc Ch
for constant U. Note that TI Th;i Tc;o represents the temperature dierence at one
end and TII Th;o Tc;i at the other end. Replacing Cc and Ch of this equation by the
values obtained from the energy balances of Eq. (3.5), we get
ln
TII 1
1
Tc;o Tc;i Th;i Th;o UA TII TI UA
TI
q
q
3:190
191
TI TII
ln TI =TII
3:191
A comparison of Eq. (3.191) with Eq. (3.179) provides the magnitude for Tm as
Tm
TI TII
Tlm
lnTI =TII
3:192
where Tlm after the second equality sign comes from the denition of Eq. (3.172). Thus
for a counterow heat exchanger, from Eq. (3.183),
F 1
3:193
A comparison of Eq. (3.192) with Eqs. (3.180) and (3.181) yields the following relationship for the counterow exchanger:
Tm Tlm
NTU1
NTU
NTU
3:194
The relationship of Eq. (3.192) or (3.193) is valid for all C* of a counterow exchanger
and hence also for the special case of C* 0. As we noted before, when C* 0, the
counterow "-NTU relationship of Eq. (3.84) is valid for all ow arrangements. Hence,
when C* 0, regardless of the ow arrangement,
F 1
3:195
This is the case when boiling or condensation takes place on one uid side in a heat
exchanger.
3.8.2
Parallelow Exchanger
From a derivation similar to the counterow exchanger, we can show for the parallelow
exchanger that
Tm Tlm
TI TII
lnTI =TII
3:196
and hence,
F 1
3:197
Here again TI represents the temperature dierence at one end of the exchanger and
TII at the other end, and they are dened in Eq. (3.174). Notice that these denitions for
the parallelow exchanger are thus dierent from those for the counterow exchanger,
Eq. (3.173). If we use the denitions of TI and TII of Eq. (3.173), it can be shown that
for the parallelow exchanger,
F
R1 1 ln1 R1 P1 =1 P1
R1 1
ln1 1 R1 P1
3:198
192
3.8.3
As noted in Eq. (3.185) or (3.186), F is a function of P1 and R1 for a given ow arrangement. Let us derive this general functional relationship in an explicit form that will have
NTU1 as an additional group. Substituting the equality of Eq. (3.181) for Tm and Eq.
(3.194) for Tlm for the counterow exchanger, Eq. (3.183) for F becomes
F
Tm
Tm
Tmax P1
Tlm Tlm cf
NTU1
NTU1
Tmax P1
3:199
cf
NTU1;cf
NTU1
3:200
Here NTU1 represents the actual number of transfer units for a given exchanger. Now
NTU1;cf from Eq. (I.1.2) in Table 3.6 can be expressed as
NTU1;cf
8
ln1 R1 =P1 =1 P1
>
>
<
1 R1
>
P
1
>
:
1 P1
for R1 6 1
3:201
for R1 1
A substitution of Eq. (3.201) into Eq. (3.200) results in the desired explicit relationship
among F, P1 , R1 , and NTU1 valid for all ow arrangements (except for parallelow).
8
ln1 R1 P1 =1 P1
>
>
>
<
NTU 1 R
1
>
>
>
:
for R1 6 1
3:202
P1
NTU1 1 P1
for R1 1
Equation (3.202) is also valid for parallelow if the F expression of Eq. (3.198) is considered instead of commonly used F 1.
From Eq. (3.202), we can also express P1 as a function of F, R1 , and NTU1 as follows.
P1
8
1 expF NTU1 1 R1
>
>
>
< R expF NTU 1 R
1
>
>
>
:
F NTU1
1 F NTU1
for R 6 1
3:203
for R 1
The relationship between F and the eectiveness can be obtained by substituting Tm
from Eqs. (3.181) and (3.180) with the rst equality sign into Eq. (3.183).
F
NTUcf
Tmax P1
Tmax "
NTU
NTU1 Tlm NTU Tlm
3:204
193
Using the relationships of Eqs. (3.99), (3.103), and (3.108), F of Eq. (3.202) can be
expressed as a function of ", NTU, and C*:
8
ln1 C*"=1 "
>
>
<
NTU1 C*
for C* 6 1
>
>
:
for C* 1
"
NTU1 "
3:205
Let us emphasize the interpretation of Eqs. (3.200) and (3.202). Since these equations
are based on the premises that P1 P1;cf , R1 R1;cf , and Tmax Tmax;cf , it means
that to achieve the same counterow eectiveness for a given exchanger, we have
F NTU1 NTU1;cf ; thus the lower the value of the F factor, the higher will be the
required NTU1 .
F/
1
NTU1
3:206
Also, since F Tlm Tm Tlm cf , and F < 1 for a noncounterow exchanger, the
log-mean temperature dierence for all other exchanger ow arrangements will be higher
than that for a counterow exchanger for specied values of NTU1 and R1 .
However, if we want to compare a given exchanger A to any other exchanger B (e.g.,
TEMA E vs. TEMA J) at the same NTU1 , R1 , and Tmax , we obtain the ratio FA =FB for
the two exchangers using Eq. (3.202) twice (in the ratio form):
8
ln1 R1 P1;A =1 P1;A
>
>
>
< ln1 R1 P1;B =1 P1;B
FA
FB >
P 1 P1;B
>
>
: 1;A
P1;B 1 P1;A
for R1 6 1
3:207
for R1 1
3:208
This means that a reduction in F translates into a reduction in P1 and vice versa, when
comparing two exchangers at the same NTU1 , R1 , and Tmax .
Although F in Eq. (3.202) is a function of three nondimensional groups P1 , R1 , and
NTU1 , we know that NTU1 is a function of P1 and R1 , as can be found from Eq. (3.110).
Thus, F is a function of only two independent nondimensional groups, P1 and R1 [as
noted in Eq. (3.185)], P1 and NTU1 , or NTU1 and R1 , for a given ow arrangement.
Based on the results of Table 3.6, P1 can be presented explicitly as a function of NTU1
and R1 for all ow arrangements considered. However, NTU1 can be presented explicitly
as a function of P1 and R1 for only a few ow arrangements, as shown in Table 3.4 in
terms of the "-NTU method. Hence, the NTU1 expression for these ow arrangements
can be substituted in Eq. (3.204) to obtain F as an explicit function of P1 and R1 , or " and
C*. All known explicit formulas for the F factor are presented in Table 3.8. For all other
ow arrangements of Table 3.6, since NTU1 cannot be expressed explicitly as a function
194
FIGURE 3.26 LMTD correction factor F as a function of P1 and R1 for a 12 TEMA E shelland-tube exchanger with the shell uid mixed (From Shah, 1983).
195
side, they would be 0.7247 0:2899 2:5 and 0.4 1=2:5, and the corresponding F factor would again be 0.7897. A careful review of Fig. 3.26 indeed indicates
this fact (although the reading accuracy is only within two digits) for the stream
symmetric exchanger considered.y
2. A very steep asymptotic curve of F vs. P1 at high values of R1 implies a large
change in F for a small change in P1 . However, it does not mean that the exchanger
will have large uctuations in the heat duty since q / F (as q UAF Tlm . All
it means is that UA will have corresponding inverse large uctuations with P1
(see the asymptotic behavior of P-NTU curves in Fig. 3.13 for the same R1 ),
and hence the changes in the product UAF will have similar changes in P1 since
q P1 C1 jT1;i T1;o j.
Example 3.6 In a 12 TEMA E shell-and-tube exchanger, water enters the shell at 218C
at a rate of 1.4 kg/s. Engine oil ows through the tubes at a rate of 1.0 kg/s. The inlet and
outlet temperatures of the oil are 1508C and 908C, respectively. Determine the surface
area of the exchanger by both the MTD and "-NTU methods if U 225 W=m2 K. The
specic heats of water and oil are 4.19 and 1.67 J/g K respectively.
SOLUTION
Problem Data and Schematic: Fluid ow rates, inlet temperatures, and hot uid outlet
temperature are provided for a one shell pass/two tube pass exchanger (Fig. E3.6A). Also
given is the overall heat transfer coecient.
Determine: The surface area of the exchanger by both the MTD and "-NTU methods.
Assumptions: The assumptions invoked in Section 3.2.1 are valid and these are no ns
on either side of the tubes.
Analysis
The MTD Method. We rst determine q and Tlm . Subsequently, we could determine P,
R, and F. Finally, we apply the MTD rate equation, Eq. (3.184) to obtain A. To nd the
Water
Ts,i = 21C
ms = 1.4 kg/s
Tt,o = 90C
Oil
U = 225 W/m2 K
Tt,i = 150C
cp,t = 1.67 J/g K
FIGURE E3.6A
y
Note that the F value calculated by the given P1 and R1 will not be equal to that for P2 P1 R1 and R2
1=R1 for a stream asymmetric exchanger as shown in Eq. (3.186). In that case, compute F using Eq. (3.202) by
replacing the subscript 1 by 2 at all places.
196
heat duty, we determine the heat capacity rate for the shell uid (water) and the tube uid
(oil).
Cs m_ cp s 1:4 kg=s 4:19 103 J=kg K 5866 W=K
Ct m_ cp t 1:0 kg=s 1:67 103 J=kg K 1670 W=K
Therefore, the heat transfer rate from the oil is
q Ct Tt;i Tt;o 1670 W=K 150 908C 100:2 103 W
Using the energy balance equation, we could also nd the water outlet temperature:
Ts;o Ts;i
q
100:2 103 W
218C
38:18C
Cs
5866 W=K
TI TII
111:9 698C
88:748C
lnTI =TII ln111:98C=698C
Now the values of tube-side P1 and R1 from Eqs. (3.96) and (3.105) are
P1
0:4651
Tt;i Ts;i 150 218C
R1
0:2847
Tt;i Tt;o Cs 5866 W=K
150C
Oil
90C
TI = 150C 38.1C = 111.9C
TII = 90C 21C = 69C
38.1C
Water
21.0C
0
I
x/L
1
II
FIGURE E3.6B
197
Therefore, from Fig. 3.26 or using the formula from Table 3.8, F 0:9776. Thus the heat
transfer area from the rate equation is
A
q
100:2 103 W
5:133 m2
2
UF Tlm 225 W=m K 0:9776 88:74 K
Ans:
The "-NTU Method. First, we determine " and C*, and subsequently, NTU and A. In
this problem, Ct < Cs , hence
C*
Ct 1670 W=K
0:2847
Cs 5866 W=K
Using the denition of the eectiveness for the tube side (Cmin side), we get
"
0:4651
Tt;i Ts;i 1508C 218C
Now we could calculate the NTU either from the formula of Table 3.4 for the 12 TEMA
E exchanger or from Fig. 3.13 with proper interpretation for ", NTU, and C*. Therefore,
NTU 0:6916. Hence, the area is
A
Cmin
1670 W=K
0:6916 5:133 m2
NTU
U
225 W=m2 K
Ans:
Discussion and Comments: The solution procedure for the sizing problem is straightforward for both the MTD and "-NTU methods. Also as expected, the calculated surface
area is identical by both methods.
Example 3.7 In an oil-to-water heat exchanger, the oil enters the exchanger at 1008C
with a heat capacity rate of 3700 W/K. Water is available at 158C and 0.6 kg/s. Determine
the exit temperatures in a 12 TEMA E shell-and-tube exchanger by the MTD method
for U 500 W=m2 K and surface area of 10 m2 . Consider cp 1:88 and 4.19 J=g K for
oil and water, respectively.
SOLUTION
Problem Data and Schematic: Fluid ow rates, inlet temperatures, and overall heat
transfer coecient are provided for a 12 shell-and-tube exchanger (Fig. E3.7). Also,
the total heat transfer area and ratio of thermal resistances of the two uids are given.
Determine: The outlet temperatures at each end of the exchanger, using the MTD
method.
Assumptions: The assumptions invoked in Section 3.2.1 are valid, wall and fouling resistances are negligible, and the ratio of thermal resistances of oil to water is uniform
throughout the exchanger.
Analysis: The MTD rate equation is
q UAF Tlm
198
Oil
Ts,i = 100C, Coil = 3700 W/K, cp,s = 1.88 J/g K
Water
U = 500 W/m2 K
A = 10 m2
Tt,i = 15C
mwater = 0.6 kg/s
cp,t = 4.19 J/g K
FIGURE E3.7
Since we do not know the outlet temperatures,{ we cannot calculate directly Tlm , q, and
F. We have to use an iterative approach to get the solution. Knowing R and NTU, we
could estimate P and F to calculate the outlet temperatures from the heat balance.
NTUs
Rs
1:351
Cs
3700 W=K
Cs
3700 W=K
1:472
Ct 0:6 kg=s4190 J=kg K
The rst estimate for P from Fig. 3.13 is 0.43. The outlet temperatures are then calculated
from the denition of P.
Ps
0:43
Ts;i Tt;i
100 158C
Cs Ts;i Ts;o
158C 1:472100 63:458C 68:698C
Ct
This rating problem can be solved straightforward, similar to Example 3.2, using the P-NTU method.
We use here two decimal places for temperatures for iterative calculations although we measure temperatures
accuracte to one decimal place.
z
199
For Ps 0:43 and Rs 1:472, from Fig. 3.26 or from the F formula in Table 3.8, we get
F 0:72. Thus the heat transfer rate is
q UAF Tlm 500 W=m2 K 10 m2 0:72 39:208C 141:9 103 W
We could then use the energy balance equations to calculate the outlet temperatures.
Ts;o Ts;i
q
141:9 103 W
1008C
61:668C
Cs
3700 W=K
Tt;o Tt;i
q
141:9 103 W
158C
71:438C
Ct
2514 W=K
Since these temperatures are dierent from those assumed above, let us iterate using
the newly calculated temperatures. The new log-mean temperature dierence, Ps , and Rs
are
Tlm
Ps
Rs 1:472
For Ps 0:451 and Rs 1:472, we get F 0:613 from Fig. 3.26 or from the F formula in
Table 3.8. Thus the heat transfer rate is
q 500 W=m2 K 10 m2 0:613 36:888C 112:9 103 W
Subsequently, new outlet temperatures from the energy balance are
Ts;o 1008C
Tt;o 158C
112:9 103 W
69:478C
3700 W=K
112:9 103 W
59:938C
2514 W=K
The outlet temperatures calculated based on the rst estimated value of P and then
for the rst two iterations are summarized as follows:
Estimation
Iteration 1
Iteration 2
Correct values
Ts;o 8C
Tt;o 8C
63.45
61.66
69.47
63.08
68.79
71.43
59.93
69.35
For this problem, if we had continued iterations, the outlet uid temperatures would
continue to diverge rather than converging to the correct value. Consequently, the application of this approach is not appropriate.
200
An alternative way for solving the rating problem is to iterate on F rather than on Ps
to ensure that the solution will converge. Assume that F 0:80, a generally minimum
value for F for a good design. Determine P1 from Fig. 3.26 for R1 1:472 and F 0:8 as
P1 0:4052 (actually, a computer program was used to determine the exact value of
P1 Ps here and in the following iterations). Then from the denition of Ps , the oil
outlet temperature is given by
Ts;o Ts;i Ps Ts;i Tt;i 1008C 0:4052100 158C 65:568C
The heat transfer rate in the exchanger is
q Cs Ts;i Ts;o 3700 W=K 100 65:568C 127:43 kW
The water outlet temperature is then
Tt;o Tt;i
q
127:43 kW
158C
65:698C
Ct
0:6 kg=s4:19 kJ=kgK
q
127:43 kW
0:6080
With this value of F, determine Ps from Fig. 3.26 for R1 1:472. Subsequently, calculate
Ts;o , q, Tt;o , and Tlm as noted above and compute the new value of F. Continue iterations in this manner until the outlet temperatures are converged within the desired
accuracy. Following are the rst 16 iterations:
F
Ps
Ts;o
Tt;o
Ps
Ts;o
Tt;o
0.8
0.6080
0.7701
0.6435
0.7477
0.6672
0.7328
0.6806
0.4052
0.4513
0.4163
0.4456
0.4234
0.4417
0.4273
0.4394
65.56
61.64
64.61
62.13
64.01
62.46
63.68
62.65
65.69
71.46
67.09
70.75
67.98
70.26
68.46
69.98
0.7243
0.6868
0.7173
0.6949
0.7125
0.6979
0.7113
0.7005
0.4291
0.4375
0.4314
0.4363
0.4322
0.4359
0.4329
0.4355
63.53
62.81
63.33
62.92
63.26
62.95
63.20
62.98
68.68
69.74
68.98
69.58
69.08
69.54
69.17
69.49
The convergence of the iterations is slow and F continues to uctuate, but it will
converge.
201
This rating problem can be made guaranteed convergent using the NewtonRaphson
method [i.e., the solution of f x 0 where the function f is nonlinear in the present
case] as follows. Since
F 1 Ps ; Rs 2 Ts;o
for this particular problem, and from the overall energy balance between the shell uid
and tube uid, we get for given Ps and Rs
Tt;o 3 Ts;o
Hence,
Thus, using the functional relationship of F and Tlm with Ts;o , we obtain the following
functional relationship for q from Eq. (3.184):
q UAF Tlm UA 2 Ts;o 5 Ts;o 6 Ts;o
Since
Hence,
In Eq. (6), the only unknown is Ts;o ; which can be determined accurately by the Newton
Raphson method of solving the f x 0 equation. Subsequently, Tt;o can be computed
from the energy balance.
Discussion and Comments: This example clearly demonstrates that the rating problem
may not converge by the simple iterative scheme on Ps . It will converge slowly by
iterating on F. One may resort to a more complex numerical convergence scheme for a
guaranteed convergence. However, this problem can be solved relatively straightforward
by the P-NTU or "-NTU method, as shown in Example 3.2 for counterow and parallelow arrangements with dierent operating conditions.
3.8.4
We consider two cases: (1) determine the F factor for an exchanger array consisting of n
identical or nonidentical exchangers, and (2) determine the F factors for an exchanger
array for specied performance data (i.e., a combined equivalent exchanger), when the
number n of identical individual exchangers is increased so that individual NTU1 =n
decreases (where NTU1 is for the combined equivalent exchanger). As we discussed in
Section 3.6.2, our description of an exchanger array also applies to a multipass exchanger
having n passes.
202
For a series-coupled exchanger array in overall counterow, Fig. 3.19, the F factor
for the combined equivalent exchanger is given by (Shah and Pignotti, 1989)
F
n
1 X
NTU1;Ai FAi
NTU1 i1
3:209
where NTU1;Ai and FAi are the NTU1 and F factors for individual exchangers in the
array. Assume that the array has all identical n exchangers:
NTU1;Ai NTU1; p
and
FA;i Fp
3:210
3:211
Thus we can conclude that the F factors for individual identical exchangers and the
combined equivalent exchangers are the same for series coupling in overall counterow
exchangers.
For exchangers series coupled in overall parallelow (see Fig. 3.21) and parallel
coupled (see Figs. 3.22 and 1.58b), no closed-form relationship exists between the individual and combined equivalent exchanger F factors. For individual exchangers, F
factors are determined for known P1;Ai and R1;Ai ; and the overall F factor is determined
using Eq. (3.202) from the known P1 , NTU1 , and R1 for the combined equivalent
exchanger. The relationships between the individual and combined equivalent exchanger
P, NTU, and R are given in Table 3.6: Eq. (IV.1) for parallel coupling and Eq. (IV.3) for
series coupling in overall parallelow.
Compared to series coupling in overall parallelow or parallel coupling of n exchangers (see Example 3.5), series coupling in overall counterow direction yields the highest
eectiveness and heat transfer rate. Hence, when an exchanger with specied ow
arrangement cannot meet the specied heat duty, one alternative is to divide the exchanger into n exchangers, with individual exchangers having NTU1; p NTU1 =n, and
arrange them such that two uids have overall counterow arrangement. In this case,
the exchanger eectiveness will approach that of a counterow exchanger as n increases.
Since individual exchanger NTU1; p will reduce as n increases, F will increase [see Eq.
(3.206)] and eventually, F ! 1. However, as noted in Eq. (3.211), the F factors for
individual and combined equivalent exchangers will be identical for a given n.
Example 3.8 It is desired to preheat the feed stream of a distillation column to 1758C.
It has been proposed to exchange the heat between the column feed and the bottom
products in an existing 12 TEMA E exchanger. Data for this problem are as follows:
Column feed : shell side; m_ s 725 kg=h; cs 3:43 kJ=kg K; Ts;i 1008C
Bottom products : tube side; m_ t 590 kg=h; ct 3:38 kJ=kg K; Tt;i 2358C
Surface area A 21:58 m2 ; Overall heat transfer coefficient U 77 W=m2 K
(a) Can the feed be heated to the desired temperature of 1758C using this exchanger?
(b) If you cannot heat the feed to the desired temperature, can you explain why?
203
(c) Now instead of one 12 exchanger having NTU calculated in part (a), consider
two 12 exchangers connected in series in overall counterow arrangement, each
having NTU exactly one-half that of part (a), so that the total NTU remains the
same. Now determine whether or not the column feed can be heated to 1758C.
(d) Compare the results of parts (a) and (c) and discuss the implications.
(e) What is the F factor for the original 12 exchanger? Outline a procedure for how
to calculate the F factor for the two 12 exchangers connected in series in part (c).
Will the F factor in the latter case be higher or lower than the single 12 exchanger
of part (a)? Why?
SOLUTION
Problem Data and Schematic: The uid ow rates, inlet temperatures, total heat transfer
area, and total heat transfer coecient are as given (see Fig. E3.8A).
Determine:
1. Can the feed be heated to the desired temperature? If not, why?
2. Can doubling the number of exchangers heat the feed to the desired temperature?
3. Compare the results and the corresponding F factors for the two different types of
arrangements.
ms = 725 kg/h
Ts,i = 100C
cs = 3430 J/kg K
Column feed
A = 21.58 m2
U = 77 W/m2 K
NTU1 = 2.406
mt = 590 kg/h
Tt,i = 235C
ct = 3380 J/kg K
Bottom Product
100C
134C
NTUp = 1.203
R1 = 1.247
235C
181C
FIGURE E3.8A
204
Assumptions: The assumptions invoked in Section 3.2.1 are valid, there are no ns on
either side of tubes, and there is perfect mixing of uid streams between exchangers when
two exchangers are connected in series.
Analysis: (a)
R1
NTU1
This part could be solved in two alternative ways. As a rst alternative, we calculate
P1 and hence outlet temperatures.
From Eq. (III.1.1) of Table 3.6 or Fig. 3.13, for R1 1:247 and NTU1 2:406, we
get
P1 0:5108
Hence,
Ts;o Ts;i P1 Tt;i Ts;i 1008C 0:5108235 1008C 169:08C
Tt;o Tt;i P1 R1 Tt;i Ts;i 2358C 0:5108 1:247235 1008C 149:08C
Thus with the 12 exchanger, the column feed cannot be heated to the desired
Ts;o 1758C temperature.
Ans.
An alternative way to solve this problem is to calculate the required P1 Ps to get
1758C.
P1
0:5556
Tt;i Ts;i 235 1008C
2
1 R1 1
R21 1=2
2
1 1:247 1 1:2472 1=2
0:5201
Thus the maximum shell uid temperature eectiveness that can ideally be achieved is
0.5201 for NTU1 ! 1, while the desired P1 is 0.5556 for a 1758C outlet temperature.
Therefore, we cannot heat the column feed to 1758C even with an innite surface
area.
Note that if we had a true counterow exchanger with the specied R1 1:247 and
NTU1 2:406, we get P1 0:6446 from Eq. (I.1.1) of Table 3.6. In this case, the shellside outlet temperature is
Ts;o Ts;i P1 Tt;i Ts;i 1008C 0:6446235 1008C 187:08C
This means that the feed can be heated to 187.08C in a pure counterow exchanger.
205
(b) The F factor for this case is 0.5023 from Eq. (III.1.3) of Table 3.6 or extrapolation
in Fig. 3.26 for P1 0:5108 and R1 1:247. It is too low and there is a temperature cross
(see Fmin line in Fig. 3.26). The exchanger ow arrangement is just not ecient to heat the
feed to 1758C with the surface area provided or even increasing A to 1.
(c) Now NTU1; p 2:406=2 1:203, and R1 1:247. Hence, P1; p 0:4553 from Eq.
(III.1.1) of Table 3.6 or 0.46 from Fig. 3.13. The overall eectiveness of two 12 exchangers in overall counterow arrangement is determined from Eq. (3.131).
P1
1 R1 P1; p =1 P1; p 2 1
1 R1 P1; p =1 P1; p R1
2
0:5999
Ts;o Ts;i P1 Tt;i Ts;i 1008C 0:5999235 1008C 181:08C
Tt;o Tt;i P1 R1 Tt;i Ts;i 2358C 0:5999 1:247235 1008C 134:08C
Using two exchangers, the feed can be heated to 1818C, which is above the required value
of 1758C.
Ans.
(d) Instead of one 12 exchanger having NTU1 2:406, if we employ two 12 exchangers in series in overall counterow direction each having NTU1 1:203, we can increase
the shell uid temperature eectiveness from 0.5108 to 0.5999. Regardless of the ow
arrangement of individual heat exchangers, when two or more exchangers are connected
in series in the overall counterow direction, the overall temperature eectiveness
approaches that of a pure counterow exchanger. Note that in the present problem,
part (c), F 0:7786 for P1; p 0:4553 and R1 1:247 using Eq. (III.1.3) of Table 3.6
(or see Fig. 3.26). Hence, there is only a small temperature cross in individual exchangers,
and the surface is used more eciently.
Note that by making two smaller frontal area 12 exchangers, the ow velocities will
go up by a factor of 2, due to the reduction in ow area by one-half. This change will also
increase individual hs and hence further increase in NTU1 , P1 , and q than what is
calculated. An important but negative impact of this design will be a signicant increase
in the pressure drop on both sides (discussed further in Chapter 6).
(e) The F factor for the original 12 exchanger for NTU1 2:406 and R1 1:247 is
0.5023, as noted in part (b). When two 12 exchangers are connected in overall counterow direction, the F factor for the two exchangers in series is the same as the F factor for
the individual exchangers [see Eq. (3.211)], since F 1 for counterow arrangement.
Hence, for two 12 exchangers in series, F 0:7786 for NTU1; p 1:203 and R1 1:247
from Eq. (III.1.3) of Table 3.6 or Fig. 3.26.
An alternative proof can be presented from the general denition of the F factor of
Eq. (3.204) as follows.
F
Tmax P1
NTU1 Tlm
This F formula is valid for one exchanger or multiple exchangers in series in overall
counterow direction as long as the appropriate values of P1 , NTU1 , and Tlm are used.
For two exchangers in series, total NTU1 2:406 and P1 0:5999, as calculated in
part (c), Tmax 2358C 1008C 1358C, and Tlm 43:238C is calculated from the
temperatures found in part (c).
206
235C
Tlm =
(54.0 34.0)C
ln (54.0C/34.0C)
= 43.23C
134C
100C
x/L
1
FIGURE E3.8B
1358C 0:5999
0:7786
2:406 43:238C
Thus, the F factor for two one-half-size 12 exchangers connected in series in counterow
will be higher than that for one full-size 12 exchanger (i.e., one having the same total
NTU).
Discussion and Comments: Several points can be observed from this example:
. When a specied exchanger cannot provide the specied heat duty, one should
investigate a higher-performance ow arrangement. One can always approach
the eectiveness of a counterow exchanger by series coupling of exchangers in
overall counterow.
. The F factors for individual identical exchangers and the equivalent combined
exchanger are the same for series coupling in overall counterow. However, the
F factor for two 12 exchangers in series in overall counterow is higher than
the F factor for one 12 exchanger having the same total NTU.
. As one exchanger is divided into n identical exchangers (NTU1 nNTU1; p and
series coupled in overall counterow, NTU1; p decreases as n increases. Similarly,
with R1 R1; p , the F of individual exchangers and hence the equivalent combined
exchanger increases with increasing n. When n ! 1, NTU1; p ! 0, "p ! 0, F ! 1;
and P1 ! P1;cf .
The overall temperature eectiveness of the two exchangers exceeded the eectiveness
required for this problem. Consequently, the NTUs of the individual exchangers could be
reduced. In doing so, the thermal and physical size of the exchanger will be reduced,
which in turn may lower the cost; however, the cost of two smaller heat exchangers will
probably be higher than the cost of a single larger exchanger. Further investigation into
the use of additional exchangers is warranted. The pressure drop constraints and external
constraints (e.g., location of pipes, compartment sizes) might limit the number of
exchangers allowed.
3.9
207
The heat transfer analysis of an exchanger can be performed by using any of the "-NTU,
P-NTU, or MTD methods described earlier. Let us rst compare the heat transfer rate
equation and the relationship of nondimensional groups associated with these methods,
presented in Table 3.9. Clearly, there are three nondimensional groups associated with
each method, and there is a direct one-to-one correspondence among the three methods;
only the algebraic forms of the resulting equations are dierent.
Now let us discuss the basic steps involved in the solution of the two specic heat
exchanger problems, the rating and sizing problems. In a sizing problem, U, Cc , Ch , and
the inlet and outlet (terminal) temperatures are specied, and the surface area A is to be
determined. The basic steps involved in the "-NTU and MTD methods are presented
below. Since the P-NTU method is closely related to the "-NTU method, the solution
procedure for the P-NTU method will be identical to that for the "-NTU method by
replacing ", NTU, and C* by P1 , NTU1 , and R1 , respectively. A computer algorithm can
readily be prepared for such solution procedures.
3.9.1
3.9.1.1
P-NTU
MTD
q P1 C1 jT1;i T2;i j
q UAF Tlm
" 1 NTU; C*
P1 2 NTU1 ; R1
F 3 P1 ; R1
Cc Cmin
Ch Cmin
P1 "
R1 C*
P1 C*"
R1 1=C*
NTUcf
Th;i Tc;i "
1
1 C*"
"
ln
!
NTU Tlm
NTU
NTU1 C*
1"
NTU1 "
C 1
!
NTU1 Tlm
NTU1 Tlm
1 P1 R1 1 NTU1 1 P1
NTU1 1 R1
208
209
. The eect of change in the inlet temperature on exchanger performance can readily
be evaluated from the denition of " since NTU and C* are not signicantly
aected by changes in the inlet temperatures (except through uid
property changes). As a result, one can calculate new outlet temperatures for
changes in the inlet temperatures for given ". This is not readily possible with the
MTD method.
. The major drawback of the "-NTU method is that one needs to keep track of the
Cmin side for a stream asymmetric heat exchanger since it will involve two dierent
formulas for ", depending on which uid side is the Cmin side.
. For the graphical presentation of "-NTU results, the abscissa ranges from 0 to 1
and hence is unbounded.
3.9.3
. If uid 1 side is the Cmin side, this method is identical to the "-NTU method for
0 R1 1; otherwise, P1 P2 R2 when 0 R2 1. The rst ve and last items of
the "-NTU method are also applicable here.
. Since the P-NTU formula is valid for 0 R1 1, one can use the P-NTU formula
for the complete operating range of R1 , regardless of whether the exchanger is
symmetric or asymmetric.
3.9.4
. The F factor represents a penalty paid in the loss of mean temperature dierence
potential for using a ow arrangement dierent from the ideal counterow.
. The F factor is not like an eciency factor. A value of F close to unity does not
represent a highly ecient heat exchanger. All it means is that the exchanger
performance is close to that of a counterow exchanger for the comparable operating conditions of ow rates and inlet uid temperatures.
. The log-mean rate equation q UAF Tlm implies that only the rate equation is
required for the heat exchanger design theory, whereas, in fact, the energy balance
equation is hidden in the F factor.
. The solution to the rating problem is iterative even when the graph of F vs. P1 with
R1 as a parameter is available.
. The simple rate equation of q UAF Tlm provides a quick feel for q if Tlm is
estimated or for Tlm if q is known. Here the value of F factor is chosen between 0.8
and 1.0 and UA is known for the application from past experience.
. In the very early design stage of a new exchanger, if the designer has a feel for the
values of Tlm and the F factor from past practices, one can estimate the surface
area required on the back of an envelope for known heat duty q and known U.
One does not need to know the ow arrangement or exchanger conguration for
a rough estimate. These types of estimates are not possible with the "-NTU or
P-NTU method.
Apart from the advantages and disadvantages of each method discussed above, it
should again be emphasized that for a given input, all of the methods above will yield
identical results within the specied convergence accuracy. Generally, the "-NTU
210
method is used by automotive, aircraft, air-conditioning, refrigeration, and other industries that design/manufacture compact heat exchangers. The MTD method is used by
process, power, and petrochemical industries that design/manufacture shell-and-tube
and other noncompact heat exchangers.
The important dimensionless groups of the aforementioned three methods ("-NTU,
P-NTU, and MTD) are P1 and P2 (or "), NTU1 and NTU2 (or NTU), R1 and R2 (or C*),
and F. Hence, if these groups are presented on one graph, the solutions to the rating and
sizing problems described above can be obtained graphically in a straightforward way,
without iterations. In this regard, two additional graphical presentation methods are
available: the -P method (Mueller charts) and the P1 -P2 method (RoetzelSpang
charts).y
For the -P method, a dierent group is introduced. It is a ratio of the true mean
temperature dierence (MTD) to the inlet temperature dierence (ITD) of the two uids:
Tm
Tm
3:212
"
P1
P2
Also substituting the value of 1/NTU1 from Eq. (3.202) into Eq. (3.213),
F as
8
>
<
>
:
FP1 1 R1
ln1 R1 P1 =1 P1
for R1 6 1
F1 P1
for R 1
3:213
is related to
3:214
3:215
y
It should be emphasized that all methods ("-NTU, P-NTU, MTD, -P, and P1 -P2 ) will yield identical results
for a given set of input variables. Thus they indeed represent dierent routes to achieving the same nal goal. The
-P and P1 -P2 methods are presented here for completeness of literature information. However, the P-NTU or "NTU method is more commonly computerized today.
211
3.10.2
212
FIGURE 3.28 P1 P2 chart for a 12 TEMA E shell-and-tube exchanger with shell uid mixed
(From Roetzel and Spang, 1993).
P1 P2 axis (the 458 line from the origin). (4) Since the NTU 1 line is shown, one can
easily recognize the region beyond NTU 1 where a set of mutually consistent heat
exchanger parameters does not exist (no design is possible). From the visual inspection
point of view, in the P1 -P2 method, (1) one cannot easily recognize direct asymptotic
behavior of P with NTU as in a P-NTU chart, and (2) one cannot recognize P decreasing
with increasing NTU after some maximum value as in the temperature cross case.
213
3.11.2
Approximate Methods
Semianalytical/numerical methods used for analyzing single-pass and multipass crossow exchangers are a collocation method, Galerkin method, and two-dimensional
Roesser discrete linear image processing method. In the collocation method applied to
a two-pass crossow exchanger, rst the Laplace transforms are applied to independent
variables. The resulting equations are coupled at the boundary conditions between passes
and lead to integral equations that can be solved exactly or approximately by the collocation and Galerkin methods. When the partial dierential equations of a heat exchanger
model are discretized and organized in matrix form, they are similar in form to a discrete
image-processing model. The important advantage of these models is their straightforward computer implementation for the solution.
3.11.3
Numerical Methods
Finite dierence, nite volume, and nite element methods have been used for obtaining
"-NTU or P-NTU relationships for single and multipass crossow exchangers and some
shell-and-tube heat exchangers.
214
3.11.4
Matrix Formalism
The matrix formalism method uses matrix algebra to obtain the P-NTU formulas for an
exchanger array (e.g., see Fig. 3.19 to 3.22) or for a complex exchanger which can be
broken into simpler subexchangers whose P-NTU formulas are known. In this method,
using the overall energy balance for each uid, any two uid temperatures (e.g., outlet
temperatures) can be presented in terms of two remaining independent uid temperatures
(e.g., inlet temperatures) (see, e.g., Table 3.10). The coecients of independent uid
temperatures can then be presented in a 2 2 linear matrix. The overall P value of an
exchanger array or a complex exchanger is eventually obtained by proper multiplication
of appropriate 2 2 matrices. Thus matrix transformation rules are applied using individual exchanger eectiveness as building blocks to obtain overall eectiveness. Refer to
Sekulic et al. (1999) for further details.
3.11.5
For a very complex ow arrangement, the matrix formalism method mentioned above
becomes too dicult to handle, due to the large number of matrices involved and the
associated matrix multiplication and transformation. Our primary interest is in obtaining
only P1 or P2 , rather than performing matrix multiplication (which involves all four
terms of each 2 2 matrix) for a complex heat exchanger conguration or an array.
Hence, only one term of the 2 2 matrix (P1 or P2 ) for the overall exchanger is evaluated
through matrix multiplication of the corresponding terms of N 2 2 matrices. In fact,
the chain rule methodology oers a scheme to relate an assembly matrix element Mi j
(such as P1 ) to individual component elements Zi j , Yi j , and so on, without the use of
direct matrix algebra or matrix multiplication. This is a powerful technique to obtain a
closed-form expression for the eectiveness of many highly complex exchanger congurations and assemblies. It can analyze exchangers with multiple (more than two) inlet or
TABLE 3.10 Formulas for Exchanger 2 Terminal Temperatures as Functions of Two Remaining
Terminal Temperatures, P1 and P2
T1;o 1 P1 T1;i P1 T2;i
T1;i
1
T P1 T2;i
1 P1 1;o
T2;o
1
fP R T 1 P1 1 R1 T2;i g
1 P1 1 1 1;o
T1;i
1
1 P1 R1 T1;o P1 T2;o
1 P1 1 R1
T1;i
1
1 P1 R1 T2;i T2;o
P1 R 1
T2;i
1
P1 R1 T1;o 1 P1 T2;o
1 P1 1 R1
T1;o
1
P T 1 P1 T2;o
P1 R1 1 2;i
T1;o
1
f1 P1 1 R1 T1;i P1 T2;o g
1 P1 R 1
T2;i
1
1 P1 T1;i T1;o
P1
T2;i
1
P1 R1 T1;i T2;o
1 P1 R1
T2;o
1
f1 P1 1 R1 T1;i 1 P1 R1 T1;o g
P1
215
outlet streams (such as 12 TEMA J or H exchangers), unmixed streams between exchangers or passes, and the exchanger assembly, which may or may not be coupled only in
series and parallel coupling (i.e., connected by some compound coupling). This chain rule
methodology with examples is presented by Pignotti and Shah (1992). A number of
formulas in Table 3.6 have been determined in the recent past using this methodology.
3.11.6
Flow-Reversal Symmetry
FIGURE 3.29 (a) 1 pass2 pass plate exchanger; (b) alternative 1 pass 2 pass plate exchanger;
(c) exchanger of (b) with both uids reversed; (d) mirror image of the exchanger of (c) (From Shah
and Pignotti, 1989).
216
Most of the congurations of Table 3.6 are geometrically symmetric (i.e., after
reversal of both uids, they coincide with the original geometry, except for some trivial
transformation, such as rotation, mirror image, etc.). For such cases, ow reversal symmetry does not provide any additional information. The geometrically asymmetric cases
are III.1, III.2, III.4, III.10, III.11, III.12, V.3, V.6, V.11, V.12, and V.13 in Table 3.6. For
these cases, two seemingly dierent geometries (one shown in the schematics of Table 3.6
and the other with both ows reversed) have the same eectiveness; and hence the
eectiveness of the seemingly dierent conguration is known instantly through the
concept of ow reversibility.
3.11.7
Rules for the Determination of Exchanger Eectiveness with One Fluid Mixed
As presented in Section 3.11.6, the principle of ow reversibility indicates that when the
directions of both uids are reversed in any two-uid individual exchangers (not an
array), the exchanger eectiveness " (and P1 and F) does not change. In contrast, consider a multipass or complex conguration exchanger having at least one uid side
perfectly mixed throughout, the other uid side being mixed, unmixed, or partially
mixed. If the eectiveness of such an exchanger is known for overall parallelow (or
counterow), the eectiveness of the same exchanger conguration with the direction of
only one uid (which one is immaterial) reversed [i.e., the resulting overall counterow
(parallelow)], can be found readily as follows. The eectiveness P1 of the original
exchanger and the eectiveness P^1 of the exchanger with one uid reversed are related
as follows (Sekulic et al., 1999):
P^1 R1 ; NTU1
P1 R1 ; NTU1
1 R1 P1 R2 ; NTU1
3:216
Here the subscript 1 refers to the uid side having no restrictions (i.e., it can be mixed,
unmixed, or split). The temperature eectiveness of the mixed uid (the subscript 2) is
then given by
P2 R2 ; NTU2
P^2 R2 ; NTU2
1 P1 R2 ; NTU2
3:217
217
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Rating
m_ 1 or C1
UA
1
2
3
4
5
6
Sizing
Problem
m_ 2 or C2
T1;i
T1;o
Dimensional Parameters
T2;i
T2;o
P1
P2
NTU1
NTU2
Dimensionless Groups
R1
R2
218
are also included with known or unknown values based on the dimensional variables. If
only one of the temperatures is known, we evaluate the specic heat at that temperature
for the determination/estimate of C and hence R; eventually, once we know both temperatures, an iteration may be needed. If specic heats are treated as constant, C1 and C2
in the equations above are interchangeable with m_ 1 and m_ 2 . Also, we have presently
considered the UA product as one of the design parameters. We discuss separating U and
A when we discuss the detailed sizing problem in Chapter 9. In the foregoing count, we
have not added q since q can readily be calculated from the overall energy balance [Eq.
(3.5)] or the rate equation [Eq. (3.6) or (3.12)]. Alternatively, if q is given, one of the
temperatures or the ow rates (heat capacity rate) could be unknown in Table 3.11. We
could classify the rst six problems as variations of the sizing problem and the next 15
problems as a variation of the rating problem. Using the P-NTU method, these 21
problems can be solved as follows.
1. For problems 1 to 6, P1 (also P2 ) and R1 are known through the heat balance.
Hence, NTU1 can be calculated from the P-NTU formula for a given ow arrangement, either straightforward or iteratively, depending on whether NTU can be
expressed explicitly (see Table 3.4) or implicitly (see Table 3.6).
2. For problem 7, since all four temperatures are known, both P1 and R1 are known
through their denitions [Eqs. (3.96) and (3.105)]. Hence, NTU1 can be calculated
as mentioned in the preceding paragraph for a given ow arrangement. Then
C1 UA=NTU1 and C2 C1 =R1 . Knowing the specic heats of the given uids,
one can determine m_ 1 and m_ 2 .
3. For problems 8 to 13, NTU1 and R1 are known. Hence, determine P1 using the
appropriate formula from Table 3.6. Now knowing P1 , R1 , and the denition of
P1 , compute the unknown temperatures using the formulas given in Table 3.10.
4. Problems 14 and 16 can only be solved iteratively, with the solution procedure
for problem 14 as follows. Assume m_ 1 (or C1 ). Calculate R2 C2 =C1 . From the
problem specications, NTU2 UA=C2 and q C2 jT2;i T2;o j are known.
Hence, knowing NTU2 and R2 for a given ow arrangement, determine P2
using the appropriate formula from Table 3.6. Subsequently, compute T1;i from
the denition of P2 , and C1 from the energy balance C1 jT1;i T1;o j q. With
this new value of C1 , repeat all the calculations above. Continue to iterate until the
successive values of C1 converge within the accuracy desired. The solution procedure for problem 16 is identical to that for problem 14 just described, starting
with assuming m_ 2 (or C2 ) and computing R1 and NTU1 . The solution procedure
for problem 15 is relatively straightforward. In this case, P2 and NTU2 are given.
From the known P-NTU formula from Table 3.6, compute R2 iteratively (such
as using the NewtonRaphson method) since it is an implicit function of P2 and
NTU2 . Similarly, P1 and NTU1 are known for problem 17 and determine
iteratively R1 from the known P-NTU formula of Table 3.6.
5. Problems 18 and 20 can only be solved iteratively, with the solution procedure for
problem 18 as follows. Assume C2 and hence determine R1 C1 =C2 . Also,
compute NTU1 UA=C1 from the input. For a given ow arrangement, determine P1 using the appropriate formula from Table 3.6. Subsequently, compute T1;i
from the denition of P1 . Finally, calculate C2 from the overall energy balance
C1 T1;i T1;o C2 T2;o T2;i . With this new value of C2 , repeat all the calculations above. Continue to iterate until successive values of C2 converge within the
SUMMARY
219
desired accuracy. The solution procedures for problem 20 is identical to that for
problem 18 just described, starting with assuming C1 and computing R2 and
NTU2 .
The solution procedure for problem 19 is relatively straightforward. In this case,
P2 and NTU1 are known. For the given ow arrangement, select the P1 -NTU1
formula from Table 3.6 and replace P1 by P2 =R1 [see Eq. (3.98)]. The resulting
equation has only one unknown, R1 , since NTU1 and P2 are known; and R1 is
implicit. It can be computed iteratively using, for example, the NewtonRaphson
method. Similarly, P1 and NTU2 are known in problem 21, and compute R2
iteratively after replacing P1 by P2 R2 in the appropriate P-NTU formula of
Table 3.6. See the footnote of Table 3.6 for how to convert P1 -NTU1 -R1 formulas
into P2 -NTU2 -R2 formulas.
SUMMARY
This chapter is very important to the book. We have presented in considerable detail the
basic thermal design theory for recuperators or exchangers with no moving parts or
periodic ows as in regenerators. Through the problem formulations, it is shown that
there are six independent and one or more dependent variables for the exchanger thermal
design problem for any ow arrangement. The solution to this problem can be presented
in terms of "-NTU, P-NTU, MTD, -P and P1 -P2 methods. The exchanger rating or
sizing problem can be solved by any of these methods and will yield the identical solution
within the numerical error of computation. Theoretical details are presented for the "NTU, P-NTU, and MTD methods in the text for an understanding of concepts, these
methods, and their advantages and disadvantages. Many idealizations are made to
simplify the complex heat exchanger design problem to obtain the solution in terms of
"-NTU, P-NTU, and MTD parameters. A thorough understanding of the concepts and
results presented in this chapter will provide a strong foundation for rating, sizing, and
analysis of industrial heat exchangers.
REFERENCES
Baclic, B. S., 1990, "-Ntu analysis of complicated ow arrangements, in Compact Heat Exchangers:
A Festschrift for A. L. London, R. K. Shah, A. D. Kraus, and D. Metzger, eds., Hemisphere
Publishing, Washington, DC, pp. 3190.
Bell, K. J., 1998, Approximate sizing of shell-and-tube heat exchangers, in Heat Exchanger Design
Handbook, G. F. Hewitt, exec. ed., Begell House, New York, Vol. 3, Sec. 3.1.4.
Kandlikar, S. G., and R. K. Shah, 1989, Asymptotic eectivenessNTU formulas for multipass
plate heat exchangers, ASME J. Heat Transfer, Vol. 111, pp. 314321.
McAdams, W. H., 1954, Heat Transmission, 3rd ed., McGraw-Hill, New York, pp. 129131.
Mueller, A. C., 1967, New Charts for True Mean Temperature Dierence in Heat Exchangers, AIChE
Paper 10, 9th National Heat Transfer Conf., Seattle, WA.
Pignotti, A., and G. O. Cordero, 1983, Mean temperature dierence in multipass crossow, ASME
J. Heat Transfer, Vol. 105, pp. 584591.
Pignotti, A., and R. K. Shah, 1992, Eectiveness-number of transfer units relationships for heat
exchanger complex ow arrangements, Int. J. Heat Mass Transfer, Vol. 35, pp. 12751291.
220
Roetzel, W., and B. Spang, 1990, Verbessertes Diagramm zur Berechnung von Warmeubertragern,
Warme-und Stoubertragung, Vol. 25, pp. 259264.
Roetzel, W., and B. Spang, 1993, Design of Heat Exchangers, in VDI Heat Atlas, VDI-Verlag
GmbH, Dusseldorf, Germany, Sect. Ca, p. 18.
Sekulic, D. P., R. K. Shah, and A. Pignotti, 1999, A review of solution methods for determining
eectivenessNTU relationships for heat exchangers with complex ow arrangements, Appl.
Mech. Rev., Vol. 52, No. 3, pp. 97117.
Shah, R. K., 1983, Heat exchanger basic design methods, in Low Reynolds Number Flow Heat
Exchangers, S. Kakac, R. K. Shah, and A. E. Bergles, eds., Hemisphere Publishing, Washington,
DC, pp. 2172.
Shah, R. K., and A. Pignotti, 1989, Basic Thermal Design of Heat Exchangers, Report Int-8513531,
National Science Foundation, Washington, DC.
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
3.1
3.2
1
1
T dA
(b) Tm
T dq
(a) Tm
A A
q q
1
1
dq
(c)
Tm
q q T
3.3
3.4
3.5
The explicit rate equation used in the heat exchanger analysis is:
(b) "h Ch "c Cc
(a) dq jCh dTh j
(c) dq U T dA
(d) q Ch Th;i Th;o
(e) q "Cmin Th;i Tc;i
3.6
The energy balance equation used in the heat exchanger analysis, from those
equations in Question 3.5, is (check as appropriate)
(a) _______ (b) _______ (c) _______ (d) _______ (e) _______
REVIEW QUESTIONS
3.7
221
Consider a heat exchanger for which C* is near unity. The hot- and cold-uid
temperatures are shown in Fig. RQ3.7 as functions of the position in the
exchanger. In this gure,
(a) sketch the variation of wall temperature with position (neglecting wall resistance) when hAh
hAc .
(b) repeat part (a) when hAc
hAh .
FIGURE RQ3.7
3.8
3.9
The perfect (ideal) heat exchanger, providing the maximum possible heat transfer
rate, has the following things in common with the actual heat exchanger with
constant cp :
(a) ow arrangement
(b) outlet temperatures (c) inlet temperatures
(d) surface area
(e) ow rates
(f) none of these
(g) all of these
3.10
A high number of transfer units NTUs (lets say, 100) is generally obtainable in
an:
(a) physically very large exchanger
(b) exchanger with high ow rates
(c) high eectiveness exchanger
(d) small exchanger
(e) cant tell
3.11
3.12
(d) 81%
222
3.13
3.14
What is the ratio of hot- and cold-side convective thermal resistances in Question
3.13?
(a) 1.33
(b) 2.67
(c) 2
(d) 0.667
3.15
The terminal temperatures of a particular heat exchanger are: hot uid: 1208C,
508C; cold uid: 408C, 808C. The eectiveness of this exchanger is approximately:
(a) 87%
(b) 50%
(c) 75%
(d) 38%
3.16
The heat capacity rate ratio C* for the heat exchanger of Question 3.15 is:
(a) 1.75
(b) 0.38
(c) 0.64
(d) 0.57
3.17
3.18
If you consider the heat exchanger of Question 3.15 as counterow, the temperature distribution in the exchanger will be:
(a) concave up
(b) convex up
(c) linear
3.19
If you consider the heat exchanger of Question 3.15 as a crossow unit with both
uids unmixed, the log-mean temperature dierence Tlm is approximately:
(a) 708C
(b) 558C
(c) 518C
(d) 408C
(e) 228C
3.20
In a hypothetical counterow gas turbine regenerator, having the same heat capacity rate for both uids, the design NTU and eectiveness are 0.5 and 33%,
respectively. If at part-load operation, NTU doubles, the corresponding eectiveness is:
(a) 66%
(b) 50%
(c) 83%
(d) 40%
3.21
Other things being equal (including identical heat transfer surfaces), a multipass
cross-counterow heat exchanger designed for a particular set of specications
(i.e., given ", C*, ow rates, and ow lengths) will have a shorter noow (stack)
height dimension than a single-pass crossow heat exchanger.
(a) It depends on the surface geometries.
(b) true
(c) false
(d) It depends on the capacity rate ratio.
3.22
3.23
REVIEW QUESTIONS
223
3.24
Suppose that we desire to increase the water outlet temperature to 408C in the heat
exchanger of Question 3.23 at the same ow rate. The surface area of this
exchanger required (considering that U constant) should be increased by a
factor of:
(a) 2.40
(b) 1.29
(c) 1.0
(d) 0.67
3.25
Suppose in Question 3.23 that the ow rate of the cooling water is doubled.
Describe qualitatively how the thermal behavior of the exchanger will change.
Idealize no subcooling of steam. Consider all the implications you can think of.
3.26
The curves in Fig. RQ3.26a and b represent the temperature proles in two
dierent counterow exchangers having the same cold uids and the same cold
uid ow rate in each exchanger. The heat transfer rate of exchanger A is as
follows compared to that of exchanger B:
(a) higher
(b) lower
(c) same
(d) cant tell
FIGURE RQ3.26
3.27
224
3.28
3.29
(a) When is it true that Tlm for counterow Tlm for parallelow for a given
duty?
(b) When is it true that Tlm T1 T2 =2 for a single-phase counterow
exchanger? (Mention the R value.)
(c) When is Tlm T1 T2 =2 true for a single-phase parallelow
exchanger?
(d) When is it true that Ta Ta;o Ta;i Tlm for a steam condenser where
the subscript a denotes the air side?
3.30
3.31
Hot (150 to 1008C) and cold (50 to 758C) uids undergo the same temperature
changes (as indicated in parentheses) in a counterow exchanger and a parallelow exchanger. For the identical temperature changes and ow rates (same U),
the heat transfer surface area required for the counterow exchanger compared to
that for the parallelow exchanger is:
(a) 1.00
(b) 1.14
(c) 0.877
(d) 2.00
3.32
Which of the following dimensionless groups can have values ranging from 0 to 1
only?
(a) "
(b) NTU1
(c) F
(d)
(f) NTU
(g) P1
(h) C*
(e) R1
3.33
3.34
3.35
For the same surface areas, uid ow rates, and inlet temperatures, arrange the
following ow arrangements from lowest to highest eectiveness:
(a) unmixedunmixed
(b) mixedmixed
(c) mixedunmixed
(d) one uid partially unmixed, other unmixed
3.36
REVIEW QUESTIONS
225
The temperature approach for a counterow exchanger of innite surface area is:
(a) indeterminate
(b) zero
(c) a very large value
(d) cant tell
3.38
In a single-pass counterow heat exchanger with a xed heat duty, the closer the
temperature approach, the more heat transfer surface area is required.
(a) true
(b) false
(c) cant tell in general
3.39
3.40
3.41
3.42
The exchanger eectiveness " can be presented as the following functional relationship for a given heat exchanger ow arrangement. Subscripts 1 and 2 refer to
uid 1 and 2 sides, respectively.
(a) " P1 ; P2
(b) " NTU1 ; NTU2
(d) " P1 ; NTU2
(c) " R1 ; R2
(e) cant tell
3.43
For a given exchanger and specied operating conditions, one can calculate
F1 P1 ; R1 or F2 P2 ; R2 . Is it correct that F1 F2 in general?
(a) yes
(b) no
(c) cant tell
3.44
In a heat exchanger, engine oil with a 0.2 kg/s ow rate and 1308C inlet temperature is being cooled by water having a 0.438 kg/s ow rate at 908C inlet. The engine
oil and water specic heats are 2.3 and 4.2 kJ/kg K, respectively. What is the
maximum possible exchanger eectiveness if it is a counterow exchanger?
(a) 100%
(b) 25%
(c) 50%
(d) 80%
(e) cant tell
3.45
3.46
226
at 658C. The length of the exchanger is 2 m. We would like to increase the outlet
temperature of the cold water to 808C by increasing the length of the double-pipe
exchanger without changing the inlet temperatures and ow rates of both uids
and also keeping constant the tube diameters of the exchanger. Make appropriate
idealizations and nd the new length of the exchanger as:
(a) 2 m
(b) 3 m
(c) 4 m
(d) 5 m
(e) cant tell
3.47
A 2-ton window air-conditioner needs to remove 8.4 kW of heat from its condenser. In the condenser, the inlet temperature of air is 388C and the refrigerant
condenses at 578C. Ignore the eect of desuperheating and subcooling in the
condenser. UA for this air-conditioner is 700 W/K. Assume the specic heat of
air as 1.0 kJ/kg K. Use any appropriate information from this or any other book,
and determine the airow rate through the condenser as:
(a) 1 kg/s
(b) 0.7 kg/s
(c) 0.5 kg/s
(d) 1.5 kg/s (e) cant tell
Hint: Assume the exponent to e to be an integer.
3.48
3.49
3.50
3.51
3.52
PROBLEMS
227
PROBLEMS
3.1
The typical temperature distributions for hot and cold uids in a counterow
exchanger are shown in the left-hand sketch of Fig. 1.50 for Ch > Cc . Using the
energy balance and rate equations, show why the temperature distributions must
be convex up. Hint: First show that Th Tc increases as x increases.
3.2
Discuss in detail where each of the assumptions listed in Section 3.2.1 is invoked in
the set of Eqs. (3.2), (3.4), (3.5), and (3.6). Alternatively, show how one or more of
these equations could be changed if those assumptions were not invoked.
3.3
The required NTU for a given exchanger is 2.0. If Cc 10; 000 W=K,
Ch 40; 000 W=K, and the thermal resistance Rw of the wall separating hot and
cold uids is 105 K/W, nd ntuh and ntuc when the convective resistances of hot
and cold sides are estimated to be equal. What would the values be if the thermal
resistance of the wall is neglected?
3.4
Explain the physical reasoning why " 1 eNTU for all ow arrangements when
C* 0.
3.5
(a) For a typical counterow heat exchanger, explain briey the change in the
eectiveness as the ow rate of the cold uid is decreased slowly from some
initial value to near zero. Note that we are not interested in transient eects,
just a series of steady-state operating conditions. See Fig. 1.50b for the initial
temperature proles.
(b) Explain briey the eect on the temperature drop of the hot uid, Th .
3.6
In a single-pass crossow exchanger with both uids unmixed, it has been determined that the hot and cold streams should leave the exchanger at the same mean
temperature (i.e. Tc;o Th;o . The following data are given for the exchanger:
Th;i 2508C, Tc;i 308C, m_ h 0:15 kg=s, m_ c 0:60 kg=s, cp;h 2000 J/kg K,
cp;c 1000 J=kg K, and U 1000 W=m2 K. Determine (a) the required heat
exchanger area (m2 ); (b) the total heat transfer rate between the hot and cold
streams (W); (c) the outlet temperatures Tc;o and Th;o (8C); (d) the eectiveness
of the heat exchanger; and (e) which (if any) of the answers to parts (a) through (d)
would be dierent if the heat exchanger were, instead, a single-pass counterow.
Explain. Be sure to address each of parts (a) through (d).
3.7
Determine the heat transfer surface area required for a heat exchanger constructed
from a 25.4 mm OD tube to heat helium from 78C to 08C using hot water at
858C. The water and helium ow rates are 0.6 and 2.4 kg/s. Helium ows within
the tubes. The overall heat transfer coecient is 120 W/m2 K. The specic heats
for water and helium are 4.18 and 5.20 kJ/kg K, respectively. Consider (a) a
counterow exchanger, and (b) a 12 TEMA E exchanger. Solve the problem
by both the "-NTU and MTD methods.
3.8
228
3.9
An automobile radiator has a heat transfer rate q 98:45 kW. Air and water ow
rates are 3.86 and 3.00 kg/s, respectively, and air and water inlet temperatures are
43 and 1158C. The specic heat for air and water are 1.00 and 4.23 kJ/kg K,
respectively. Consider the radiator as a crossow heat exchanger with both uids
unmixed.
(a) Determine the exchanger eectiveness ", the number of transfer units NTU,
and UA.
(b) Determine the true mean temperature dierence Tm for heat transfer.
(c) Determine the F factor. Is the calculated value of F reasonable? If not, why?
(d) Describe by means of a thermal circuit the makeup of (UA1 , and provide
your estimates of the component resistances expressed as a percentage of the
overall resistance UA1 .
3.10
Derive Eq. (3.131) for a four-pass arrangement and demonstrate that it reduces to
P1
8
>
<
>
:
4Pp
1 3Pp
for C * 1
1 eNTU
for C * 0
3.11
Three exchangers identical to the crossow exchanger of Problem 3.6 are placed in
a multipass overall counterow arrangement with the same inlet temperatures and
ow rates. Find the overall eectiveness of this arrangement and the outlet uid
temperatures.
3.12
In the text, the exchanger overall eectiveness " (P1 ) for a multipass overall
counterow exchanger is given by Eq. (3.131) for the case when the NTU per
pass is equal. The objective of this problem is to calculate " for a multipass overall
counterow exchanger when the NTU per pass is not equal. Consider a counterow exchanger with crossow headers as a three-pass exchanger (Fig. P3.12).
(a) Derive an algebraic expression for an overall eectiveness
" "1 ; "2 ; "3 ; C *
Then show that for C* 1 and "1 "3 , it reduces to
"
FIGURE P3.12
PROBLEMS
229
(b) Harrison Model TR regenerator (for a 12,000-hp GE Frame 3000 marine gas
turbine for a Chevron tanker) may be considered as having NTU 4 and
C* 1. Obtain its eectiveness considering it as a true counterow heat
exchanger. Next approximate it as having two crossow (unmixedunmixed)
headers with a counterow core, as shown in part (a) with NTU1 NTU3 1
and NTU2 2. Evaluate the overall eectiveness. How good is this approximation?
3.13
Consider a single-pass crossow exchanger with one uid mixed and NTU 5
and C* 1.
(a) Divide this exchanger into two equal passes (so that NTUp 2:5) and arrange
the two passes into overall counterow arrangement and obtain the exchanger
overall eectiveness. Repeat this procedure by dividing the original exchanger
into three passes and arranging them in an overall counterow arrangement,
and subsequently obtain the exchanger overall eectiveness. Repeat this
procedure by dividing the original exchanger into four and ve passes and
subsequently, obtain the exchanger eectiveness for overall counterow
arrangement. Make a plot of this crossow exchanger eectiveness as a
function of the number of passes. Compare this curve with the eectiveness
of a single-pass counterow exchanger having NTU 5 and C* 1. Discuss
the results.
(b) Repeat the complete part (a) using an arrangement of n passes 2 n 5 in
overall parallelow arrangement. Plot the results on the same graph as that of
part (a) and compare them with the eectiveness of a single-pass parallelow
exchanger having NTU 5 and C* 1. Discuss the results.
3.14
FIGURE P3.14
230
3.15
Given Th;i 608C, Th;o 43:38C, Tc;i 26:78C, and Tc;o 40:68C, calculate the
true-mean temperature dierence for (a) parallelow, (b) counterow, (c) singlepass crossow with both uids unmixed, (d) single-pass crossow with cold uid
unmixed and hot uid mixed, and (e) a 12 TEMA E exchanger. Also determine
the exchanger eectiveness for each case.
3.16
R 1 ln1 RP=1 P
R 1 ln1 1 RP
3.18
A shell-and tube exchanger is required to cool the hot oil from 2408C to 908C.
After some process, the same oil returns to the exchanger at 408C and leaves at
1908C. Assume a negligible change in the specic heat of oil in the temperature
range 110 to 1708C.
(a) If you design one 12 TEMA E shell-and-tube exchanger, would there be a
temperature cross?
(b) Is it possible to design one 12 TEMA E exchanger to meet the required
specications, and why?
(c) What minimum number of 12 TEMA E exchangers in series would you
require to eliminate the temperature cross? Hint: Use an appropriate
equation from Eqs. (3.131)(3.139). Specify any idealization you may
make.
(d) What would be the overall arrangement of two uids in part (c)? Tell whether
or not you could use an alternate (opposite) ow arrangement and why.
(e) What would be the F factor for part (c)?
3.19
Two identical single-pass counterow exchangers are used for heating water
cp 4:2 kJ=kg K) at 258C with the hot oil cp 2:1 kJ=kg K) at 1208C. The
water and oil ow rates are 1 and 4 kg/s, respectively. The heat exchangers are
connected in series on the water side and in parallel on the oil side, as shown in
PROBLEMS
231
FIGURE P3.19
Fig. P3.19. The oil ow rate splits equally between the two exchangers at the
inlet and rejoins at the exit. Use U 420 W=m2 K and Apass 10 m2 for each
exchanger. Determine the outlet temperatures of the water and oil.
The design theory for heat exchangers developed in Chapter 3 is based on the set of
assumptions discussed in Section 3.2.1. That approach allows relatively straightforward
solution of the corresponding design problems. In many applications, such design theory
suces and is used extensively. Still, some applications do require inclusion of additional
eects i.e., relaxation of a number of assumptions. In all these situations, however, the
conventional theory generally fails. So, additional assumptions are necessary to modify
the simplied approach or to devise a completely new design methodology.
In industry, heat exchanger design and analysis calculations are performed almost
exclusively using commercial and/or proprietary computer software. These tools are
equipped with sophisticated routines that can deal with real engineering designs,
although they do not possess the transparency necessary for clearly guiding an engineer
through the design process. To assess the order of magnitude of various inuences, to
analyze preliminary designs in a fast and exible manner, and to involve engineering
judgment in a most creative way, analytical and/or back-of-the-envelope approaches
would be very helpful. These also require, however, insights into the additional inuences
beyond the basic assumptions mentioned above. Thus, it would be necessary to develop
ways of assessing the eects not included in the basic design procedure covered in
Chapter 3. In this chapter we consider the following enhancements to the basic design
procedure: (1) longitudinal wall heat conduction eects in Section 4.1, (2) nonuniform
heat transfer coecients in Section 4.2, and (3) complex ow distributions in shell-andtube heat exchangers in Section 4.4.
Additional considerations for completion of either a simplied design approach or
one based on the relaxed assumptions are still necessary. Among these, the most important is the need to take into account the n eciency of extended heat transfer surfaces
commonly used in compact heat exchangers and some shell-and-tube heat exchangers.
Hence, considerable theory development and discussion is devoted to n eciency in
Section 4.3.
4.1
In a heat exchanger, since heat transfer takes place, temperature gradients exist in both
uids and in the separating wall in the uid ow directions. This results in heat conduc232
233
tion in the wall and in uids from the hotter to colder temperature regions, which may
aect the heat transfer rate from the hot uid to the cold uid.
Heat conduction in a uid in the uid ow direction is negligible for Pe > 10 and
x* 0:005, where the Peclet number Pe Re Pr um Dh = and x* x=Dh Pe; the
signicance and meaning of Pe and x* are presented in Section 7.2. For most heat
exchangers, except liquid metal heat exchangers, Pe and x* are higher than the values
indicated above. Hence, longitudinal heat conduction in the uid is negligible in most
applications and is not covered here.
If a temperature gradient is established in the separating walls between uid ow
streams in a heat exchanger, heat transfer by conduction takes place from the hotter
to colder region of the wall, attens the wall temperature distribution, and reduces the
performance of the exchanger. For example, let us review typical temperature distributions in the hot uid, cold uid, and wall for a counterow heat exchanger as shown in
Fig. 4.1. Dashed lines represent the case of zero longitudinal heat conduction, and solid
lines represent the case of nite longitudinal conduction [ 0:4, dened in Eq. (4.13)].
From the gure it is clear that longitudinal conduction in the wall attens the temperature distributions, reduces the mean outlet temperature of the cold uid, and thus reduces
the exchanger eectiveness " from 90.9% to 73.1%. This in turn produces a penalty in the
exchanger overall heat transfer rate. The reduction in the exchanger eectiveness at a
specied NTU may be quite signicant in a single-pass exchanger having very steep
temperature changes in the ow direction (i.e., large Th =L or Tc =L). Such a situation
arises for a compact exchanger designed for high eectiveness (approximately above
80%) and has a short ow length L. Shell-and-tube exchangers are usually designed
for an exchanger eectiveness of 60% or below per pass. The inuence of heat conduction in the wall in the ow direction is negligible for such eectiveness.
Since the longitudinal wall heat conduction eect is important only for high-eectiveness single-pass compact heat exchangers, and since such exchangers are usually
designed using "-NTU theory, we present the theory for the longitudinal conduction
eect by an extension of "-NTU theory. No such extension is available in the MTD
method.
The magnitude of longitudinal heat conduction in the wall depends on the wall heat
conductance and the wall temperature gradient. The latter in turn depends on the
thermal conductance on each side of the wall. To arrive at additional nondimensional
groups for longitudinal conduction eects, we can work with the dierential energy and
rate equations of the problem and can derive the same Eq. (4.9) mentioned later. For
example, see Section 5.4 for the derivation of appropriate equations for a rotary regenerator. However, to provide a feel to the reader, a more heuristic approach is followed
here. Let us consider the simple case of a linear temperature gradient in the wall as shown
in Fig. 4.1. The longitudinal conduction rate is
qk kw Ak
Tw;1 Tw;2
L
4:1
where Ak is total wall cross-sectional area for longitudinal conduction. The convective
heat transfer from the hot uid to the wall results in its enthalpy rate drop as follows,
which is the same as enthalpy rate change (convection rate) qh :
m_ hh Ch Th;1 Th;2 qh
4:2
234
FIGURE 4.1 Fluid and wall temperature distributions in a counterow exchanger having zero
and nite longitudinal wall heat conduction.
Similarly, the enthalpy rate rise of the cold uid as a result of convection from the wall
to the uid is
m_ hc Cc Tc;1 Tc;2 qc
4:3
qh
LCh Th;1 Th;2
4:4
qk kw Ak Tw;1 Tw;2
qc
LCc Tc;1 Tc;2
4:5
The resulting new dimensionless groups in Eqs. (4.4) and (4.5) are dened as
h
kw A k
LC
c
h
kw Ak
LC c
4:6
where for generality (for an exchanger with an arbitrary ow arrangement), the subscripts h and c are used for all quantities on the right-hand side of the equality sign of s.
235
Th o hA*Tc
1 o hA*
4:7
where
o hA*
o hAc
o hAh
4:8
4:9
It should be added that the same holds for a parallelow exchanger, but as shown in
Section 4.1.3, longitudinal wall conduction eects for a parallelow exchanger are negligible. Note that for a counterow exchanger,
Lh Lc L and
Ak;h Ak;c Ak
4:10
kw Ak
LCc
4:11
kw Ak
LCh
c
and therefore,
h Cc
c Ch
C*
for Cc Cmin
1=C*
for Cc Cmax
4:12
Thus for a counterow exchanger, h and c are not both independent parameters since
C* is already included in Eq. (4.9); only one of them is independent. Instead of choosing
236
kw A k
LCmin
4:13
and Eq. (4.9) takes the following form for a counterow exchanger:
" NTU; C*; ; o hA*
4:14
However, for exchangers other than counterow and parallelow, Eq. (4.9) is the correct
general relationship.
Multiplying by (Th;i Tc;i ) the numerator and denominator of Eq. (4.13), can be
interpreted as the ratio of longitudinal heat conduction rate in a counterow heat
exchanger (having NTU 1 and C* 1) to the thermodynamically maximum possible
heat transfer rate.
Next, we summarize the longitudinal conduction eects for various exchangers.
4.1.1
Exchangers with C* 0
The operating condition having C * 0 usually occurs when condensing or boiling takes
place on one side of a two-uid heat exchanger. The thermal resistance (1/hA) in such a
case on the two-phase (Cmax ) side is generally very small, and hence the wall temperature
is close to the uid temperature of the Cmax side and almost constant (i.e., the wall has a
very small temperature gradient, if any, in the longitudinal direction). As a result, longitudinal heat conduction in the wall is negligible and its eect on the exchanger eectiveness is also negligible for the C * 0 case.
4.1.2
The counterow exchanger problem with nite longitudinal wall conduction has been
analyzed by Kroeger (1967), among others, and extensive graphical results are available
for the exchanger ineectiveness (1 ") for a wide range of NTU, C*, and . Kroeger
showed that the inuence of (o hA)* on " or (1 ") is negligible for 0:1 o hA* 10,
and hence the longitudinal wall conduction eect can be taken into account by only one
additional parameter [see Eq. (4.14)] besides NTU and C*. The penalty to exchanger
eectiveness due to longitudinal wall conduction increases with increasing values of
NTU, C*, and , and is largest for C* 1. Kroegers solution for C* 1 and
0:1 o hA* 10 is
"1
1
1 NTU1 =1 NTU
4:15
NTU
1 NTU
1=2
4:16
for 0
for 1
237
4:17
Note that for ! 1, the counterow exchanger eectiveness " from Eq. (4.17) is
identical to " for a parallelow exchanger (see Table 3.4). This is expected since the
wall temperature distribution will be perfectly uniform for ! 1, and this is the case
for a parallelow exchanger with C* 1.
For NTU ! 1, Eq. (4.15) reduces to
"1
1 2
4:18
The results from Eq. (4.15) are presented in Fig. 4.2 in terms of ineectiveness (1 ")
as a function of NTU and . The concept of ineectiveness is useful particularly for
high values of eectiveness because small changes in large values of eectiveness are
magnied. For example, consider an exchanger to be designed for " 98% and
C* 1. From Fig. 4.2, NTU 49 when 0, and NTU 94 for 0:01. Thus longitudinal heat conduction increases the NTU required by 92%, a signicant penalty in
required surface area due to longitudinal conduction. Alternatively, for NTU 49 and
0:01, from Fig. 4.2, 1 " 100 3% or " 97%. Thus, the result is a 1%
decrease in exchanger eectiveness for 0:01! A direct expression for the reduction
in the exchanger eectiveness due to longitudinal wall heat conduction is
" "0 "60
NTU
"0
"
1 NTU1
4:19
FIGURE 4.2 Counterow heat exchanger ineectiveness 1 " as a function of NTU and for
C* 1. (From Kroeger, 1967.)
238
for NTU large. In Eq. (4.19), the term on the right-hand side of the second equal sign is
obtained from using Eq. (4.15) for 6 0 and Eq. (4.17) for 0, and the last term on
the right-hand side is obtained from Kays and London (1998).
The exchanger ineectiveness for C* < 1 has been obtained and correlated by
Kroeger (1967) as follows:
1 C*
expr1 C*
4:20
1 C*NTU
1 NTU C*
4:21
1"
where
r1
1 *
1 *
1 C* 1
1 C* 1
*
1=2
1
1
1= 2
NTU C*
4:22
4:23
Note that here and are local dimensionless variables as dened in Eq. (4.23). The
values of are shown in Fig. 4.3. The approximate Eq. (4.20), although derived for
o hA*=C * 1, could be used for values of this parameter dierent from unity. For
0:5 < o hA*=C* 2, the error introduced in the ineectiveness is within 0.8% and
4.7% for C* 0:95 and 0.8, respectively.
FIGURE 4.3 Function of Eq. (4.20) for computation of single-pass counterow exchanger
ineectiveness, including longitudinal wall heat conduction eects. (From Kroeger, 1967.)
239
In the case of a parallelow exchanger, the wall temperature distribution is always almost
close to constant regardless of the values of C* and NTU. Since the temperature gradient
in the wall is negligibly small in the uid ow direction, the inuence of longitudinal wall
conduction on exchanger eectiveness is negligible. Hence, there is no need to analyze or
take this eect into consideration for parallelow exchangers.
4.1.4
In this case, the temperature gradients in the wall in the x and y directions of two uid
ows are dierent. This is because Tw;1 and Tw;2 (Tw;1 is the temperature dierence
in the wall occurring between inlet and outlet locations of uid 1; similarly Tw;2 is
dened for uid 2) are in general dierent, as well as Lh 6 Lc in general. Hence, h
and c are independent parameters for a crossow exchanger, and " is a function of
ve independent dimensionless groups, as shown in Eq. (4.9), when longitudinal wall
heat conduction is considered. For the same NTU and C* values, the crossow exchanger eectiveness is lower than the counterow exchanger eectiveness; however, the wall
temperature distribution is two-dimensional and results in higher temperature gradients
for the crossow exchanger compared to counterow. Hence, for identical NTU, C*, and
, the eect of longitudinal conduction on the exchanger eectiveness is higher for the
crossow exchanger than that for the counterow exchanger. Since crossow exchangers
are usually not designed for " 80%, the longitudinal conduction eect is generally small
and negligible compared to a counterow exchanger that is designed for " up to 98 to
99%. Since the problem is more complicated for the crossow exchanger, only the
numerical results obtained by Chiou (published in Shah and Mueller, 1985) are presented
in Table 4.1.
4.1.5
Multipass Exchangers
The inuence of longitudinal conduction in a multipass exchanger is evaluated individually for each pass, depending on the ow arrangement. The results of preceding
sections are used for this purpose. Thus although the overall exchanger eectiveness
may be quite high for an overall counterow multipass unit, the individual pass eectiveness may not be high, and hence the inuence of longitudinal conduction may not be
signicant for many multipass exchangers.
240
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
0.5
1.0
2.0
0.5
NTU
o hAx
o hAy
x
y
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
" for
x 0
0.0027
0.0048
0.0076
0.0095
0.0111
0.0124
0.0245
0.0310
0.0029
0.0055
0.0078
0.0097
0.0113
0.0125
0.0245
0.0310
0.0032
0.0053
0.0080
0.0099
0.0114
0.0127
0.0246
0.0311
0.005
0.0051
0.0092
0.0147
0.0184
0.0214
0.0238
0.0444
0.0542
0.0055
0.0097
0.0151
0.0188
0.0217
0.0240
0.0445
0.0543
0.0062
0.0103
0.0156
0.0192
0.0220
0.0243
0.0446
0.0543
0.010
"=" for x :
0.0074
0.0134
0.0213
0.0268
0.0309
0.0343
0.0613
0.0731
0.0079
0.0141
0.0220
0.0273
0.0314
0.0347
0.0614
0.0731
0.0089
0.0150
0.0227
0.0279
0.0319
0.0351
0.0616
0.0732
0.015
0.0095
0.0173
0.0277
0.0346
0.0398
0.0440
0.0761
0.0891
0.0102
0.0182
0.0284
0.0353
0.0404
0.0446
0.0763
0.0892
0.0115
0.0194
0.0294
0.0360
0.0411
0.0451
0.0764
0.0893
0.020
0.0116
0.0211
0.0336
0.0420
0.0482
0.0532
0.0893
0.1032
0.0123
0.0221
0.0346
0.0428
0.0489
0.0538
0.0895
0.1033
0.0139
0.0236
0.0357
0.0437
0.0497
0.0545
0.0897
0.1034
0.025
0.0135
0.0247
0.0393
0.0490
0.0561
0.0617
0.1013
0.1158
0.0143
0.0258
0.0404
0.0499
0.0569
0.0624
0.1015
0.1159
0.0162
0.0276
0.0418
0.0510
0.0578
0.0633
0.1017
0.1160
0.030
0.0170
0.0313
0.0499
0.0619
0.0706
0.0774
0.1223
0.1377
0.0180
0.0327
0.0512
0.0630
0.0716
0.0782
0.1226
0.1378
0.0204
0.0350
0.0530
0.0644
0.0728
0.0793
0.1229
0.1379
0.040
0.0232
0.0432
0.0685
0.0844
0.0956
0.1041
0.1563
0.1725
0.0244
0.0449
0.0702
0.0857
0.0968
0.1052
0.1566
0.1727
0.0276
0.0481
0.0726
0.0877
0.0984
0.1066
0.1569
0.1729
0.060
0.0285
0.0533
0.0844
0.1033
0.1164
0.1262
0.1831
0.1997
0.0296
0.0553
0.0863
0.1049
0.1178
0.1274
0.1834
0.1999
0.0336
0.0592
0.0892
0.1072
0.1197
0.1290
0.1838
0.2001
0.080
TABLE 4.1 Reduction in Crossow Exchanger Eectiveness (e=e) Due to Longitudinal Wall Heat Conduction for C* 1.
0.0330
0.0621
0.0982
0.1196
0.1342
0.1450
0.2050
0.2219
0.0341
0.0643
0.1002
0.1213
0.1356
0.1462
0.2053
0.2221
0.0387
0.0689
0.1036
0.1239
0.1377
0.1480
0.2057
0.2223
0.100
0.0489
0.0938
0.1468
0.1760
0.1949
0.2083
0.2754
0.2928
0.0490
0.0959
0.1491
0.1779
0.1965
0.2096
0.2758
0.2933
0.0558
0.1026
0.1535
0.1810
0.1988
0.2115
0.2765
0.2942
0.200
0.0652
0.1274
0.1971
0.2328
0.2548
0.2698
0.3401
0.3600
0.0634
0.1286
0.1991
0.2344
0.2560
0.2708
0.3405
0.3619
0.0720
0.1367
0.2036
0.2372
0.2580
0.2724
0.3427
0.3666
0.400
241
1.0
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
1.00
2.00
4.00
6.00
0.00
10.00
50.00
100.00
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
0.5
1.0
2.0
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
0.0020
0.0034
0.0053
0.0066
0.0076
0.0085
0.0170
0.0218
0.0020
0.0034
0.0053
0.0066
0.0076
0.0085
0.0170
0.0218
0.0020
0.0034
0.0053
0.0066
0.0076
0.0085
0.0170
0.0218
0.0039
0.0067
0.0103
0.0129
0.0148
0.0165
0.0316
0.0395
0.0038
0.0066
0.0103
0.0129
0.0149
0.0165
0.0316
0.0395
0.0039
0.0067
0.0103
0.0129
0.0148
0.0165
0.0316
0.0395
0.0057
0.0098
0.0152
0.0189
0.0217
0.0241
0.0445
0.0543
0.0055
0.0097
0.0152
0.0189
0.0217
0.0241
0.0445
0.0543
0.0057
0.0098
0.0152
0.0189
0.0217
0.0241
0.0445
0.0543
0.0074
0.0128
0.0198
0.0246
0.0283
0.0313
0.0562
0.0673
0.0072
0.0127
0.0198
0.0246
0.0283
0.0313
0.0562
0.0673
0.0074
0.0128
0.0198
0.0246
0.0283
0.0313
0.0562
0.0673
0.0090
0.0157
0.0243
0.0302
0.0346
0.0382
0.0667
0.0789
0.0088
0.0156
0.0243
0.0301
0.0346
0.0382
0.0667
0.0789
0.0090
0.0157
0.0243
0.0302
0.0346
0.0382
0.0667
0.0789
0.0106
0.0185
0.0287
0.0355
0.0406
0.0448
0.0765
0.0894
0.0103
0.0183
0.0286
0.0354
0.0406
0.0448
0.0765
0.0894
0.0106
0.0185
0.0287
0.0355
0.0406
0.0448
0.0765
0.0894
0.0136
0.0238
0.0369
0.0456
0.0520
0.0571
0.0940
0.1080
0.0132
0.0236
0.0368
0.0455
0.0519
0.0571
0.0940
0.1080
0.0136
0.0238
0.0369
0.0456
0.0520
0.0571
0.0940
0.1080
0.0190
0.0336
0.0520
0.0637
0.0723
0.0790
0.1233
0.1385
0.0183
0.0331
0.0517
0.0636
0.0722
0.0789
0.1233
0.1385
0.0190
0.0336
0.0520
0.0637
0.0723
0.0790
0.1233
0.1385
0.0237
0.0423
0.0653
0.0797
0.0900
0.0979
0.1474
0.1632
0.0228
0.0417
0.0650
0.0795
0.0898
0.0978
0.1473
0.1632
0.0237
0.0423
0.0653
0.0797
0.0900
0.0979
0.1474
0.1632
0.0280
0.0501
0.0773
0.0940
0.1057
0.1145
0.1677
0.1840
0.0268
0.0493
0.0769
0.0936
0.1054
0.1143
0.1677
0.1840
0.0280
0.0501
0.0773
0.0940
0.1057
0.1145
0.1677
0.1840
0.0609
0.1154
0.1753
0.2070
0.2270
0.2410
0.3092
0.3273
0.0567
0.1125
0.1742
0.2064
0.2266
0.2407
0.3090
0.3270
0.0609
0.1154
0.1753
0.2070
0.2270
0.2410
0.3092
0.3273
(continued)
0.0436
0.0803
0.1229
0.1472
0.1634
0.1751
0.2378
0.2546
0.0412
0.0786
0.1220
0.1466
0.1630
0.1749
0.2377
0.2546
0.0436
0.0803
0.1229
0.1472
0.1634
0.1751
0.2378
0.2546
242
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
1.0
2.0
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
0.4764
0.6147
0.7231
0.7729
0.8031
0.8238
0.9229
0.9476
" for
x 0
0.0016
0.0027
0.0041
0.0051
0.0058
0.0065
0.0130
0.0169
0.0015
0.0026
0.0040
0.0050
0.0057
0.0064
0.0130
0.0169
0.0014
0.0024
0.0039
0.0049
0.0057
0.0063
0.0130
0.0168
0.005
0.0032
0.0053
0.0080
0.0099
0.0114
0.0127
0.0246
0.0311
0.0029
0.0050
0.0078
0.0097
0.0113
0.0125
0.0245
0.0310
0.0027
0.0048
0.0076
0.0095
0.0111
0.0124
0.0245
0.0310
0.010
"=" for x :
0.0047
0.0078
0.0119
0.0146
0.0168
0.0186
0.0350
0.0434
0.0042
0.0074
0.0115
0.0144
0.0166
0.0184
0.0350
0.0434
0.0039
0.0070
0.0112
0.0141
0.0163
0.0182
0.0349
0.0433
0.015
1.00
2.00
4.00
6.00
8.00
10.00
50.00
100.00
0.5
2.0
NTU
o hAx
o hAy
x
y
0.0062
0.0103
0.0156
0.0192
0.0220
0.0243
0.0446
0.0543
0.0055
0.0097
0.0151
0.0188
0.0217
0.0240
0.0445
0.0543
0.0051
0.0092
0.0147
0.0184
0.0214
0.0238
0.0444
0.0542
0.020
0.0076
0.0127
0.0192
0.0236
0.0270
0.0298
0.0534
0.0642
0.0067
0.0119
0.0186
0.0231
0.0266
0.0295
0.0533
0.0641
0.0063
0.0113
0.0181
0.0227
0.0262
0.0291
0.0531
0.0641
0.025
0.0089
0.0150
0.0227
0.0279
0.0319
0.0351
0.0616
0.0732
0.0079
0.0141
0.0220
0.0273
0.0314
0.0347
0.0614
0.0731
0.0074
0.0134
0.0213
0.0268
0.0309
0.0343
0.0613
0.0731
0.030
0.0115
0.0194
0.0294
0.0360
0.0411
0.0451
0.0764
0.0893
0.0102
0.0182
0.0284
0.0353
0.0404
0.0446
0.0763
0.0892
0.0095
0.0173
0.0277
0.0346
0.0398
0.0440
0.0761
0.0891
0.040
0.0162
0.0276
0.0418
0.0510
0.0578
0.0633
0.1017
0.1160
0.0143
0.0258
0.0404
0.0499
0.0569
0.0624
0.1015
0.1159
0.0135
0.0247
0.0393
0.0490
0.0561
0.0617
0.1013
0.1158
0.060
0.0204
0.0350
0.0530
0.0644
0.0728
0.0793
0.1229
0.1379
0.0180
0.0327
0.0512
0.0630
0.0716
0.0782
0.1226
0.1378
0.0170
0.0313
0.0499
0.0619
0.0706
0.0774
0.1223
0.1377
0.080
0.0241
0.0418
0.0632
0.0766
0.0862
0.0936
0.1410
0.1566
0.0213
0.0391
0.0611
0.0749
0.0848
0.0924
0.1407
0.1564
0.0203
0.0375
0.0596
0.0736
0.0837
0.0914
0.1404
0.1563
0.100
0.0387
0.0689
0.1036
0.1239
0.1377
0.1480
0.2057
0.2223
0.0341
0.0643
0.1002
0.1213
0.1356
0.1462
0.2053
0.2221
0.0330
0.0621
0.0982
0.1196
0.1342
0.1448
0.2050
0.2219
0.200
0.0558
0.1026
0.1535
0.1810
0.1988
0.2115
0.2765
0.2942
0.0490
0.0959
0.1491
0.1779
0.1965
0.2096
0.2758
0.2933
0.0489
0.0938
0.1468
0.1760
0.1949
0.2083
0.2754
0.2928
0.400
243
Example 4.1 A gas-to-air crossow waste heat recovery exchanger, having both uids
unmixed, has NTU 6 and C* 1. The inlet uid temperatures on the hot and cold
sides are 3608C and 258C, respectively. Determine the outlet uid temperatures with
and without longitudinal wall heat conduction. Assume that c h 0:04, and
o hAh =o hAc 1.
SOLUTION
Problem Data and Schematic: NTU, ratio of heat capacity rates, inlet temperatures, and
hot- and cold-side s are given for a crossow gas-to-air waste heat recovery exchanger
(Fig. E4.1).
Determine: The outlet temperatures with and without longitudinal wall heat conduction.
Assumptions: Fluid properties are constant and the longitudinal conduction factor is
also constant throughout the exchanger.
Analysis: In the absence of longitudinal conduction (i.e., x 0), we could nd the
eectiveness " 0:7729 from Table 4.1 or Eq. (II.1) of Table 3.6. Using the denition
of ", the outlet temperatures are
Th;o Th;i "Th;i Tc;i 3608C 0:7729360 258C 101:18C
Ans:
Ans:
"
1
"0 1 0:04450:7729 0:7377
"
NTU = 6
C* = 1
c = h = 0.04
360C
(ohA)h/(ohA)c = 1
25C
FIGURE E4.1
244
Ans:
Ans:
Discussion and Comments: For this particular problem, the reduction in exchanger eectiveness is quite serious, 4.6%, with corresponding dierences in the outlet temperatures.
This results in a 4.6% reduction in heat transfer or a 4.6% increase in fuel consumption
to make up for the eect of longitudinal wall heat conduction. This example implies that
for some high-eectiveness crossow exchangers, longitudinal wall heat conduction may
be important and cannot be ignored. So always make a practice of considering the eect
of longitudinal conduction in heat exchangers having " 75% for single-pass units or for
individual passes of a multiple-pass exchanger.
4.2
In "-NTU, P-NTU, and MTD methods of exchanger heat transfer analysis, it is idealized
that the overall heat transfer coecient U is constant and uniform throughout the
exchanger and invariant with time. As discussed for Eq. (3.24), this U is dependent on
the number of thermal resistances in series, and in particular, on heat transfer coecients
on uid 1 and 2 sides. These individual heat transfer coecients may vary with ow
Reynolds number, heat transfer surface geometry, uid thermophysical properties,
entrance length eect due to developing thermal boundary layers, and other factors.
In a viscous liquid exchanger, a tenfold variation in h is possible when the ow pattern
encompasses laminar, transition, and turbulent regions on one side. Thus, if the individual h values vary across the exchanger surface area, it is highly likely that U will not
remain constant and uniform in the exchanger.
Now we focus on the variation in U and how to take into account its eect on
exchanger performance considering local heat transfer coecients on each uid side
varying slightly or signicantly due to two eects: (1) changes in the uid properties or
radiation as a result of a rise in or drop of uid temperatures, and (2) developing thermal
boundary layers (referred to as the length eect). In short, we relax assumption 8
postulated in Section 3.2.1 that the individual and overall heat transfer coecients are
constant.
The rst eect, due to uid property variations (or radiation), consists of two components: (1) distortion of velocity and temperature proles at a given free ow cross
section due to uid property variations (this eect is usually taken into account by the
property ratio method, discussed in Section 7.6); and (2) variations in the uid temperature along axial and transverse directions in the exchanger, depending on the exchanger
ow arrangement; this eect is referred to as the temperature eect. The resulting axial
changes in the overall mean heat transfer coecient can be signicant; the variations in
Ulocal could be nonlinear, depending on the type of uid. While both the temperature and
thermal entry length eects could be signicant in laminar ows, the latter eect is
generally not signicant in turbulent ow except for low-Prandtl-number uids.
It should be mentioned that in general the local heat transfer coecient in a heat
exchanger is also dependent on variables other than the temperature and length eects,
such as ow maldistribution, fouling, and manufacturing imperfections. Similarly, the
245
overall heat transfer coecient is dependent on heat transfer surface geometry, individual Nu (as a function of relevant parameters), thermal properties, fouling eects, temperature variations, temperature dierence variations, and so on. No information is
available on the eect of some of these parameters, and it is beyond the scope of this
book to discuss the eect of other parameters. In this section we concentrate on nonuniformities in U due to temperature and length eects.
To outline how to take temperature and length eects into account, let us introduce
specic denitions of local and mean overall heat transfer coecients. The local overall
heat transfer coecient U x*1 ; x*2 ; T , is dened as follows in an exchanger at a local
position x* x=Dh Re Pr, subscripts 1 and 2 for uids 1 and 2] having surface area
dA and local temperature dierence Th Tc T:
U
dq
dA T
4:24
o hm Ac
Um A o hm Ah
4:25
Here fouling resistances and other resistances are not included for simplifying the discussion but can easily be included if desired in the same way as in Eq. (3.24). In Eq. (4.25),
the hm s are the mean heat transfer coecients obtained from the experimental/empirical
correlations, and hence represent the surface area average values. The experimental/
empirical correlations are generally constant uid property correlations, as explained
in Section 7.5. If the temperature variations and subsequent uid property variations are
not signicant in the exchanger, the reference temperature T in Um T for uid properties
is usually the arithmetic mean of inlet and outlet uid temperatures on each uid side for
determining individual hm s; and in some cases, this reference temperature T is the logmean average temperature on one uid side, as discussed in Section 9.1, or an integralmean average temperature. If the uid property variations are signicant on one or both
uid sides, the foregoing approach is not adequate.
^
A more rigorous approach is the area average U used in the denition of NTU [see the
rst equality in Eq. (3.59)], dened as follows:
^
1
A
Ux; y dA
A
4:26
This denition takes into account exactly both the temperature and length eects for
counterow and parallelow exchangers, regardless of the size of the eects. However,
there may not be possible to have a closed-form expression for Ux; y for integration.
Also, no rigorous proof is available that Eq. (4.26) is exact for other exchanger ow
arrangements.
When both the temperature and length eects are not negligible, Eq. (4.24) needs to be
integrated to obtain an overall U (which takes into account the temperature and length
eects) that can be used in conventional heat exchanger design. The most accurate
approach is to integrate Eq. (4.24) numerically for a given problem. However, if we
can come up with some reasonably accurate value of the overall U after approximately
246
integrating Eq. (4.24), it will allow us to use the conventional heat exchanger design
methods with U replaced by U.
Therefore, when either one or both of the temperature and length eects are not
negligible, we need to integrate Eq. (4.24) approximately as follows. Idealize local
Ux; y; T Um T f x; y and U x*1 ; x*2 ; T Um T f x*1 ; x*2 ; here Um T is a pure
temperature function and f x; y f x*1 ; x*2 is a pure position function. Hence, Eq.
(4.24) reduces to
dq
dA T
4:27
f x*1 ; x*2 dA
4:28
Um T f x*1 ; x*2
and integrate it as follows:
dq
Um T T
An overall heat transfer coecient U~ that takes the temperature eect into account
exactly for a counterow exchanger is given by the rst equality of the following equation, obtained by Roetzel as reported by Shah and Sekulic (1998):
1
1
U~ ln TII ln TI
ln TII
ln TI
dln T
1
UT
ln TII ln TI
ln TII
ln TI
dln T
Um T
4:29
Note that UT Um T in Eq. (4.29) depends only on local temperatures on each uid
side and is evaluated using Eq. (4.25) locally. The approximate equality sign in Eq. (4.29)
indicates that the counterow temperature eect is valid for any other exchanger ow
arrangement considering it as hypothetical counterow, so that TI and TII are
evaluated using Eq. (3.173).
The overall heat transfer coecient Um T on the left-hand side of Eq. (4.28) depends
on the temperature only. Let us write the left-hand side of Eq. (4.28) by denition in the
following form:
dq
1
Um T T U~
dq
T
4:30
Thus, this equation denes U~ which takes into account the temperature eect only. The
integral on the right-hand side of Eq. (4.30) is replaced by the denition of true mean
temperature dierence (MTD) as follows:
dq
q
T Tm
4:31
Integration of the right-hand side of Eq. (4.28) yields the denition of a correction factor
that takes into account the length eect on the overall heat transfer coecient.
1
A
f x*1 ; x*2 dA
4:32
where x*1 and x*2 are the dimensionless axial lengths for uids 1 and 2, as noted earlier.
247
Finally, substituting the results from Eqs. (4.30)(4.32) into Eq. (4.28) and rearranging yields
q U~ A Tm UA Tm
4:33
Thus, an overall heat transfer coecient U which takes into account both the temperature U~ and length eects () is given by
U
1
Um T f x; y dA U~
A
4:34
U U~ U
4:35
of UT or
with U dened by Eq. (4.26) and U~ dened by Eq. (4.29) with Ux; T instead
^
Um T. Hence, for evaluating U for a counterow exchanger, one can use U , which is a
function of only the area (ow length), such as for laminar gas ows, or U~ , which is a
function of the temperature only such as for turbulent liquid ows. These dierent
denitions of overall heat transfer coecients are summarized in Table 4.2.
TABLE 4.2 Denitions of Local and Mean Overall Heat Transfer Coecients
Symbol
Denition
Comments
dq
U
dA T
Um
1
1
1
Rw
Um A o hm Ah
o hm Ac
1
U
A
UA dA
A
U~
U~ ln T2 ln T1
U U~
ln T
ln T1
dln T
UT
1
248
Now we discuss methods that take temperature and length eects into account, to
arrive at U for the exchanger analysis.
4.2.1
Temperature Eect
4:36
where UI and UII are overall heat transfer coecients determined at the exchanger hot
and cold terminals, and TI and TII are given by Eq. (3.173). Note that U Tm
cannot be equal to U~ Tm . In this case, from Eq. (4.36), we get the exchanger heat
transfer rate as q U Tm A. Equation (4.36) represents a good approximation for
very viscous liquids and partial condensation, and further discussion will be provided
later with an example.
An alternative approach to take into account the temperature eect on U is to use the
approximate method of integration by evaluating local U at specic points in the exchanger or perform a numerical analysis. Since such methods are more general, they are
discussed next for all other exchanger ow arrangements.
4.2.1.2 Other Exchangers. We rst illustrate the concept of how to include the eect
of variable UA for a counterow exchanger and then extend it to all other ow arrangements. To nd out whether or not variations in UA are signicant with temperature
variations, rst evaluate UA at the two ends of a counterow exchanger or a hypothetical counterow for all other exchanger ow arrangements. If it is determined that
variations in UA are signicant, such as shown in Fig. 4.4, the average value U~ A can be
determined by approximate integration of the variations in UA [i.e., Eq. (4.29) with the
rst equality sign], by the three-point Simpson method as follows (Roetzel and Spang,
1993):
1
1 1
2 1
1 1
* A 6 UII A
U~ A 6 UI A 3 U1=2
4:37
where
U *1=2 A U1=2 A
T1=2
T *1=2
4:38
and
4:39
249
where the subscripts I and II correspond to terminal points at the end sections, and the
subscript 12 corresponds to a point in between dened by the second equation, respectively. Here Th;1=2 and Tc;1=2 are computed through the procedure of Eqs. (4.43)(4.45).
Usually, uncertainty in the individual heat transfer coecient is high, so that the
three-point approximation may be sucient in most cases. Note that for simplicity, we
have selected the third point as the middle point in the example above. The middle point
is dened in terms of TI and TII [dened by Eq. (3.173)] to take the temperature eect
properly into account; it is not a physical middle point along the length of the exchanger.
The step-by-step procedure that involves this approach is presented in Section 4.2.3.1.
4.2.2
Length Eect
The heat transfer coecient can vary signicantly in the entrance region of laminar ow.
This eect is negligible for turbulent ows. Hence, we associate the length eect to
laminar ow. For hydrodynamically developed and thermally developing ow, the
local and mean heat transfer coecients hx and hm for a circular tube or parallel plates
are related as follows (Shah and London, 1978):
hx 23 hm x*1=3
4:40
where x* x=Dh Re Pr. Using this variation in h on one or both uid sides, counterow and crossow exchangers have been analyzed, and the correction factors are
presented in Fig. 4.5 and Table 4.3 as a function of 1 or 2 , where
1 o;2 hm;2 A2
1
2Rw
o;1 hm;1 A1
2 Rw
1
1
o hm A1 o hm A2
1
4:41
The value of is 0.89 for 1 1, (i.e., when the exchanger has the hot- and cold-side
thermal resistances approximately balanced and Rw 0). Thus, when a variation in the
heat transfer coecient due to the thermal entry length eect is considered, U U~ or Um
since U U~ from Eq. (4.34). This can be explained easily if one considers the thermal
resistances connected in series for the problem. For example, consider a very simplied
problem with the heat transfer coecient on each uid side of a counterow exchanger
250
FIGURE 4.5 Length eect correction factor for one and both laminar streams based on
equations in Table 4.3 (From Roetzel, 1974).
TABLE 4.3 Length Eect Correction Factor When One or Both Streams Are in Laminar Flow for
Various Exchanger Flow Arrangements
One stream laminar
counterow, parallelow,
crossow, 12n TEMA E
Both streams laminar
Counterow
1
Rw
a1
0:65 0:23Rw a1 a2
4:1 a1 =a2 a2 =a1 3Rw a1 a2 2R2w a1 a2
0:44 0:23Rw a1 a2
4:1 a1 =a2 a2 =a1 3Rw a1 a2 2R2w a1 a2
4
8
3
1 2 1 2 22 ln 1
3
9
22
Parallelow
1
1
Crossow
1 a2
4
8
3
1 1 1 1 21 ln 1
3
9
21
2
1
Rw
1=a1 1=a2
a1 o hm A1
a2 o hm A2
4.2.3
251
Combined Eect
A specic step-by-step procedure is presented below to take into account the combined
temperature and length eects on U; the reader may refer to Shah and Sekulic (1998) for
further details. First, we need to determine heat transfer coecients on each uid side by
taking into account uid property variations due to two eects: (1) distortion of velocity
and temperature proles at a given ow cross section due to uid property variations,
and (2) variations in uid temperature along the axial and transverse directions in the
exchanger. In general, most correlations for the heat transfer coecient are derived
experimentally at almost constant uid properties (because generally, small temperature
dierences are maintained during experiments) or are theoretically/numerically obtained
for constant uid properties. When temperature dierences between the uid and wall
(heat transfer surface) are large, the uid properties will vary considerably across a given
cross section (at a local x) and will distort both velocity and temperature proles. In that
case, the dilemma is whether to use the uid bulk temperature, wall temperature, or
something in between for uid properties to determine hs for constant property correlations. Unless a specic heat transfer correlation includes this eect, it is commonly taken
into account by a property ratio method using both uid bulk temperatures and wall
temperature, as discussed in Section 7.6. Hence, it must be emphasized that the local heat
transfer coecients at specic points needed in the Simpson method of integration must
rst be corrected for the local velocity and temperature prole distortions by the property
ratio method and then used as local h values for the integration. The net eect on U~ due
to these two temperature eects can be signicant, and U~ can be considerably higher or
lower than Um at constant properties.
The individual heat transfer coecients in the thermal entrance region could be
generally high. However, in general it will have less impact on the overall heat transfer
coecient. This is because when computing Ulocal by Eq. (4.25), with Um and hm s
replaced by corresponding local values [see also Eq. (3.20) and the subsequent discussion], its impact will be diminished, due to the presence of the other thermal resistances in
the series that are controlling (i.e., having a low hA value). It can also be seen from Fig.
4.5 that the reduction in Um due to the entry length eect is at the most 11% (i.e., the
minimum value of 0:89. Usually, the thermal entry length eect is signicant for
laminar gas ow in a heat exchanger.
4.2.3.1 Step-by-Step Procedure to Determine U. A step-by-step method to determine
U~ A for an exchanger is presented below based on the original work of Roetzel and
Spang (1993), later slightly modied by Shah and Sekulic (1998). In this method, not
only the variations in individual hs due to the temperature eect are taken into
account, but the specic heat cp is considered temperature dependent.
1. Hypothesize the given exchanger as a counterow exchanger (if it is different from
a counterow exchanger), and determine individual heat transfer coefcients and
enthalpies at the inlet and outlet of the exchanger. Subsequently, compute the
overall conductances UI A and UII A at inlet and outlet of the exchanger by using
Eq. (3.24).
2. To consider the temperature-dependent specic heats, compute the specic enthalpies h of the Cmax uid (with a subscript j) at the third point (designated by 1/2 as a
subscript, see Section 4.2.1.2) within the exchanger from the following equation
252
* TII
T 1=2
TI TII
4:42
* is given by
where T 1=2
* TI TII 1=2
T 1=2
4:43
4:44
4:45
3. For a counterow exchanger, the heat transfer coefcient hj;1=2 on each uid side at
the third point is calculated based on the temperatures Tj;1=2 determined in the
preceding step. For other exchangers, compute hj;1=2 at the following corrected
reference (Roetzel and Spang, 1993):
3
1F
Th;1=2;corr Th;1=2 Th;1=2 Tc;1=2
2=3
2
1 Rh
4:46
3
1F
Tc;1=2;corr Tc;1=2 Th;1=2 Tc;1=2
2=3
2
1 Rc
4:47
4:48
253
4. Calculate the apparent overall heat transfer coefcient at this third point using Eq.
(4.38):
T1=2
* A U1=2 A
U1=2
4:49
*
T 1=2
5. Find the mean overall conductance for the exchanger (taking into account the
temperature dependency of the heat transfer coefcient and heat capacities)
from the equation
1
1 1
2 1
1 1
4:50
* A 6 UII A
U~ A 6 UI A 3 U1=2
6. Finally, the true mean heat transfer coefcient U that also takes into account the
laminar ow entry length effect is given by
UA U~ A
4:51
where the entry length eect factor 1 is given in Fig. 4.5 and Table 4.3.
Example 4.2 In a liquid-to-steam two-uid heat exchanger, the controlling thermal
resistance uid side is the liquid side. Lets assume that the temperature of the steam
stays almost constant throughout the exchanger (Tsteam 1088C) while the liquid
changes its temperature from 26.78C to 93.38C. The heat transfer coecient on the
steam side is uniform and constant over the heat transfer surface (12,200 W/m2 K),
while on the liquid side, its magnitude changes linearly between 122 W/m2 K (at the cold
end) and 415 W/m2 K (at the hot end). Determine the heat transfer surface area if the
following additional data are available. The mass ow rate of the liquid is 1.682 kg/s. The
specic heat at constant pressure of the liquid is 1,967.8 J/kg K. The heat exchanger is a
double-pipe design with the inner tube inside diameter 52.6 mm and the inner tube
outside diameter 60.4 mm. The thermal conductivity of the tube wall is 60.58 W/m K.
Assume that no fouling is taking place.
SOLUTION
Problem Data and Schematics: Data for the double-pipe heat exchanger are provided in
Fig. E4.2.
Determine: The heat transfer area of the heat exchanger.
Assumpions: All the assumptions described in Section 3.2.1 hold except for a variable
heat transfer coecient on the liquid side. Also assume that there is no thermal entry
length eect on U. To apply a conventional design method (say, the MTD method;
Section 3.7) a mean value of the overall heat transfer coecient should be dened (see
Section 4.2.3.1).
Analysis: The heat transfer surface area can be calculated from
A
q
U Tlm
254
Steam
hl,i
hl,o
Tsteam
Liquid
in
Tl,i
Liquid
out
Tl,o
di
do
2
2
hl,i = 122 W/m K hl,o = 415 W/m K
2
hsteam = 12200 W/m K
II
Tl,i = 26.7C
Tl,o = 93.3C
Tsteam = 108C
108C
93.3C
T
26.7C
I
II
FIGURE E4.2
where U U~ represents an average overall heat transfer coecient dened by Eq. (4.50).
The heat transfer rate can be calculated from the enthalpy change of the cold uid
(liquid):
q m_ cp Tliquid 1:682 kg=s 1967:8 J=kg K 93:3 26:78C
220; 435 W 220:4 kW
The log-mean temperature dierence by denition [Eq. (3.172)] is equal to
Tlm
TI TII
81:38C 14:78C
38:98C
lnTI =TII ln81:38C=14:78C
where
TI 1088C 26:78C 81:38C
and I and II denote the terminal points of the heat exchanger (cold and hot ends of the liquid
side, respectively). Now let us calculate UI and UII (step 1 of Section 4.2.3.1) from the
information given.
255
1
1
d lndo =di
1
i
UI hsteam do =di
2kw
hliquid;i
12200 W=m2
1
52:6 103 m ln60:4 mm=52:6 mm
2 60:58 W=m K
K60:4 mm=52:6 mm
1
122 W=m2 K
* TII
T 1=2
TI TII
where
* TI TII 1=2 81:38C 14:78C1=2 34:68C
T 1=2
Therefore,
Tliquid;1=2 93:38C 26:78C 93:38C
34:68C 14:78C
73:48C
81:38C 14:78C
and
T1=2 Tsteam Tliquid;1=2 1088C 73:48C 34:68C
It is specied that the liquid-side heat transfer coecient varies linearly from 122 to 415
W/m2 K with the temperature change from 26.78C to 93.38C. Hence from a linear
interpolation, hliquid;1=2 at 73.48C is
hliquid;1=2 122 W=m2 K
Now changing 122 W/m2 K to 327.5 W/m2 K in the last term of the 1=UI equation
above, we get
U1=2 314:0 W=m2 K
256
The apparent overall heat transfer coecient at this third point is given by Eq. (4.49) as
* U1=2
U1=2
T1=2
34:68C
314:0 W=m2 K
314:0 W=m2 K
*
34:68C
T 1=2
Finally, the mean overall heat transfer coecient can be calculated, using Eq. (4.50), as
1 1
1
2
1
1
1
3 314:0 W=m2 K
6 393:6 W=m2 K
U~ 6 120:1 W=m2 K
3:934 103 m2 K=W
Since the thermal entry length is negligible for liquids and zero for condensing steam,
0. Hence,
U~ U 254:2 W=m2 K
The heat transfer surface area of the exchanger is now
A
q
U Tlm
220; 435 W
22:29 m2
254:2 W=m2 K 38:9 K
Discussion and Comments: This simple example illustrates how to determine U and the
heat transfer surface area for a counterow exchanger. For other exchangers, with both
uids as single-phase, the procedure will be more involved, as outlined in Section 4.2.3.1.
Calculation of the mean overall heat transfer coecient by determining the arithmetic
mean value of the local values of terminal overall heat transfer coecients, [i.e.,
U~ 12 UI UII ] will result in this case in a heat transfer surface area about 1% smaller
than the value determined using the elaborate procedure above. It should be noted that
the local heat transfer coecient on the liquid side changes linearly. Often, the changes of
heat transfer coecients are not linear and the values calculated for the mean overall heat
transfer coecient determined using various approaches may be substantially dierent
(Shah and Sekulic, 1998). In those situations, a numerical approach is the most reliable.
4.2.3.2 Numerical Analysis. In the foregoing section, the methodology was presented
to take into consideration variations in U due to the temperature eect, the length
eect, or both. As mentioned earlier, other factors also play a role in making U nonuniform. In addition, a number of other factors that could violate the assumptions (see
Section 3.2.1) are built into the basic "-NTU, P-NTU, or MTD methods, such as
nonuniform velocity and temperature distributions at the exchanger inlet, and highly
variable uid properties. All these eects can be taken into account by a numerical
analysis.
To illustrate the principles, consider an unmixedunmixed single-pass crossow
exchanger. Divide this exchanger into m n segments, as shown in Fig. 4.6, with the
hot-uid passage having m segments, the cold-uid, n segments. The size of individual
segments is chosen suciently small so that all uid properties and other variables/
parameters could be considered constant within each segment. Fluid outlet temperatures
from each segment are indexed as shown in Fig. 4.6. Energy balance and rate equations
for this problem are given in Table 11.2 for an unmixedunmixed case. For the j; k
257
Tc,1,n Tc,2,n
Tc, m,n
k(n segments)
y
(1,n)
Th,m,n
Th,0,n
Tc, j,k
(1,k)
Th,0,k
j(m segments)
Th,m,k
(1,2)
Th,0,2
(1,1)
Th,0,1
( j,1)
(m,1)
Tc, m,0
(2,1)
Th,m,2
Th,m,1
A
Tc,1,0 Tc,2,0
4:52
c; j;k c; j;k1
h; j;k h; j;k1 c; j;k c; j;k1
NTUc; j;k
2
2
*
4:53
c; j;0 0
4:54
where * x=L1 and * y=L2 C*. Equations (4.52) and (4.53) have two unknowns
(the two outlet temperatures), and hence the solution is straightforward at each element if
the inlet temperatures are known for that element. The NTUs required in Eqs. (4.52) and
(4.53) are based on the overall U for individual local segments and are to be evaluated
locally taking into account all the eects for the local element, such as uid property
variations and ow maldistribution. Flow maldistribution translates into dierent local
m_ values for each element in the transverse direction, such as that m_ h could be
dierent for each element at the entrance to the rst element (left most vertical line
AA in Fig. 4.6). Generally, we assume that m_ h remains the same all along the row
specied. If the inlet temperature is not uniform, the boundary conditions of Eq. (4.54)
will need to be changed accordingly. Once the local velocity and temperature distributions are known on each uid side, individual heat transfer coecients can be calculated
using the appropriate correlations: subsequently, UA and local NTUs for individual
elements.
258
For this particular exchanger, the analysis procedure is straightforward since it represents an explicit marching procedure analysis. Knowing the two inlet temperatures for
the element (1,1), two outlet temperatures can be calculated. For the rst calculation, all
uid properties can be calculated at the inlet temperature. If warranted, in the next
iteration, they can be calculated on each uid side at the average temperature of the
preceding iteration. Once the analysis of the element (1, 1) is completed, analyze element
(1, 2) in the same manner since inlet temperatures (Th;0;2 and Tc;1;1 ) for this element are
now known. Continue such analysis for all elements of column 1. At this time, the hotuid temperatures at the inlet of the second column are known as well as the cold-uid
outlet temperature from the rst column. Continue such analysis to the last column, after
which all outlet temperatures are known for both hot and cold uids.
The example we considered was simple and did not involve any major iteration. If the
temperature of one of the uids is unknown while starting the analysis, the numerical
analysis method will become iterative, and perhaps complex, depending on the exchanger
conguration, and one needs to resort to more advanced numerical methods.
Particularly for shell-and-tube exchangers, not only do the baes make the geometry
much more complicated, but so do the leakage and bypass ows (see Section 4.4.1) in the
exchanger. In this case, modeling for evaluating the leakage and bypass ows and their
eects on heat transfer analysis needs to be incorporated into the advanced numerical
methods.
259
Next, we derive an expression for n eciency. The analysis presented next is valid for
both n cooling and n heating situations.
4.3.1
y
x
z
T0
T ,h
T
T0
Ak
qconv
q0
Lf
q0
(a)
qx +dx
qx
x x + dx
(b)
Thin fin
of variable
cross section
260
dT
dx
dAf = Pdx
Ak(x)
(x)
Lf
qx + dx = kf Ak,x
dT + d A dT dx
dx
dx k,x dx
dx
FIGURE 4.8 Energy terms associated with the dierential element of the n.
4:55
The two conduction rate equations associated with conduction through the n and one
convection rate equation for convection to the surroundings for this dierential element
are given by
qx kf Ak;x
dT
dx
4:56
dT
d
dT
Ak;x
dx
qxdx kf Ak;x
dx dx
dx
4:57
4:58
Substituting these rate equations into Eq. (4.55) and simplifying, we get
d
dT
kf
Ak;x
dx hP dxT T1
dx
dx
4:59
T T 1 0
kf Ak;x
dx2 Ak;x dx dx
4:60
or
d 2 T d ln Ak;x dT
m2 T T1 0
dx
dx
dx2
261
4:61
where
m2
hP
kf Ak;x
4:62
where both P and Ak will be functions of x for a variable cross section. Note that m has
units of inverse length. To simplify further, dene a new dependent variable, excess
temperature, as
x Tx T1
4:63
{
We assume that the ambient temperature T1 is a constant, so that d=dx dT=dx and
Eq. (4.61) reduces to
d 2 d ln Ak;x d
m2 0
dx
dx
dx2
4:64
2hLf
2h
kf Lf
kf
4:65
for Lf
. Since Ak;x is constant, dln Ak;x =dx 0, and m2 is constant. Equation (4.64)
reduces to
d2
m2 0
dx2
4:66
4:67
where C1 and C2 (the local nomenclature only), the constants of integration, remain to be
established by the boundary conditions discussed in the following subsection.
{
262
4:68
At the n tip (x ), there are ve possible boundary conditions, as shown in Fig. 4.10.
case 1: long, thin fin. As shown in Fig. 4.10a on page 264, the n is very long
compared to its thickness (=
! 1), so that T T1 at x ! 1. Hence,
1 0
4:69
case 2: thin fin with an adiabatic tip. As shown in Fig. 4.10b on page 264, the n
tip is considered adiabatic and hence the heat transfer rate through the n tip is zero.
Therefore,
q qjx kf Ak
or
dT
d
kf Ak
0
dx x
dx x
d
dx
4:70
0
4:71
case 3: thin fin with convective boundary at the fin tip. As shown in Fig. 4.10c
on page 264, there is a nite heat transfer through the n tip by convection and hence
q kf A k
dT
he Ak T T1
dx x
4:72
Or in terms of ,
d
dx
x
he
kf
4:73
Here we have explicitly specied the n tip convection coecient he as dierent from h
for the n surface. However, in reality, he is not known in most applications and is
considered the same as h (i.e., he h).
case 4: thin fin with finite heat transfer at the fin tip. As shown in Fig. 4.10d
on page 264, the nite heat transfer through the n tip is shown as q since it could be
conduction to the neighboring primary surface (not shown in the gure).
kf A k
or
dT
q
dx x
d
q
kf A k
dx x
4:74
4:75
263
Primary surface
T ,h
Ak
Fin
Lf
q0
T0
P = 2(Lf + )
Ak = Lf
x
case 5: thin fin with fin tip temperature specified. As shown in Fig. 4.10e on
page 264, the n is not very long and the n tip temperature specied as T is constant,
so that
jx T T1
4:76
All these boundary conditions are summarized in Table 4.4. In Fig. 4.10, the temperature distribution in the n near the n tip region is also shown for the rst three
boundary conditions. The common trend in all three temperature distributions is that the
temperature gradient in the n decreases continuously with increasing x when the n
convects heat to the ambient (T > T1 ). This is due to less heat available for conduction
as x increases because of convection heat transfer from the n surface. The reverse will be
true if the n receives heat from the ambient. Later we discuss temperature distributions
for the last two cases of Fig. 4.10 by Eq. (4.104) and Fig. 4.11, respectively.
Total Fin Heat Transfer. The total convective heat transfer from the n can be found
by computing convective heat transfer from the n surface once the specic temperature
distribution within the n is found from Eq. (4.67) after applying boundary conditions
from the preceding section.
The convective heat transfer rate from the dierential element dx, from Fig. 4.8, is
dqconv :
dqconv hP dx T T1 hP dx
4:77
Hence, the total convective heat transfer rate qconv from the constant cross section n,
excluding the heat exchanged from the tip (to be included later), is found by integrating
264
T(x)
T(x)
dT
=0
dx
qx = = kf Ak dT = qe
dx
(b)
(c)
q
(d)
qx = = kf Ak dT = 0
dx
(a)
qe = he Ak(T T)
Insulated
x
/
T(x)
T
T
x
(e)
FIGURE 4.10 Five boundary conditions at x for the thin-n problem: (a) very long n; (b)
adiabatic n tip, (c) convection at the n tip; (d) nite heat leakage at the n tip, (e) temperature
specied at the n tip.
dx
4:78
dT
d
kf Ak
dx x0
dx x0
4:79
4:80
4:81
From here onward, we consider T0 > T1 , so that q0 is coming from the base to the n
(i.e., in the positive x direction) by conduction and ql is leaving n (again in the positive x
direction). For the boundary conditions of Fig. 4.10a and b, q 0. Hence
qconv q0
4:82
265
Long, thin n
Case
cosh m x B sinh m x
0
cosh m B sinh m
d
h
e
kf
dx
cosh m x q m=hP0 sinh mx
0
cosh m
sinh m x =0 sinh mx
0
sinh m
d
q
kf Ak
dx
he
mkf
cosh m x
0
cosh m
d
0
dx
where B
emx
0
Temperature
Distribution in the Fin
0
Boundary
Condition
at x
f
tanh m
m
hP cosh m 0 =
m
sinh m
hP =0 cosh m 1
m 0
sinh m
q
hP cosh m =0
m 0
sinh m
q0
1 q m=hP0 sinh m
0
cosh m
hP
tanh m q
m 0
tanh m B
B m1 B tanh m
q0
f
1
0 cosh m B sinh m
hP sinh m B cosh m
m 0 cosh m B sinh m
1
q he A k 0
cosh m B sinh m
q0
1
0 cosh m
hP
hP cosh2m 1
tanh m
m 0
m 0 sinh2m
q 0
q0
hP
kf Ak m0
m 0
1
q 0 0 f
m
q0 qconv
TABLE 4.4 Boundary Conditions, Temperature Distributions, and Heat Transfer Rates for the Rectangular Prole Thin Fin Problem
266
and the n heat transfer can be obtained either by integrating the temperature distribution along the n length [Eq. (4.78)] or by dierentiating the temperature distribution and
evaluating the derivative at the n base [Eq. (4.79)]!
For the boundary condition of Fig. 4.10c and from Eq. (4.81),
q0 qconv qe
4:83
If qe is the convection heat transfer from the n tip, q0 represents the total convection
heat transfer through the n surface, including the n tip.
For the boundary conditions of Fig. 4.10d and e, q can be positive, zero, or negative,
and one needs to apply Eq. (4.81) to determine qconv .
Now let us derive specic temperature distributions for the ve boundary conditions.
case 1: long thin fin =
! 1. Substituting the boundary conditions of Eqs. (4.68)
and (4.69) into the general solution of Eq. (4.67), we obtain
C1 C2 0
C1 0 C2 1 0
4:84
4:85
The equality of Eq. (4.85) will hold only if C2 0, and hence from Eq. (4.84), C1 0 .
Thus, the specic solution of Eq. (4.67) has
C1 0
C2 0
4:86
and
emx
0
4:87
As noted before, the total n heat transfer rate can be obtained by integrating this
temperature prole as indicated by Eq. (4.78) or by dierentiating it and using Eq. (4.79),
and we get
q0 qconv
hP
kf Ak m0
m 0
4:88
q 0
4:89
case 2: thin fin with an adiabatic tip. Substituting the boundary conditions of Eqs.
(4.68) and (4.71) into the general solution, Eq. (4.67), we obtain
C1 C2 0
mC1 em mC2 em 0
4:90
4:91
267
Solving for C1 and C2 from these equations and substituting them into Eq. (4.67), after
simplication, we get
cosh m x
0
cosh m
4:92
hP
tanh m
m 0
4:93
0 cosh m
q 0
4:94
case 3: thin fin with convective boundary at the fin tip. In this case, the boundary conditions are given by Eqs. (4.68) and (4.73). Substituting them into the general
solution, Eq. (4.67), we obtain
C1 C2 0
mC1 em mC2 em
4:95
he
kf
4:96
Solving for C1 and C2 from these equations and substituting them into Eq. (4.67), after
some algebraic manipulation we get
cosh m x B sinh m x
o
cosh m B sinh m
4:97
and after nding d=dx from Eq. (4.97), we get q0 from Eq. (4.79) and q from Eq. (4.80):
q0
hP sinh m B cosh m
m 0 cosh m B sinh m
q he Ak 0
4:98
1
cosh m B sinh m
4:99
0 cosh m B sinh m
4:100
where
B
*f
he
2
Bi*
Bi*
mkf m
Bi*
he
2kf
*f
4:101
268
Here Bi* is the Biot number at the n tip; it is the ratio of conduction resistance within
the n 1=fkf =
=2g to convection resistance at the n tip 1=he . *f is the n aspect
ratio as dened.
case 4: thin fin with finite heat transfer at the fin tip. Substituting the boundary
conditions of Eqs. (4.68) and (4.75) into the general solution of Eq. (4.67), we get
C1 C2 0
mC1 em mC2 em
4:102
q
kf Ak
4:103
Solving for C1 and C2 from these equations and substituting them into Eq. (4.67), after
some algebraic manipulation, yields
cosh m x q m=hP0 sinh mx
cosh m
0
4:104
hP
tanh m q
m 0
4:105
0
cosh m
4:106
case 5: thin fin with fin tip temperature specified. Substituting the boundary
conditions of Eqs. (4.68) and (4.76) into the general solution, Eq. (4.67), we obtain
C1 C2 0
4:107
C1 em C2 em
4:108
Solving for C1 and C2 from these equations and substituting them into Eq. (4.67), after
some algebraic manipulation, we get
sinh m x =0 sinh mx
0
sinh m
4:109
Subsequently, the heat transfer rates at x 0 and are obtained for Eqs. (4.79) and
(4.80) as
q0
hP
cosh m =0
m 0
sinh m
4:110
hP
1 =0 cosh m hP 0 = cosh m
m 0
sinh m
m
sinh m
4:111
269
e2mX
em =0
=0 em
4:112
where X denotes the value of x where the temperature gradient within the n becomes
zero, and this location is shown in Fig. 4.11 for four possible temperature distributions
(Prasad, 1996). If 0 , then X =2 as expected, and this case is identical to the case
of a thin n with an adiabatic tip discussed previously. From Eq. (4.112), we nd the zero
heat ux location X as
X<
for
1
>
0 cosh m
4:113
for
1
0 cosh m
4:114
X>
for
1
<
0 cosh m
4:115
Equations (4.114) and (4.115) correspond to the temperature distributions of Fig. 4.11d
and b, respectively. The location X of zero heat transfer rate for Fig. 4.11a and c can be
found from Eq. (4.112) and will depend on the value of =0 specied. Note that
=0 < 1 for all four cases of Fig. 4.11.
The solutions for the temperature distributions of the foregoing boundary conditions,
along with the expressions for q0 , q , and are summarized in Table 4.4.
4.3.1.3 Thin Circular Fin of Uniform Thickness. Next, we derive a solution for the
temperature distribution in another important n geometry, the circular n, also
referred to as the disk n, radial n, or circumferential n, shown in Fig. 4.12. An
energy balance for a dierential element dr of the n of uniform thickness presented
in Fig. 4.12 is given by as follows.
q 00 2r
jr q 00 2r
jrdr 2h2r drT T1 0
4:116
where q 00 kf dT=dr represents conduction heat ux at a given location. After simplication, Eq. (4.116) becomes
d
r q 00 hrT T1 0
dr
2
4:117
270
T0
T0
T
T
T
X
T
(a)
(b)
T0
T0
T
T, T
T
X
, X
(c)
(d)
FIGURE 4.11 Temperature distributions in the thin n when the temperatures are specied
at both ends as in Fig. 4.10e: (a) T0 > T > T1 ; (b) T0 > T > T1 ; (c) T0 > T1 > T ;
(d) T0 > T T1 . (From Prasad, 1996).
kf
4:118
d 2 1 d
m2 0
dr2 r dr
4:119
or
4:120
271
where C3 and C4 are constants of integration, and Io and Ko are the modied zero-order
Bessel functions of the rst and second kinds, respectively.
Similar to the uniform thickness straight-n case, the boundary condition for the
circular n at r ro is simply T T0 . There are ve boundary conditions at r re ,
similar to those shown in Fig. 4.10. Since the boundary condition of an adiabatic n
tip at r re (similar to that of Fig. 4.10b) represents a good approximation in many
practical applications, we will obtain a solution of Eq. (4.120) for this case only.
0
d
0
dr
at r ro
4:121
at r re
4:122
4:123
0 C3 I1 mre C4 K1 mre
4:124
where I1 and K1 are the modied rst-order Bessel functions of the rst and second kinds,
respectively. Solving for C3 and C4 from these two algebraic equations and substituting
them into Eq. (4.120) yields
K mre Io mr I1 mre Ko mr
1
0 K1 mre Io mro I1 mre Ko mro
4:125
d
dr rro
4:126
272
After dierentiating Eq. (4.125), evaluate d=drrro and substitute it in Eq. (4.126). The
result is
q0 kf 2ro
mB1 0
4ro h
B1 0
m
4:127
where
B1
4.3.2
4:128
Fin Eciency
For extended surface heat exchangers, we are interested in actual heat transfer by the ns.
Regardless of whether the n is heating or cooling the ambient uid, all heat must pass
through the n base for the rst three boundary conditions of Fig. 4.10 or Table 4.4. The
heat transfer rate q0 through the n base can be calculated in two dierent ways as shown
before, either by integrating or by dierentiating the temperature prole for these three
boundary conditions.
This heat transfer rate q0 can be presented in dimensionless form by the concept of n
eciency f , dened as
f
q0
qmax
4:129
Here our meterbar (yardstick) for comparison is a perfect n having (1) the same
geometry as the actual n; (2) the same operating conditions, T0 , T1 , h, and he ; and
(3) innite thermal conductivity kf of the n material. Under these circumstances, the
perfect n is at the uniform base temperature T0 , as shown in Fig. 4.13 on page 274. The
heat transfer from the n base, qmax (the n is considered to be of uniform cross section),
is
qmax hP T0 T1 he Ak T0 T1 hAf he Ak T0 T1 hAf he Ak 0
4:130
Thus the heat transfer rate through the n is then, from Eq. (4.129),
q0 f hAf he Ak T0 T1
4:131
Rf
8
1
>
>
>
>
< f hAf he Ak
>
>
>
>
:
1
f hAf
if he 6 0
4:132a
if he 0
4:132b
273
However, for a nned surface in an exchanger, we need to take the primary surface into
consideration, and we will use the thermal resistance of an extended surface given by Eq.
(4.164).
We now derive expressions for the n eciency for some n geometries of importance.
4.3.2.1 Thin, Straight Fin of Uniform Rectangular Cross Section. The temperature
distributions through this n for the rst three boundary conditions in Table 4.4 are
presented by Eqs. (4.87), (4.92), and (4.97) or in Table 4.4. The actual heat transfer
rates q0 through the n base are given by Eqs. (4.88), (4.93), and (4.98) for these
boundary conditions. Using these values of q0 and qmax from Eq. (4.130), the f values
are obtained as follows using Eq. (4.129). For completeness, the expression for q0 is also
provided below in terms of f .
For the long, thin n (case 1),
f
1
m
q0 f hAf T0 T1
4:133
tanh m
m
q0 f hAf T0 T1
4:134
For the thin n with a convection boundary at the n tip B he =mkf 2Bi*=m
; case 3),
f
4:135
where B *f Bi*=m from Eq. (4.101) is substituted to get the expression after the
second equality in Eq. (4.135). For this case, q0 is given by Eq. (4.131). If the convected
heat from the n tip is small, f can be approximately computed from Eq. (4.134) with
replaced by
=2. This approximation is referred to as the HarperBrown approximation.
The n eciency of case 4 of Table 4.4 is the same as that in Eq. (4.135) if we assume
that q at the n tip represents convection heat transfer; otherwise, f cannot be dened
since q can be any positive or negative value and q0 cannot be evaluated explicitly [see
Eq. (4.81)]. The n eciency for case 5 may not readily be dened; it is discussed later in a
subsection on p. 278.
The expression for the n eciency becomes increasingly complicated in Eqs. (4.133)
(4.135). It can be shown that
f of Eq: 4:135 < f of Eq: 4:134 < f of Eq: 4:133
4:136
A simple criterion can be set up for the use of Eq. (4.133) for Eq. (4.134) and the use of
Eq. (4.134) for Eq. (4.135) as follows.
Note that if q0 instead of T0 is specied at x 0 for a case alternative to case 2 (and q 0 at x ), it can be
shown that the expression for f is identical to that in Eq. (4.134).
274
T0
T(x) for the actual fin
T
T(x) for the perfect fin
T
0
0
T0
x
q0
0
0
qe
T0
qe
q0
T0
FIGURE 4.13 Temperature distributions for actual and perfect thin ns: (a) n is being cooled;
(b) n is being heated (From Shah, 1983).
0:98
f of Eq: 4:134 1 B1 0:761B 0:761
for B 0:045
4:137
Hence for m 1, B 0:045 will provide an error of less than 2% in f by the use of Eq.
(4.134) instead of Eq. (4.135).
As shown later, the f formula of Eq. (4.134) with an adiabatic tip is applicable to
most two-uid plate-n heat exchangers. This f as a function of m is shown in Fig. 4.14
on page 276 for a straight n designated by re =ro 1. It is found that f increases with
decreasing m, but the variation is nonlinear. Exploitation of this nonlinear behavior by
industry is explained in Example 4.3. Since m2 2h=kf
, m decreasing means or h
decreasing or kf or
increasing. This means that f will increase with decreasing n
length , decreasing heat transfer coecient h (or decreasing Bi), increasing n thermal
conductivity kf , or increasing n thickness
. From an engineering viewpoint, to attain
high n eciency: (1) maintain low n conduction length or plate spacing b (see Fig. 4.15
on page 277); (2) use ns when the heat transfer coecients are low, such as for forced
convection with gases, oils, or a free convection situation; (3) use low height ns when the
heat transfer coecients are high, such as for forced convection with water and other
liquids or phase-change uids; (4) use n material as aluminum or copper instead of
stainless steel and other low-thermal-conductivity materials; and (5) increase the n
thickness to a practical value. However, a decrease in the n thickness is generally
275
pursued by industry due to the nonlinear behavior of f vs. m. The decrease in f will be
much less than the decrease in the n weight and n material cost, as illustrated in
Example 4.3.
Example 4.3 A plate-n exchanger has 24-mm-high 0.12-mm-thick rectangular ns
(Fig. 4.15a on page 277) with a n density of 600 ns/m. The heat transfer coecient
for airow over the ns is 170 W/m2 K. Determine the n eciency if the ns are made
from copper with a thermal conductivity of 401 W/m K. If the n thickness is reduced to
0.06 mm, what happens to the n eciency and associated heat transfer rate? How would
you change the n density to bring the same level of heat transfer performance with the
thinner ns if you idealize that there is no change in heat transfer coecient with the
change in n density? Now change the n material from copper to aluminum with
the thermal conductivity as 237 W/m2 K. Discuss the implication on n eciency and
n heat transfer of changing the material from copper to aluminum. The mass densities
of these materials are 8933 and 2702 kg/m3.
SOLUTION
Problem Data and Schematic: The n is shown in Fig. 4.15a. The following information is
provided.
12 mm (b 24 mm)
h 170 W/m2 K
Determine:
(a) The change in the n efciency and heat transfer rate if the copper n thickness is
changed from 0.12 mm to 0.06 mm.
(b) The change in n density for the same heat transfer rate when reducing the n
thickness from 0.12 to 0.06 mm.
(c) How the n efciency is changed if the n material is changed from copper to
aluminum, keeping the same n geometry. What are other design implications?
Assumptions: The heat transfer coecient does not change with the change in n density
and the assumptions of Section 4.3.1.1 are valid here.
Analysis: Let us compute the n eciency using Eq. (4.134) for original copper ns.
Using Eq. (4.65) or (4.147) yields
!1=2 "
#1=2
2h
2 170 W=m2 K
m
84:0575 m1
kf
0:759
m
1:0087
1:0087
Ans:
276
2h
kf
!1=2
#1=2
2 170 W=m2 K
4
401 W=m K 0:06 103 m
118:875 m1
Therefore,
tanh m tanh1:4265
0:625
m
1:4265
Ans:
Thus, the n eciency is reduced from 0.759 to 0.625, about 18% reduction. This in turn
will reduce n heat transfer by about 18%. For simplicity here we do not include the
eect of the primary surface (which is considered in Section 4.3.4).
Fin heat transfer can be kept approximately constant by increasing the n surface
area by 18% (i.e., increasing the n density from 600 ns/m to 729 ns/m). Here we have
idealized that the heat transfer coecient h does not change with the change in n
density. In reality, h will decrease with an increase in the n density in general.
Now changing the n material from copper to aluminum, m, m, and f for the
0.12 mm thick ns will be
277
FIGURE 4.15 (a) Idealized plain rectangular n (b) plain triangular n (From Shah, 1981).
2h
kf
"
!1=2
#1=2
2 170 W=m2 K
109:339 m1
237 W=m K 0:12 103 m
tanh m tanh1:3121
0:659
m
1:3121
Ans:
Thus, we can see that by changing the material from copper to aluminum reduces f from
0.759 to 0.659. However, if we compare 0.12-mm-thick aluminum n with 0.06-mm
copper n, the n eciency is about the same (0.659 vs. 0.625), and hence from the
heat transfer point of view, these two ns are equivalent. Now let us compare the material
use for these two ns (0.12 mm aluminum ns vs. 0.06 mm copper ns) at the same n
density. Since the mass densities of aluminum and copper are 2702 and 8933 kg/m3,
respectively, the ratio of the n material is
Al fin material
Cu fin material
0:12 mm
0:06 mm
2702 kg=m3
8933 kg=m3
!
0:60
This indicates that aluminum n material use by weight is about 40% less than that for
copper for about the same heat transfer performance despite the double n thickness for
aluminum.
Discussion and Comments: This example clearly indicates that reducing the n thickness
will reduce n heat transfer as expected, but in a nonlinear way, as shown in Fig. 4.14. In
the present case, reducing the n thickness from 0.12 mm to 0.06 mm (50% reduction)
reduces the n eciency and heat transfer by 18%. This reduction can be compensated
by increasing the n density by about 18%. Thus there is a net gain of about 32%
reduction in weight for the same heat transfer. Although the example above is not for
an automobile radiator, this was the direction taken by the automobile industry for the
278
radiator n. The n gauge was reduced over the years from 0.20 mm to 0.046 mm as the
manufacturing technology improved to make ns with thin gauges, along with increased
n density from about 400 ns/m to 800 ns/m. When the n thickness was reduced to
about 0.046 mm, it was so thin that copper could not handle the corrosion common in the
application, and the radiator life was reduced signicantly. Changing the material from
copper to aluminum and increasing the n thickness from 0.046 mm to 0.075 mm, the
durability was brought back to the desired radiator life with thicker aluminum ns, and
at the same time the total material mass was reduced, as was the cost of material. In
summary, the material mass for ns can be minimized by reducing the n thickness as
permitted by the design and manufacturing considerations and increasing the n density
to maintain the desired n heat transfer. This is primarily because ns are used for
increasing the surface area for convection heat transfer and thickness comes into picture
for conduction heat transfer through the n, which can be maintained smallest by
increasing the n density somewhat.
To repeat, the important variable for convection heat transfer through the n surface
is the n surface area, and the important variable for conduction heat transfer from the
base along the n length is the n thickness. The n thickness is chosen the thinnest and is
of secondary importance for industrial heat exchangers due to a limited range of permissible n sizes. That is the reason that ns have one-third to one-fourth primary surface
thickness to cost eectively utilize the material. Thus the modern n designs have the
largest possible n density and smallest possible n thickness permitted by design to
make the most use of the n material for ns in a heat exchanger. This trend will continue
with the improvements in material and manufacturing technologies.
Fin Heat Transfer for a Specied Fin Tip Temperature. For this case, the boundary
conditions of case 5 shown in Fig. 4.10e, q0 and q are given by Eqs. (4.110) and (4.111).
Substituting them in Eq. (4.81) with 0 = 1, we get the total n transfer rate as
(Prasad, 1996)
hP
cosh m 1 hP
m
tanh
m 0
sinh m
m 0
2
tanhm=2
hAf ;1=2 0 f ;1=2
h Af ;1=2 0
m=2
qconv q0 q
4:138
hP
m hP
cosh m 1
tanh
m 0
2
m 0 sinh m
4:139
As shown clearly shown in Fig. 4.11, the reason is that the adiabatic plane will not be at
x =2, but can be found from Eq. (4.112) for the specied temperatures and 0 . Thus
we nd that q0;1=2 of Eq. (4.139) is dierent from the actual q0 given by Eq. (4.110). Thus,
279
the concept of n eciency has real limitations for this n problem and similarly for
the n problem of case 4, Fig. 4.10d. Further implications are discussed by Prasad
(1996).
Dimensionless Groups. A careful review of Eqs. (4.133)(4.135) reveals that f is a
function of m and B or Bi*. Employing the denitions of m and B yields
f Ak ; P; ; kf ; h; he
4:140
4:141
f
8
>
Bi 1=2 Bi* coshf*Bi 1=2 Bi 1=2 sinhf*Bi 1=2
>
>
>
>
1=2
1=2
1=2
*
*
>
>
< f Bi Bi Bi* sinhf*Bi Bi coshf*Bi
>
>
Bi 1=2
>
>
* 1=2
>
> * Bi tanhf Bi
>
: f
for Bi* 0
4:142a
4:142b
where
Bi
Bi
1 Bi=4
f*
4:143
A comparison of Eqs. (4.134) and (4.142b) reveals that the thin n approximation of Eq.
(4.134) introduces a maximum error of 0.3% for Bi 0:01 and f* < 100; otherwise, use
Eq. (4.142b) for a thick n.
280
b
b
2
2
or
b 2
b
2
2
4:145
where the last approximate expression in each formula is valid for b (or )
. If is not
very large compared to
, then f > 0:95, and either of the approximations will not
signicantly aect the value of f . The n eciency of this plain rectangular n is then
tanh m
m
!#1=2
"
#1=2 "
h 2Lf 2
2h
1
kf
Lf
kf L f
f
hP
kf Ak
!1=2
4:146
2h
kf
!1=2
4:147
Here review Fig. 4.9 for substituted values of P and Ak in Eq. (4.147). The last approximate term on the right-hand side is valid when
Lf .
The plain triangular n geometry of constant cross section is shown in Fig. 4.15b.
Here the corners are shown rounded with a braze or solder llet. The size of this llet will
depend on the radius of the corner and the manufacturing process. The n conduction
length for the f evaluation is half of the distance 2 shown in this gure. The n
eciency is computed from Eq. (4.146) with the value of the n parameter m obtained
by Eq. (4.147).
281
Now let us summarize the f evaluation for other plate-n surfaces of Fig. 1.29. The
oset strip n is similar to the plain rectangular n except that the n length Lf is not
continuous. The length Lf is the sum of many strip lengths s . The n eciency is
determined by the use of Eq. (4.146) with and m calculated from Eqs. (4.145) and
(4.147), respectively, except that Lf in Eq. (4.147) is replaced by s .
In a louver n, the louver cuts are ideally parallel to the heat conduction path direction from the base to the center of the n, and the louvers do not extend to the base of the
n in order to maintain the required n structural strength. Hence, ideal heat transfer
through individual louvers is identical to the case of plain triangular ns of Fig. 4.15b,
and the applicable n eciency formula is Eq. (4.146). The exposed edge area of the
louver is part of the n surface area Af and should be included in the determination Af
for consistency. However, the common industrial practice is to ignore the louver edge
area (see Example 8.2) in the Af determination and also in the calculation of m from Eq.
(4.147). Thus for the f evaluation from Eq. (4.146), the louver n is treated as the plain
triangular n with m 2h=kf
1=2 .
The wavy and corrugated ns are treated as either rectangular or triangular ns,
depending on the cross-section shape for the f evaluation. Ambiguity exists for the
determination of Af and for perforated ns for the determination of f . The common
practice is to ignore the perforations and determine f as if the n were unperforated. As
long as the heat transfer coecient is determined experimentally using this assumption
and the n eciency is calculated subsequently the same way for its use in the design of a
heat exchanger, the error introduced may not be signicant. For pin ns, f is evaluated
from Eq. (4.146) with
b
b
d0
2
2
"
hd0
m
kf d02 =4
#1=2
4h
kf d0
!1=2
4:148
282
thermal conductivity of n material as 190 W/m K for aluminum and heat transfer
coecient is 120 W/m2 K. Fin base temperature is 2008C and uid temperature 308C.
SOLUTION
Problem Data and Schematic: Fin geometry and material properties and heat transfer
coecients are provided for a single-pass crossow plate-n exchanger (Fig. E4.4) on
page 284. Fluids are unmixed on both sides. The n base temperature as well as the
ambient temperature are also provided.
Determine: The n eciency and heat transfer rate per unit area for:
(a) A n in the center of the core (n 2 in Fig. E4.4)
(b) A n at one end of the core (n 1 in Fig. E4.4) for which heat source is only on one
side
Assumptions: The assumptions of Section 4.3.1.1 are invoked here.
Analysis: (a) For a n at the center of the core, the heat source is on both sides. Hence, the
adiabatic plane in the n will be at the center of the n, as shown in Fig. 4.15a. For this
case, 6:5 mm. Now from Eq. (4.147),
1=2
2h
1
kf
Lf
"
#1=2
2 120 W=m2 K
0:15 mm
1
91:79 m1
300 mm
190 103 W=m K0:15 103 m
and hence
m 91:79 m1 6:5 103 m 0:5966
f
tanh m tanh0:5966
0:896
m
0:5966
Ans:
The heat transfer rate is just the product of the n eciency and the theoretical maximum
heat transfer rate:
00
f hT0 T1 0:896 120 W=m2 K 200 308C
q000 f qmax
18,278 W=m2
Ans:
(b) When the heat source is only on one side of the n, the heat conduction length for
the n will be at least 2 (see Fig. 4.15a), and there will be heat transfer from that end to
the braze llet area and the plate. Let us designate the appropriate m and for this case
with the subscript 1:
m1 m 91:79 m1
1 2 13 mm 0:13 m
283
TABLE 4.5 Fin Eciency for Plate-Fin and Tube-Fin Geometries of Uniform Fin Thickness
Geometry
tanhmi i
1 i
mi
Ei
i 1; 2; 3
Lf
kf
i
m i i
f E1
1
f
b
1
2
sinhm1 1
hA1 T0 T1
qe
m 1 1
T T1
coshm1 1 hA1 T0 T1 qe 0
T 1 T1
f E1
1 =2
f
E 1 1 E 2 2
1
1 2 1 m21 E1 E2 1 2
1 b
s =2 2 3 pf =2
1
2
3
s
tanhm
m
4h 1=2
m
kf do
f
f
b
do
2
8
< ame b
: tanh
a r*0:246
do
2
me r*n
n exp0:13me 1:3863
0:9107 0:0893r*
for r* 2
b
0:9706 0:17125 n r* for r* > 2
2h 1=2
e f
r* de o
m
kf
m
f
tanhme
me
1=2
2h
1
kf
e f
de do
2
284
Fin 1
L3
Fluid 1
kf = 190 W/m K
h = he = 120 W/m2 K
Fin 2
T0 = 200C
L1
T = 30C
L2
FIGURE E4.4
he
120 W=m2 K
0:0069
tanh m1 1 B
0:8316 0:0069
0:695
B m1 1 1 B tanh m1 1 0:0069 1:1933 1 0:0069 0:8316
Ans:
tanh m1 1 0:8316
0:697
m1 1
1:1933
Ans:
which is within 0.3% of f calculated by Eq. (4.135). The heat transfer rate could be
calculated by Eq. (4.131), per unit of n surface area Af as
A
00
q000 f qmax
f h he k T0 T1
Af
Since Ak =Af
=2 1, we neglect the last term. Therefore, the heat transfer rate is
q000 0:697 120 W=m2 K 200 308C 14,219 W=m2
Discussion and Comments: Two points of interest from this example may be observed:
(1) decreasing the n conduction length from 13 mm to 6.5 mm increased the n eciency
from 0.697 to 0.896 but not by a factor of 2; and (2) the heat transfer rate per unit n area
increased from 14,219 to 18,278 W/m2. This is because the addition of the n surface
area from the conduction length of 6.5 mm to 13 mm (for only one heat source) is not
285
quite as eective as that from 0 to 6.5 mm (having a heat source on both sides in Fig.
4.15a).
We can consider this problem in a dierent way: The results obtained in this example
are applicable to ns with a heat source on both sides and having n lengths 2 13 mm
and 26 mm. In this comparison, the heat transfer rate ratio of the two ns is
00
q0 2 1 6:5 mm
q6:5 mm
00
q13 mm
q0 2 1 13 mm
18,278 W=m2
14,219 W=m2
!
6:5 mm
0:643
13 mm
Thus reducing the n height by 50% (i.e., reducing the n material mass by 50%) reduces
the n heat transfer rate by 35.7%. Even increasing the n density (ns per meter or inch)
by 35.7% and assuming no change in the heat transfer coecient, there will be 14.3%
material savings. However, more important is the fact that the shorter n has much
higher column structural strength and that increasing the n density will increase the
column strength. Hence, a shorter n is more desirable from both the heat transfer and
mechanical strength points of view. In reality, reduction in the heat transfer coecient
due to shorter and denser ns should be taken into account as well as a potential fouling
problem before making a nal selection on the n height.
Inuence of Violation of Basic Assumptions for the Fin Analysis. The basic assumptions
made for heat transfer analysis of ns are presented in Section 4.3.1.1. The inuence of
a violation of assumptions 5 to 8 is found as follows for the thin straight n of a
rectangular prole (Huang and Shah, 1992).
. A 10% linear variation in n thermal conductivity will increase or decrease f by up
to 1.7%, depending on whether the n is being cooled or heated, respectively, when
f > 80%. For a composite n, the low thermal conductivity layer plays a dominant
role in the heat transfer mechanism; a reduction of heat ow by the insulating layer
due to frost can be signicant, about 8% at a typical f 96%.
. The assumption of uniform heat transfer coecient may lead to gross errors in
predicting n heat transfer. It was found that for a particular linear variation of h,
f was lower by 6% and 16% for f 90% and 80%. In reality, since h for a given
heat transfer surface is determined experimentally based on f for constant h, the
assumption of constant h for f for the design of the exchanger would not introduce
a signicant error, particularly for high f , such as f > 80%. However, one needs to
be aware of the impact of nonuniform h on f if the heat exchanger test conditions
and design conditions are signicantly dierent.
. Nonuniform ambient temperature T1 has less than a 1% eect on the n eciency
for f > 60%, and hence this eect can be neglected.
. Longitudinal heat conduction in the n in the Lf direction in Fig. 4.9 aects f less
than 1% for f > 10%, and hence this eect can be neglected.
. The thermal resistance between the n and the base can have a signicant impact on
n heat transfer, and hence care must be exercised to minimize or eliminate it.
It should be emphasized that for most plate-n heat exchangers, the n eciency is
maintained at 80% and higher, and more typically at 90% and higher. One way of
thinking of a n having f 80% means that 80% of that n material is as good as
286
the primary surface from the convection heat transfer point of view; thus, eectively 20%
of the material cost is wasted. A typical automotive radiator has a n eciency of over
95% at a 80- to 90-km/h automobile speed. If f is maintained high, as pointed out
above, most of the assumptions made for the n heat transfer analysis are quite adequate.
4.3.2.3
Thin Circular Fin of Uniform Thickness. The temperature distribution for this n (see
Fig. 4.12) is given by Eq. (4.125) and the heat ow q0 through the base as Eq. (4.127).
The n surface area and qmax for this n are given by
Af 2 r2e r2o
qmax 2 r2e r2o hT0 T1 2 r2e r2o ho
4:149
4:150
where the n tip surface area is neglected but can be added if desired. Finally, using the
denition of Eq. (4.129), the n eciency is given by
f
q0
2ro B1
4:151
where B1 is given by Eq. (4.128). Since B1 involves evaluating six Bessel functions,
approximations have been proposed using simpler expressions for hand calculations.
Such a formula is given in Table 4.5, which is accurate within 1% for most useful ranges
of mro re and re =ro .
If the radial n tip is not adiabatic and heat transfer at the n tip is small, it can be
taken into account by the HarperBrown approximation by replacing re with re
=2
in Eq. (4.151).
Although it is not obvious from Eq. (4.151), a review of the formula in Table 4.5
shows that for the thin circular n,
f mre ro ; r*
4:152
for an adiabatic n tip. Hence, f for the thin circular n is shown in Fig. 4.14 as a
function of mro re and r* re =ro . Note that for re =ro 1, the circular n becomes a
straight n. Also note that f for the circular n is less than f for a straight n.
The n eciency for most of the n geometries shown in Fig. 1.32 can be evaluated by
Eq. (4.151) except for studded, serrated, and wire form ns. Those ns having slots and
cuts are treated as if they were plain circular ns. Studded and serrated ns are treated as
straight ns, and the f formula is given in Table 4.5 as a last item. A wire form n is
treated as a pin n with the f given by Eq. (4.146) with de do =2 and m given by
Eq. (4.148).
Flat Fins on an Array of Circular Tubes. The n eciency for at plain ns on inline
and staggered tube arrangements may be obtained approximately by an equivalent
annulus method or by a more accurate sector method. Based on the arrangement of
the tubes, rst the idealized adiabatic planes are located. Such planes for inline and
staggered tube arrangements are shown in Fig. 4.16a and b by dashed lines, resulting in
a rectangle or a hexagon around the tube. In an equivalent annulus method, the
287
rectangular or hexagonal n around the tube (see Fig. 4.16a and b) is represented
hypothetically as a circular n (see Fig. 4.12) having the same n surface area. The
n eciency of this n is then computed using Eq. (4.151). In the sector method, the
smallest representative segment of the n of Fig. 4.16a is shown in Fig. 4.16c, which is
divided into two parts, OAB and OBC. The part OAB (having the subtended angle 0 )
is then divided into m equal-angle ( 0 =m) segments. Similarly, the part OBC
(having the subtended angle 0 ) is then divided into n equal-angle ( 0 =n) segments. The outer radius of each circular sector is determined by equating the area of the
sector with the area of the equivalent annular sector (Kundu and Das, 2000). Thus for
the inline tube arrangement of Fig. 4.16a and c, it is given by
re;i
Xt tani tani 1 1=2
2
re; j
X tan j tan j 1 1=2
2
4:153
The smallest representative segment of the staggered tube arrangement of Fig. 4.16b is
shown in Fig. 4.16d, which is divided into two parts, OAD and ODF; and ODF is divided
into two equal parts, ODE and OEF, as shown in Fig. 4.16e. The part OAD (having the
subtended angle 0 ) is then divided into m equal angle ( 0 =m) segments. Similarly,
each part (ODE and OEF, each having the subtended angle 0 ) is then divided into n
equal angle ( 0 =n) segments. Here again, re;i and re; j of ith and jth segments of
OAD and ODE are given by
Flow
Flow
2r0 = d0
Xt
Xt
X
X
(a)
(b)
A
re,i
ith sector
A
re,i
0
0
re, j
B
0
re,j
jth
sector
d0
0
0
C
(c)
E
re, j
d0
F
(d)
FIGURE 4.16 Flat n over (a) an inline, and (b) a staggered tube arrangement. The smallest
representative shaded segment of the n for (c) an inline, and (d) a staggered tube arrangement.
288
re;i
tani tani 1 1=2
X
re; j
X2 Xt =22
1=2
tan j tan j 1 1=2
4:154
The n eciency of each sector is then determined by the circular n of constant cross
section [Eq. (4.151)]. Once f for each sector is determined, f for the entire n is the
surface-area weighted average of f s for each sector.
f
Pm
Pn
f ; j Af ; j
i1 f ;i Af ;i a
Pm
Pj1
n
i1 Af ;i a
j1 Af ; j
4:155
Here a 1 for inline arrangement (Fig. 4.16c) for segment OBC and n 2 for staggered
arrangement (Fig. 4.16d) for two equal segments, ODE and OEF. This approximation
improves as the number of sectors m, n ! 1. However, in reality, only a few sectors m
and n will suce to provide f within the desired accuracy (such as 0.1%). An implicit
assumption made in this method is that the heat ow is only in the radial direction and
not in the path of least thermal resistance. Hence, f calculated by the sector method will
be lower than that for the actual at n, a conservative value. However, the equivalent
annulus method yields an f that may be considerably higher than that by the sector
method, particularly when the n geometry around the circular tube departs more and
more from a square n; as a result, the heat transfer computed will be too high.
4.3.3
Fin Eectiveness
q0
hAk;0 0
4:156
This concept is useful for the rst three boundary conditions of Fig. 4.10, and particularly for the most common boundary condition of an adiabatic n tip. In the denition of
" , it is idealized that the heat transfer coecient at the n base when the n is removed is
identical to the heat transfer coecient at the n surface. Using the q0 expression of Eq.
(4.134) for the thin straight n with an adiabatic n tip, " of Eq. (4.156) is related to f as
"
2 Lf
Af
2
f
f
Lf
Ak;0
4:157
289
qp qf
qtotal
qmax h Ap Af T0 T1
4:158
In this equation the denition of qmax is similar to that dened by Eq. (4.130) for an
adiabatic n tip with an added heat transfer component for the primary surface. We have
also redened Af and qf q0 as total values for all nned surfaces; we also redene f
using the values of Af and qf for all nned surfaces instead of a single n. Also, Ap in Eq.
(4.158) represents the total primary surface area associated with the particular uid side
of concern in the exchanger. We have idealized heat transfer coecients for the primary
and n surface areas as the same in Eq. (4.158). Now the total heat transfer rate qtotal is
given by
qtotal qp qf hAp T0 T1 hAf f T0 T1 hAp f Af T0 T1
4:159
Substituting Eq. (4.159) into Eq. (4.158), and simplifying yields
o 1
{
Af
1 f
A
4:160
For the case of nite heat transfer at the n tip with he as the actual or equivalent heat transfer coecient, replace
Af with Af h=he Ak in Eqs. (4.158), (4.160), (4.162), and (4.163), and Af with Af Ak in Eq. (4.158).
290
FIGURE 4.17 Extended surface heat exchanger: (a) plate-n, (b) tube-n (From Incropera and
DeWitt, 2002).
where A Ap Af . Note that we considered f 100% for the primary surface in Eq.
(4.159). Hence, one can readily see that{
o f
4:161
qtotal o hAT0 T1
4:162
and
In the derivation of Eq. (4.159), we have neglected the thermal contact resistance that
could be present in the exchanger when the ns are attached by the mechanical t. The
thermal circuit for heat transfer in such an extended surface is shown in Fig. 4.18 on page
292. Here hcont is the contact conductance and Ak;0 is the cross-sectional area of all ns at
the n base. The total thermal resistance for the extended surface from this circuit is given
by
1
1
1
hAp
hAp
Rt
1=hcont Ak;0 1=f hAf
1=hcont Ak;0 Rf
4:163
1
1
hAp f hAf o hA
4:164
Thus, Rt represents the combined thermal resistance of the n and primary surface area.
If we want to include the thermal contact resistance, o of Eq. (4.164) is not given by
{
291
4:165
where
C1 1
f hAf
hcont Ak;0
4:166
The thermal resistance term of Eq. (4.164) is used for thermal resistances (Rh and Rc )
on the hot- and cold-uid sides in Eq. (3.23). It must thus be emphasized that the
extended surface eciency o appears in the overall resistance equation, such as Eq.
(3.24) and not f . In research articles, is sometimes used for o and is designated as
n eciency, although it is really the extended surface eciency.
In some industries, the eective surface area Aeff is used instead of the extended
surface eciency, dened as
Aeff Ap f Af o A
4:167
q hAeff To T1
4:168
so that
292
r0
q0
q0
T0
T0
qp
qp
(hAp)1
qtotal
T0
(o, cont hA)1
T,h
FIGURE 4.18 Thermal circuit for an extended surface with nite thermal contact resistance
(From Incropera and DeWitt, 2002).
clearances are required for the construction of the exchanger and the shell uid leaks or
bypasses through these clearances with or without owing past the tubes (heat transfer
surface). Three clearances associated with a plate bae are tube-to-bae hole clearance, bundle-to-shell clearance, and bae-to-shell clearance. In a multipass unit, the
tube layout partitions may create open lanes for bypass of the crossow stream.
The total shell-side ow distributes itself into a number of distinct partial streams due
to varying ow resistances, as shown in Fig. 4.19. This ow model was originally proposed by Tinker (1951) and later modied by Palen and Taborek (1969) for a segmental
bae exchanger.
Various streams in order of decreasing inuence on thermal eectiveness are as
follows:{
. B stream: crossow stream owing over the tubes (and ns, if any) between successive windows. This stream is the desired stream and is considered
fully eective for both heat transfer and pressure drop.
. A stream: tube-to-bae hole leakage stream through the annular clearance
between the tubes and bae holes of a bae. This stream is created
by the pressure dierence on the two sides of the bae. As heat transfer
coecients are very high in the annular spaces, this stream is considered
fully eective.
. C stream: bundle-to-shell bypass stream through the annular spaces (clearances)
between the tube bundle and shell. This bypass ow area exists because
the tube holes cannot be punched close to the tubesheet edge, due to the
structural strength requirement. The C stream ows between successive
bae windows. This stream is only partially eective for heat transfer,
as it contacts only those tubes near the circumference.
. E stream: shell-to-bae leakage stream through the clearance between the edge of
a bae and the shell. This stream is least eective for heat transfer,
particularly in laminar ow, because it may not come in contact with
any tube.
{
Note that there is no D stream since D is used for the shell diameter.
293
4.4.1.2 Flow Fractions for Each Shell-Side Stream. Each of the streams has a certain
ow fraction Fi of the total ow such that the total pressure drop is the same for each
stream from the entrance to the exit of the exchanger. Each stream undergoes dierent
acceleration/deceleration and frictional processes and inuences heat transfer in dierent ways.
The design of the plate-baed shell-and-tube exchanger should be such that most of
the ow (ideally, about 80%) represents the crossow B stream. However, this is rarely
achieved in practice. The narrow bae spacing results in a higher p for the B stream
and forces more ow into the A, C, and E streams. If the computed values of the B stream
are lower than those indicated, the bae geometry and various clearances should be
checked.
Since the A stream is eective from a heat transfer point of view, it is not of great
concern if its ow fraction is large, as in the case of a narrow bae spacing. If the tube-tobae hole clearance is plugged due to fouling, the shell-side pressure drop generally
increases. The ow fraction of the A stream generally decreases for increasing values
of multiple-segmental baes.
Since C and F streams are only partially eective, the design of the tube bundle should
be such that it minimizes the ow fraction for each of these streams to below 10%.
Sealing devices are used for this purpose.
The E stream does not contact the heat transfer area and is ineective from the heat
transfer viewpoint. It mixes only poorly with other streams. Since the bae-to-shell
clearances are dictated by TEMA standards, if the ow fraction computed for the E
stream is found to be excessive (15% or more), the designer should consider multiplesegmental baes instead of a single-segmental bae. This is because the total shell-side
pressure drop is lower for the multiple-segmental bae case, thus forcing more ow to
the B, A, and C streams.
294
Based on extensive test data, Palen and Taborek (1969) arrived at the ow fractions of
Table 4.6, on page 296, for various streams. It is surprising to note from this table that the
B stream may represent only 10% of the total ow for some exchangers. Even for a good
design, it represents only 65% of the total ow in turbulent ow. Hence the performance
predicted based on the conventional MTD method will not be accurate in general. As a
result, there is no need to compute very accurate values of the MTD correction factor F
for various exchanger congurations.
4.4.1.3 The BellDelaware Method
In this method the ow fraction for each stream on the shell side is found by knowing the
corresponding ow areas and ow resistances. The heat transfer coecient for ideal
crossow is then modied for the presence of each stream by correction factors. These
correction factors are based on experimental data obtained on units of 200 mm diameter
TEMA E shell and segmental baes by Bell (1988) in the 1950s at the University of
Delaware. The shell-side heat transfer coecient hs is given by
hs hid Jc J Jb Js Jr
4:169
where hid is the heat transfer coecient for the pure crossow stream (B stream) evaluated at a Reynolds number at or near the centerline of the shell. Jc is the correction
factor for bae conguration (bae cut and spacing) and takes into account the heat
transfer in the window. J is the correction factor for bae leakage eects and takes into
account both the shell-to-bae (E stream) and tube-to-bae hole (A stream) leakages. Jb
is the correction factor for bundle and pass partition bypass (C and F) streams and is
dependent on the ow bypass area and number of sealing strips. Js is the correction factor
for bae spacing that is larger at the inlet and outlet sections than in the center. Jr is the
correction factor for the adverse temperature gradient in laminar ows (at low Reynolds
numbers). These correction factors are discussed further with Eq. (9.50) and their eects
are discussed in Section 9.5.1.1. These correction factors in equation form are presented
in Table 9.2.
4.4.1.4 The Stream Analysis Method. The conventional MTD method cannot be
applied to a shell-and-tube exchanger to determine the actual heat transfer rate. This
is because of the existence of crossow, leakage, and bypass streams on the shell side;
the leakage and bypass streams are idealized as zero in the MTD analysis. Each stream
ows through dierent paths, has dierent heat transfer rates, and has dierent temperature proles, depending on the mixing.
As an illustration, consider the temperature distributions of the tube-side uid and of
various uid streams on the shell side, shown in Fig. 4.20. The shell-side exit temperature
Ts;o represents a mixed mean temperature of these streams, depending on the heat transfer eectiveness and heat capacity rate of each stream. Whereas Ts;o is used to calculate
the log-mean temperature dierence, TB;o (the exit temperature of the crossow B
stream) denes the crossow driving potential for heat transfer.
Since the apparent temperature prole is dierent from an idealized temperature
prole (B stream), this eect must be considered in the determination of the true mean
temperature dierence. To obtain Tm from Tlm , Palen and Taborek (1969) proposed
a correction factor
s , the delta factor, in addition to the F factor, as follows:
Tm
s F Tlm
4:170
295
TABLE 4.6 Flow Fractions (%) for Various Shell-Side Flow Streams
Flow Stream
Crossow stream B
Tube-to-bae leakage stream A
Bundle-to-shell bypass stream C
Bae-to-shell leakage stream E
Turbulent Flow
Laminar Flow
3065
923
1533
621
1050
010
3080
648
so that
q UA Tm UA
s F Tlm
4:171
s FE ;
Ts;o Ts;i
where FE is the bae-to-shell leakage stream ow fraction, Re is the Reynolds number
for crossow, and the subscripts t and s for temperatures are for tube and shell uids,
respectively. They found that FE is the only important ow fraction aecting
s in most
heat exchangers; all other streams mix with the crossow stream in the exchanger. The
bae-to-shell leakage stream also becomes well mixed at high Reynolds numbers, and
s
approaches unity as Re becomes large. The qualitative behavior of
s at low Reynolds
numbers is shown in Fig. 4.21;
s decreases with decreasing Tt;i Ts;o =Ts;o Ts;i and
increasing FE . Palen and Taborek found
s varying from 0.4 to 1.0 for their units. Thus,
s can be a large correction compared to F, and hence the minimum possible bae-toshell clearances are recommended.
In the stream analysis method, rst ow fractions for each stream are calculated
considering the total pressure drop for each stream the same. Next, the heat transfer
296
eectiveness is assigned to each stream and appropriate correction factors are developed.
Finally, an equation of the type Eq. (4.171) is used for an overall rate equation for rating
or sizing problems.
4.4.2
4:173
where
m1 ; m 2
NTUt
f
Rt 2 12 41 1=2 Rt 2 1g
2
NTUpf
NTUt
Rt
Ct
Cs
4:174
4:175
Here NTUpf represents the NTU on the tube side of the parallelow pass and
NTUt NTUpf NTUcf is the total NTU of the exchanger on the tube side.
Equation (4.173) represents an excellent approximation for a 1n (n even) exchanger
for NTUt 2, with being not close to zero. If is close to zero, the appropriate
formulas are given by Roetzel and Spang (1989). Refer to Roetzel and Spang (1989)
for formulas for unequal passes for 13 and 1n (n even) exchangers. The following are
the general observations that may be made from the results above.
. As expected, F factors are higher for K > 1:0 than for the K 1 (balanced pass)
case for given P and R, where K UAcf =UApf 1 = and the subscripts cf
and pf denote counterow and parallelow passes, respectively.
. As K increases, P increases for specied F (or NTU) and R.
. The F factors for the 12 exchanger are higher than those for the 14 exchanger for
specied values of P, R, and K.
. As the number of passes is increased, the F factors (or P) continue to approach that
of a crossow exchanger with both uids mixed, and the advantage of unbalanced
passes over balanced passes becomes negligible.
. Although not evaluated specically, the unbalanced UA (i.e., K > 1) exchanger will
have a higher total tube-side pressure drop and lower tube-side h than those for the
balanced UA (i.e., K 1) exchanger.
297
Since the analysis was based on the value of K Ucf Acf =Upf Apf , it means that not
only the inuence of unequal tube pass area can be taken into account, but also the
unequal tube-side overall heat transfer coecients. Similarly, it should be emphasized
that, if properly interpreted, the results for nonuniform UA presented in Section 4.2 can
also apply to unequal surface areas in dierent passes. As noted above, higher exchanger
performance can be achieved with higher values of K Ucf =Upf for equal pass areas.
Hence, the shell inlet nozzle should be located at the stationary head when heating the
tube uid and at the oating head when cooling the tube uid. This is because higher
temperatures mean higher heat transfer coecients. It should be emphasized that Ucf
and Upf represent mean values of U across the counterow and parallelow tube passes
and not at the inlet and outlet ends.
4.4.3
Assumption 12 in Section 3.2.1 indicates that the number of baes used is very large and
can be assumed to approach innity. Under this assumption, the temperature change
within each bae compartment is very small in comparison with the total temperature
change of the shell uid through the heat exchanger. Thus the shell uid can be considered as uniform (perfectly mixed) at every cross section (in a direction normal to the
shell axis). It is with this model that the mean temperature dierence correction factor for
the exchanger eectiveness is normally derived for single-phase exchangers. In reality, a
nite number of baes are used and the condition stated above can be achieved only
partially. Shah and Pignotti (1997) have made a comprehensive review and obtained new
results as appropriate; they arrived at the following specic number of baes beyond
which the inuence of the nite number of baes on the exchanger eectiveness is not
signicantly larger than 2%.
.
.
.
.
Nb
Nb
Nb
Nb
298
SUMMARY
Many assumptions are made to simplify the complex heat exchanger design problem
to obtain the solution in terms of "-NTU, P-NTU, and MTD parameters. Sucient
information is provided in this chapter for relaxing these assumptions. These include
approaches to design that include longitudinal wall heat conduction, variable local
overall heat transfer coecient, and specic eects (the inuence of ow bypassing
and leakages, unequal pass areas, and a nite number of baes) in a shell-and-tube
heat exchanger. For an extended heat transfer surface exchanger, careful determination
of the n eciencies of the extended surfaces must be accomplished. Sucient details
are provided in this chapter for the most important extended surfaces used in heat
exchangers by industry. A thorough understanding of concepts and results presented
in this chapter and Chapter 3 will provide a strong foundation for rating, sizing, and
analysis of industrial heat exchangers.
REFERENCES
Bell, K. J., 1988, Delaware method for shell-side design, in Heat Transfer Equipment Design, R. K.
Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere Publishing, Washington, DC,
pp. 145166.
Colburn, A. P., 1933, Mean temperature dierence and heat transfer coecient in liquid heat
exchangers, Ind. Eng. Chem., Vol. 25, pp. 873877.
Huang, L. J., and R. K. Shah, 1992, Assessment of calculation methods for eciency of straight ns
of rectangular prole, Int. J. Heat Fluid Flow, Vol. 13, pp. 282293.
Incropera, F. P., and D. P. DeWitt, 1996, Fundamentals of Heat and Mass Transfer, 4th ed., Wiley,
New York.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger Publishing, Malabar, FL.
Kraus, A. D., A. Aziz, and J. R. Welty, 2001, Extended Surface Heat Transfer, Wiley, New York.
Kroeger, P. G., 1967, Performance deterioration in high eectiveness heat exchangers due to axial
heat conduction eects, Adv. Cryogen. Eng., Vol. 12, pp. 363372.
Kundu, B., and P. K. Das, 2000, Performance of symmetric polygonal ns with and without tip
lossa comparison of dierent methods of prediction, The Canadian J. Chem. Eng., Vol. 78,
pp. 395401.
Palen, J. W., and J. Taborek, 1969, Solution of shell side ow, pressure drop and heat transfer by
stream analysis method, Chem. Eng. Prog. Symp. Ser. 92, Vol. 65, pp. 5363.
Prasad, B. S. V., 1996, Fin eciency and mechanisms of heat exchanger through ns in multi-stream
plate-n heat exchangers: formulation, Int. J. Heat Mass Transfer, Vol. 39, pp. 419428.
Roetzel, W., 1974, Heat exchanger design with variable transfer coecients for crossow and mixed
arrangements, Int. J. Heat Mass Transfer, Vol. 17, 10371049.
Roetzel, W., and B. Spang, 1989, Thermal calculation of multipass shell and tube heat exchangers,
Chem. Eng., Res. Des., Vol. 67, pp. 115120.
REVIEW QUESTIONS
299
Roetzel, W., and B. Spang, 1993, Design of heat exchangers, Sect. Cb, Heat transfer, VDI Heat
Atlas, VDI-Verlag, Dusseldorf, Germany.
Shah, R. K., and A. L. London, 1978, Laminar Flow Forced Convection in Ducts, Supplement 1 to
Advances in Heat Transfer, Academic Press, New York.
Shah, R. K., 1981, Compact heat exchanger design procedures, in Heat Exchangers: ThermalHydraulic Fundamentals and Design, S. Kakac. A. E. Bergles and F. Mayinger, eds., Hemisphere
Publishing, Washington, DC, pp. 495536.
Shah, R. K., 1983, Heat exchanger basic design methods, in Low Reynolds Number Flow Heat
Exchangers, S. Kakac, R. K. Shah and A. E. Bergles, eds., Hemisphere Publishing, Washington,
DC., pp. 2172.
Shah, R. K., and A. C. Mueller, 1985, Heat exchanger basic thermal design methods, in Handbook of
Heat Transfer Applications, 2nd ed., W. M. Rohsenow, J. P. Hartnett, and E. N. Ganic, eds.,
McGraw-Hill, New York, pp. 41 to 477.
Shah, R. K., and A. Pignotti, 1997, The inuence of a nite number of baes on the shell-and-tube
heat exchanger performance, Heat Transfer Eng., Vol. 18, No. 1, pp. 8294.
Shah, R. K., and D. P. Sekulic, 1998, Nonuniform heat transfer coecients in conventional heat
exchanger design theory, ASME J. Heat Transfer, Vol. 119, pp. 520525.
Tinker, T., 1951, Shell side characteristics of shell and tube heat exchangers, General Discussion on
Heat Transfer, Proc. Institution of Mechanical Engineers, London, UK, pp. 89116.
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
4.1
1
1
1
1
dq
T dA (b) Tm
T dq (c)
(a) Tm
A A
q q
Tm
q q T
4.2
h ho
1
1
1
1
(a) U i
U dA
(c)
(b) U
dA
2
A A
U A AU
4.3
4.4
A loss in counterow exchanger eectiveness (and not the absolute value of ") due
to longitudinal heat conduction increases with:
(a) decreasing value of
(b) increasing value of C*
(c) increasing value of NTU
4.5
300
4.6
m2 T T1 0
dx
dx
dx2
It is valid for thin ns of the following cross sections:
(a) triangular straight ns
(c) rectangular constant
cross-sectional ns
4.7
4.8
4.9
4.10
For a plate-n heat exchanger, the total surface eciency of the nned surface side
is given by tanhm=m.
(a) true
(b) false
(c) It depends on the n geometry.
(d) It depends on the boundary conditions.
4.11
4.12
REVIEW QUESTIONS
4.13
(e) T
(f) T
(g) T
F
F
301
It is proposed to add pin ns to the outside surfaces of tubes in a tube bank over
which air ows. Water ows inside the tubes fast enough to be in turbulent ow, so
that hwater hair .
(a) Sketch the variation of n eciency f as a function of n length for a
constant value of m (Fig. RQ4.13a).
(b) For xed water and air temperatures, sketch the total heat transfer rate as a
function of the n length (Fig. RQ4.13b).
FIGURE RQ4.13
4.14
Plate-n exchanger A has plain ns whereas exchanger B has oset strip ns. For
equal total surface area, frontal area, free-ow area, and ow rates, which has a
higher n eciency?
(a) A
(b) B
4.15
The thermal conductivity of plastics is about 1000 times lower than that for the
aluminum. If the n eciency is 95% for an aluminum n of 6 mm height, the
height required for a similar (same thickness and cross section) plastic n for 95%
n eciency would be:
(a) 6 mm
(b) 0.006 mm
(c) 0.2 mm
(d) 0.5 mm
(e) cant tell
Hint: Do not calculate the n eciency. Assume the heat transfer coecient as
constant and same.
4.16
4.17
302
4.18
4.19
4.20
Arrange uids for case (a) and materials for case (b) below having from the highest
to lowest magnitude of the n eciency for identical corrugated plain n geometry
in a given heat exchanger.
(a) air, water, viscous oil
(b) stainless steel, copper, aluminum
PROBLEMS
4.1
4.2
PROBLEMS
303
(c) Based on the results of the foregoing two values of U (lumped mean value and
integrated mean value), determine the percentage increase or decrease in the
integrated mean value of U compared to the lumped mean value of U.
(d) Now discuss the results. Will the overall heat transfer coecient increase,
decrease, or remain the same when you take into account the inuence of
thermal entry length eect on U? Explain the physical reason on whatever
you found out.
4.3
Heat is transferred from water to air through a 0:305 m 0:305 m brass wall
(thermal conductivity 77.9 W/m K). The addition of straight ns of rectangular
cross section is contemplated. These ns are 0.76 mm thick, 25.4 mm long and
spaced 12.7 mm apart. Consider the water- and air-side heat transfer coecients
as 170 and 17.0 W/m2 K, respectively.
(a) Determine f and o if the ns are added only on (i) the air side and (ii) the
water side.
(b) Calculate the gain in heat transfer rate achieved by adding ns to (i) the air
side, (ii) the water side, and (iii) both sides. If you try to make the thermal
resistances more balanced, on which side would you add ns? Why? Consider
the temperature drop through the wall negligible and the n tip as adiabatic for
f calculations.
4.4
A solid copper rod is used to cool a hot spot inside a box of airborne electronic
equipment (Fig. P4.4). The shell of the box is double walled and has Freon boiling
at 308C serving as a sink for the thermal energy generated inside the box. The
ambient temperature inside the box is 908C. Because of good thermal contact at
the box wall, the rod temperature is 408C at the rod-to-wall junction. The ultimate
purpose of the analysis is to establish q0 and q , heat ows in and out of the rod.
Do the following to this end:
(a) Picture how you expect Tx to vary relative to the other temperatures noted
in the sketch and text of this problem.
(b) Derive from basic considerations, using a thermal circuit, an appropriate
dierential equation for Tx and formulate the boundary conditions
explicitly.
30C
boiling
freon
Box
90C
x=
x
Hot
spot
q
q0
200C
A(x) and
perimeter
and constant
with x
40C
FIGURE P4.4
304
(c) Write down a general solution for the dierential equation and indicate how
you would reduce it for this particular problem.
(d) Given the solution of your dierential equation and boundary conditions for
Tx, how would you evaluate q0 and q ?
4.5
q0 coshm x q cosh mx
kf Ak sinh m
Also locate the plane, a value of x, in terms of q0 , q , m, and where the temperature gradient dT=dx 0. Show that x =2 for q0 q .
q0
x=
x=0
O2
N2
Air
4.6
q
FIGURE P4.5
T2
T1 > T2 > T
x=
x=0
FIGURE P4.6
4.7
E1 1 E2 2
1
1 2 1 m21 1 2 E1 E2
PROBLEMS
where
tanh mi i
Ei
mi i
mi
2h
kf
i
305
1=2
i 1; 2
Perform the analysis from the solutions presented in the text for thin n with (a) an
adiabatic tip, and (b) nite heat transfer at the n tip without solving any dierential equations. Make an appropriate energy balance at x 1 . Mention
explicitly any additional assumptions that you may make.
FIGURE P4.7
4.8
A heat exchanger design is generally considered good if o hA on hot and cold uid
sides are about the same. Because of very low values of heat transfer coecients
with gas ows compared to those for liquid ows, a considerable amount of surface area is needed on the gas side. It can be achieved by increasing either the n
density or the n height, or both. High n height may be structurally very weak.
An alternative way is to put two layers of ns in between liquid tubes as shown in
the Fig. P4.8a. Figure P4.8b represents a general unit n section. Show that the n
eciency for this n is
f
2E1 1 E2 2 E3 3
1
21 2 3
1 m21 =2E1 1 E2 2 E3 3
where
Ei
tanh mi i
mi i
mi
2h
kf
i
1=2
i 1; 2; 3
Perform the analysis using the solutions presented in the text for thin ns with
(a) an adiabatic n tip, and (b) nite heat transfer at the n tip without solving any
FIGURE P4.8
306
4.10
1
T1
T dx
0 1
Using the results of part (a), show that
m2 2
T0 T1 T T1 1
3
T0
T
T
T
x
T
T
T0
x
dT = 0
dx
FIGURE P4.10
PROBLEMS
307
1
1 m2 2 =3
tanh m
m
In a round tube and at n exchanger, the air on the n side is heated with water in
the tubes. The use of turbulators is contemplated on the tube side to augment the
exchanger performance. The following experimental data have been obtained at
various air and water ows with and without turbulators.
o hAwater (W/K)
Test Airow Rate
Point
(m3/s)
1
2
3
4
0.118
0.024
0.118
0.024
Water Flow
Rate (L/s)
6:3 105
6:3 105
4:8 104
4:8 104
Without
o hAair
(W/K) Turbulators
395.6
182.0
395.6
182.0
110.8
110.8
264.8
264.8
With
Turbulators
395.6
395.6
817.6
1345.1
(a) For the same Tm , determine an increase in the heat transfer rate due to the
turbulators for each of the test points. Consider the wall thermal resistance as
2:27 105 K/W.
(b) Discuss at what airows and water ows the turbulators provide a signicant
increase in performance, and why.
(c) For what design points would you recommend the use of turbulators?
In this chapter our objective is to present comprehensive thermal design theory for rotary
and xed-matrix regenerators. Denitions, types, operation, and applications of regenerators are described in Sections 1.1 and 1.5.4. Here in Section 5.1, basic heat transfer
analysis is presented, including a list of assumptions made in the analysis and development of governing equations. Two methods have been used for the regenerator thermal
performance analysis: "-NTUo and - methods, respectively, for rotary and xedmatrix regenerators. These are discussed in Sections 5.2 and 5.3. The inuence of
longitudinal wall heat conduction is important in high-eectiveness regenerators and is
discussed in Section 5.4. The inuence of transverse conduction may be important in lowthermal-conductivity materials and in thick-walled regenerators and is presented in
Section 5.5. The inuence of pressure and carryover leakages is important in regenerators, particularly those operating at high eectivenesses. A detailed procedure to take
these eects into account is presented in Section 5.6. Finally, the inuence of matrix
material, size, and arrangement is summarized in Section 5.7.
5.1
The following assumptions are built into the derivation of the governing equations
presented in Section 5.1.3 for rotary and xed-matrix regenerators.
1. The regenerator operates under quasi-steady-state or regular periodic-ow
conditions (i.e., having constant mass ow rates and inlet temperatures of both
uids during respective ow periods).
2. Heat loss to or heat gain from the surroundings is negligible (i.e., the regenerator
outside walls are adiabatic).
3. There are no thermal energy sources or sinks within the regenerator walls or
uids.
308
309
310
uid stream. The pressure leakage will depend on the pressure dierence between the hot
and cold gases. Flow bypassing on each uid side in the gap between the rotor and the
housing will depend on its pressure drop in the matrix. Fluid carryover leakage will occur
of the uids trapped in ow passages during the switch from hot to cold uid, and vice
versa, during matrix rotation or valve switching. This eect will depend on the matrix
rotational speed and the void volume of the matrix and headers. Similarly, for a xedmatrix regenerator, the pressure leakage will occur across the valves and cracks in the
housing, and the carryover leakage will occur when switching the uids. The eects of
various pressure and carryover leakages are discussed in Section 5.6.
The inuence of matrix material (specic heat cw and packing density ) and of the
size and arrangement of the layers in a multilayer regenerator are discussed in Section
5.7. The governing equations based on the aforementioned assumptions are developed in
Section 5.1.3 for the regenerator heat transfer analysis.
The matrix and uid temperatures depend on x and coordinates in a xed-matrix
regenerator. In a rotary regenerator, the uid temperatures are functions of the axial
coordinate x and the angular coordinate for a stationary observer looking at the
regenerator. Based on the foregoing assumptions, for an observer riding on the matrix,
the uid temperatures in a rotary regenerator are also functions of x and . Thus, we
consider the uid temperatures Th and Tc as functions of x and for both types of
regenerators.
5.1.2
In this section we dene and describe the heat capacity rates, heat capacitances, heat
transfer areas, porosity, and volumetric heat capacity for rotary and xed-matrix regenerators. Let us rst dene heat capacitance terms Ch ; Cc ; Cr;h ; Cr;c , and Cr for the uids
and the matrix, and their relationship to the heat capacity rates (Ch and Cc ) and the heat
transfer areas Ah and Ac before setting up the energy balance and rate equations. Ch is the
hot-uid heat capacity rate and Ch is the hot-uid heat capacitance within the regenerator. The same set of entities can be dened for cold uid as well, i.e., Cc and Cc . Their
denitions and relationships are as follows:
Cj m_ j cp; j
CL
um j
j h or c
5:1
Here the subscript j h for the hot uid and j c for the cold uid; m_ j is the mass ow
rate; Mj is the mass of the j-uid contained in the regenerator matrix at any instant of
time; cp; j , um; j , and d; j are the specic heat, mean uid axial velocity, and uid dwell
time (or residence time), respectively; and Lj is the regenerator matrix length. The regenerator matrix wall heat capacitance Cr and the matrix wall heat capacity rate Cr are
dened and related as follows:
Cr Mw cw
8
>
< Mw cw N Cr N
Cr M c
C
>
: w w r
Pt
Pt
rotary regenerator
5:2
fixed-matrix regenerator
Here Mw is the mass of all matrices (disks), cw is the specic heat of the matrix material, N
is the rotational speed for a rotary regenerator, and Pt is the total period for a xed-
311
matrix regenerator. This Pt is the interval of time between the start of two successive
heating periods and is the sum of the hot-gas ow period Ph , cold-gas ow period Pc , and
reversal period Pr (i.e., the time required to switch from the hot to the cold gas period,
and vice versa):
Pt Ph Pc Pr
5:3
5:4
with r as the sector angle covered by the radial seals shown Fig. 5.13. The periods and
sector angles are related as
Pj j
Pt t
j h or c
5:5
Now the matrix wall heat capacitances Cr; j , j h or c, are related to the total matrix
heat capacitance as
Pj
Cr; j Cr
Pt
j
or Cr
t
j h or c
5:6
The matrix wall heat capacity rates during the hot and cold periods, using the denition,
are
Cr; j
Cr; j Cr
Cr N
Pj
Pt
j h or c
5:7
where the second equality is from Eq. (5.6) and the third equality from Eq. (5.2). Thus,
Cr; j Cr;h Cr;c Cr
5:8
The heat transfer areas Ah and Ac are related to the total heat transfer area A of all
matrices of a xed-matrix regenerator as
Aj
APj VPj
Pt
Pt
j h or c
5:9
Aj Vj
t
t
j h or c
5:10
Here is the heat transfer surface area density or packing density, and V is the total
volume of all matrices.
312
At this time, it may again be pointed out that we have selected the reference coordinate system as (x; ) for both rotary and xed-matrix regenerators (see Fig. 5.1). Hence,
even for a rotary regenerator, we will use Eq. (5.9) and the pertinent expression in Eqs.
(5.2), (5.6), and (5.7) in terms of Pt , Ph , and Pc .
Porosity and Volumetric Heat Capacity of the Matrix. The core or matrix porosity is a
ratio of the void volume to the total core or matrix volume. If the heat transfer surface
is made of continuous ow passages (see Fig. 1.43), the porosity is the ratio of the ow
area to the frontal area of the core. If the heat transfer surface has interruptions (such
as perforations) or is made up of porous materials, the porosity is the ratio of the void
volume to the total core volume. Thus, the porosity is dened as follows:
8
Ao
>
>
>
>
< Afr
>
Vvoid
>
>
>
:A L
5:11
fr
Dh
>
> 4Vvoid
:
A
Governing Equations
On the basis of the foregoing assumptions, let us derive the governing equations and
boundary conditions. Consider the counterow regenerator of Fig. 5.1. A rotary regenerator is shown in Fig. 5.1a. For clarity, only one regenerator elemental ow passage and
the associated ow matrix are shown in Fig. 5.1b during the hot-gas ow period and in
Fig. 5.1c during the cold-gas ow period. In fact, in the derivation of the governing
dierential equations, all quantities (surface area, ow area, ow rate, etc.) associated
with a complete cross section of the regenerator at x and x dx are considered. The
reference coordinate system considered is x; , so that Figs. 5.1b and c are valid for
a rotary regenerator having an observer riding on the matrix. Figure 5.1b and c are
also valid for a xed-matrix regenerator with the observer standing on the stationary
matrix. To show clearly that the theoretical analysis is identical for rotary and xedmatrix regenerators, we consider variables and parameters associated with a complete
313
FIGURE 5.1 (a) Rotary regenerator showing sections x and dx; (b) regenerator elemental ow
passage and associated matrix during the hot-gas ow period; (c) same as (b) during the cold-gas
ow period (From Shah, 1981).
314
dqh = hh
dx
ChTh
( )
Ch dx Th
L h
( )
dqh = hh Ahdx
L
Ch Th +
Th
dx
x
Ahdx
(Th Tw)
L
( )
Tw
Cr,h dx
L h
w /2
dx
(Th Tw)
(b)
(a)
FIGURE 5.2 Energy rate terms associated with the elemental passage dx (a) of uid, and (b) of
matrix at a given instant of time during the hot-gas ow period (From Shah, 1981).
regenerator in this section. That means that we will consider the heat transfer surface
area, ow rates, and so on, associated with all matrices of a xed-matrix regenerator.
Hot Period: Fluid. The dierential uid and matrix elements of the hot-gas ow period
are shown in Fig. 5.2 with the associated energy transfer terms at a given instant of
time. In Fig. 5.2a, during its ow through the elemental passage, the hot gas transfers
heat to the wall by convection, resulting in a reduction in its outlet enthalpy and
internal thermal energy storage. Applying the energy balance, the rst law of thermodynamics, to this elemental passage, we get
@T
A dx
dx @Th
Ch Th Ch Th h dx hh h Th Tw;h Ch
@x
L
L @h
5:13
Substituting the value of Ch from Eq. (5.1) into this equation and simplifying, we get
@Th
L @Th hAh
Tw;h Th
@h d;h @x
Ch d;h
5:14
Hot Period: Matrix. With zero longitudinal and innite transverse wall heat conduction, the heat transferred from the hot uid to the matrix wall is stored in the wall in the
form of an increase in wall enthalpy. An energy balance on the matrix wall elemental
passage is (see Fig. 5.2b)
dx @Tw;h
A dx
5:15y
hh h
Cr;h
Th Tw;h
@h
L
L
{
This equation is accurate for a xed-matrix wall. For a rotary regenerator, the temperature time derivative
@Tw;h =@h may be interpreted as a substantial derivative (i.e., a material derivative), DTw;h =Dh . This is because
a xed reference frame is used and the wall temperature appears to be a function of both time and angular
coordinate, which are not independent variables. Consequently, Eq. (5.15) can be represented accurately as
follows for a rotary regenerator matrix:
dx DTw;h
A dx
Cr;h
hh h
5:16
Th Tw;h
Dh
L
L
Because the angular velocity is constant, Eq. (5.16) reverts to Eq. (5.15).
315
T Tw;h
@h
Cr;h Ph h
5:17
Cold Period: Fluid and Matrix. For the cold-gas ow period, a pair of equations similar
to Eqs. (5.14) and (5.17) results:
@Tc
L @Tc hAc
Tc Tw;c
@c d;c @x
Cc d;c
5:18
@Tw hAc
Tw;c Tc
@c
Cr;c Pc
5:19
The boundary conditions are as follows. The inlet temperature of the hot gas is constant
during the hot-gas ow period, and the inlet temperature of the cold gas is constant
during the cold-gas ow period:
Th 0; h Th;i constant
for 0 h Ph
5:20
Tc L; c Tc;i constant
for 0 c Pc
5:21
for 0 x L
for 0 x L
5:22
5:23
Since the regenerator is in periodic equilibrium, Eqs. (5.20)(5.23) are valid for
nPt , where n is an integer, n 0.
The boundary conditions of Eqs. (5.20) and (5.21) are simplest for the analysis. The
corresponding analytical models can be solved using analytical and seminumerical
methods. In applications, the hot- and cold-uid inlets to the regenerator may have
nonuniform temperature proles. Then the solution can only be obtained by a numerical
analysis.
Based on the foregoing dierential equations and boundary conditions [Eqs. (5.14)
and (5.17)(5.23)], the dependent uid and matrix temperatures are functions of the
following variables and parameters:
Th ; Tc ; Tw x; h ; c ; Th;i ; Tc;i ; Ch ; Cc ; d;h ; d;c ; Cr ; hAh ; hAc ; L; Ph ; Pc
|{z}
|{z} |{z} |{z}
dependent
independent
parameters under designers control
operating condition
variables
variables
variables
5:24
Neither Cr;c nor Cr;h is included in the foregoing list since Cr;c Cr;h Cr [see Eq. (5.8)].
The regenerator uid and wall temperatures are thus dependent on 14 independent
variables and parameters (since is the independent time variable designated with the
subscripts h and c for ease of understanding) . Through nondimensionalization we obtain
316
four independent and one dependent dimensionless groups. The specic form of these
groups is optional to some extent. Two such sets have been used for regenerator analysis
leading to two design methods. The eectivenessnumber of transfer units ("-NTUo )
method is generally used for rotary regenerators. The reduced lengthreduced period
(-) method is generally used for xed-matrix regenerators. It was shown by Shah
(1981) that both methods are equivalent. These methods are presented in the following
sections, with solutions for counterow and parallelow regenerators. Note that for
regenerators, there are no counterparts to the other ow arrangements used in recuperators, such as crossow and multipass cross-counterow.
5.2
This "-NTUo method was developed by Coppage and London (1953). The dimensionless
groups in this method are rst formulated in Section 5.2.1 such that most of the important groups parallel those of the recuperators discussed in Section 3.3. In Sections 5.2.2
and 5.2.3, the physical signicance of the additional dimensionless groups (compared to
that for a recuperator) is discussed. In Sections 5.2.4 and 5.2.5, the "-NTUo results for
counterow and parallelow regenerators are presented.
5.2.1
Dimensionless Groups
There are a number of dierent ways to formulate dimensionless groups. In Section 3.3.1
on recuperators, we made a list of all possible dimensionless groups from the energy
balances and rate equations and then eliminated those which were dependent. Now we
use a dierent approach to illustrate an alternative way to formulate dimensionless
groups for the regenerator problem. We derive dimensionless groups by making dierential equations and boundary conditions nondimensional.
Introduce the following dimensionless independent variables X* and * asy
x
L
h* h
Ph
X*
5:25
c*
c
Pc
5:26
Th Tc;i
Th;i Tc;i
Tc*
Tc Tc;i
Th;i Tc;i
Tw*
Tw Tc;i
Th;i Tc;i
5:28
{
In the case of cryogenics and Stirling engine regenerators, the rotational speed or the valve switching frequency is
so high that d;h and d;c may not be so small as to be negligible compared to Ph and Pc . Consideration of this eect
is beyond the scope of this chapter. However, in that case, *h and *c are dened as
*h
1
x
h d;h
Ph
L
*c
1
x
c d;c
Pc
L
5:27
317
hAh
Ch
C*r ;h
Cr;h
Ch
ntuc
*
Cr;c
hAc
Cc
5:29
Cr;c
Cc
5:30
With these nondimensional groups, Eqs. (5.14), (5.17), (5.18), and (5.19) reduce as
follows:
@Th*
ntuh Tw* Th*
@X *
5:31
@Tw* ntuh
Th* Tw*
*
Cr;h
@h*
5:32
@Tc*
ntuc Tc* Tw*
@X *
5:33
@Tw* ntuc
Tc* T*w
@c*
C *r;c
5:34
for 0 h* 1
5:35
Tc* 1; c* 0
for 0 c* 1
5:36
* X *; h* 1 Tw;c
* X *; c* 0
Tw;h
for 0 X * 1
5:37
* X *; c* 1
T *w;h X *; h* 0 Tw;c
for 0 X * 1
5:38
It is clear from Eqs. (5.31)(5.38) that the dependent temperatures are functions of
Th*; Tc*; Tw* X *; h*; c*; ntuh ; ntuc ; C *r;h ; C*r;c
5:39
Thus we are able to reduce independent variables and parameters from 14 to 6 (considering only one period at a time).
For overall regenerator performance, we are interested in determining average uid
outlet temperatures. In a rotary regenerator, the outlet temperatures vary as a function of
the angular coordinate . If h and c represent the angles for the sectors through which
hot and cold gases ow, respectively, the space average outlet temperatures are
1
Th;o
h
h
0
Th;o d;
1
Tc;o
c
c
0
Tc;o d
5:40
where Th;o and Tc;o represent the angular coordinate-dependent uid temperatures
at the regenerator outlet. However, for an observer riding on the rotary regenerator
318
matrix, the uid outlet temperatures are functions of the time . In this case, the timeaveraged outlet temperatures are
1
Th;o
Ph
Ph
0
Th;o d;
1
Tc;o
Pc
Pc
0
Tc;o d
5:41
Here Th;o and Tc;o are the time-dependent uid temperatures at the regenerator
outlet. Th;o and Tc;o in Eqs. (5.40) and (5.41), respectively, represent the space- and timeaveraged uid temperatures at the regenerator outlet. Equation (5.41) is valid for both
the xed-matrix and rotary regenerators, provided that an observer rides on the matrix
for a rotary regenerator.
Thus, the functional relationship of Eq. (5.39) for the dependent regenerator average
outlet temperatures is
T *h;o ; T *c;o ntuh ; ntuc ; C *r;h ; C*r;c
5:42
A bar on these dimensionless as well as dimensional [of Eq. (5.41)] average outlet temperatures is eliminated for convenience (like in a recuperator) in the rest of the chapter,
except where such terminology could create ambiguity.
These outlet temperatures are conveniently expressed by the regenerator eectiveness
" q=qmax . Using the outlet temperatures dened by Eq. (5.41), the actual heat transfer
rate in the rotary regenerator is
q Ch Th;i Th;o Cc Tc;o Tc;i
5:43
In this section we have considered either a rotary regenerator or all matrices of a xedmatrix regenerator as a system. The hot and cold uids ow continuously in and out of
such a system. To determine qmax for such a system, we dene a perfect heat exchanger
as dened for recuperators after Eq. (3.37). This perfect heat exchanger is a counterow
recuperator of innite surface area, zero longitudinal wall heat conduction, and zero ow
leakages from one uid to another uid, operating with uid ow rates and uid inlet
temperatures the same as those of an actual regenerator; uid properties are considered
constant for both exchangers. qmax for this perfect heat exchanger as in Eq. (3.42) is
qmax Cmin Th;i Tc;i
where Cmin is the minimum of Ch and Cc . The regenerator eectiveness is thus
Ch Th;i Th;o
C T Tc;i
q
c c;o
"
qmax Cmin Th;i Tc;i
Cmin Th;i Tc;i
5:44
5:45
Then for Cc Cmin , comparing " with T*h;o 1 "C* and T*c;o " from the denitions
in Eq. (5.28) for the outlet temperatures, we get
"
1 T*h;o
T*c;o
C*
5:46
5:47
319
Since these independent dimensionless groups, dened in Eqs. (5.29) and (5.30), are
not parallel to those of a recuperator (a direct-transfer exchanger), let us dene a related
set as follows:
1
1
Cmin 1=hAh 1=hAc
5:48
C*
Cmin
Cmax
5:49
C*r
Cr
Cmin
5:50
5:51
NTUo
hA*
Since Eqs. (5.48)(5.51) utilize only four independent dimensionless groups [to be shown
related to those of Eq. (5.47)], it is then valid to recast Eq. (5.47) as
" NTUo ; C*; C*r ; hA*
5:52
Here NTUo is the modied number of transfer units. Since there is no direct heat transfer
from the hot uid to the cold uid in a regenerator (similar to that in a recuperator), UA
does not come into the picture directly for the regenerator. However, if the bracketed
term of Eq. (5.48) is designated as Uo A, with Uo termed as a modied overall heat transfer
coecient, then
1
1
1
Uo A hAh hAc
and
NTUo
Uo A
Cmin
5:53
A comparison of this expression with Eq. (3.24) or (3.20) reveals that Uo A is the same as
UA when the wall thermal resistance and fouling resistances are zero. Note that the entire
surface in a regenerator is primary surface (no ns), and hence the overall extended
surface eciency o 1. Thus the denition of NTUo Uo A=Cmin parallels that of
NTU of Eq. (3.59), and in the limiting case of C*r 1, the numerical solutions demonstrate that the regenerator has the same performance as a recuperator with its NTUo
identical to NTU provided that the pressure and carryover leakage eects are neglected.
The newly dened dimensionless groups of Eq. (5.52) are related to those of Eq. (5.47) as
follows for Cc Cmin :
1
C*r;h =C*r;c =ntuh 1=ntuc
5:54
C*r;h
C*r;c
5:55
C*r C*r;c
5:56
NTUo
C*
hA*
ntuc C *r;h
ntuh C*r;c
5:57
320
A comparison of Eq. (5.52) with Eq. (3.50) reveals that the eectiveness of a regenerator is dependent on two additional parameters, C*r and (hA)*. Since the thermal
energy is stored and delivered by the matrix wall periodically, the wall temperature is
going to be dependent on (1) the storage heat capacity rate of the matrix wall, and (2) the
thermal conductances between the matrix wall and the hot uid, hAh , and between
the matrix wall and the cold uid, hAc . And as a result, two additional dimensionless
groups, C*r and (hA)*, come into the picture; they are discussed further next.
5.2.2
Heat transfer in the regenerator from the hot gas to the matrix surface is by convection/
radiation during the hot gas ow period, depending on the applications, and that from
the matrix surface to the cold gas is by convection/radiation during the cold-gas ow
period. With all other variables/parameters the same, this heat transfer rate is greatest
when the temperature dierence between the gas and surface is the greatest. For this
reason, the matrix surface is not allowed to be heated to the hot-uid inlet temperature or
cooled to the cold-uid temperature; otherwise, this would result in zero temperature
potential and zero heat transfer near the end of the hot- or cold-gas ow period. The
temperature swing at the exit of the regenerator is reduced with large C*r , which translates
into relatively fast rotational speeds for rotary regenerators or fast valve switching in
xed-matrix regenerators. Hence, from the heat transfer point of view, the dimensionless
group C*r which takes the rotational speed/valve switching frequency into account, can
have a large impact on the design of the regenerator.
When the cold-gas period starts, the cold gas gets heated at the entrance from the heat
rejected by the local matrix elements. If the blow period is long (i.e., low values of C*r ), the
local matrix elements at the entrance will be cooled to the cold-uid inlet temperature,
and hence no heat transfer will take place in that region subsequently. A similar phenomenon will take place during the hot-gas period, where the entrance region will be
heated to the hot-uid inlet temperature, and hence no heat transfer will take place in
that region subsequently. This eect propagates downstream, depending on the length of
the period until the uid ow switches. This phenomenon is sometimes quantied by an
exhaustion coecient which is inversely proportional to C*r .
Figure 5.3 clearly demonstrates that the regenerator eectiveness " increases with C*r
for given values of NTUo and C*. However, higher rotational speeds/valve switching
frequency would induce larger carryover leakage and eect higher seal wear and tear
(thus increasing the seal leakage), both of which will reduce the regenerator eectiveness.
The range of the optimum value of C*r is between 2 and 4 for optimum regenerator
eectiveness and seal life (Shah, 1988), although many rotary regenerators are designed
with C*r larger than 4.
5.2.3
As mentioned earlier, the wall temperature prole in a regenerator (in the absence of
longitudinal wall heat conduction) is going to be dependent on the thermal conductances
hAh and hAc between the matrix wall and the hot/cold uids. For a high-temperature
regenerator, the thermal conductance will not only include convection conductance but
also radiation conductance. The dimensionless group that takes into account the eect of
the convection conductance ratio is (hA)*, as dened by Eq. (5.51). Lambertson (1958)
and others have shown through a detailed analysis that (hA)* has a negligible inuence
321
100
90
80
%
1.0
1.5
70
2.0
5.0
C*r =
60
0.25 (hA)* 4
50
0
10
NTUo
FIGURE 5.3 Counterow regenerator " as a function of NTUo and for C* 1 (Kays and
London, 1998).
on the regenerator eectiveness for the range 0:25 hA* < 4. Since most regenerators
operate in this range of hA*, fortunately, the eect of hA* on the regenerator eectiveness can usually be ignored.
5.2.4
5:58
322
where
B 33 134 305 6
R 2 34 535 46
A 3 33 534 45 126
N 4 24 35 6 352
E 2 4 6 2 52 32 6 23 4 5 43
N1 4 4 25 6
5:59
252
i 2; 3; . . . ; 6
and
Vi x; y expx y
n=2
1
X
p
n
y
In 2 xy
i
1
x
ni1
i 2; 3; . . . ; 6
5:60
In these equations, all variables/parameters are local except for NTUo , C*r , and ".
Here In represents the modied Bessel function of the rst kind and nth (integer)
order. The regenerator eectiveness of Eq. (5.58) is presented in Table 5.1 for
0:5 NTUo 500 and 1 C*r 1, and some typical results are presented in Fig.
5.3. The values for C*r 1 can be calculated from an asymptotic expression
" NTUo =1 NTUo .
Now let us discuss further the reasons for choosing the set of dimensionless independent groups of Eq. (5.52) instead of those of Eq. (5.47):
1. For specied NTUo , C*, and C*r , the effectiveness " generally decreases with
decreasing values of (hA)*, and the reverse occurs for large values of NTUo and
C* 1. However, the inuence of (hA)* on " is negligibly small for
0:25 hA* 4, as shown by Lambertson, among others. A maximum error
of 0.5 point on % " occurs for C* 0:9, C*r 1, and NTUo 9. A maximum
error of 2 points on % " occurs at C* 0:7, C*r 1, and NTUo 9. The
maximum error of 5 points on % " occurs for C* 0:1, C*r 1, and
NTUo 3. Since in most regenerators, C * > 0:8, we can effectively eliminate
(hA)* from Eq. (5.52). Hence,
" NTUo ; C*; C*r
5:61
2. When C*r ! 1, the effectiveness " of a regenerator approaches that of a recuperator. The difference in " for C*r 5 and that for C*r 1 is negligibly small and
may be ignored for the design purpose.
323
0.3321
0.4665
0.5477
0.6006
0.6385
0.6672
0.6900
0.7086
0.7242
0.7375
0.7491
0.7592
0.7682
0.7762
0.7834
0.7900
0.7959
0.8014
0.8065
0.8111
0.8155
0.8195
0.8233
0.8268
0.8302
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
10.0
10.5
11.0
11.5
12.0
12.5
1.0
0.5
1.0
1.5
2.0
2.5
NTUo
0.8733
0.8774
0.8813
0.8848
0.8882
0.8468
0.8531
0.8588
0.8641
0.8689
0.8029
0.8139
0.8236
0.8322
0.8399
0.7117
0.7374
0.7583
0.7756
0.7903
0.3282
0.4845
0.5757
0.6359
0.6791
1.5
0.8886
0.8924
0.8960
0.8993
0.9204
0.8635
0.8695
0.8749
0.8799
0.8844
0.8206
0.8315
0.8410
0.8494
0.8568
0.7279
0.7544
0.7757
0.7933
0.8080
0.3304
0.4912
0.5861
0.6490
0.6940
2.0
0.8956
0.8992
0.9026
0.9058
0.9087
0.8713
0.8771
0.8824
0.8872
0.8916
0.8290
0.8398
0.8492
0.8575
0.8648
0.7356
0.7625
0.7840
0.8017
0.8165
0.3315
0.4943
0.5910
0.6553
0.7011
2.5
0.8999
0.9035
0.9068
0.9099
0.9127
0.8759
0.8817
0.8869
0.8917
0.8960
0.8339
0.8447
0.8540
0.8622
0.8695
0.7399
0.7670
0.7887
0.8065
0.8213
0.3320
0.4960
0.5937
0.6587
0.7051
3.0
0.9029
0.9064
0.9097
0.9128
0.9156
0.8790
0.8848
0.8900
0.8947
0.8990
0.8369
0.8477
0.8571
0.8653
0.8726
0.7425
0.7698
0.7916
0.8094
0.8243
0.3324
0.4971
0.5954
0.6608
0.7075
3.5
0.9050
0.9086
0.9119
0.9149
0.9177
0.8811
0.8869
0.8921
0.8968
0.9011
0.8390
0.8498
0.8592
0.8674
0.8747
0.7442
0.7716
0.7935
0.8114
0.8264
0.3326
0.4978
0.5965
0.6622
0.7091
4.0
TABLE 5.1 Counterow Regenerator e as a Function of NTUo and for C* 1.0 and (hA)* 1
0.9065
0.9101
0.9134
0.9165
0.9193
0.8827
0.8885
0.8937
0.8984
0.9026
0.8404
0.8513
0.8607
0.8689
0.8762
0.7454
0.7729
0.7948
0.8128
0.8278
0.3328
0.4982
0.5972
0.6631
0.7102
4.5
0.9077
0.9113
0.9146
0.9176
0.9205
0.8838
0.8896
0.8948
0.8995
0.9038
0.8415
0.8524
0.8618
0.8701
0.8773
0.7463
0.7738
0.7958
0.8138
0.8288
0.3329
0.4986
0.5977
0.6638
0.7109
5.0
0.9117
0.9153
0.9186
0.9217
0.9245
0.8876
0.8934
0.8987
0.9034
0.9077
0.8450
0.8559
0.8654
0.8737
0.8811
0.7491
0.7768
0.7989
0.8171
0.8322
0.3332
0.4996
0.5994
0.6659
0.7134
10.
0.9130
0.9167
0.9200
0.9231
0.9259
0.8889
0.8947
0.9000
0.9048
0.9091
0.8462
0.8571
0.8667
0.8750
0.8824
0.7500
0.7778
0.8000
0.8182
0.8333
0.3333
0.5000
0.6000
0.6667
0.7143
324
0.8489
0.8532
0.8571
0.8607
0.8640
0.8775
0.8874
0.8951
0.9014
0.9065
0.9109
0.9180
0.9235
0.9280
0.9317
0.9348
0.9522
0.9602
0.9650
0.9684
16.0
17.0
18.0
19.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
60.0
70.0
80.0
90.0
100.0
200.0
300.0
400.0
500.0
0.9742
0.9820
0.9847
0.9861
0.9869
0.9594
0.9642
0.9677
0.9704
0.9725
0.9328
0.9412
0.9475
0.9524
0.9562
0.9065
0.9106
0.9143
0.9176
0.9207
0.8913
0.8942
0.8970
0.8996
0.9020
0.9791
0.9860
0.9884
0.9896
0.9903
0.9662
0.9704
0.9734
0.9758
0.9776
0.9427
0.9502
0.9558
0.9600
0.9634
0.9192
0.9229
0.9262
0.9292
0.9320
0.9053
0.9079
0.9105
0.9128
0.9151
0.9782
0.9841
0.9861
0.9871
0.9877
0.9670
0.9707
0.9733
0.9754
0.9770
0.9461
0.9529
0.9578
0.9616
0.9646
0.9245
0.9279
0.9310
0.9338
0.9363
0.9114
0.9140
0.9163
0.9186
0.9207
0.9784
0.9837
0.9854
0.9863
0.9869
0.9682
0.9716
0.9740
0.9758
0.9773
0.9487
0.9551
0.9597
0.9632
0.9660
0.9280
0.9313
0.9343
0.9370
0.9394
0.9154
0.9178
0.9202
0.9223
0.9243
a
Neglecting the eects of longitudinal wall heat conduction, uid bypass and carryover.
Source: Data from Shah (1988).
0.8333
0.8363
0.8391
0.8417
0.8442
13.0
13.5
14.0
14.5
15.0
0.9798
0.9848
0.9864
0.9873
0.9878
0.9700
0.9732
0.9755
0.9773
0.9787
0.9510
0.9572
0.9617
0.9651
0.9678
0.9307
0.9340
0.9369
0.9395
0.9419
0.9182
0.9207
0.9230
0.9251
0.9271
0.9814
0.9863
0.9879
0.9888
0.9893
0.9717
0.9749
0.9772
0.9789
0.9803
0.9530
0.9591
0.9636
0.9670
0.9696
0.9328
0.9360
0.9389
0.9415
0.9439
0.9203
0.9228
0.9250
0.9272
0.9292
0.9828
0.9877
0.9894
0.9902
0.9907
0.9732
0.9764
0.9787
0.9804
0.9817
0.9545
0.9607
0.9651
0.9685
0.9711
0.9344
0.9376
0.9405
0.9431
0.9455
0.9219
0.9243
0.9266
0.9287
0.9307
0.9840
0.9889
0.9905
0.9914
0.9919
0.9744
0.9776
0.9799
0.9816
0.9829
0.9557
0.9619
0.9663
0.9697
0.9723
0.9356
0.9388
0.9417
0.9443
0.9467
0.9231
0.9255
0.9278
0.9299
0.9319
0.9885
0.9934
0.9950
0.9959
0.9964
0.9788
0.9820
0.9843
0.9860
0.9874
0.9600
0.9662
0.9707
0.9740
0.9767
0.9397
0.9430
0.9459
0.9485
0.9509
0.9271
0.9296
0.9319
0.9340
0.9360
0.9901
0.9950
0.9967
0.9975
0.9980
0.9804
0.9836
0.9859
0.9877
0.9890
0.9615
0.9677
0.9722
0.9756
0.9783
0.9412
0.9444
0.9474
0.9500
0.9524
0.9286
0.9310
0.9333
0.9355
0.9375
325
Thus by the selection of the Eq. (5.52) set, we have demonstrated the similarities and
dierences between a regenerator and a recuperator in the absence of ow leakages and
carryover.
The following observations may be made by reviewing Fig. 5.3, the results of Table
5.1 for C * 1, and those for C * < 1 by Kays and London (1998):
1. For specied C*r and C *, the heat exchanger effectiveness increases with increasing
NTUo . For all C*r and C*, " ! 1 as NTUo ! 1.
2. For specied NTUo and C*, " increases with increasing values of C*r and
approaches asymptotically the value for a counterow recuperator.
3. For specied NTUo and C*r , " increases with decreasing values of C*. The percentage change in " is largest in the lower NTUo range, and this percentage change in "
increases with increasing values of C*r .
4. For " < 40% and C*r > 0:6, C* and C*r do not have a signicant inuence on the
exchanger effectiveness.
Now let us present approximate formulas to compute " for a wide range of C*r and C*.
The inuence of C*r on " can be presented by an empirical correlation for " 90% by
Kays and London (1998) as
"
#
1
5:62
" "cf 1
9C*r 1:93
where "cf is the counterow recuperator eectiveness as follows:
"cf
1 expNTUo 1 C*
NTUo
!
1 C expNTUo 1 C * C 1 1 NTUo
5:63
Equation (5.62) agrees within 1% with the tabular results of Lambertson (1958) for
C* 1 for the following ranges: 2 < NTUo < 14 for C*r 1:5, NTUo 20 for
C*r 2, and a complete range of NTUo for C*r 5. For decreasing values of C*, the
error due to the approximation increases with lower values of C*r . For example, to obtain
accuracy within 1%, C*r 1:5 for C* 0:9, and C*r 2:0 for C * 0:7.
The following approximate procedure is proposed by Razelos (1980) to calculate the
regenerator eectiveness " for the case of C * < 1. For the known values of NTUo , C*,
and C*r , calculate equivalent values of NTUo and C*r for a balanced regenerator
(C * 1), designated with a subscript m, as follows:
2NTUo C*
1 C*
2C*C *
C*r;m r
1 C*
5:64
NTUo;m
5:65
With these values of NTUo;m and C *r;m , obtain the value of "r using Eq. (5.58) with " "r
or from the approximate equation
NTUo;m
"r
1 NTUo;m
"
1
1
9C*r;m 1:93
#
5:66
326
5:67
A comparison of " from this procedure with that of Eq. (5.62) shows that the Razelos
approximation yields more accurate values of " compared to that from Eq. (5.62) for
C* < 1. It can be shown that "s of Eqs. (5.62) and (5.67) are identical for C*r 1. Note
that either by employing the foregoing approximate method or a direct use of Eq. (5.62),
we at most need Table 5.1 or Fig. 5.3; thus the tabular data for C * < 1 are not needed.
5.2.5
The dierential equations and boundary conditions for the parallelow regenerator are
the same as those of Eqs. (5.31)(5.38) except for Eq. (5.36). The boundary condition of
Eq. (5.36) for this case is
Tc*0; c* 0
for 0 c* 1
5:68
The solution may be presented in terms of " as a function of the same four dimensionless
groups as for a counterow regenerator, Eq. (5.52). Theoclitus and Eckrich (1966)
obtained the solution numerically by a nite dierence method. They covered the following ranges of the parameters: 1 NTUo 10, 0:5 C* 1:0, 0:2 C*r 1, and
0:25 hA* 1. Their results for C * 1 are presented in Fig. 5.4.
From a review of Fig. 5.4, it is interesting to note that the eectiveness of a parallelow regenerator oscillates above and below that for a parallelow recuperator
80
Cr* = Cr/Cmin = 1.00
70
1.25
0.60
60
%
1.50
5.00,
50
2.00
0.40
40
30
10
NTUo
FIGURE 5.4 Parallelow regenerator " as a function of NTUo and for C*r 1 and (hA* 1.
327
("max 50%). The oscillations decrease in amplitude as Cr* increases. The maximum
eectiveness for the parallelow regenerator is reached at C*r 1, and it exceeds the
eectiveness of a parallelow recuperator. These results contrast with those for the
counterow regenerator where the limiting eectiveness, represented by the counterow
recuperator, is never exceeded, but is approached asymptotically as C*r increases.
Similar to a recuperator, parallelow is sometimes preferred to counterow in some
regenerator applications for the rst three reasons mentioned for a recuperator in Section
1.6.1.2.
Example 5.1 A boiler is equipped with a regenerator rotating at 4 rpm and having
the ue gas (hot) and air (cold) ow areas with disk sector angles of 1958 and 1658,
respectively. The rotor with 2 m diameter and 0.4 m height is turned by a shaft of 0.2 m
diameter. The matrix material has the following characteristics:
Density 7800 kg=m3
Porosity 0:76
The ue gas and airstream owing in counterow have the following characteristics:
Flue Gas
Air
60
1.11
8.3
320
50
1.005
6.9
26
Assuming that 10% of the rotor face is covered by radial seals, calculate the regenerator eectiveness and heat transfer rate from the ue gas stream to the airstream.
SOLUTION
Problem Data and Schematic: The heat transfer coecients, uid ow rates, and inlet
temperatures are provided in Fig. E5.1 for both the hot and cold uid streams. In
4 rpm
Seal coverage
10%
Hot gas
Th,i = 320C
195
m
h = 8.3 kg/s
165
0.2 m
0.4 m
Matrix
w = 7800 kg/m3
cw = 0.44 kJ/kg K
= 1200 m2/m3
= 0.76
Gas/air
hh = 60 W/m 2 K, cp,h = 1.11 kJ/kg K
hc = 50 W/m 2 K, cp,c = 1.005 kJ/kg K
FIGURE E5.1
328
addition, the regenerator dimensions and rotational speed are specied. Also specied
are the physical properties of both hot and cold uid streams as well as the matrix
material.
Determine: The regenerator eectiveness and the heat transfer rate from the ue gas to
the air.
Assumptions: The assumptions of Section 5.1.1 are invoked here.
Analysis: The regenerator eectiveness " is a function of four dimensionless groups as
given in Eq. (5.52):
" NTUo ; C *; C*r ; hA*
To determine whether the eect of the convection conductance ratio (hA)* is
negligible, we rst establish which side has the lower heat capacity rate. To that end,
Ch m_ h cp;h 8:3 kg =s 1:11 kJ=kg K 9:21 kW=K
Cc m_ c cp;c 6:9 kg=s 1:005 kJ=kg K 6:93 kW=K
Thus,
Cmax Ch 9:21 kW=K
C*
0:752
Cmax 9:21 kW=K
Now
hA*
hA on Cmin side
hA on Cmax side
Noting that the heat transfer surface areas on the two sides are in proportion to the disk
sector angles, we have
hA*
50 W=m2 K 1658
0:71
60 W=m2 K 1958
As pointed out by Lambertson (1958), (hA)* has a negligible inuence on the regenerator eectiveness " in the range 0:25 hA* 4. Hence, in the present case with
hA* 0:71, we have
" NTUo ; C*; C*r
To determine C*r , we rst determine matrix mass as
Mw rotor cross-sectional area rotor height matrix material density
matrix solidity
22 0:22 m2 0:4 m 7800 kg=m3 1 0:76 2328:9 kg
4
Knowing the matrix mass, its heat capacity rate is computed as
329
4
rev=s 68:3 kW=K
60
Cr
68:3 kW=K
9:86
Cmin 6:93 kW=K
To nd the hot- and cold-side heat transfer surface areas, we note that the total matrix
surface area A is given by
A rotor cross-sectional area rotor height matrix packing density
fraction of rotor face area not covered by radial seals
22 0:22 m2 0:4 m 1200 m2 =m3 1 0:1 1343:6 m2
4
The hot- and cold-gas-side surface areas are proportional to the respective sector angles,
so that
1958
1958
A
1343:6 m2 727:8 m2
3608
3608
1658
1658
Ac
A
1343:6 m2 615:8 m2
3608
3608
Ah
Thus,
hAh 60 W=m2 K 727:8 m2 43,668 W=K 43:67 kW=K
hAc 50 W=m2 K 615:8 m2 30; 790 W=K 30:79 kW=K
From the foregoing values in conjunction with Eq. (5.48), we get
NTUo
1
1
1
1
2:606
Cmin 1=hAh 1=hAc 6:93 kW=K 1=43:67 1=30:79 K=kW
Knowing C* and NTUo , the counterow recuperator eectiveness "cf is determined from
Eq. (5.63). Thus
"cf
1 expNTUo 1 C*
1 exp2:6061 0:752
0:7855
Knowing C*r and "cf , the regenerator eectiveness " can be calculated from Eq. (5.62)
valid for " 0:9:
" "cf
1
1
9C*r 1:93
0:7855 1
1
9 9:861:93
0:7844
Ans:
330
By denition,
"
q
q
Hence,
q "Cmin Th;i Tc;i 0:7844 6:93 kW=K 320 26 K 1598 kW
Ans:
We now demonstrate the use of Razelos method (1980) to evaluate " since C* is
considerably less than unity. First compute the parameters NTUo;m , C*r;m , and "r of
Eqs. (5.64)(5.66).
2NTUo C* 2 2:606 0:752
2:237
1 0:752
1 C*
2C*C* 2 9:86 0:752
C *r;m r
8:46
1 0:752
1 C*
NTUo;m
1
2:237
1
1
0:6898
1
"r
1:93
1 NTUo;m
1 2:237
9C *r;m
9 8:461:93
NTUo;m
In terms of these parameters, the regenerator eectiveness " is given in Eq. (5.67) as
"
Ans:
This value of " is very close to the value (0.7844) calculated by Eq. (5.62). Hence, the
value of q calculated above will remain virtually the same.
Discussion and Comments: Determination of the regenerator eectiveness and subsequent heat transfer rate for a regenerator is straightforward, as was the case with the
recuperator. The only dierence is that the regenerator eectiveness depends on C*r and
(hA)* in addition to NTUo and C*. Only the latter two dimensionless groups are used for
the recuperator eectiveness determination.
A comparison of Kays and London (1998) and Razelos methods for the determination of " shows that both methods yield virtually the same eectiveness, as expected. The
Razelos method would be more accurate for low values of C*.
As mentioned in the text, the regenerator eectiveness approaches the recuperator
eectiveness for higher values of C*r . It can be seen here that when C*r 9:86, " "cf
(0.784 vs. 0.785). When we consider the seal and carryover leakages, the actual " will be
lower than 0.784, depending on the amount of leakage. These leakage eects are discussed in Section 5.6.
Example 5.2 A rotary regenerator, with a rotational speed of 10 rpm, is used to recover
energy from a gas stream at 2508C owing at 10 kg/s. This heat is transferred to the
airstream at 108C, also owing at 10 kg/s. The wheel depth is 0.22 m and diameter 1.6 m,
331
so that its face area is approximately 1.8 m2. The mass of the matrix is 150 kg with a
surface-to-volume ratio of 3000 m2/m3, and the mean specic heat of the matrix material
is 0.8 kJ/kg K. The heat transfer coecient for both uid streams is 30 W/m2 K. The
mean isobaric specic heat of the gas is 1.15 kJ/kg K and that of air is 1.005 kJ/kg K.
The ow split gas : air 50% : 50%. For a counterow arrangement, calculate the
following values:
(a) The regenerator effectiveness
(b) The rate of heat recovery and the outlet temperatures of air and gas
(c) The rate of heat recovery and the outlet temperatures of air and gas if the
rotational speed of the wheel is increased to 20 rpm
(d) The rate of heat recovery and the outlet temperatures of air and gas if the
rotational speed of the wheel is reduced to 5 rpm
SOLUTION
Problem Data and Schematic: The heat transfer coecients, uid ow rates, and inlet
temperatures are provided in Fig. E5.2 for both uid streams. In addition, the regenerator dimensions and rotational speed are specied. Also specied are the physical
properties of both hot and cold uid streams as well as the matrix material.
10 rpm
Hot gas
Th,i = 250C
m
h = 10 kg/s
Matrix
Mw = 150 kg/150 kg
0.22 m
cw = 0.8 kJ/kg K
= 3000 m2/m3
Afr = 1.8 m2
Gas/air
hh = hc = 30 W/m 2 K
1.6 m
Cold air
Tc,i = 10C
mc = 10 kg/s
FIGURE E5.2
Determine: (a) The heat exchanger eectiveness, (b) the heat recovery rate and air outlet
temperature at the given wheel speed, (c) the heat recovery rate and air outlet temperature at the increased wheel speed, and (d) the heat recovery rate and air outlet temperature at the reduced wheel speed.
Assumptions: The assumptions of Section 5.1.1 are invoked here. Seal face coverage is
assumed negligible.
332
Analysis: (a) The regenerator eectiveness " can be calculated using Eq. (5.62), valid for
" 0:9, in conjunction with Eq. (5.63). For this purpose, we rst need to determine C*,
C*r , and NTUo . Let us rst calculate various heat capacity rates.
Ch m_ h cp;h 10 kg=s 1:15 kJ=kg K 11:5 kJ=s K 11:5 kW=K
Cc m_ c cpc 10 kg=s 1:005 kJ=kg K 10:05 kJ=s K 10:05 kW=K
Cmax Ch 11:5 kW=K
C*
0:8739
Cmax
11:5 kW=K
10
rev=s 20 kW=K
60
Cr
20 kW=K
1:99
Cmin 10:05 kW=K
The modied overall heat transfer coecient Uo , with negligible wall thermal resistance
and fouling resistance, is given as
1
1
1
1
Uo Ah Uo Ac hh Ah hc Ac
With equal hot- and cold-side heat transfer areas (since ow split 50 : 50; i.e., Ah Ac ),
Uo is given by
1
1
1
1
1
1 2
m K=W
Uo hh hc 30 W=m2 K 30 W=m2 K 15
or
Uo 15 W=m2 K
The total heat transfer area A is expressible as
A matrix volume surface area density
matrix face area depth surface area density
1:8 m2 0:22 m 3000 m2 =m3 1188 m2
Therefore,
Ah Ac 594 m2
Thus, NTUo is determined as
NTUo
Uo A 15 W=m2 K 594 m2
0:8866
Cmin
10:05 103 W=K
333
Now "cf and " are evaluated from Eqs. (5.63) and (5.62) as
1 expNTUo 1 C*
1 exp0:88661 0:8739
0:4840
1 C* expNTUo 1 C* 1 0:8739 exp0:88661 0:8739
1
1
0:4840
1
0:4697
Ans:
" "cf 1
9 1:991:93
9C*r 1:93
"cf
(b) By denition,
"
q
q
Hence,
q "Cmin Th;i Tc;i 0:4697 10:05 kW=K 250 10 K 1133 kW
Ans:
To nd the outlet temperatures of the cold and hot uid streams, we have from the
energy balance
q Cc Tc;o Tc;i Ch Th;i Th;o
Thus,
Tc;o Tc;i
q
1133 kW
108C
122:78C
Cc
10:05 kW=8C
Ans:
Th;o Th;i
q
1133 kW
2508C
151:58C
Ch
11:5 kW=8C
Ans:
20
rev=s 40 kW=K
60
Cr
40 kW=K
3:98
Cmin 10:05 kW=K
In this case, "cf will still remain the same as in part (a) since it is not aected by the rpm.
Then " is given by
" "cf 1
1
1:93
9C*r
!
0:4840 1
1
9 3:981:93
!
0:4803
Ans:
Tc;o Tc;i
q
1158 kW
108C
125:28C
Cc
10:05 kW=8C
Ans:
Th;o Th;i
q
1158 kW
2508C
149:38C
Ch
11:5 kW=8C
Ans:
334
5
rev=s 10 kW=K
60
Cr
10 kW=K
0:995
Cmin 10:05 kW=K
!
1
1
0:4840
1
" "cf 1
0:4297
9 0:9951:93
9C*r 1:93
C*r
Ans:
Tc;o Tc;i
q
1036 kW
108C
113:18C
Cc
10:05 kW=8C
Ans:
Th;o Th;i
q
1036 kW
2508C
159:98C
Ch
11:5 kW=8C
Ans:
Discussion and Comments: This example illustrates the eect of the rotational speed on
the heat recovery rate and the cold- and hot-stream outlet temperatures. As seen from the
following list, the higher the rotational speed, the higher the heat recovery rate and the
cold-stream outlet temperature and the lower the hot-stream outlet temperature.
However, the rate of change decreases with increasing C*r when comparing the results
for 5, 10, and 20 rpm. The rotational speed (in low C*r operating range) is clearly an
important variable in controlling all three quantities.
Wheel Rotational
Speed (rpm)
Heat
Recovery Rate
(kW)
Cold Stream
(Air) Outlet
Temperature (8C)
Hot Stream
(Gas) Outlet
Temperature (8C)
"
5
10
20
1036
1133
1158
113.1
122.7
125.2
159.9
151.5
149.3
0.4297
0.4697
0.4803
335
Physical properties
cp;air 1.050 kJ/kg K
cp;gas 1.084 kJ/kg K
cw 1.130 kJ/kg K
SOLUTION
Problem Data and Schematic: The heat transfer coecients, uid ow rates, and inlet
temperatures are provided in Fig. E5.3 for both hot and cold streams. In addition, the
regenerator dimensions and rotational speed are specied. Also specied are the physical
properties of both hot and cold uid streams as well as the matrix material.
15 rpm
Seal coverage
7%
Hot gas
Th,i = 960C
0.076 m
0.0715 m
Matrix (2 disks)
Mw = 34.93 kg
cw = 1.130 kJ/kg K
= 5250 m2/m3
Gas/air
hh = 240.5 W/m 2 K
0.683 m
hc = 220.5 W/m 2 K
Cold air
Tc,i = 480C
FIGURE E5.3
Determine: The regenerator eectiveness and outlet temperatures of both hot and cold
uids.
Assumptions: The assumptions of Section 5.1.1 are invoked here.
Analysis: Let us rst evaluate the dimensionless groups NTUo , C*, C*r , and (hA)* for the
determination of ". To that end, rst we need to compute the frontal area and volume of
both disks:
frontal area disk area hub area 1 seal face coverage number of disks
0:6832 0:0762 m2 1 0:07 2 0:673 m2
4
matrix volume frontal area flow length disk height
0:673 m2 0:0715 m 0:0481 m3
The total heat transfer area is given by
A V 5250 m2 =m3 0:0481 m3 252:53 m2
336
Since the ow split is 50 : 50, the hot and cold side areas are equal, so that
Ah Ac
252:53 m2
126:26 m2
2
Knowing the heat transfer areas and heat transfer coecients for both sides, we have
hAh 240:5 W=m2 K 126:26 m2 30,366 W=K 30:37 kW
hAc 220:5 W=m2 K 126:26 m2 27,840 W=K 27:84 kW=K
hA*
0:917
hAh 30:37 kW=K
Knowing the ow rates and specic heats of the two uids, we get
Ch m_ h cp;h 2:094 kg=s 1:084 kJ=kg K 2:270 kW=K
Cc m_ c cp;c 2:029 kg=s 1:050 kJ=kg K 2:130 kW=K
Cmin 2:130 kW=K
C*
0:9383
Cmax Ch 2:270 kW=K
15
rev=s 9:868 kW=K
60
Cr
9:868 kW=K
4:63
Cmin 2:130 kW=K
1
1
1
1
NTUo
6:819
Cmin 1=hAh 1=hAc 2130 kW=K 1=30:37 1=27:84 K=kW
With the foregoing values of NTUo , C*, C*r , and (hA)*, we determine the regenerator
eectiveness " using Eqs. (5.63) and (5.62).
"cf
1 expNTUo 1 C *
1 exp6:8191 0:9383
0:8945
1
1:93
*
9Cr
0:8945 1
1
9 4:631:93
0:8893
Ans:
Knowing the heat exchanger eectiveness ", the average outlet temperatures of the uids
are readily computed based on the denition of ":
Th;o Th;i "C* Th;i Tc;i 9608C 0:8893 0:9383960 4808C 559:58C Ans:
Ans:
Tc;o Tc;i " Th;i Tc;i 4808C 0:8893960 4808C 906:98C
337
Discussion and Comments: This example is very similar to Example 5.1 or 5.2 except that
the regenerator consists of two disks. We have opted for the analysis that includes both
disks together as a single entity; we could have performed the analysis considering a
single disk and would have obtained the same results. Note that the design rotational
speed of the regenerator is suciently high, as reected by C*r 4:63 to yield the regenerator eectiveness " 0:8893, within 0.58% of the pure counterow recuperator
"cf 0:8945.
5.3
This method for determining regenerator performance is due to Hausen (1929, 1983).
He analyzed a xed-matrix regenerator starting with Eqs. (5.14) and (5.17)(5.19). He
dened spatial and time-independent nondimensional variables such that Eqs. (5.31)
(5.34) became parameter-free. Hausen introduced the following, now referred to as
Schumann dimensionless independent variables{:
j
hA x
C jL
j
hA
j
Cr j
j h or c
5:69
Here A represents the total heat transfer area of the matrix in the jth period, Mw represents its mass, and Cr; j Mw cw j Mw cw based on Eqs. (5.2) and (5.8). Substituting
these variables into Eqs. (5.14) and (5.17)(5.19) yield the following. For the hot-gas ow
period,
@Th
Tw Th
@ h
@Tw
Th Tw
@h
5:71
@Tw
Tc Tw
@c
5:72
Note that these equations are now parameter-free. The boundary and periodic-ow
conditions are still those of Eqs. (5.20)(5.23).
The independent variables and of Eq. (5.69) are represented irrespective of the
period as
hA
x bx / x
CL
hA
c /
Cr
5:73
Here b and c are the constants since h, A, C, L, and Cr all are constants due to the
assumptions listed in Section 5.1.1. For this reason, the variables and are interpreted
as the reduced length and reduced period variables, respectively.
{
If the dwell periods d;h and d;c are not negligible, h and c are expressed as
h A
x
h h h h d;h
L
Cr;h
hA
x
c c c c 1
L d;c
Cr;c
5:70
338
5:74
For overall regenerator performance, we are interested in determining the time-averaged uid outlet temperatures. These temperatures are obtained from their denitions
[Eq. (5.41)] after suitable modications based on Eq. (5.69), with replaced by h and c
for the hot- and cold-gas ow periods, respectively. These temperatures are expressed by
the regenerator eectiveness. Since we have considered only one matrix at a time in the
foregoing analysis, the hot and cold uid ows are intermittent. In an ideal steady-state
periodic condition, the actual heat transfer ( joules or Btu) during one hot- or cold-gas
ow period will be
Q Ch Ph Th;i Th;o Cc Pc Tc;o Tc;i
5:75
This means that the regenerator is in a cyclic equilibrium because the heat transferred to
the matrix during the hot period equals the heat transferred to the cold uid during the
cold period. The maximum possible heat transfer will be in a counterow regenerator of
innite surface area having the same uid ow rates and uid inlet temperatures. Thus
this maximum possible heat transfer is
Qmax CPmin Th;i Tc;i
5:76
Thus, the eectiveness for a xed-matrix regenerator (with two matrices) is dened as
follows:
"
5:77
Qmax;c Cc Pc Th;i Tc;i
5:78
Thus, the regenerator eectiveness during hot- and cold-gas ow periods are
"h
1 T*h;o
Qmax;h Ch Ph Th;i Tc;i
Th;i Tc;i
5:79
"c
T*c;o
Qmax;c Cc Pc Th;i Tc;i Th;i Tc;i
5:80
where the last equalities in Eqs. (5.79) and (5.80) are obtained using the denitions of Eq.
(5.28). Note that the eectiveness in Eqs. (5.79) and (5.80) is similar to the temperature
eectiveness for a recuperator [Eqs. (3.51) and (3.52)]. Razelos (1979) dened the overall
339
Q h Qc
2Q
"r
5:81
2Cc =Ch
"
1 Cc =Ch c
5:82
Comparing Eqs. (5.77) and (5.82), it can be shown for a two-matrix regenerator that
1 C*
5:83
"
"r
2C*
The independent dimensionless variables of Eqs. (5.69), collocated at x L, h Ph
or c Pc , become the dimensionless parameters as follows{:
j j L
hA
C j
j j Pj
hA
Pj
Cr j
j h or c
5:84
5:86
From the rst equation of Eq. (5.84), j j L; and from Eq. (5.73), L bL, we get
bL
5:87
Similarly, from the second equation of Eq. (5.84), j j Pj ; and from Eq. (5.73),
j Pj cPj . Hence, we get
cPh
or
cPc
5:88
Since b and c are constants in Eqs. (5.87) and (5.88), and are designated as reduced
length and reduced period, respectively, for the regenerator. The reduced length also
designates the dimensionless heat transfer or thermal size of the regenerator [see, e.g., Eq.
(5.107)], and this method is referred to as the - method. It has been used primarily for
the design of xed-matrix regenerators, but of course it can also be used for the design of
rotary regenerators. In the - method, several dierent designations are used to classify
regenerators, depending on the values of and . Such designations and their equivalent
dimensionless groups of the "-NTUo method are summarized in Table 5.2.
For the eectiveness of the most general unbalanced and unsymmetrical regenerator,
Razelos (1979) proposed an alternative set of four dimensionless groups, instead of those
of Eq. (5.86):
"; "r ; "h ; "c m ; m ;
; R*
{
5:89
If the dwell periods d;h and c;h are not negligible, h and c would be
h A
h h Ph h h Ph d;h
Cr;h
hA
c c Pc c c Pc d;c
Cr;c
5:85
340
- Method
"-NTUo Method
Balanced
Unbalanced
Symmetric
Unsymmetric
Symmetric and balanced
Unsymmetric but balanced
Long
h =h c =c or
1
h =h 6 c =c
h c or R* 1
h 6 c
h c , h c
h =h c =c
= > 5
C* 1
C* 6 1
hA* 1
hA* 6 1
hA* 1, C* 1
hA* 6 1, C* 1
C*r > 5
where m and m are the mean reduced length and mean reduced period, respectively.
They have been proposed by Hausen (1983) as the harmonic means in the following
sense:
1
1 1
1
m 2 h c
5:90
1
1
h c
m 2m h c
5:91
c =c
C P Mw;h cw Cc
c c
h =h Mw;c cw Ch Ph
Ch
5:92
c hAc Mw;h cw Pc
hAc
5:93
and
and R* are dened as
R*
In Eqs. (5.92) and (5.93), note that Mw;h cw =Ph Cr;h , Mw;c cw =Pc Cr;c , and
Cr;h Cr;c Cr , from Eq. (5.8). Razelos (1979) also showed that the inuence of R*
on "r is negligible for 1 R* 5:{ Thus,
"r m ; m ;
5:94
"r m ; m ;
"r m ; m ; 1=
5:95
341
ratio of the heat that the matrix (brick) actually absorbs or gives up during one period to
the heat it would have absorbed or given up if it had innite transverse thermal conductivity. The amount of heat that the matrix absorbs or gives up during one period is the
same as the amount of heat transferred from the hot gas to the matrix or the amount
given up from the matrix to cold gas during one period.
Ch Ph Th;i Th;o
Cc Pc Tc;o Tc;i
"
c "c
c
C*r
Mw cw h Th;i Tc;i
Mw cw c Th;i Tc;i
5:96
Thus it is clear that the higher the value of C*r , the lower is the utilization coecient,
indicating the lower amount of heat stored in the regenerator.
Before presenting the specic results, let us compare the dimensionless groups of the
- method with those of the "-NTUo method.
5.3.1
The functional relationships for these methods are given by Eqs. (5.52) and (5.86) or
(5.89). For comparison purposes, we consider Cc Cmin . The regenerator eectiveness "
is related to "r , "h , and "c as
8
1"r
>
>
>
>
2
<
" "h
>
>
>
>
:
"c
5:97
The independent variables of Eqs. (5.52) and (5.89) are related as follows for Cc Cmin :
C*
1
ntuh ntuc
NTUo
C*
C*r
hA*
1
m 1
5:98
5:99
m 1
2
m
5:100
1
R*
5:101
Similarly, the independent variables of Eqs. (5.86) and (5.47) are related as follows for
Cc Cmin :
342
hA
1
ntuh C * 1
NTUo
C h
hA*
hA
c
ntuc 1 hA*NTUo
C c
hA
ntuh
1
1
h
1
Ph
NTUo
hA*
C *r;h C*r
Cr h
hA
ntuc
1
1 hA*NTUo
c P c
C*r;c C*r
Cr c
h
5:102
5:103
5:104y
5:105
TABLE 5.3 General Functional Relationships and Basic Denitions of Dimensionless Groups for
e-NTUo and - Methods for Counterow Regenerators
- Methoda
"-NTUo Method
q "Cmin Th;i Tc;i
" NTUo ; C*; C*r ; hA*
Ch Th;i Th;o
Cc Tc;o Tc;i
"h
1
1
Cmin 1=hAh 1=hAc
"c
C*
Cmin
Cmax
"r
C*r
Cr
Cmin
1
"r
"
NTUo
hA*
1
m
Qh Qc
2Q
"
"r
2 "h "c
2
1 1
1
1
1
h c
2 h c m 2m h c
c =c
h =h
c
hAc
Cc
h
c
hA
h
Cr h
R*
hAh
Ch
hA
c
Cr c
h
343
TABLE 5.4 Relationship between Dimensionless Groups of e-NTUo and - Methods for
Cc Cmin a
"-NTUo
NTUo
hA*
h C* 1
m 1
c =c
1=h 1=c
4
C*
C*r
-
c =c
h =h
1
NTUo
hA*
c 1 hA*NTUo
m 1
c
c
2
m
h
1
1
NTUo
1
C*r
hA*
1
c
R* h
c
1
1 hA*NTUo
C*r
Rotary Regenerator
Fixed-Matrix Regenerator
NTUo
hc Ac
h h Ah
Cc hh Ah hc Ac
hc Ac
hh Ph
Cc hh Ph hc Pc
C*
Cc
Ch
Cc P c
Ch Ph
C*r
Mw cw N
Cc
Mw cw
Cc Pc
hA*
hc Ac
hh Ah
hc Pc
hh Ph
1
m
C h Cc
4
1
m
Mw cw N
2
Cc
Ch
Cc P c
Ch Ph
R*
hh Ah
hc Ac
hh Ph
hc Pc
1
1
hh Ah hc Ac
1
1
hh Ah hc Ac
Ch Ph Cc Pc
1
1
4A
hh Ph hc Pc
Mw cw
1
1
2A
hh Ph hc Pc
344
5.3.2
Hausen (1929) obtained the uid temperature distributions for a balanced and symmetric
counterow regenerator [C* 1 and hA* 1 in the form of innite series by using
the method of separation of variables for solving sets of governing equations (5.71)
(5.72) with the boundary and periodic-ow conditions of Eqs. (5.20)(5.23).
Subsequently, integrating exit temperatures over the entire duration of the period,
average exit temperatures and ultimately the eectiveness can be determined. In this
case,
" ;
5:106
where h c and h c . His results are shown in Fig. 5.5. From the
relationships of Eqs. (5.102)(5.105), we have
2NTUo
2NTUo
C*r
5:107
Using the results of Table 5.1 and Eq. (5.107), the eectiveness of Fig. 5.5 can be
determined for the balanced and symmetric C* 1, hA* 1 regenerators.
For a general (balanced/unbalanced and symmetric/asymmetric) counterow regenerator, Dragutinovic and Baclic (1998) have presented exact analytical relations and
computational algorithms for evaluation of temperature distributions and regenerator
eectiveness in terms of four dimensionless groups; they refer to them as the utilization
factor U1 , reduced length 1 , unbalance factor , and asymmetry factor , dened as
follows.
CP1
hA
U
CP1
hA
C
U1 1
1
1
1
1 Mw cw
U2 CP2
2
C 1
C 1 hA 2
5:108
1.0
0.8
=0
20
10
30
0.6
40
50
0.4
= reduced length
= reduced period
0.2
0.0
10
20
30
40
50
FIGURE 5.5 Eectiveness chart for a balanced and symmetric counterow regenerator (From
Hausen, 1983).
345
1.0
w=0
Counterflo
w=
Counterflo
30
=
40
20
0.8
=1
0.6
=0
0.4
= reduced length
= reduced period
0.2
0.0
0
=5
10
15
20
25
30
35
40
45
50
FIGURE 5.6 Eectiveness chart for a balanced and symmetric parallelow regenerator (From
Hausen, 1983).
They provided comprehensive tabular results for the counterow regenerator eectiveness for the following ranges of the dimensionless groups: 1 1 to 1; U1 0 to 2,
1 0 to 1, and 0 to 1.
5.3.3
Hausen (1983) also obtained the solution for a balanced and symmetric parallelow
regenerator as shown in Fig. 5.6. The oscillations in " above and below " 0:5 are clearly
observed in this gure (see Fig. 5.4 for similar trends, although clearly not seen due to the
shortened abscissa, NTUo 10 ! 20). Comparing the results for the eectiveness
of a counterow regenerator (Fig. 5.5) and parallelow (Fig. 5.6), it becomes obvious
that parallelow provides signicantly smaller eectiveness for the same regenerator
length and period duration. The performance of a regenerator reduces to the performance of a parallelow recuperator in the limit of an innitely short period ( 0).
Example 5.4 A rotary regenerator has a circular matrix of 2 m diameter and 1 m depth
rotating at 1 rpm. The metal matrix weighs 3500 kg and has a total surface area of
7900 m2. Its specic heat is 0.43 kJ/kg K. Flue gas at 5008C ows through the matrix
at the rate of 52 kg/s and air at 208C ows at the same ow rate in a countercurrent
fashion. Determine the air and ue gas outlet temperatures assuming that the isobaric
specic heat is 1.05 kJ/kg K and the heat transfer coecient is 130 W/m2 K for both the
ue gas and air. Idealize the regenerator as symmetric and balanced.
SOLUTION
Problem Data and Schematic: The heat transfer coecients, uid ow rates, and inlet
temperatures are provided in Fig. E5.4 for both hot and cold streams. In addition, the
346
1 rpm
Hot gas
Th,i = 500C
mh = 52 kg/s
Matrix
Mw = 3500 kg
1m
cw = 0.43 kJ/kg K
Ah + Ac = 7900 m2
Gas/air
2m
Cold air
Tc,i = 20C
hh = hc = 130 W/m 2 K
mc = 52 kg/s
regenerator dimensions and rotational speed are specied. Also specied are the physical
properties of both hot and cold uid streams as well as the matrix material.
Determine: The cold- and hot-uid outlet temperatures.
Assumptions: The assumptions of Section 5.1.1 are invoked here.
Analysis: For a symmetric and balanced regenerator, the reduced length and the
reduced period are equal on the hot and cold sides:
h c m
hA
ntu
m_ cp
h c m
hAP
Mw c w
Note that in Eq. (2), P is the total period if Mw is the total matrix mass of the regenerator;
otherwise, P Pc or Ph when Mw Mw;c or Mw;h .
The heat transfer coecients, ow rates, and specic heats are equal on the hot and
cold uid sides. Based on Eqs. (1) and (2), for a symmetric and balanced regenerator
(h c , h c ), this implies that the ow split between the hot and cold sides must be
50 : 50 and the heat transfer surface area must be equal. Now let us rst determine the
reduced length and the reduced period from using Eqs. (1) and (2).
hA
0:130 kW=m2 K 7900=2 m2
9:4
hAP
0:130 kW=m2 K 3950 m2 60 s
20:47
Mw cw
3500 kg 0:43 kJ=kg K
347
With 9:4 and 20:47, the regenerator eectiveness is obtained as " 0:42
from Fig. 5.5. Knowing " together with the inlet temperatures Th;i and Tc;i of the hot
and cold uid streams, the cold-uid outlet temperature Tc;o and the hot-uid temperature outlet Th;o are determined from Eqs. (5.79) and (5.80).
Tc;o Tc;i " Th;i Tc;i 208C 0:42500 208C 221:68C
Th;o Th;i " Th;i Tc;i 5008C 0:42500 208C 298:48C
Ans:
Ans:
Cmin
1
Cmax
1
rev=s 25:08 kW=K
Cr Mw cw N 3500 kg 0:43 kJ=kg K 60
C*r
Cr
25:08 kW=K
0:459
Cmin
54:6 kW=K
1
1
1
1
4:702
Cmin 1=hAh 1=hAc 54:6 kW=K 1=513:5 1=513:5 K=kW
Now the regenerator eectiveness " could have been determined by using Fig. 5.3 if the
graph for C*r 0:459 or a close value (0.50) would have been available. Hence, we will
resort to Eqs. (5.62) and (5.63) for the determination of ". Using Eq. (5.63) for C * 1,
"cf
NTUo
4:702
0:8246
1 NTUo 1 4:702
" 0:8246 1
1
9 0:4591:93
0:413
The outlet temperatures, using the denition of " from Eq. (5.45), are
Tc;o Tc;i " Th;i Tc;i 208C 0:413500 208C 218:28C
Th;o Th;i " Th;i Tc;i 5008C 0:413500 208C 301:88C
Ans:
Ans:
Discussion and Comments: First it is shown how to solve the problem by the - method.
Since Fig. 5.5 is valid only for the balanced and symmetric regenerator, should be
evaluated carefully using Eq. (2), where A represents the heat transfer surface area on
only one uid side (hot or cold), not the total surface area; however, P=Mw can be
evaluated either for the total matrix or for the hot (or cold) gas side. The rest of the
procedure for determining " is straightforward.
348
The same problem is solved by the "-NTUo method for illustration. The regenerator
eectiveness is determined graphically for the - method and by an empirical formula
for the "-NTUo method; they are found to be close enough within the reading accuracy
of the graphical results. Thus, as expected, the results should be identical regardless of
which method is used for the analysis.
5.4
Longitudinal heat conduction in the wall may not be negligible, particularly for a higheectiveness regenerator having a short ow length L. Longitudinal wall heat conduction
reduces the exchanger eectiveness as discussed in Section 4.1 with Fig. 4.1, hence it is
important that its inuence on " be determined quantitatively. It should be emphasized
that one end of the matrix (regenerator) in a counterow regenerator is always hotter
than the other end during both hot and cold blow periods. Hence, longitudinal heat
conduction in the matrix wall occurs in the same direction through both periods.
In Section 4.1, we used a heuristic approach to derive the dimensionless groups
associated with the longitudinal wall heat conduction eect; here we use a more rigorous
approach. The basic dierential energy balance equations, Eqs. (5.14) and (5.18) for the
hot and cold uid sides, do not change, but those for the wall change. For nite axial heat
conduction in the wall during the hot-gas ow period, the model is modied by adding
the corresponding heat conduction terms as shown in Fig. 5.7 (compare with Fig. 5.2b).
Applying the energy balance to the dierential element of the wall in Fig. 5.7, we get
kw Ak;h
@Tw
@Tw @ 2 Tw
Ah dx
dx @Tw
kw Ak;h
dx
h
T
C
T
h
h
w
r;h
@x
@x
L
L @h
@x2
5:109
Upon simplication,
kw Ak;h L @ 2 Tw
@Tw hAh
Th Tw
@h
@x2
Cr;h
Cr;h
5:110
( )
dqh = hh
kw Ak, h Tw
x
Ah dx
(Th Tw)
L
( )
Tw
Cr,h dx
L h
kw Ak, h
Tw 2Tw
dx
+
x
x2
dx
FIGURE 5.7 Energy rate terms associated with the element dx of the matrix during the hot-gas
ow period; longitudinal heat conduction has been included (From Shah, 1981).
349
Using the previous denitions of X*, h*, Th*, Tw*, ntuh , and C*r;h [Eqs. (5.25)(5.30)],
Eq. (5.110) is made dimensionless:
@Tw* ntuh
@ 2 Tw*
Th* Tw* h
@h*
C *r;h
C *r;h @X *2
5:111
where
h
kw Ak;h
LCh
5:112
Similarly, the governing dierential equation for the matrix wall temperature during the
cold-gas ow period is
@Tw* ntuc
@ 2 Tw*
Tw* Tc* c
@c*
C*r;c
C*r;c @X *2
5:113
where
c
kw Ak;c
LCc
5:114
Since Eqs. (5.111) and (5.113) are second-order partial dierential equations with
respect to X*, we need to dene four boundary conditions for the matrix wall temperatures in order to get a particular solution: two during the hot-gas ow period and two
during the cold-gas ow period. The realistic boundary conditions are adiabatic boundary conditions for each period at X * 0 and 1:
8
@Tw*
>
>
>
0
>
< @X * X 1
@Tw*
@X * X 0 >
>
@Tw*
>
>
0
:
@X X 1
for 0 h* 1
5:115
for 0 c* 1
5:116
Thus, the inclusion of the eect of longitudinal heat conduction adds two dimensionless
groups, h and c , on which the exchanger eectiveness " would depend. Bahnke and
Howard (1964) suggested an alternative set of two dimensionless groups:
kw Ak;t
LCmin
Ak*
5:117
5:118
Note that and Ak* are related to h and c as follows for Cc Cmin :
c
h
C*
Ak*
Ak;c
C* c
Ak;h
h
5:119
350
This choice of dimensionless groups oers the advantage that the resulting " is not
aected signicantly by Ak* for 0:25 Ak* 1 (Bahnke and Howard, 1964; Skiepko
and Shah, 1994). Thus, the eect of longitudinal heat conduction in the wall is taken
into account by and is added to the functional relationship for " of Eq. (5.61):
" NTUo ; C*; C*r ;
5:120
In order to obtain an exact solution for this problem, Eqs. (5.31), (5.33), (5.111), and
(5.113) need to be solved using the boundary conditions and periodic equilibrium conditions of Eqs. (5.35) through (5.38), (5.115), and (5.116). The rst closed-form analytical
solution to these equations was obtained by Skiepko (1988). Bahnke and Howard
(1964) obtained a numerical solution by a nite dierence method. They determined
the exchanger eectiveness and longitudinal conduction eect over the following
ranges of dimensionless parameters: 1 NTUo 100, 0:9 C* 1, 1 C*r 1,
0:01 0:32, 0:25 hA* 1, 0:25 Ak* 1. The ineectiveness (1 ") as a function of NTUo and is shown in Fig. 5.8 for C * 0:95 and C*r > 5. Similar results for
C* 1 and C*r > 5 are already shown in Fig. 4.2.
Bahnke and Howards results are correlated by Shah (1975) as follows{:
!
"
#
1
C
5:121
1
" "cf 1
2 C*
9C*r 1:93
where
C
and
1
1
1 NTUo 1 =1 NTUo 1 NTUo
NTUo
1 NTUo
NTUo
1 NTUo
1=2
(
tanh
NTUo
NTUo =1 NTUo 1=2
5:122
)
5:123a
1=2
for NTUo 3
5:123b
351
10
8
6
= 0.008
Ineffectiveness, (1 ) (%)
5
4
3
0.04
1.0
0.8
0.6
0.5
10
=0
20
30
0.004
0.02
0.01
40 50 60
80 100
200
300
500
NTU or NTUo
FIGURE 5.8 Inuence of longitudinal wall heat conduction on the performance of storage and
direct-transfer counterow exchangers; C* 0:95. (From Kays and London, 1998.)
This procedure yields ", which is accurate within 1% for 1 NTUo 20 for C*r 1
when compared to the results of Bahnke and Howard.
A careful review of Figs 5.8 and 4.2 and Eq. (5.121) reveals that longitudinal heat
conduction in the wall reduces the regenerator eectiveness. Thus, similar to a recuperator, due to longitudinal heat conduction in the wall, the regenerator eectiveness "
decreases with increasing values of NTU, C*, and , and the decrease in " is largest for
C* 1. However, the eect of increasing C*r when NTU, C*, and are kept constant is
complicated. For given values of NTUo , C*, and , increasing C*r increases "=" at small
values of NTUo for 0:32; at intermediate values of NTUo ( 4 to 5), increasing C*r
from 1 to 5 (and higher) rst increases and then decreases "="; at large values of NTUo
( 9), increasing C*r from 1 to 10 decreases "=". Bahnke and Howards results show
that Kays and Londons approximation (1998)
"
"
5:124
is a very good engineering approximation for NTUo > 10 and < 0:1.
Longitudinal heat conduction can have a serious impact on the regenerator eectiveness or NTU for an ultra high-eectiveness regenerator. For example, a Stirling engine
regenerator may require 350 ideal NTU to get 200 usable NTU due to longitudinal heat
conduction. As a result, such regenerators may require a stack of high thermal conductivity (copper or aluminum) perforated plates (Venkatarathnam, 1996) or wire screens,
alternating with low thermal conductivity spacers made up of plastic, stainless steel,
352
and so on. Such a design would signicantly reduce longitudinal heat conduction or the
stack conduction.
No detailed temperature distributions for uids and wall were obtained by either
Lambertson (1958) or Bahnke and Howard (1964). Mondt (1964) obtained these temperature distributions by solving the dierential equations numerically for some values of
the associated dimensionless groups. Illustrative results are shown in Figs. 5.95.11.
In Fig. 5.9, the matrix wall temperatures Tw* at x 0, x L=2, and x L are shown
as functions of a dimensionless time for the 0 case. Also shown are the hot- and coldgas inlet and outlet temperatures. Experimental points shown for T *c;o are in good agreement with the theoretical predictions. The wall temperatures are linear with time except
for the sections of hot- and cold-uid inlets.
In Fig. 5.10, hot-gas and matrix temperatures are shown as a function of the ow
length X* for h 0; Ph =2, and Ph . Here, again, these temperature distributions are linear
except for the regenerator ends.
In Fig. 5.11, the matrix wall temperatures are shown at switching time Ph and Pc .
The reduction in matrix wall temperature gradients at X * 0 due to longitudinal heat
conduction is evident. Note that the time average T *c;o is reduced, which in turn indicates
that the exchanger eectiveness " is reduced due to longitudinal wall heat conduction, as
expected.
Skiepko (1988) presented three-dimensional temperature charts demonstrating how
the longitudinal matrix heat conduction aects the matrix as well as the gas temperature
distributions, shown as dependent on coordinate x and time .
Example 5.5 Determine the reduction in the regenerator eectiveness of Example 5.3
due to longitudinal wall heat conduction given that the thermal conductivity of the
T *h,i
T *w,i
1.0
at x = 0
T *c,o
T *w,m
T*
0.5
at x = L/2
Ph
Pc
T *h,o
T *w,o
at x = L
0.0
0.0
0.5 _
hA/Cr
1.0
T *c,j
FIGURE 5.9 Cyclic temperature uctuations in the matrix at the entrance, midway, and exit of a
regenerator. (From Mondt, 1964.)
353
1.0
T *h
T *w
Ph
T*
0.5
Ph/2
Hot fluid
Matrix
0.0
0.0
h = 0
0.5
X*
1.0
FIGURE 5.10 Fluid and matrix wall temperature excursion during hot-gas ow period. (From
Mondt, 1964.)
1.0
1/C *r
T *h
T*
0.5
Solid
T *w,m
Fluid
T *c
0.0
0.0
0.5
X*
1.0
ceramic matrix is 0.69 W/m K and its porosity is 0.7. Use all other pertinent information
from Example 5.3.
SOLUTION
Problem Data and Schematic: The heat transfer coecients, uid ow rates, and inlet
temperatures are provided in Fig. E5.5 for both hot and cold streams. In addition, the
354
15 rpm
Seal coverage
7.%
Hot gas
Th,i = 960C
0.076 m
0.0715 m
Matrix
Mw = 34.93 kg
cw = 1.130 kJ/kg K
= 5250 m2/m3
= 0.76, kw = 0.69 W/m K
Gas/air
0.683 m
hh = 240.5 W/m 2 K
Cold air
hc = 220.5 W/m 2 K
Tc,i = 480C
m
c = 2.029 kg/s (2 disks)
FIGURE E5.5
regenerator dimensions and rotational speed are specied. Also specied are the physical
properties of both hot- and cold-uid streams as well as the matrix material.
Determine: The regenerator eectiveness and outlet temperatures of both hot and cold
uids.
Assumptions: The assumptions of Section 5.1.1 are invoked here.
Analysis: To assess the impact of the longitudinal wall heat conduction on the regenerator eectiveness, we must rst compute the total conduction area Ak;t as the dierence
between the eective frontal area and the free-ow area. Thus,
Ak;t Afr Ao Afr 1 0:673 m2 1 0:70 0:2019 m2
Introducing this value of Ak;t together with the value of Cmin 2130 W/K computed
in Example 5.3, we obtain from Eq. (5.117) the value of the longitudinal wall conduction
parameter as
9:133 104
LCmin
0:0715 m 2130 W=K
Knowing and NTUo 6:819 from Example 5.3, the dimensionless parameter ,
dened in Eq. (5.123b) for NTUo 3, can be determined as
NTUo
1 NTUo
1=2
1=2
0:0787
Knowing , NTUo , and , the parameter C , dened in Eq. (5.122), can be determined as
355
1
1
1 NTUo 1 =1 NTUo 1 NTUo
1
1
1 6:8191 9:133 104 0:0787=1 9:133 104 6:819 1 6:819
6:860 104
From Example 5.3 we have "cf 0:8945, C*r 4:63, and C* 0:9383. Hence, the
value of the regenerator eectiveness with longitudinal conduction from Eq. (5.121) is
C
2 C*
1
6:860 104
0:8945 1
1
2 0:9383
9 4:631:93
" "cf 1
1
9C*r 1:93
1
0:8888
On comparing this value with the value of " 0:8893 calculated in Example 5.3
without longitudinal heat conduction, we notice that longitudinal heat conduction
accounts for only a 0.06% decrease in the regenerator eectiveness.
Discussion and Comments: This example was intended to demonstrate how to evaluate
the eect of longitudinal heat conduction on the regenerator eectiveness. The result
shows that this eect is negligible for this case primarily because of the very low value of
. However, this may not be the case for matrices made up of metals, thick walls, or short
regenerator ow lengths. The recommended practice is to include the eect of longitudinal wall heat conduction in a computer program and thus always consider it, no
matter of how small or large.
5.5
One of the idealizations (assumption 9 in Section 5.1.1) made in the foregoing regenerator design theory is that the wall thermal resistance is zero. This assumption is invoked
in deriving Eqs. (5.17) and (5.19). The temperature gradient in the wall thickness (
w =2)
direction in Fig. 5.2b is zero. It was also shown that the wall thermal resistance in Uo A of
Eq. (5.53) is zero. The zero wall thermal resistance represents a good approximation for
metal matrices having high thermal conductivity. For those matrices having thick walls
or low thermal conductivity, such as ceramic matrices for xed-matrix regenerators, the
wall resistance may not be negligible.
5.5.1
Simplied Theory
A simplied method is now outlined to include the inuence of wall thermal resistance
on regenerator eectiveness. An additional but essential assumption made for the
analysis{ is: The temperatures of hot and cold gases and the wall at any cross section
{
In this section, the analysis is made for nite thermal resistance in the wall thickness direction and innite
thermal resistance in the longitudinal direction (the zero-longitudinal heat conduction case).
356
in the regenerator are linear with time, and the numerical value of this time derivative of
temperature at any point in the wall is the same. For the wall temperature, this means
@Tw @Tw;o @Tw;m
constant
@
@
@
5:125
Here Tw;o and Tw;m are the surface wall temperature and mean wall temperature at a
given instant in time. This linear temperaturetime relationship represents a good
approximation to the actual wall temperature prole in the greater part of either hotor cold-gas ow period along most of the regenerator length, as shown in Fig. 5.9.
Now let us consider a dierential element of the regenerator matrix wall as shown
in Fig. 5.12 during the hot-gas ow period. Here Aw represents the conduction area for
the wall for heat conduction in the y direction. For continuous ow passages, Aw A
(the convective heat transfer surface area), and Aw A for noncontinuous ow passages.
The energy balance on the element of Fig. 5.12b yields the well-known one-dimensional
transient conduction equation valid at each x coordinate.
w
@ 2 Tw @Tw
@
@y2
5:126
where w kw =w cw is the thermal diusivity of the matrix material. The appropriate
boundary conditions are (see Fig. 5.12)
Tw Tw;o
at y 0 and w
5:127
dy
Section x
dQ
dA
dx
Tw,o
Tw,o
Tw,o
Tw,m
x
Tw
Th,m
Tw,o
Th,m
Tw
y
w
(a)
Tw
kw Aw
y
(c)
T
(wAwdycw) w
Tw 2Tw
kw Aw
+
dy
y
y2
dy
(b)
FIGURE 5.12 (a) Matrix wall of thickness
w , (b) energy transfer terms across the dierential wall
element; (c) parabolic temperature distribution in the wall at a given x (From Shah, 1981).
357
Now let us rst obtain the temperature distribution in the wall by a double integration
of Eq. (5.126) using the boundary conditions of Eq. (5.127) and the assumption of Eq.
(5.125). We get
1 @Tw
Tw Tw;o
w yy
5:128
2w @
Hence, Tw and Tw;o of Eq. (5.128) must be time-averaged temperatures, and they are
designated by a bar in Eq. (5.128) to denote them as time-averaged quantities. The
temperature prole of Eq. (5.128) is parabolic at each x as shown in Fig. 5.12c.
The mean wall temperature Tw:;m is obtained by integrating Eq. (5.128) with respect to
y across the wall thickness as
2 @Tw
Tw;m Tw;o w
12w @
5:129
Thus, the dierences between the mean wall temperature and the surface wall temperature during hot- and cold-gas ow periods are
2
@Tw
5:130
Tw;m Tw;o h w
12w @ h
2
@Tw
Tw;m Tw;o c w
5:131
12w @ c
The mean wall temperatures during the hot- and cold-gas ow periods must by equal for
the idealized true periodic ow conditions (i.e., Tw;m;h Tw;m;c ). Subtracting Eq. (5.131)
from Eq. (5.130), we get
Tw;o
h Tw;o
c
2
w
12w
@Tw
@Tw
@ c
@ h
5:132
Tw;o
1
1 dQ
w;o
w
T
h
c
6kw Ph Pc dA
5:134
358
Now the rate equations during hot- and cold-gas ow periods are
dQh hh dATh Tw;o h Ph
5:135
where Th and Tc are the hot- and cold-uid time-averaged temperatures at a section x
during the hot- and cold-gas ow periods, respectively. Substituting Tw;o h and Tw;o c
from these equations into Eq. (5.134) and noting that dQh dQc dQ, rearrangement
yields
1
1
1
1
1
dQ
5:136
dATh Tc
hh Ph hc Pc 6kw Ph Pc
Since Q Qh Qc represents total heat transfer per cycle in time Ph Pc , the average
heat transfer rate during one cycle is
q
Q
Ph Pc
5:137
dq
dQ
Ph Pc
5:138
and hence
5:139
1
1
1
1
1
Ph Pc
Uo
hh Ph hc Pc 6kw Ph Pc
5:140
where
5:141
where
Tlm
Th;i Tc;o Th;o Tc;i
lnTh;i Tc;o =Th;o Tc;i
5:142
The bar on T represents the corresponding (hot or cold)-period time-averaged temperatures. A in Eq. (5.141) represents the total surface area (Ah Ac ), in contrast to either Ah
or Ac for a recuperator. This is because dq in Eq. (5.138) is over one cycle.
Now dividing both sides of Eq. (5.140) by the total surface area A and introducing the
denitions of Ah and Ac from Eq. (5.9), we get
1
1
1
1
1
5:143
where the relationships Aw;h Ah and Aw;c Ac , as noted just before Eq. (5.126), are
used in Eq. (5.143). Now this equation is valid for both the rotary and xed-matrix
regenerators as long as proper values of Ah and Ac from Eqs. (5.9) and (5.10) are
359
used. From a comparison of this equation with that for a recuperator, the equivalent wall
thermal resistance for the regenerator is
Rw
w
1
1
5:144
Now consider a rotary or xed-matrix regenerator with two matrices and the special case
of a 50 : 50 split of ow areas in a rotary regenerator or Ph Pc in a xed-matrix
regenerator. For this regenerator, Aw;h Aw;c Aw . Thus, for this case, Eq. (5.144)
reduces to
Rw
w
3kw Aw
5:145
w
kw A w
5:146
From a comparison of Eqs. (5.145) and (5.146), it is evident that the wall thermal
resistance of a regenerator is one-third that of an equivalent recuperator.
Alternatively, the regenerator is equivalent to a recuperator of one-third wall thickness.
The qualitative reason for the lower wall thermal resistance is that the thermal energy is
not transferred through the wall in a regenerator; it is stored and rejected.
One of the basic assumptions made in the foregoing analysis is that the temperatures
Th , Tc , and Tw are all linear with time. Since this is not true at the switching moment and
near the regenerator inlet and outlet, Hausen (1983) in 1942 suggested modifying Rw of
Eq. (5.145) by a factor *, so that
1
1
1
Rw *
Uo A hh Ah hc Ac
5:147
1
<
60w Ph Pc
*
1=2
>
2
1
1
>
>
: 2:142 0:3 w
2w Ph Pc
w2
1
1
10
2w Ph Pc
2
1
1
10
for w
2w Ph Pc
for
5:148
Generally, the wall thermal resistance is much smaller than the hot- or cold-gas lm
resistance. Hence the wall thermal resistance formula of Eq. (5.144) is adequate for rating
and sizing problems of most applications. This correction factor * is not adequate to
accurately determine the temperature distribution in the wall immediately after the
changeover and near the regenerator ends. In those cases, a numerical method of
Heggs and Carpenter (1979) is suggested to take into account the wall thermal resistance
eect. Heggs and Carpenter designate this eect the intraconduction eect.
360
5.6
In both rotary and xed-matrix regenerators, ow leakages from cold to hot gas streams,
and vice versa, occur due to pressure dierences and carryover (as a result of the matrix
rotation or valve switching). Pressure leakage is dened as any leakage due to the pressure dierence between the cold and hot gases in a regenerator. In a rotary regenerator,
this leakage occurs at face seals, through the pores in the matrix, or through the circumferential gap between the disk and the housing. In a xed-matrix regenerator, this leakage
occurs at valves as well as to and from the cracks in the housing (which is usually made of
bricks) in high-temperature applications. In such an application, the operating pressure
on the hot-gas matrix is below ambient pressure (i.e., under vacuum) to avoid the
leakage of poisonous ue gas from the matrix to the surrounding room; and in this
case, the pressure leakage is from outside to the matrix through cracks in the walls.
For further details, refer to Shah and Skiepko (1998).
Carryover leakage in a rotary regenerator is dened as the transport of the trapped gas
in the voids of the matrix to the other uid side just after the switching from uid 1 stream
to uid 2 stream, and vice versa, due to matrix rotation. The carryover leakage is
unavoidable and its mass ow rate is a function of total void volume, rotor speed, and
uid density. In a xed-matrix regenerator, the carryover leakage in one cycle from uid
1 to uid 2 is the gas in the void volume of one matrix of a two-matrix regenerator and
associated header volume.
In the following, a model is presented to take into account the pressure and carryover
leakages in a rotary regenerator (Shah and Skiepko, 1997). Refer to Shah and Skiepko
(1998) for modeling leakages for a xed-matrix regenerator.
5.6.1
In a rotary regenerator, pressure leakage occurs at face seals separating the hot and cold
gas sides, through the pores in the matrix itself (which is neglected in the present analysis), or through the circumferential gap between the disk (rotor) and housing. The circumferential leakage is sometimes referred to as the side bypass leakage, and the pressure
leakage at the face seals is referred to as the cross bypass leakage. The inuence of the
total pressure leakage on the regenerator eectiveness is modeled and summarized next.
However, this net ow leakage also represents a loss in the heated cold-gas ow rate to
the process or the thermodynamic system. This loss may also have a substantial inuence
on the process or cycle eciency. For example, in a gas turbine power plant, a 6% cold
high-pressure air leak to the exhaust gas stream in a regenerator means a 6% reduction in
the net power outputa signicant penalty! However, unless there are signicantly large
leakages, the regenerative cycle has substantially higher thermal eciency than a simple
cycle for the gas turbine power plant.
A Ljungstrom rotary regenerator with radial, peripheral, and axial seals is shown in
Fig. 5.13. Radial seals prevent leakage of high-pressure gas to low-pressure gas at the
inlet and outlet faces of the regenerator. The axial seals prevent the leakage of highpressure gas to low-pressure gas in the circumferential direction in the gap between the
housing and rotor. The peripheral seals prevent the ow bypass from the regenerator
inlet to regenerator outlet side on each gas side (in the axial direction) of the regenerator
in the gap between the housing and rotor if the axial seals are perfect (zero leakage); the
leakage between the axial seals (since there are usually 12 or fewer axial seals along the
perimeter of the rotor) also eventually has to go through the peripheral seals. Note that in
361
FIGURE 5.13 A Ljungstrom rotary regenerator with radial, peripheral, and axial seals. (From
Shah and Skiepko, 1997.)
some regenerator applications, such as for a gas turbine, the axial seals are not used; only
peripheral seals are used to prevent/minimize the ow leakage bypasses in axial and
circumferential directions.
A model for regenerator thermal design consists of an internal (or ideal) regenerator
(represented by the rotating disk with no leakage streams within its boundary as marked
by dashed lines in Fig. 5.14), and an actual regenerator (that is considered as the internal
regenerator with its housing, and radial, peripheral, and axial seals that include all
leakages and bypass ows). The concept of these two regenerators is used for the thermal
design procedure. It is easily seen from Fig. 5.14 that there are no leakages within the
362
FIGURE 5.14 Regenerator gas ow network with all categories of leakages. (From Shah and
Skiepko, 1997.)
region bounded by dashed lines (i.e., the internal regenerator). In Fig. 5.14, all leakage
streams are shown by medium-thickness lines with arrows indicating the ow directions;
the exception is the axial leakage stream, which is shown as a large-dashed line through
the internal regenerator but which in reality ows circumferentially through the gap
between the rotor and housing. The ow rates associated with each stream are identied
as follows: The rst subscript, h or c, designates the hot or cold gas stream. The second
subscripts r, p, ax, and co designate the radial, peripheral, axial, and carryover leakage
streams. The superscripts (H) and (C) denote the hot end (hot-gas inlet side) and cold
end (cold-gas inlet side) of the regenerator disk. The temperatures associated with
each stream are identied as follows: The rst subscript, h or c, designates the hot
or cold stream. The second subscripts, i and o, refer to the inlet and outlet gas sides of
the regenerator; those with a prime mark denote inlet and outlet from the internal
regenerator, and those without a prime mark denote inlet and outlet from the actual
regenerator.
All radial, peripheral, and axial seal leakages are modeled as a ow through an orice
as follows:
m_ seal Cd Ao;s Y
p
2 p
5:149
363
Here we have assumed that the thermodynamic process involved is isentropic, the pressure leakages through seals can be evaluated by applying Bernoulli and continuity equations, and the thermodynamic relations are used for gas ow through the leakage area. In
Eq. (5.149), Cd is the coecient of discharge, assumed to be 0.80 (Shah and Skiepko,
1997); Ao;s is the seal gap ow area; Y is the expansion factor for the compressibility
eect, assumed to be 1; is the gas density before the seal; and p is the pressure drop in
the seal. The leakage ow rate terms, p, and inlet density for each leakage stream are
summarized in Table 5.6.
The carryover leakage from hot to cold gas in one rotation will correspond to the hot
void volume times the average density of the gas:
m_ h;co Afr L
h N
5:150
where h is determined at arithmetic averages of inlet and outlet gas temperature and
pressure. A similar expression can be written for the carryover leakage m_ c;co from the cold
side to the hot side. In some regenerator applications, heat transfer surface of dierent
porosity
is used along the ow length, in which case L in Eq. (5.150) is replaced by
P
i Li i . In the Ljungstrom regenerator and some other regenerators, the matrix ow
length is L, and the radial ribs (separating heat transfer baskets) are slightly higher by
L=2 on the hot and cold ends on which the radial seals scrub (see Fig. 5.13). Hence, the
gas trapped in this header volume is also carried over to the other side. Hence, the
carryover leakage term in general for each uid side is given by
"
m_ j;co Afr N
j
n
X
#
Li i L
j h or c
5:151
i1
TABLE 5.6 Rotary Regenerator Pressure Leakage and Carryover Flow Rates, Pressure Drops, and
Inlet Density for the Orice Analysis
Symbols
Pressure
Drop
Density
m_ H
c;r
pc;o ph;i
c;o
m_ C
c;r
pc;i ph;o
c;i
m_ H
c; p
ph;i ph*
ph* ph;i
pc;o p*c
p*c pc;o
m_ h; p
ph* ph;o
h*
m_ C
c; p
m_ c;ax
pc;i p*c
p*c ph*
c;i
*c
Leakage Term
m_ h; p
m_ h;co
m_ c;co
c
364
Generally, the carryover leakage is very small and its inuence on the regenerator eectiveness is also negligibly small for most rotary regenerator applications, except for the
cryogenics and Stirling engine regenerators, which have eectively a very high value of
switching frequency.
The leakage modeling of an actual regenerator involves the determination of all nine
leakage ow rates mentioned in Table 5.6, plus mass ow rates, pressures, and temperatures at the mixing points A, B, . . . , F in Fig. 5.14. To be more specic, four additional
mass ow rate terms are hot- and cold-gas mass ow rates through the regenerator disk
m_ h ; m_ c , and hot- and cold-gas mass ow rates at the outlet of the actual regenerator
m_ h;o ; m_ c;o ; four pressures and temperatures at points B and E of Fig. 5.14: ph*, pc*, Th*,
Tc* (note these are dimensional values); and three temperatures from the actual regen0
erator to the outside or to an internal regenerator: Th;o , Tc;o , and Th;i
. Thus it represents a
total of 20 unknowns for the actual regenerator. Determination of various leakage ow
rates of Table 5.6 represents nine equations [seven of Eq. (5.149) and two of Eq. (5.150)].
The mass ow rate balances at six junction points A, B, . . . , F in Fig. 5.14 represent six
equations presented in Table 5.7. The energy balances at ve points, A, B, C, E, and F in
Fig. 5.14, provide an additional ve equations, as summarized in Table 5.8. Thus a total
of 20 linear/nonlinear equations are available to determine 20 unknowns mentioned
above. If m_ C
c;r leaks from point C to D in Fig. 5.14, one additional energy balance
equation at point D is available in Table 5.8, and the additional unknown is Tc;i0 .
To investigate the inuence of leakage distribution on regenerator heat transfer performance, we also need to consider heat transfer and pressure drop modeling of the
internal regenerator. Heat transfer analysis includes applying "-NTUo theory, including
longitudinal and transverse wall heat conduction eects, discussed in Sections 5.2, 5.4,
and 5.5 for the internal regenerator. Here the appropriate mass ow rates and inlet
temperatures are determined for the internal regenerator from an analysis of the actual
regenerator as mentioned above. The internal regenerator heat transfer analysis will then
yield temperatures and the internal regenerator eectiveness "i;b . The pressure drop
analysis for the hot- and cold-gas sides of the internal regenerator is also straightforward
and is presented in Chapter 6. Appropriate ow rates within the internal regenerator and
TABLE 5.7 Mass Flow Rate Balances at Mixing Locations in the Regenerator
Mixing Location
A
Pressure Condition
_ h; p m_ h;co m_ h 0
m_ h;i m_ H
c;r m
m_ h;i
m_ H
c;r
m_ h; p m_ h;co m_ h 0
m_ h; p m_ c;ax m_ h; p 0
m_ h; p m_ c;ax m_ h; p 0
C
_ h; p m_ c;co m_ h;o 0
m_ h m_ C
c;r m
_ C
_c 0
m_ c;i m_ c;co m_ C
c; p m
c;r m
_ H
_ c;ax 0
m_ C
c; p m
c; p m
C
_ c;ax 0
m_ c; p m_ H
c; p m
_ H
_ c;o 0
m_ c m_ h;co m_ H
c; p m
c;r m
H
H
_
_
_
_
mc mh;co mc; p mc;r m_ c;o 0
365
Energy Balance
H
0
_ h; p m_ h;co m_ h Hh;i
0
m_ h;i Hh;i m_ H
c;r Hc;o m
H
H
0
*
0
m_ h;i Hh;i m_ c;r Hc;o m_ h; p Hh m_ h;co m_ h Hh;i
H
C
0
m_ c;ax H*c m_ h; p Hh* 0
m_ h; p Hh;i
*
*
Th Tc
0
*
_ h Hh;o
_ C
m_ h;o Hh;o 0
m_ c;co m_ C
c;r Hc;i m
h; p Hh m
0
If m_ C
c;r from D to C: Tc Tc;i
C
0
0
0
_ C
_ c Hc;i
If m_ c;r from C to D: m_ c;co Hc;i
m_ C
m_ c;i Hc;i 0
c;r Hh;o m
c; p Hc;i m
_ H
_ c;ax H*c 0
m_ C
c; p Hc;i m
c; p Hc;o m
*
Tc Tc;i
0
0
_ H
_ c;o Hc;o 0
m_ h;co Hh;i
m_ H
m_ c Hc;o
c; p m
c; r m
0
0
H
H
_
_
_
_
mc Hc;o mh;co Hh;i mc; p Hc;i mc;r m_ c;o Hc;o 0
temperatures (for density calculations) are needed. This analysis yields outlet pressures
ph;o and pc;o .
Thus heat transfer and pressure drop modeling of the internal regenerator has ve
0
0
, Tc;o
, ph;o , and pc;o . Correspondingly, there are ve equations: the
unknowns: "i;b , Th;o
denition of "i;b , two equations for outlet temperatures (for known inlet temperatures,
"i;b and Cmin ), and two pressure drop equations. The modeling of actual and internal
regenerators thus requires solving 25 nonlinear equations iteratively for 25 unknowns
using, for example, an iterative NewtonRaphson method. Such a large set of equations
can only be solved by a computer program. Specic illustrative results for 5 and 10%
leakages individually through radial, peripheral, and axial seals for a gas turbine regenerator problem are presented by Shah and Skiepko (1997). The following are specic
conclusions from that study.
. Any radial seal leakage location (at the hot or cold end of a regenerator) has a
negligible eect on actual heat transfer to the cold stream.{ Hence, one can assume
50% radial seal leakage at the hot end and 50% at the cold end. For one specic
example of a gas turbine regenerator, 5% and 10% radial seal leakages reduced the
useful thermal energy transfer (heat recovery) to the cold gas in the outlet duct by
3.2% and 6.9%, respectively. These 5% and 10% leakages reduced the cold-gas
side pressure drop by 1.4% and 3.2% and increased the hot-gas side pressure drop
by 4.5% and 9.5%, respectively. If a conservative (i.e., high) value of the pressure
drop is desired, consider total pressure leakage of the cold gas at the regenerator hot
end only.
{
This is often true for high-eectiveness " regenerators. If " 40 to 50%, the hot-end leakage of cold gas can
dilute the hot-gas temperature entering the actual regenerator, thus having some impact on Th;i Tc;i and
reduction in heat transfer.
366
. The eects of the peripheral seal leakage on heat transfer and cold and hot gas side
pressure drops are similar to those for radial seal leakage.
. Four cases of leakage across axial seals have been considered for the following ow
paths in Fig. 5.14: DEBC, DEBA, FEBC, and FEBA. The eects on heat transfer
and pressure drops are very sensitive to the leakage distribution (ow path): from
about the same eect as for the radial seals to detrimental eect over a factor of
two.
. Since all leakages occur simultaneously in a regenerator, even a moderate or small
leak through individual seals can have a signicant eect on regenerator performance. It is highly desirable to have all radial, peripheral, and axial seals as tight as
possible. However, it is shown that the axial seals should be tightened as much as
possible because of the most detrimental eects for the gas turbine regenerator
problem considered.
5.7
In a xed-matrix regenerator, the outlet temperature of the cold gas (air) decreases as a
function of time during the cold-gas ow period. The dierence in the outlet temperature
from 0 to that at Pc is referred to as the temperature swing. The swing should be
minimized so that the heated air from the regenerator is at a relatively constant temperature for the process downstream. The design of three and four stove systems, as
mentioned in Section 1.5.4.2, has been developed to minimize this temperature swing.
Since the temperature swing depends on the heat capacity of the matrix material, it can be
minimized by employing a high-volumetric-heat-capacity (w cw ) material in the matrix,
such as silicon carbide or corundum instead of reclay. However, the high-heat-capacity
material is much more expensive. Use of proper matrix material size and arrangement
along with dierent materials in a two- or three-zoned regenerator results in an optimum
regenerator.
Heggs and Carpenter (1978), among others, proposed the use of high-heat-capacity
material near the outlet end of the cold-gas ow path and low-heat-capacity material
near the inlet end of the cold-gas ow path. Only 10% of the heat transfer surface area is
required to have higher-heat-capacity material, the other 90% with lower-heat-capacity
material. Heggs and Carpenter also found that thick bricks can be used near the outlet
end and thin bricks near the inlet end of the cold-gas ow path in a two-zoned regenerator.
Either of the alternatives will reduce the temperature swing. The design of a two-zoned
regenerator is carried out by considering two zones in series; the outlet uid temperature
from the rst zone is the inlet uid temperature of the second zone. Alternatively, an
equivalent height for one zone is computed in terms of the second-zone heat transfer
coecient, surface area, and packing density, as shown in Example 5.6.
The steam boiler regenerator utilizes two layers (each having dierent surfaces) in the
matrix for the following reasons: (1) To get higher heat transfer, the hot end has a higher
performance surface of rather complex geometry, which is unsuitable for the cold end
due to the possibility of plugging by particulate (a plain surface is often used so that
particulates can ow through easily); and (2) near the cold end, the hot ue gas is cooled
enough so that there is a possibility of sulfuric acid corrosion. Hence, the sheet metal
thickness of the surface near the cold end is higher than that for the hot matrix layer,
where corrosion is not a problem. Also, due to the condensed water vapor near the cold
367
end, the surface is wet and hence fouling is a serious problem. Hence, the matrix surface
ow passages are straight (duct ow with lower heat transfer coecients) to get easier self
cleaning and forced blow-o of the matrix. At the cold end, ceramic elements are sometimes applied to the matrix surface, or the steel sheet surface is covered with ceramic
material to resist corrosion.
Example 5.6 A rotary regenerator matrix used to preheat air in a thermal power plant
is made of two layers (each having dierent surfaces) in the air/gas ow direction. The
upper layer near the hot end of the regenerator has an enhanced surface for higher
performance. The lower layer near the cold end has a plain surface with higher sheet
metal thickness (compared to that for the upper layer) to minimize and accommodate
corrosion and fouling, as mentioned in the text above. The geometries, heat transfer
coecients, and physical properties of the matrix material for the upper and lower layers
are as follows:
Regenerator Parameters and
Fluid Properties
Height of the individual layer (m)
Hydraulic diameter (mm)
Sheet metal thickness (mm)
Packing density (m2/m3)
Porosity
Flue gasside heat transfer coecient (W/m2 K)
Air-side heat transfer coecient (W/m2 K)
Density w (kg/m3)
Specic heat cw (J/kg K)
Thermal conductivity kw (W/m K)
Upper Layer
(Near the
Hot End)
Lower Layer
(Near the
Cold End)
1.300
8.54
0.7
402.3
0.859
82.2
72.7
7841
456
50.3
0.300
7.08
1.0
440.4
0.780
33.0
29.3
7841
456
50.3
The frontal area of the rotor, excluding the shaft but including the seal coverage, is
19.1 m2. The rotor turns at 2.32 rpm with the ue gas (hot) side spanning 1958 and the air
(cold) side spanning 1658 of the rotor face. The radial seals cover 10% of the rotor face
area. The mass ow rate of the ue gas stream entering at 3398C is 51.1 kg/s, and it ows
in countercurrent to the air stream at 298C, owing at 42.4 kg/s. The isobaric specic heat
of gas and air are 1.11 and 1.02 kJ/kg K respectively.
Determine the regenerator eectiveness and the heat transfer rate from the ue gas to
the airstream.
SOLUTION
Problem Data and Schematic: The rotor geometry comprising two layers of the same
material but with dierent physical characteristics is specied. In addition, the rotor
dimensions and rotational speed are specied. The uid ow rates and inlet temperatures
are provided for both the hot and cold uid streams. The physical properties of matrix
material and uids are specied (see Fig. E5.6).
Determine: The regenerator eectiveness and the heat transfer rate from the ue gas to
the air.
368
2.32 rpm
Effective face
area =19.1 m2
Hot gas
Th,i = 339C
mh = 51.1 kg/s
195
165
1.3 m
Upper
layer
7841
0.456
Lower
layer
7841
0.456
402.3
440.4
0.859
0.780
hh (W/m 2) K
82.2
33.0
hc (W/m 2) K
72.7
29.3
Matrix
w (kg/m3)
cw (kJ/kg K)
(m2/m3)
0.3 m
Gas/air
Assumptions: The assumptions of Section 5.1.1 are invoked here, and the uid temperature distributions after the rst layer and entering the second layer are uniform.
Analysis: To solve this problem, compute the height of the equivalent lower layer to
represent it as if it were an equivalent upper layer maintaining the same actual heat
transfer rate of the lower layer. Once this equivalent upper layer height is obtained for
the lower layer, add it up with the upper layer height to obtain a single height (in terms of
the upper layer geometry and other characteristics) of the equivalent regenerator individually for hot gas and cold air. So this equivalent regenerator may have dierent
heights for the gas and air sides for performance calculation purposes. Then the solution
procedure of this single-layer regenerator follows the same steps as those presented in
Section 5.2.
Let us calculate the height of the equivalent lower layer. Noting that the heat transfer
area A of individual layers is given as
A rotor frontal area Afr layer height L packing density
we can write, based on Eq. (5.141), the following expression for the heat transfer rate q in
the lower layer:
q Uo;lo Alo Tlo Uo;lo Afr Llo lo Tlo
Similarly, using Eq. (5.141), we can write an expression for the heat transfer rate q in the
lower layer in terms of the height of the equivalent upper layer as follows:
q Uo;up Afr Leq;up up Tup
Equating the two equations for the lower layer heat transfer rates and taking cognizance
of the fact that the temperature potential for heat transfer in the original and equivalent
369
layers is equal (Tlo Tup ), we obtain an equivalent upper layer height of the same
heat transfer duty as the lower layer:
Leq;up
Uo;lo lo
L
Uo;up up lo
We use Eq. (5.143) to calculate the overall heat transfer coecients Uo;lo and Uo;up .
Neglecting the heat conduction term in Eq. (5.143), we get
1
1
1
Uo hh Ah =A hc Ac =A
where Ah Afr;h L and Ac Afr;c L. However, the frontal areas are given as
Afr;h 0:9Afr h =3608
and
A
h c
and
Ac
c
A
h c
where h and c are specied as 1958 and 1658 respectively. Using Ah =A and Ac =A in the
formula for Uo , we get the overall heat transfer coecients in the lower and upper
layers as
1
1
1
Adding the equivalent height to the actual upper layer height of 1.300 m, we obtain
the eective height for the purpose of heat transfer as Leff 1:300 0:132 m
1:432 m. Note that this eective height is lower than the actual physical height of the
two layers, which is 1:300 0:300 m 1:600 m. Now we can treat this two-layer regenerator as a single-layer regenerator with a height of 1.432 m, having all the characteristics
of the upper layer only.
370
Now let us calculate hA values on each uid side. Frontal areas for the hot and cold
sides are calculated knowing the fraction of frontal area not covered by seals, total
frontal area, and the sector angle divided by 3608:
Afr;h 0:9Afr h =3608 0:9 19:1 m2 1958=3608 9:311 m2
Afr;c 0:9Afr c =3608 0:9 19:1 m2 1658=3608 7:879 m2
Knowing individual frontal areas and eective heights, the heat transfer areas for the hot
and cold sides are found as
Ah Afr;h Leff h 9:311 m2 1:432 m 402:3 m2 =m3 5364 m2
Ac Afr;c Leff h 7:879 m2 1:432 m 402:3 m2 =m3 4539 m2
Knowing the heat transfer areas and heat transfer coecients, we obtain
hAh 0:0822 kW=m2 K 5364 m2 440:9 kW=K
hAc 0:0727 kW=m2 K 4539 m2 330:0 kW=K
Hence, the ratio of the convection conductances is found as
hA*
0:748
hAmax hAh 440:9 kW=K
Thus hA* is within the range 0.25 to 4.0, and hence its eect is negligible on the
regenerator ".
To determine the regenerator eectiveness, we calculate various heat capacity rates:
Ch m_ h cp;h 51:1 kg=s 1:11 kJ=kg K 56:72 kW=K
Cc m_ c cp;c 42:4 kg=s 1:02 kJ=kg K 43:25 kW=K
Cmax Ch 56:72 kW=K
C*
0:7625
Cmax 56:72 kW=K
For the matrix heat capacity rate, we must rst determine the matrix mass as follows.
Mw rotor frontal area rotor height matrix solidity matrix material density
19:1 m2 1:3 m 1 0:859 0:3 m 1 0:780 7841 kg=m3 37,336 kg
Knowing the matrix mass, its heat capacity rate is computed as
Cr Mw cw N 37,336 kg 0:456 kJ=kg K
C*r
Cr
658:3 kW=K
15:22
Cmin 43:25 kW=K
2:32
rev=s 658:3 kW=K
60
SUMMARY
371
From the foregoing values, we can compute NTUo , given in Eq. (5.48), as
NTUo
1
1
1
1
4:364
Cmin 1=hAh 1=hAc 43:25 kW=K 1=440:9 1=330:0 K=kW
Now determine the regenerator " using Eqs. (5.62) and (5.63):
1 expNTUo 1 C*
1 exp4:3641 0:7625
0:8845
1 C* expNTUo 1 C* 1 0:7625 exp4:3641 0:7625
!
1
1
0:8845
1
0:8840
" "cf 1
9 15:221:93
9C*r 1:93
"cf
It can be shown that " calculated by the Razelos method [Eqs. (5.64)(5.67)] would have
been identical to 0.8840. The heat transfer rate is then determined as
q "Cmin Th;i Tc;i 0:8840 43:25 kW=K 339 298C 11,852 kW
Ans:
Knowing the heat transfer rate, the hot- and cold-uid outlet temperatures are calculated as
Th;o Th;i
q
11,852 kW
3398C
130:08C
Ch
56:72 kW=K
Ans:
Tc;o Tc;i
q
11,852 kW
298C
303:08C
Cc
43:25 kW=K
Ans:
SUMMARY
Regenerators dier from recuperators in that the heat is transferred intermittently from
the hot uid to the cold uid via periodic thermal energy storage and release from the
heat transfer surface (matrix). As a result, two additional parameters enter in the analysis
of regenerators: the storage heat capacity rate of the matrix wall and the ratios of the
thermal conductances between the wall and hot and cold uids [in a dimensionless form
C*r and (hA)*, respectively], with the latter group of less importance in most industrial
regenerators. The two most commonly used methods for the design and analysis of
372
regenerators described are the "-NTUo and - methods. The details of these methods
with the basic concepts and advantages/disadvantages are presented. In addition to
longitudinal conduction in the wall (as in recuperators) at high eectivenesses, the transverse conduction in the wall can also be important for ceramic and low-thermalconductivity materials. The design theory is presented to take these eects into account.
In addition, pressure and carryover leakages can reduce the regenerator eectiveness
signicantly, depending on the operating conditions. A design theory is presented to
take these eects into account. Since the last two eects are too complex and interdependent on the regenerator geometry and operating conditions, the design theory
involves iterative calculations. The details presented in the text are essential for any
computer analysis of regenerator rating and sizing.
REFERENCES
Baclic, B. S., 1985, The application of the Galerkin method to the solution of the symmetric and
balanced counterow regenerator problem, ASME J. Heat Transfer, Vol. 107, pp. 214221.
Bahnke, G. D., and C. P. Howard, 1964, The eect of longitudinal heat conduction on periodic-ow
heat exchanger performance, ASME J. Eng. Power, Vol. 86, Ser. A, pp. 105120.
Coppage, J. E., and A. L. London, 1953, The periodic-ow regenerator: a summary of design theory,
Trans. ASME, Vol. 75, pp. 779787.
Dragutinovic, G. D., and B. S. Baclic, 1998, Operation of Counterow Regenerators, Vol. 4, Computational Mechanics Publications, WIP Press, Southampton, UK.
Hausen, H., 1929, Uber die Theorie von Warmeaustauches in Regeneratoren, Z. Angew. Math.
Mech., Vol. 9, pp. 173200.
Hausen, H., 1983, Heat Transfer in Counterow, Parallel Flow and Cross Flow, McGraw-Hill, New
York.
Heggs, P. J., and K. J. Carpenter, 1978, The eects of packing material, size and arrangement of the
performance of thermal regenerators, Heat Transfer 1978, Proc. 6th Int. Heat Transfer Conf.,
Vol. 4, Hemisphere Publishing, Washington, DC, pp. 321326.
Heggs, P. J., and K. J. Carpenter, 1979, A modication of the thermal regenerator innite conduction model to predict the eects of intraconduction, Trans. Inst. Chem. Eng., Vol. 57, pp. 228
236.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger Publishing, Malabar, FL.
Kroger, P. G., 1967, Performance deterioration in high eectiveness heat exchangers due to axial
heat conduction eects, Advances in Cryogenic Engineering, Vol. 12, pp. 363372.
Lambertson, T. J., 1958, Performance factors of a periodic-ow heat exchanger, Trans. ASME, Vol.
80, pp. 586592.
Mondt, J. R., 1964, Vehicular gas turbine periodic-ow heat exchanger solid and uid temperature
distributions, ASME J. Eng. Power, Vol. 86, Ser. A, pp. 121126.
Razelos, P., 1979, An analytic solution to the electric analog simulation of the regenerative heat
exchanger with time-varying uid inlet temperatures, Warme-und Stoubertragung, Vol. 12,
pp. 5971.
Razelos, P., 1980, Personal communication, Department of Applied Science, City University of New
York, Staten Island, NY.
Saunders, O. A., and S. Smoleniec, 1951, Heat transfer in regenerators, IMechEASME General
Discussion on Heat Transfer, London, pp. 443445.
REVIEW QUESTIONS
373
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
5.1
What is the dierence between heat capacitance and the heat capacity rate of a
uid? Explain with equations.
5.2
How do you dene the heat capacity rate of a matrix? How is it related to the heat
capacity of the wall for a rotary regenerator and for a xed-matrix regenerator?
5.3
In a storage-type heat exchanger, when the hot uid ows through the matrix, the
temperature of the entire matrix rises to the hot-uid temperature. When cold
uid ows through it, its temperature drops to the cold-uid temperature. Check
the appropriate answer for this phenomenon to occur.
(a) It depends on the matrix NTU.
(b) true
(c) false
(d) It depends on the temperature levels.
5.4
374
5.6
5.7
The following factors make analysis of regenerators more complicated than for
recuperators:
(a) Thermal resistance of solid material is important.
(b) Thermal capacitance of solid material is important.
(c) Heat transfer coecients vary with position.
(d) Heat transfer from/to uid streams occurs independently.
(e) Compactness (m2/m3) is higher.
5.8
5.9
5.10
5.11
The following things occur when the thickness (ow length) of an ideal counterow regenerator matrix is doubled with all other operating conditions remaining
unchanged:
(a) Wall thermal resistance eects become more important.
(b) Exchanger eectiveness increases (ideal regenerator).
(c) Longitudinal conduction eects become more important.
(d) Capacity rate ratio C* increases.
(e) Pressure drop increases.
5.12
REVIEW QUESTIONS
375
How would " change for a counterow regenerator (having C*r 5 and C * 1)
by increasing the disk speed from 3 rpm to 20 rpm?
(a) less than 1%
(b) about 5%
(c) more than 10%
5.14
In a vehicular gas turbine rotary regenerator, the terminal temperatures are: hot
uid: 7608C, 2978C; cold uid: 2048C, 6678C. The rotational speed is such that
C*r 10. The eectiveness of this regenerator is approximately:
(a) 83%
(b) 75%
(c) 100%
(d) 64%
5.15
The number of transfer units NTU for the regenerator of Question 5.14 is
approximately:
(a) 10
(b) 5
(c) 8
(d) 1
(e) 7.33
5.16
5.17
5.18
5.19
5.20
For identical surface geometry, matrix surface thickness and rpm, which matrix
material will require the least disk depth for the same "? Assume xed frontal area
and velocities.
(a) Aluminum: w 2702 kg=m3 , cw 903 J=kg K, kw 237 W=m K
(b) Stainless steel: w 8238 kg=m3 , cw 468 J=kg K, kw 13:4 W/m K
Ignore longitudinal and transverse conduction as well as pressure and carryover
leakages.
5.21
Consider two counterow rotary regenerators, one with C*r 5 and the other with
C*r 2. Otherwise, the exchangers are identical, with the same inlet temperatures
and ow rates. Sketch the temperature variation with position of the hot-uid
outlet temperature during one blow period in Fig. RQ5.21a for each exchanger.
Repeat in Fig. RQ5.21b for the cold-uid outlet temperature variation. For
convenience, let C* 1.
376
FIGURE RQ5.21
PROBLEMS
5.1 Following are the design data for a ceramic counterow disk rotary regenerator.
Two such disks are used in a truck gas turbine engine.
Regenerator operating conditions
Airow rate 1.400 kg/s
Gas ow rate 1.414 kg/s
Disk speed 15 rpm
Air inlet temperature 2048C
Gas inlet temperature 8668C
Air inlet pressure 393 kPa
Gas inlet pressure 105 kPa
Regenerator leakage
5.15% of airow
Physical properties
Matrix geometry
cp;air 1093 J/kg K
Surface area density 6463 m2/m3
cp;gas 1118 J/kg K
Hydraulic diameter Dh 0:44 mm
cw at 4008C 1093 J/kg K
Porosity 0:71
Cell count N 2:17 mm2
Matrix wall density 2214 kg/m3
Matrix k 0:19 W/m K
Wall thickness 0.076 mm
Triangular passages
Heat transfer coecients
hair 409 W/m2 K
hgas 425 W/m2 K
Calculate the regenerator eectiveness and outlet temperatures in the absence of
longitudinal conduction, wall thermal resistance, and pressure and carryover leakage eects. Then determine separately the inuences of longitudinal wall heat conduction and wall thermal resistance, again neglecting the eect of pressure and
carryover leakages on regenerator eectiveness. What will be the outlet temperatures in the latter case? Discuss your results.
5.2 The rotary air preheaters at the Moss Landing power plant have the following
geometry and operating conditions: 14.63-m disk diameter with Ah Ac
31,120 m2 /unit, m_ gas m_ air 340 kg/s per preheater, eectiveness " 65%,
PROBLEMS
377
Mw 1121 kg/m3 of total volume, csteel 460:6 J/kg K, cair cgas 1026
J/kg K, and the matrix air ow length 1.52 m. The preheater rotates
at 48 rev/h. What are the disk heat capacity rate ratio C*r and the disk NTUo ?
What is the average air-side convective conductance? Assume that hc hh .
Explain why you would neglect longitudinal conduction in the wall for this
problem. In this case, the C of Eq. (5.122) is assumed as zero.
5.3 Consider a rotary regenerator of compactness m2 =m3 , disk diameter D, thickness (ow length) L, hot and cold gas ow ratio m_ h and m_ c , respectively, and
rotating speed N. For what ow split (a fraction of a cycle in which an element
of matrix is in the hot blow divided by the fraction in the cold blow) will the
exchanger eectiveness be maximum? Justify your answer briey. Assume that
the heat transfer coecients hh and hc are equal and do not depend on the ow
split. Neglect any seal area. Hint: Use Eq. (5.10) and the denition of NTUo .
5.4 Consider a xed-matrix regenerator used as a thermal storage device with the
following data: m_ h m_ c 0:156 kg/s, cp;c cp;h 1011:0 J/kg K, hh hc
50:23 W/m2 K, A 5:8 m2 , Mw 904:8 J/kg K, and cw 920 J/kg K. The
cycle period Pt is specied as very short. Determine the mean outlet temperatures
of the hot and cold gases if Th;i 808C and Tc;i 108C and the hot and cold gases
are in counterow.
5.5 An industrial water tube boiler produces 2.52 kg/s of steam at 5.17 MPa. The ue
gases leave the boiler at 3218C (55.68C above the steam saturation temperature). To
increase the boiler eciency, a rotary counterow regenerator will be installed to
recover heat from 3218C gases to preheat 17:88C combustion air. The boiler is oil
red, and the combustion products, which contain sulfur, have a dew point of
1358C. To avoid corrosion, the gas temperature leaving the regenerator is limited
to 1438C. The following design data are provided: Air and gas ow rates are 2.39
and 2.55 kg/s, respectively, and their specic heats are 1005 and 1089 J/kg K. As a
good design practice, consider C*r 5. The regenerator matrix is made up of lowalloy steel with 8009 kg/m3 and cw 461 J/kg K. The heat transfer coecients
for air and gas sides are 190 and 207 W/m2 K, respectively, and the split of gas to
air ow is 50 : 50%. Determine the total heat transfer surface area required for the
regenerator. Neglect longitudinal conduction, carryover, and pressure leakage
eects. Once you determine the total area required, how would you determine
the disk diameter and ow length? Answer the last question qualitatively only.
Fluids need to be pumped through the heat exchanger in most applications. It is essential
to determine the uid pumping power required as part of the system design and operating
cost analysis. The uid pumping power is proportional to the uid pressure drop, which
is associated with uid friction and other pressure drop contributions along the uid ow
path. The uid pressure drop has a direct relationship with exchanger heat transfer,
operation, size, mechanical characteristics, and other factors, including economic considerations. The objective of this chapter is to outline the methods for pressure drop
analysis in heat exchangers and related ow devices. In this chapter, we start with the
importance of pressure drop evaluation, pumping devices used for uid ow, and the
major contributions to the total pressure drops associated with a heat exchanger in
Section 6.1. Two major contributions to pressure drop are associated with (1) core or
matrix and (2) ow distribution devices. Then a detailed derivation for various contributions to the core pressure drop is outlined for a plate-n exchanger, followed by the
pressure drop equations for all other major types of exchangers, in Sections 6.2 through
6.5. For most other exchangers, the ow distribution devices are of varied type (ducting,
pipe bends, valves and tting, etc.), and the pressure drop associated with them is calculated separately; appropriate methods for their computation are summarized in Section
6.6. Next, the presentation of pressure drop data in nondimensional and dimensional
forms is outlined for the design and analysis of heat exchangers in Section 6.7. Finally,
pressure drop dependence on geometry and uid properties is outlined in Section 6.8.
6.1
INTRODUCTION
First we outline why pressure drop is important for gases vs. liquids, what devices are
used for pumping uids in the exchanger, and the major components of the pressure drop
in a heat exchanger. Assumptions for the pressure drop analysis are also presented in
Section 6.1.4 before we start the pressure drop analysis in the following subsections.
6.1.1
The determination of pressure drop p in a heat exchanger is essential for many applications for at least two reasons: (1) The uid needs to be pumped through the exchanger,
which means that uid pumping power is required. This pumping power is proportional
to the exchanger pressure drop. (2) The heat transfer rate can be inuenced signicantly
378
INTRODUCTION
379
V_ p m_ p
p
p
6:1
where V_ is the volumetric ow rate and p is the pump/fan eciency. Now introduce the
following relationships:
m_ GAo
p f
4L G2
Dh 2gc
Re
GDh
6:2
where G is referred to as the core mass velocity G um , Ao is the minimum free ow
area, f is the Fanning friction factor,y and Re is the Reynolds number as dened in Eq.
(6.2). The p expression in Eq. (6.2) is for the core frictional pressure drop and is derived
later as Eq. (6.29). Substituting the expressions of Eq. (6.2) into Eq. (6.1) and simplifying
results in
8
1 4L m_ 2
>
>
>
>
< 2gc p 2 Dh Dh Ao f Re
m_ p
P
>
p
>
0:046 0:2 4L m_ 2:8
>
>
:
0:2
2gc p 2 Dh A1:8
o Dh
6:3a
Here f 0:046Re0:2 [see Eq. (7.72) in Table 7.6] is used in the derivation of the righthand-side expression of Eq. (6.3b) for fully developed turbulent ow.z Note also that
f Re in Eq. (6.3a) is constant for fully developed laminar ow, as discussed in Section
7.4.1.1. To determine the order of magnitude for the uid pumping power requirement
for gas vs. liquid ow, let us assume that the ow rate and ow passage geometry are
given (i.e., m_ , L, Dh , and Ao are specied). It is evident from Eq. (6.3) that P / 1=2 (i.e.,
strongly dependent on in laminar and turbulent ows); P / (i.e., strongly dependent
on in laminar ow); and P / 0:2 (i.e., weakly dependent on in turbulent ow). For
high-density moderate-viscosity liquids, the pumping power is generally so small that it
has only a minor inuence on the design. For laminar ow of highly viscous liquids in
large L=Dh exchangers, the uid pumping power is an important constraint. In addition,
the pumping power is an important consideration for gases, in both turbulent and
laminar ow, because of the great impact of 1/2 . For example, the density ratio for
liquid water vs. air at ambient conditions is approximately 800 : 1, which indicates that
the pumping power for airow will be much higher than that for water if p is to be kept
y
380
the same. Hence, typical design values of p for water and air as working uids in a heat
exchanger are 70 kPa (10 psi) (a typical value in shell-and-tube exchangers) and 0.25 kPa
(1 in. H2 O) (for compact exchangers with airows near ambient pressures), respectively,
to maintain the low uid pumping power requirement for exchanger operation. Refer
also to the rst footnote in Section 13.1 and the associated discussion.
6.1.2
The most common uid pumping devices are fans, pumps, and compressors. A fan is a
low-pressure air- or gas-moving device, which uses rotary motion. There are two major
types of fans: axial and radial (centrifugal), depending on the direction of ow through
the device. Fans may be categorized as blowers and exhausters. A blower is a centrifugal
fan when it is used to force air through a system under positive pressure, and it develops a
reasonably high static pressure (500 Pa or 2.0 in. H2O). An exhauster is a fan placed at the
end of a system where most of the pressure drop is on the suction side of the fan. A pump
is a device used to move or compress liquids. A compressor is a high-volume centrifugal
device capable of compressing gases [100 to 1500 kPa (15 to 220 psi) and higher].
Fans and pumps are volumetric devices and are commonly used to pump uids
through heat exchangers. This means that a fan will develop the same dynamic head
[pressure rise per unit uid (gas) weight across the fan; Eq. (6.4)] at a given capacity
(volumetric ow rate) regardless of the uids handled, with all other conditions being
equal. This means that the pressure rise across a fan will be proportional to the uid
density at a given volumetric ow rate for all other conditions being equal. Note that the
head, dynamic head or velocity head is referred to as the kinetic energy per unit weight of
the uid pumped, expressed in units of millimeters or inches (feet). Thus the pressure rise
across a fan (which is mainly consumed as the pressure drop across a heat exchanger) can
be expressed in terms of the head H as follows:
p
u2
H m
2g
g=gc
6:4
Since fans and pumps are generally head limited, the pressure drop in the heat exchanger
can be a major consideration.
6.1.3
The pressure drop associated with a heat exchanger is considered as a sum of two major
contributions: pressure drop associated with the core or matrix, and pressure drop
associated with uid distribution devices such as inlet/outlet headers, manifolds, tanks,
nozzles, ducting, and so on. The purpose of the heat exchanger is to transfer thermal
energy from one uid to the other; and for this purpose, it requires pressure dierence
(and uid pumping power) to force the uid ow over the heat transfer surface in the
exchanger. Hence, ideally most of the pressure drop available should be utilized in
the core and a small fraction in the manifolds, headers, or other ow distribution devices.
However, this ideal situation may not be the case in plate heat exchangers and other heat
exchangers in which the pressure drop associated with manifolds, headers, nozzles, and
so on, may not be a small fraction of the total available pressure drop.
If the manifold and header pressure drops are small, the core pressure drop dominates. This results in a relatively uniform ow distribution through the core. All heat
381
transfer and core pressure drop analyses outlined here and in preceding chapters presume
that the ow distribution through the core is uniform. A serious deterioration in performance may result for a heat exchanger when the ow through the core is not uniformly
distributed. This topic is covered in Chapter 12.
The core pressure drop is determined separately on each uid side. It consists of one
or more of the following contributions, depending on the exchanger construction: (1)
frictional losses associated with uid ow over the heat transfer surface (this usually
consists of skin friction plus form drag), (2) momentum eect (pressure drop or rise due
to the uid density changes in the core), (3) pressure drop associated with sudden contraction and expansion at the core inlet and outlet, and (4) gravity eect due to the
change in elevation between the inlet and outlet of the exchanger. The gravity eect is
generally negligible for gases. For vertical liquid ow through the exchanger, the pressure
drop or rise due to the elevation change is given by
p
m gL
gc
6:5
where the + sign denotes vertical upow (i.e., pressure drop), the sign denotes
vertical downow (i.e., pressure rise or recovery), g is gravitational acceleration, L is the
exchanger length, and m is the mean uid mass density calculated at bulk temperature
and mean pressure between the two points where the pressure drop is to be determined.
The rst three contributions to the core pressure drop are presented for extended surface
exchangers, regenerators, and tubular and plate heat exchangers in Sections 6.2 through
6.5. Since the manifolds are integral parts of a PHE, the pressure drop associated with
manifolds is also included in p calculations for a PHE in Section 6.5.
6.1.4
The following are the major assumptions made for the pressure drop analysis presented
in this chapter.
1. Flow is steady and isothermal, and uid properties are independent of time.
2. Fluid density is dependent on the local temperature only or is treated as a constant
(inlet and exit densities are separately constant).
3. The pressure at a point in the uid is independent of direction. If a shear stress is
present, the pressure is dened as the average of normal stresses at the point.
4. Body forces are caused only by gravity (i.e., magnetic, electrical, and other elds
do not contribute to the body forces).
5. If the ow is not irrotational, the Bernoulli equation is valid only along a streamline.
6. There are no energy sinks or sources along a streamline; ow stream mechanical
energy dissipation is idealized as zero.
7. The friction factor is considered as constant with passage ow length.
6.2
The pressure drop analysis is presented now for plate-n and tube-n heat exchangers.
382
6.2.1
One ow passage in a plate-n heat exchanger is shown in Fig. 6.1 along with uid ow
and static pressure distribution along the ow path. The incoming ow to the passage is
assumed to be uniform. As it enters the passage, it contracts due to the free-ow area
change. Flow separation takes place at the entrance followed by irreversible free expansion. In the core, the uid experiences skin friction; it may also experience form drag at
the leading and trailing edges of an interrupted n surface; it may also experience internal
contractions and expansions within the core, such as in a perforated n core. If heating or
cooling takes place in the core, as in any heat exchanger, the uid density and mean
velocity change along the ow length. Thus, the uid within the ow passage accelerates
or decelerates depending on whether it is being heated or cooled. At the core exit, ow
separation takes place followed by an expansion due to the free-ow area change. Then
the total pressure drop on one side of the exchanger, from Fig. 6.1, is
p p12 p23 p34
6:6
Here the subscripts 1, 2, 3, and 4 represent locations far upstream, passage entrance,
passage exit, and far downstream, respectively, as shown in Fig. 6.1. The p12 is the
pressure drop at the core entrance due to sudden contraction, p23 the pressure drop
within the core (also simply referred to as the core pressure drop), and p34 the pressure
rise at the core exit. Usually, p23 is the largest contribution to the total pressure drop,
and we evaluate it rst before the other two contributions.
6.2.1.1 Core Pressure Drop. The pressure drop within the core consists of two contributions: (1) the pressure loss caused by uid friction, and (2) the pressure change
due to the momentum rate change in the core. The friction losses take into account
both skin friction and form drag eects. The inuence of internal contractions and
FIGURE 6.1 Pressure drop components associated with one passage of a heat exchanger (From
Shah, 1983; modied from Kays and London, 1998).
383
expansions due to ow area changes, if present, is also lumped into the core friction loss
term. Consider a dierential element of ow length dx in the core as shown in Fig. 6.1.
Various force and momentum rate terms in and out of this element are shown in Fig.
6.2.y
Applying Newtons second law of motion, we have
G2 Ao 1 d 1
G2 Ao
dp
pAo p dx Ao w P dx
dx
gc dx
gc
dx
6:7
Here w is the eective wall shear stressz due to skin friction, form drag, and internal
contractions and expansions, if any. P is the wetted perimeter of the uid ow passages of
heat exchanger surface. Rearranging and simplifying Eq. (6.7), we get
dp G2 d 1
P
w
dx gc dx
Ao
6:8
Note that we use the mass velocity G m_ =Ao as the ow variable for the exchanger p
analysis. This is because G is constant for a constant-steady state uid ow rate m_ and
constant Ao } even though both and um in G um vary along the ow length in the heat
exchanger.
Now dene the Fanning friction factor f } as the ratio of wall shear stress w to the
ow kinetic energy per unit volume.
f
w
w
6:9
FIGURE 6.2 Force and momentum rate terms for a dierential element of a heat exchanger core.
y
While w P dx is shown acting on both top and bottom surface in Fig. 6.2, in reality it acts along the entire surface
P dx.
z
w is dependent on the ow passage geometry and size, uid velocity, uid density and viscosity, and surface
roughness, if any.
}
The minimum free-ow area Ao is constant in most heat exchangers, including that for ow over the tube banks
and ow in the regenerator matrix, made up of solid objects such as bricks, rocks, pebbles, and so on.
}
The friction factor is either derived experimentally for a surface or derived theoretically for laminar ow and
simple geometries. It is discussed further in Chapter 7.
384
As discussed in Section 7.2.1.2, w represents the eective wall shear stress, and is the
uid mass density determined at the local bulk temperature and mean pressure. Also
dene the hydraulic radius rh as
rh
Ao
P
6:10
Note that the hydraulic diameter Dh 4rh ; it was dened earlier by Eq. (3.65) and in the
footnote in Section 1.4.
Substituting Eqs. (6.9) and (6.10) into Eq. (6.8) and using d1= 1=2 d, we
get, upon simplication,
dp G2
2 d
1
2
f
dx 2gc
rh
dx
6:11
G2
L
1
2 i 1 f i
2gc i
o
rh
m
6:12
where the mean specic volume with respect to the ow length, (1=m , is dened as
1
1 L dx
m L 0
6:13
Here, the uid mean specic volume 1=m can be expressed as follows [see also Eq.
(9.18)]. For a liquid with any ow arrangement, or for an ideal gas with C* 1 and any
ow arrangement except for parallelow,
1
v vo 1 1
1
vm i
2
m
2 i o
6:14
Here v denotes the specic volume in m3 =kg or ft3 /lbm; vi and vo are evaluated at inlet
and outlet temperatures and pressures, respectively. Note that, in general,
1
1
6
m m
where m
i o
2
6:15
However, 1=m 1=m is a good approximation for liquids with very minor changes in
density with temperatures and small changes in pressure. For a perfect gas with C* 0
and any exchanger ow arrangement,
R~
1
T
m pave lm
6:16
385
Note the core pressure drop [Eq. (6.12)] has two contributions: The rst term represents the momentum rate change or the ow acceleration (deceleration) eects due to the
uid heating (cooling); its positive value represents a pressure drop for ow acceleration
and the negative value a pressure rise for ow deceleration. The second term represents
the frictional losses and is the dominating term for p.
6.2.1.2 Core Entrance Pressure Drop. The core entrance pressure drop consists of two
contributions: (1) the pressure drop due to the ow area change, and (2) the pressure
losses associated with free expansion that follow sudden contraction. To evaluate the
core entrance losses, it will be assumed that the temperature change at the entrance is
small and that the uid velocity is small compared to the velocity of sound. Thus the uid
is treated as incompressible. The pressure drop at the entrance due to the area change
alone, for a frictionless incompressible uid, is given by the Bernoulli equation as
p1
p20
i
u22
u2
1
2gc 2gc
"
2 #
i u22
u
1 1
2gc
u2
6:17
where i is the uid density at the core inlet and i 1 2 in Fig. 6.1; and p20 is the
hypothetical static pressure at section 2 in Fig. 6.1 if the pressure drop would have been
alone due to the area change. From the continuity equation,
i Ao;1 u1 i Ao;2 u2
6:18
Introduce as the ratio of core minimum free-ow area to frontal area and G as the core
mass velocity:
Ao;2 Ao;3
Ao;1 Ao;4
6:19
m_
Ao;2
6:20
G i u2
Substituting Eqs. (6.18)(6.20) into Eq. (6.17), the pressure drop at the core entrance due
to the area change alone is
p1 p20
G2
1 2
2gc i
6:21
The second contribution to the pressure drop at the entrance is due to the losses
associated with irreversible free expansion that follows the sudden contraction. A region
of ow separation and secondary ows (as shown in Fig. 6.1 at the vena contracta)
produces irreversible pressure losses, and the change in the momentum rate (due to
any nonuniform ow) will also produce pressure losses. The resulting pressure change
is due to the change in the momentum rate downstream of the vena contracta. Pressure
drop due to these losses is taken into account by the contraction loss coecient Kc
multiplied by the dynamic velocity head at the core inlet as follows:
ploss Kc
i u22
G2
Kc
2gc
2gc i
6:22
386
FIGURE 6.3 Entrance and exit pressure loss coecients for (a) a multiple circular tube core,
(b) multiple-tube at-tube core, (c) multiple square tube core, and (d) multiple triangular tube core
with abrupt contraction (entrance) and abrupt expansion (exit). (From Kays and London, 1998.)
387
G2
1 2 Kc
2gc i
6:23
6.2.1.3 Core Exit Pressure Rise. The core exit pressure rise p4 p3 is divided into
two contributions idealizing the uid as incompressible at the core exit 3 4 o
in Fig. 6.1. The rst contribution is the pressure rise due to the deceleration associated
with an area increase and it is given by an expression similar to Eq. (6.21):
prise
G2
1 2
2gc o
6:24
The second contribution is the pressure loss associated with the irreversible free expansion and momentum rate changes following an abrupt expansion, and it is similar to Eq.
(6.22).
ploss Ke
3 u23
G2
Ke
2gc
2gc o
6:25
Note that Ke is based on the dynamic velocity head at the core outlet. The exit loss
coecient Ke is a function of the expansion ratio 1=, the Reynolds number Re, and the
ow cross-sectional geometry. Values of Ke for four dierent ow passage geometries
are presented in Fig. 6.3.
It should be emphasized that two eects are lumped into dening Ke : (1) pressure loss
due to the irreversible free expansion at the core exit, and (2) pressure rise due to the
momentum rate changes, considering partially or fully developed velocity prole at
the core exit and uniform velocity prole far downstream at section 4 in Fig. 6.1.
Hence, the magnitude of Ke will be positive or negative, depending on whether the
sum of the foregoing two eects represents a pressure loss or a pressure rise.
The net pressure rise at the core exit, p34 p4 p3 , from Eqs. (6.24) and (6.25), is
p34
G2
1 2 K e
2gc o
6:26
388
6.2.1.4 Total Core Pressure Drop. The total core pressure drop on one uid side of a
plate-n exchanger is given by Eq. (6.6) as
p p12 p23 p34
6:27
1
2
1
f
1
K
c
e
pi
rh i m
2gc i pi 6
o
7
4|{z}
5
|{z}o
|{z}
|{z}
entrance effect
momentum effect
core friction
6:28
exit effect
Generally, the core frictional pressure drop is a dominating term, about 90% or more of
p for gas ows in many compact heat exchangers. The entrance eect represents the
pressure loss, and the exit eect in many cases represents a pressure rise; thus the net
eect of entrance and exit pressure losses is usually compensating.
The entrance and exit losses are important when their values with respect to the core
friction term in the brackets of Eq. (6.28) is nonnegligible. Reviewing the terms in the
brackets of Eq. (6.28), it is clear that the entrance and exit eects may be nonnegligible
when and L are small, rh (or Dh ) is large, and f is small. Small values of f for a given
surface are usually obtained at high values of Re (such as in turbulent ow). Thus, the
entrance and exit losses are important at small values of and L (short cores), large
values of Dh and Re, and for gases; they are generally negligible for liquids because
the total p of Eq. (6.28) is small compared to that for gases. Note that the small values
of are obtained in a plate-n exchanger (1) if the passages are small and the plates/ns
are relatively thick, and/or (2) a large portion of the frontal area (on the uid side of
interest) is blocked by the ow passages of the other uid.
The values of Kc and Ke presented in Fig. 6.3 apply to long tubes for which ow is
fully developed at the exit. For partially developed ows, Kc is lower and Ke is higher
than that for fully developed ows, due to the associated momentum rate changes, as
discussed before. For interrupted surfaces, ow is hardly ever fully developed but may be
periodic. For highly interrupted n geometries, the entrance and exit losses are generally
small compared to a high value of the core pressure drop, and the ow is mixed very well;
hence, Kc and Ke for Re ! 1 should represent a good approximation. For many
enhanced and compact heat exchangers, ow passages are nonsmooth and uninterrupted
(such as wavy, ribbed, stamped, etc.) or interrupted with ows partially developed,
periodic or with ow separation, attachment, recirculation, vortices, and so on. For
ows through such passages, the estimate of Kc and Ke from Fig. 6.3 may not be
accurate. However, if the entrance and exit losses are only a small fraction of the core
pressure drop p, the error in the calculation of p due to large errors in Kc and Ke will
still be small.
The core frictional pressure drop, being the major contribution in the total core
pressure drop of Eq. (6.28), may be approximated as follows in dierent forms:
p
4fLG2 1
4fLm_ 2
4L m u2m
4L G2
4L m_
f
f
f Re
2
2gc Dh m 2gc Ao m Dh
Dh 2gc
Dh 2gc 2gc D2h Ao
6:29
389
where it is idealized that 1=m 1=m 1=. Corresponding uid pumping power P is
P
1
m_ p GAo 4fLG2
fAG3
6:30
where the last term is obtained after substituting Dh 4Ao L=A in the preceding term
and simplifying it. Equations (6.29) and (6.30) will be considered later for comparing,
assessing, and evaluating the merits of dierent heat exchanger surfaces.
We can evaluate the ow area ratio for two dierent surfaces on one uid side for a
given application (specied mass ow rate and pressure drop on one uid side) using
Eq. (6.29) as follows:
Ao;1
f L=Dh 1
1
Ao;2
f2 L=Dh 2
6:31
Example 6.1 A gas-to-air single-pass crossow plate-n heat exchanger has overall
dimensions of 0.300 m 0:600 m 0:900 m and employs strip ns on the air side. The
following information is provided for the air side:
Geometrical properties
Fin density 0:615 mm1
Plate spacing 6:35 mm
Fin oset length 3:18 mm
Airow length 0:6 m
Hydraulic diameter 0:002383 m
Fin metal thickness 0:15 mm
Minimum free-ow area 0:1177 m2
Free-ow area/frontal area 0:437
Operating conditions
Volumetric airow rate 0:6 m3 =s
Reynolds number 786
Fanning friction factor 0.0683
Inlet pressure 110 kPa
Inlet temperature 48C
Outlet temperature 194:58C
Gas constant for air
287:04 J=kg K
FIGURE E6.1
390
Determine: The air-side pressure drop for this plate-n heat exchanger.
Assumptions: The ow distribution through the heat exchanger is uniform, and air is
treated as an ideal gas.
Analysis: To compute the pressure drop for a plate-n heat exchanger using Eq. (6.28),
rst we need to determine the inlet, outlet, and mean air densities in the core as well as the
core mass velocity G. Considering air as an ideal gas, the inlet density is given by
a;i
pa;i
110 103 Pa
1:3827 kg=m3
~
RTa;i 287:04 J=kg K 4:0 273:15 K
Note that we converted the inlet temperature to an absolute temperature scale. Similarly,
the air density at the core outlet is given by
a;o
pa;o
110 103 Pa
0:8195 kg=m3
R~Ta;o 287:04 J=kg K 194:5 273:15 K
Note that we have considered here the outlet pressure as 110 kPa since the pressure drop
across the core is usually very small, and hence it is neglected in the rst trial. The mean
density is the harmonic mean value given by Eq. (6.14) as
1
1 1
1
1
1
1
m3
1
0:9717
3
3
kg 1:0291 kg=m3
m 2 i o
2 1:3827 kg=m
0:8195 kg=m
Since the airow is given as the volumetric ow rate at the inlet, let us calculate the mass
velocity as
7:0486 kg=m2 s
Ao
0:1177 m2
Now let us calculate the entrance and exit pressure loss coecients so that we can
compute the pressure drop using Eq. (6.28). Since strip ns are used, the ow is well
mixed. Hence, Kc and Ke are evaluated at Re ! 1 using Fig. 6.3. Reviewing Fig. 6.3b
and c for parallel plates and square ducts, while it is found that Kc and Ke are dependent
on the aspect ratio of the rectangular passages, it is also found that Kc and Ke are
identical for Re ! 1 (i.e., independent of the aspect ratio). Hence, even though one
can easily calculate the aspect ratio of the rectangular passages formed by the oset strip
n from the given geometry data, there is no need to compute it. Then from Fig. 6.3b or c,
for a 0:437, we get
Kc 0:33
Ke 0:31
The core pressure drop for the air side is then given by Eq. (6.28) as
391
p
G2
L
1
1 2 Kc 2 i 1 f i
1 2 K e i
2gc i pi
o
o
pi
rh
m
7:0486 kg=m2 s2
2 1 110 103 Pa 1:3827 kg=m3
1:3827 kg=m3
1 0:4372 0:33 2
1
0:8194 kg=m3
0:0683 0:6 m 1:3827 kg=m3
1:3827 kg=m3
2
1
0:437
0:31
0:002383=4 m 1:0291 kg=m3
0:8194 kg=m3
p
110 kPa 0:01536 1:69 kPa
pi
Ans:
Note that the pressure drop on the air side (1.69 kPa) is 1.5% of the inlet pressure
(110 kPa). Hence, our assumption of po pi to calculate o is good. Otherwise, once the
pressure drop is computed, determine the outlet pressure and then iterate once more with
the newly computed outlet density.
Discussion and Comments: As one can see, the determination of the pressure drop is
straightforward. In this example, the core pressure drop is dominant, consisting of
98.2% of the total pressure drop. The combined entrance and exit losses are 0.3%
1:2 0:9 of the pressure drop. Since the core frictional pressure drop contribution
for this example is so large that reducing the core depth by 50% would not have made
any signicant dierence in the contribution of the combined entrance and exit losses.
However, modern automotive compact heat exchangers have an airow length of about
12 to 50 mm only. In this case, the entrance and exit losses may become a nonnegligible
fraction of the total core pressure drop; and the approximation of the entrance and exit
losses for Re ! 1 may not be a good approximation. However, until better information
is available, the current approach of using Kc and Ke from Kays and London (1998) is
recommended to obtain at least a good approximate correction.
6.2.2
6.2.2.1 Tube Inside. The pressure drop inside the tubes is determined in the same
manner as that for plate-n surfaces using Eq. (6.28). Appropriate values of the f factor
and Kc and Ke are used in this expression for ow inside the tubes with or without ns.
6.2.2.2 Tube Outside with Fins. The three types of ns on tubes (Section 1.5.3.2) are:
normal ns on individual tubes, longitudinal ns on individual tubes, and at ns on an
array of tubes. For the rst two types of nned tubes, the tube outside ow in each tube
392
row experiences a contraction and an expansion. Thus, the pressure losses associated
with a tube row within the core are of the same order of magnitude as those at the
entrance with the rst tube row and those at the exit with the last tube row.
Consequently, the entrance and exit pressure drops are not calculated separately, but
they are generally lumped into the friction factor (which is generally derived experimentally) for individually nned tubes and longitudinally nned tubes. Then the total
pressure drop associated with the core, from Eq. (6.28), becomes
p
G2
2gc i pi
pi
f
L
1
i
2 i 1
o
rh
m
6:32
It should be emphasized that the friction factor in Eq. (6.32) is based on the hydraulic
diameter. If, instead, the pressure drop correlation is available in terms of an average
Euler number Eu per tube row [see Eq. (7.22) for the denition], the pressure drop will be
p
G2
1
EuNr i
2 i 1
2gc i pi
o
pi
m
6:33y
where Nr represents the number of tube rows. Thus, entrance and exit pressure losses are
eectively lumped into the friction factor f by eliminating them from the p equation.
For at ns on an array of tubes (Fig. 1.31b), the components of the total core
pressure drop on the n side are all the same as those for plate-n surfaces. The only
dierence is that the ow area at the entrance and exit is between the ns and is independent of the tube arrangement.
To obtain the entrance and exit losses based on the ow area at the leading edge, rst
apply the continuity equation as follows:
m_ um Ao leading edge um Ao core
6:34
Introducing G 0 um leading edge and 0 Ao;leading edge =Afr in this equation, we get
G 0 0 G
6:35
Thus, Kc and Ke are evaluated for 0 from Fig. 6.3. The total pressure drop for this
geometry (at ns on an array of tubes) is then given by
p
G2
L
1
G02
f i
2 i 1
1 02 Kc 1 02 Ke i
2gc i pi rh
o
2gc i pi
o
pi
m
6:36
6.3
393
as that for the plate-n exchanger [Eq. (6.28)]. For a xed-matrix regenerator matrix
made up of any porous material (such as randomly packed screens, cross rods, bricks,
tiles, spheres, copper wool, etc.), the entrance and exit pressure drops are included in the
experimental friction factors. Thus Eq. (6.32) applies for the pressure drop of xedmatrix regenerators.
6.4
6.4.1
The pressure drop on the tube side is determined from Eq. (6.28). The pressure drop
associated with ow over the tube banks consists of the same contributions as that for the
plate-n exchanger, except that the entrance and exit pressure drops are included in the
friction factors. Hence, the total pressure drop on the outside of a tube bank is given by
Eq. (6.32).
6.4.2
Shell-and-Tube Exchangers
6.4.2.1 Tube Side. The pressure drop inside the tube is determined from Eq. (6.28)
with proper values of Kc , Ke , and f. However, in shell-and-tube exchangers, the
entrance and exit pressure drops for the tube ow are generally neglected since their
contribution is small compared to the losses associated with inlet and outlet nozzles and
chambers. If U-tubes or hairpins are used in a multipass unit, additional pressure drop
due to the 1808 bend needs to be included. The pressure drop associated with such a
bend is discussed in Section 6.6.3.
6.4.2.2 Shell Side. The pressure drop evaluation on the shell side of a shell-and-tube
heat exchanger is complicated due to the presence of bypass and leakage streams
(discussed briey in Section 4.4.1.1) in addition to the crossow stream. In this case,
the pressure drop is evaluated rst for an ideal crossow section and an ideal window
section. Correction factors are then applied for the leakage and bypass streams. The
total pressure drop is then the sum of the pressure drops for each window section and
each crossow section (Bell, 1988). In this section we just provide the empirical correlations for the pressure drop for the ideal crossow and window sections. The expression
for the total shell-side pressure drop is presented where correction factors for leakage
and bypass streams are dened in Section 9.5.1.2.
The pressure drop associated with liquid ow in an ideal crossow section between
two baes is
pb;id Eu
0:25
0:25
G2c
4f G2
2 Nr;cc
Nr;cc w
id c Nr;cc w
Hg
2gc s
m
2gc s
m
gc do2
6:37y
The viscosity correction factor (w =m 0:25 is considered only for liquids. Replace this term from Eq. (6.37) with
(Tw =Tm m for a gas on the shell side. See Tables 7.12 and 7.13 for the values of m.
394
where
Eu
pb;id
1
u2m =2gc Nr;cc
Eu 4fid
and
Hg 32 Re
6:38
Hence, where Eu is the average Euler number per tube row, Nr;cc is the number of
eective tube rows crossed during ow through one crossow section; Gc is the crossow
mass velocity, a ratio of total mass ow rate divided by the minimum free-ow area Ao;c
over the tubes at or near the shell centerline for one crossow section; is the liquid
viscosity evaluated at the tube wall (w) or bulk/mean (m) temperatures; and Hg is the
Hagen Number per tube row dened by Eq. (7.23). The Euler number and Hagen
number are determined from the correlations for ow normal to a tube bank of a
specied arrangement. Correlations in terms of the Hagen number are summarized in
Section 7.5.1. It should be emphasized that Eu 4fid in Eq. (6.37), and fid is the ideal
Fanning friction factor per tube row as dened by (with data provided by) Bell (1988).
This denition of fid is used only here and in Section 9.5.1.2 while discussing the Bell
Delaware method. At all other places in the book, the denition of the Fanning friction
factor used is given by Eqs. (7.17) and (7.18).
The pressure drop associated with an ideal one-window-section pw;i depends on the
shell-side Reynolds number Red Gdo = uc do =, where uc is evaluated at or near the
shell centerline in one crossow section as mentioned above. It is given by
pw;id
8
>
G2w
>
>
0:6N
>
r;cw
>
2gc s
<
>
>
26Gw s
>
>
>
: g
c s
Nr;cw
L
b
pt do D2h;w
G2w
gc s
6:39a
6:39b
where pt is the tube pitch, Nw is the number of eective crossow tube rows in each
window, Lb is the bae spacing, and Dh:w (the hydraulic diameter of the window section)
and uz (and Gw ) are given by
Dh;w
4Ao;w
do Nt;w Ds b =360
G2w
m_
m_
u2z uc uw
Ao;cr s Ao;w s
2s
6:40
6:41
where m_ is the total shell-side ow rate, Ao;cr and Ao;w are the ow areas for the crossow
and window sections, respectively, uc and uw are the ideal crossow and window mean
velocities, and b in Eq. (6.40) is in degrees. The s m is the mean density of the shellside uid.
The combined pressure drop associated with the inlet and outlet sections on the shell
side is given by
Nr;cw
pio 2 pb;id 1
Nr;cc b s
6:42
395
The total pressure drop on the shell side is the sum of the pressure drop associated
with each crossow section between baes, the pressure drop associated with each
window section, and the pressure drop for crossow sections on each end between the
rst (and last) bae and the tubesheet. Since the pb;id , pw;id , and pio of Eqs. (6.37),
(6.39), and (6.42) are for ideal conditions, they must be corrected for the presence of the
bypass and leakage streams. The total pressure drop on the shell side, excluding the p
associated with the entrance and exit nozzles and headers, is
ps pcr pw pio Nb 1 pb;id b Nb pw;id
Nr;cw
2pb;id 1
Nr;cc b s
6:43
where Nb is the number of baes, b is the pressure drop correction factor for bypass
ows (C and F streams in Fig. 4.19), l is the pressure drop correction factor for both
bae-to-shell (E stream) and tube-to-bae (A stream) leakage streams, and s is the
pressure drop correction factor for the unequal bae spacings for inlet and exit bae
sections. The values of b , l , and s are presented later in Table 9.3.
Example 6.2 A shell-and-tube heat exchanger is designed to cool the shell-side lubricating oil from 658C to 608C. Following are the specications for the shell-and-tube heat
exchanger.
Tube outside diameter 19 mm
Tube pitch 25 mm, square layout
Crossow area near the shell centerline
0:04429 m2
Flow area through the window zone
0:01261 m2
Number of eective tube rows
bae section 9
Factors for various leakage and bypass ows for the pressure drop correction are (1) 0.59
for bae-to-shell and tube-to-bae leakage streams, (2) 0.69 for bae-to-shell bypass
stream, and (3) 0.81 for unequal bae spacing on inlet and exit bae sections. Calculate
the shell-side pressure drop.
SOLUTION
Problem Data and Schematic: All necessary geometrical information and operating conditions are given for the oil side, as listed below, to compute the shell-side oil pressure
drop. A schematic of a shell-and-tube heat exchanger is shown in Fig. 1.5a.
Geometry: do 19 mm, pt 25 mm, Ao;c 0:04429 m2 , Ao;w 0:01261 m2 , Nr;cc 9,
Nr;cw 3:868, w 1:2 mm, Nb 14
Operating conditions and oil density: m_ 36:3 kg/s, fid 0:23, Res 242, 849 kg=m3
Correction factors: l 0:59; b 0:69; s 0:81
396
Determine: The oil-side pressure drop for this shell-and-tube heat exchanger.
Assumption: The assumptions adopted for the pressure drop analysis made in Section
6.1.4 are invoked here. Fluid density on hot-water side is treated as constant.
Analysis: To compute the pressure drop for the shell side using Eq. (6.43), let us rst
compute individual pressure drop components using Eqs. (6.37) and (6.39a). The mass
velocity
Gc
pb;id
36:3 kg=s
m_
819:60 kg=m2 s
Ao;c 0:04429 m2
0:14
4fid G2c
4 0:23 819:60 kg=m2 s2
Nr;cc w
9 10:14 3275:6 Pa
2gc s
m
2 1 849 kg=m3
Note that we have not included the viscosity correction primarily because no data are
given (due to the small temperature drop specied). Since the shell-side Reynolds number
is given as 242, the appropriate equation for the window zone pressure drop is Eq.
(6.39a). Let us rst calculate the velocity uz using Eq. (6.41):
u2z uc uw
m_
m_
Ao;c s Ao;w s
36:3 kg=s
36:3 kg=s
3:2732 m2 =s2
0:04429 m2 849 kg=m3 0:01261 m2 849 kg=m3
m u2z
849 kg=m3 3:2732 m2 =s2
2 0:6 3:868
6003:6 Pa
2gc
21
We are now ready to compute the shellside pressure drop using Eq. (6.43) as
Nr;cw
ps Nb 1 pb;id b Nb pw;id 2pb;id 1
Nr;cc b s
14 1 3275:6 Pa 0:69 14 6003:6 Pa 0:59 2 3275:6 Pa
3:868
1
0:69 0:81
9
17; 335 49; 590 5235 Pa 72,160 Pa 72:2 kPa
24%
69% 7%
Ans:
Discussion and Comments: Since all the data were provided, the computation of the shellside pressure drop is straightforward. Note that the total pressure drop contribution
associated with the crossow streams is 24 7 31% and the pressure drop contribution
associated with the window zone is 69%. It should be clear that the pressure drop in the
window zone can be signicant and can also increase substantially with smaller bae
397
cuts and bae spacings. In Chapters 8 and 9, we show how to obtain some of the input
data of this problem.
6.5
Pressure drop in a plate heat exchanger consists of three contributions: (1) pressure drop
associated with the inlet and outlet manifolds and ports, (2) pressure drop within the core
(plate passages), and (3) pressure drop due to the elevation change for a vertical ow
exchanger. The pressure drop in the manifolds and ports should be kept as low as
possible (generally < 10%, but may be as high as 25 to 30% or higher in some designs).
Empirically, it is calculated as approximately 1.5 times the inlet velocity head per pass.
Since the entrance and exit losses in the core (plate passages) cannot be determined
experimentally, they are included in the friction factor for the given plate geometry.
Although the momentum eect [see Eq. (6.28)] is negligibly small for liquids, it is also
included in the following p expression. The pressure drop (rise) caused by the elevation
change for liquids is given by Eq. (6.5). Summing all contributions, the pressure drop on
one uid side in a plate heat exchanger is given by
p
1:5G2p np 4fLG2 1
1
1 G2 m gL
2gc i
2gc De m
o i g c
gc
6:44
where Gp m_ ==4D2p is the uid mass velocity in the port, np is the number of passes on
the given uid side, De is the equivalent diameter of ow passages (usually, De equals
twice the plate spacing), and o and i are uid mass densities evaluated at local bulk
temperatures and mean pressures at outlet and inlet, respectively.
Example 6.3 A 1-pass 1-pass plate heat exchanger with chevron plates is being used to
cool hot water with cold water on the other uid side. The following information is
provided for the geometry and operating conditions: number of ow passages 24 on
the hot-water side, plate width 0.5 m, plate height 1.1 m, port diameter 0.1 m, channel
spacing 0.0035 m, equivalent diameter 0.007 m, hot-water ow rate 18 kg/s, mean
dynamic viscosity 0.00081 Pa s, and mean density 995.4 kg/m3 for both manifolds and
core. The hot water is owing vertically upward in the exchanger. The friction factor for
the plates is given by f 0:8 Re0:25 , where Re GDe = is the Reynolds number.
Compute the pressure drop on the hot-water side.
SOLUTION
Problem Data and Schematic: The following information is provided:
np 1
Np 24
De 0:007 m
w 0:5 m
m_ 18 kg=s
L 1:1 m
Dp 0:1 m
0:00081 kg=m s
b 0:0035 m
995:4 kg/m3
f 0:8 Re0:25
A typical plate heat exchanger is shown in Fig. 1.16.
Determine: The hot-water-side pressure drop for this plate heat exchanger.
398
Assumptions: All assumptions of Section 6.1.4 involved in the pressure drop evaluation
are invoked here.
Analysis: We compute the pressure drop using Eq. (6.44). The mass velocity through the
port is given by
Gp
18 kg=s
m_
2291:83 kg=m2 s
2
=4Dp =40:1 m2
18 kg=s
m_
428:57 kg=m2 s
Ao 0:042 m2
where
Ao Np w b 24 0:5 m 0:0035 m 0:042 m2
Next, compute the Reynolds number and the friction factor as follows:
Re
3704
0:00081 kg=m s
1:5G2p np 4fLG2
2gc i
2gc De
1
gL
m
m
gc
2 1 0:007 m
2 1 995:4 kg=m3
1
995:4 kg=m3 9:87 m=s2 1:1 m
1
995:4 kg=m3
Ans:
Note that we did not include the momentum eect term of the pressure drop above since
the inlet and outlet densities were not given primarily because the inlet and outlet density
dierence will be negligible for water.
Discussion and Comments: The pressure drop evaluation is straightforward for this
exchanger. For this problem, the pressure drop due to the elevation change is quite
signicant. However, if it had been a two-pass exchanger, the pressure drop and pressure
rise due to the elevation change would have been almost canceled out for vertical ows in
the exchanger. Note also that the port pressure drop is about the same order of magnitude as the core pressure drop. The port pressure drop is generally not a small fraction of
the total pressure drop in plate heat exchangers. Care should be exercised in the design of
the exchanger to minimize the port pressure drop so that more allowable pressure drop is
399
available for the core and that will also result in more uniform ow through the plate
passages.
Pipe Losses
The pressure drop associated with a pipe of constant cross section, due to wall friction, is
given by Eq. (6.29) as
p f
4L u2m
Dh 2gc
6:45
where f is the Fanning friction factor, generally dependent on the Reynolds number and
ow cross-section geometry. In turbulent ow, f is also dependent on the surface roughness of the pipe. Fluid mass density is evaluated at local bulk temperature and mean
pressure.
The Fanning friction factor as a function of Re and e=di e=Dh for a circular tube is
presented in Fig. 6.4. Here e is the surface roughness magnitude (average height) and
di Dh is the tube inside diameter. The results are valid for fully developed laminar and
turbulent ows. Notice that the surface roughness has no inuence on the f factors in
laminar ow. If the ordinate of Fig. 6.4 is changed to 4f Darcy friction factor [see Eq.
(7.20)], the resulting gure is referred to as the Moody diagram. The turbulent ow f
factors of Fig. 6.4 are also valid for noncircular pipes provided that a proper value of the
hydraulic diameter is used in Eq. (6.45). The laminar friction factors f are dependent on
the cross-section geometry and are presented in Chapter 7. Further explanation of this
gure and the theory are presented in a subsection on Circular Tube with Surface
Roughness in Section 7.4.1.3.
6.6.2
400
FIGURE 6.4 Fanning friction factors for smooth and rough circular tubes (From Bhatti and
Shah, 1987.)
applying the momentum equation and Bernoulli equation across the sudden expansion
(sections 3 and 4 in Fig. 6.1), it can be shown rigorously that
Ke 1 2
6:46
6:47
This represents a 25% higher value of Kc than the more conservative value of Kays
and London (1998) in Fig. 6.3a for Re 1. The pressure drops due to sudden contraction and expansion are then determined from Eqs. (6.22) and (6.25), respectively, with the
appropriate values of the density and mean velocities.
Example 6.4 Determine the eect of a change in the cross-sectional area on the pressure
drop of a square pipe. The original square pipe has a cross section side length of 70.7 mm.
401
To reduce the pressure drop, an engineer decided to double the pipe cross section with a
side length of 141.4 mm over the pipe length of 1 m. Consider air owing at 0.05 kg/s at
278C. The air density and dynamic viscosity are 1.1614 kg/m3 and 184.6 107 Pa s,
respectively. For fully developed turbulent ow for rectangular (and square) ducts,
consider
f 0:0791Re0:25 1:0875 0:1125
*
where
* is the aspect ratio of the rectangular ow passage.
Problem Data and Schematic: The schematics of the pipe cross sections are shown in Fig.
E6.4. The following data are provided:
a1 b1 0:0707 m
* 1
a2 b2 0:1414 m
1:1614 kg=m3
L 1m
m_ 0:05 kg=s
1m
(a)
70.7 mm
141.4 mm
b1
b2
a1
a2
1m
a2 = b2 = 141.4 mm
(b)
Determine: The increase or decrease in pressure drop due to the sudden expansion and
sudden contraction associated with the square pipe.
Assumptions: It is assumed that the ow entering the small cross-sectional-area pipe is
fully developed turbulent ow and is isothermal throughout the ow length.
Analysis: Let us rst compute the necessary geometry and other information to determine the pressure drop desired.
Flow area before the sudden expansion:
402
Ao;1 5 103 m
0:25
Ao;2
0:02 m
The hydraulic diameter of the ow passage for a square duct is the length of its side:
Dh;1 0:0707 m
Dh;2 0:1414 m
The mass velocities G based on the ow area before and after the sudden expansion with
subscripts 1 and 2 are
G1
0:05 kg=s
m_
10 kg=m2 s
Ao;1 5 103 m2
and
G2 2:5 kg=m2 s
GDh
10 kg=m2 s 0:0707 m
38,304
1
184:6 107 kg=m s
GDh
2:5 kg=m2 s 0:1414 m
19,152
2
184:6 107 kg=m s
Using the equation given, let us compute the friction factor for the small- and large-crosssectional-area ducts
f1 0:0791Re0:25 1:0875 0:1125
*
0:0791 38,3040:25 1:0875 0:1125 1 0:005513
f2 0:0791Re0:25 1:0875 0:1125
*
0:0791 19,1520:25 1:0875 0:1125 1 0:006556
Now we determine the sudden expansion and contraction losses for the large pipe using
Fig. 6.3. In both cases, we calculate the mass velocity for the small-cross-sectional-area
pipe. The sudden expansion and sudden contraction coecients for the square pipe for
Re 38304 and 0:25 from Fig. 6.3 are
Ke 0:55
Kc 0:47
Hence, the pressure rise due to sudden expansion at the entrance to the large-crosssectional-area pipe, using Eq. (6.26), is given by
pexp
2
10 kg=m2 s
G2
2
1 Ke
403
The pressure drop due to sudden contraction at the exit of the large-cross-sectional-area
pipe, from Eq. (6.23), is
pcon
2
10 kg=m2 s
G2
2
1 Kc
2
2:5 kg=m2 s
4L G2
4 1m
0:499 Pa
0:006556
Dh;2 2gc
0:1414 m 2 1 1:1614 kg=m3
Thus the total pressure drop associated with the large cross-sectional area pipe due
sudden expansion, friction and sudden contraction is
pt pexp pfr pcon 16:68 0:499 60:59 Pa 44:41 Pa
For a straight small cross-sectional area pipe of 1 m length, the associated pressure
drop is only for the friction component and is given by
pfr f1
2
10 kg=m2 s
4L G21
4 1m
13:43 Pa
0:005513
Dh;1 2gc
0:0707 m 2 1 1:1614 kg=m3
Thus, the pressure drop for the pipe with large cross-sectional area and 1 m length is
44.41 Pa; while the pressure drop for the straight pipe without any change in cross section
is 13.43 Pa.
Ans.
Discussion and Comments: From this example it is found that the frictional pressure
drop in the small-cross-sectional-area pipe is increased by a factor of about
27 (13.43 Pa/0.499 Pa) compared to that for a pipe with four times the ow area and
double the hydraulic diameter. Despite this signicant increase in the frictional component, there are no other pressure losses, whereas for the large-cross-sectional-area pipe,
the sudden expansion and contraction losses are signicantly larger than the frictional
pressure loss contribution. Hence, the increase in cross-sectional area for supposedly
reducing pressure drop, in fact, increases the pressure drop.
However, for this example, if the length of the large cross-sectional pipe were
increased over 3.4 m (by comparing the total pressure drop for each pipe), its total
pressure drop would be lower than that for the small-cross-sectional-area pipe. This is
because the expansion and contraction losses remain constant irrespective of the pipe
length.
6.6.3
Bend Losses
In a number of applications, the inlet and outlet pipes that carry the uids into and out of
heat exchangers have various bends. These pipes may have a circular or rectangular cross
section with a certain bend deection angle b (see Fig. 6.5) or a miter bend (see inset in
Fig. 6.12) with a circular cross section.
404
u2m
2gc
6:48
where is the uid density evaluated at local bulk temperature and mean pressure, um the
mean axial velocity (both and um evaluated at the entrance of the bend), and Kb;t is the
total pressure drop coecient due to the bend. The total pressure drop for a bend consists
of two contributions: (1) the pressure drop for the bend due to the curvature eect,
the ow development eect in the outlet pipe, and the surface roughness eect, and (2)
the pressure drop associated with the outlet straight pipe of specied surface roughness.
The pressure loss coecients Kb and Kf for these two contributions to the pressure drop
are strongly dependent on the ow Reynolds number. Thus,
Kb;t Kb Kf Kb f
4L
Dh
6:49
405
where the Fanning friction factor f for the outlet pipe is determined from Fig. 6.4 for the
appropriate surface roughness size, and the bend pressure loss coecient Kb is given by
Miller (1990) as
Kb K*b CRe Cdev Crough
6:50
Here K*b is the bend pressure loss coecient evaluated at Re um di = 106 , CRe is the
correction factor for the actual Reynolds number for a given application, Cdev is the
correction factor for ow development in the outlet pipe, and Crough is the correction
factor for the pipe surface roughness. Crough is given by
Crough
frough
fsmooth
6:51
where fsmooth is the friction factor for a hydraulically smooth pipe and frough is the
friction factor for the given or assumed roughness for the pipe bend. Both these friction
factors can be obtained from Fig. 6.4. Values of K*b , CRe , and Cdev are provided next for
pipe bends of three dierent geometries.
6.6.3.1 Pipe Bends with a Circular Cross Section. The pipe bend is shown in Fig. 6.5
with important geometrical parameters as the pipe inside diameter di , the radius of
curvature rc of the bend, the bend deection angle b , and the outlet pipe length L. The
bend total pressure loss coecient Kb;t is computed using Eq. (6.49) with Kb from Eq.
(6.50), where K*b at Re 106 is given in Fig. 6.6. Since the actual Reynolds number for
a given application in general will be dierent, a correction factor CRe for the Reynolds
number is applied. CRe is presented in Fig. 6.7 as a function of Re for various values of
rc =di . For bends with rc =di < 1, CRe is strongly dependent on Re. It is calculated as
follows (Miller, 1990):
1. For 0:7 < rc =di < 1 or K*b < 0:4, use CRe from Fig. 6.7 for bends with rc =di 1:
2. For other rc =di < 1 and K*b 0:4; compute CRe from the equation
CRe
K*b
0 0:2
K*b 0:2CRe
6:52
0
where CRe
is CRe from Fig. 6.7 for rc =di 1.
The outlet pipe length correction Cdev is presented in Fig. 6.8 as a function of L=di and
K*b . For rc =di > 3 and/or b > 1008; Cdev 1 (Miller, 1990). Refer to Miller (1990) for K*b
values for short outlet pipes.
The pb of Eq. (6.48) is not underestimated using the pressure loss coecient Kb;t of
Eq. (6.49) if the inlet pipe lengths (Li ) comply with the following conditions (Miller,
1990): (1) Li =di > 2 when the bend follows another component which has a pressure loss
coecient of less than 0.25 at Re 106 , and (2) Li =di > 4 when the bend follows another
component which has a pressure loss coecient greater than 0.5 at Re 106 .
Example 6.5 In a 908 circular bend, water at 258C ows at 2 kg/s. The pipe inside
diameter is 25 mm, the radius of curvature of the bend is 150 mm, and the average height
of the pipe inside surface roughness is 0.025 mm. The downstream straight pipe length is
0.25 m. Assume that the ow entering the pipe is fully turbulent. Compute the pressure
406
FIGURE 6.6 Bend pressure loss coecient Kb* at Re 106 for circular cross section bends. (From
Miller, 1990.)
FIGURE 6.7 Reynolds number correction factor CRe as a function of Re and rc =di . (From Miller,
1990.)
407
FIGURE 6.8 Outlet pipe length correction factor Cdev as a function of the outlet pipe L=di and
Kb*. (From Miller, 1990.)
drop associated with the bend and the downstream pipe. Use the following properties
for water: density 997 kg/m3 and a dynamic viscosity of 0.000855 Pa s.
SOLUTION
Problem Data and Schematic: The schematic of the pipe bend is shown in Fig. E6.5
together with the pipe data.
L = 0.25 m
Flow
rc = 150 mm
di = 25 mm
di = 0.025 m
rc = 0.150 m
b = 90
e = 0.000025 m
L = 0.25 m
= 997 kg/m2
= 0.000855 Pa s
FIGURE E6.5
Determine: Bend and downstream pipe pressure drops for the specied pipe.
Assumptions: The ow entering the bend is fully developed turbulent ow.
408
Analysis: Let us rst compute the necessary geometry data and other information to
determine the desired pressure drops.
2
d 0:0252 0:0004909 m2
4 i
4
2 kg=s
m_
Mass velocity : G
4074:15 kg=m2 s
Ao 0:0004909 m2
Pipe flow area : Ao
Flow velocity : um
2 kg=s
m_
4:086 m=s
Ao 997 kg=m3 0:0004909 m2
Reynolds number : Re
855 106 kg=m s
frough 0:0056
The bend pressure drop is computed using Eqs. (6.48)(6.50). The bend loss pressure
loss coecient Kb from Eq. (6.50) is
Kb K*b CRe Cdev Crough
For rc =di 0:150 m=0:025 m 6 and b 908, we get K*b 0:20 from Fig. 6.6. For
Re 0:119 106 and rc =di 6, we get CRe 1:48 from Fig. 6.7. For K*b 0:20 and
L=di 0:25 m=0:025 m 10, we get Cdev 0:86 from Fig. 6.8. The correction factor for
the pipe surface roughness, from Eq. (6.51), is
Crough
frough
0:0056
1:30
fsmooth 0:0043
4L
4 0:25 m
0:331 0:0056
0:555
Dh
0:025 m
u2m
997 kg=m3 4:086 m=s2
0:555
4619 Pa 4:62 kPa
2gc
21
Ans:
For comparison purposely let us compute the pressure drop associated with the
straight pipe of the same length and surface roughness. The straight length equals the
409
length of 908 bend with the radius of curvature as 150 mm plus the straight pipe of 0.25 m
downstream.
Leq;st
rc L 0:150 m 0:25 m 0:486 m
2
2
Using the friction factor for a rough pipe as 0.0056 from the above,
p
4Leq;st
u2
4 0:486 m
997 kg=m3 4:086 m=s2
f m
0:0056
Dh
2gc
21
0:025 m
* a=b 0:5, 1, and 2, respectively. The magnitude of CRe , the Reynolds number
correction factor (for actual Re 6 106 ), is the same as that given for the circular crosssection bend in Section 6.6.3.1. The outlet pipe length correction Cdev is determined from
Fig. 6.8 for the circular cross section bend with the following modications:
1.
2.
3.
4.
The values for K*b for short outlet pipes are given by Miller (1990). If the inlet pipe
length exceeds 4Dh , the pressure loss computed for a rectangular bend with the foregoing
methodology will probably be conservative.
6.6.3.3 Miter Bends. For the miter bends of circular and rectangular cross sections
(see inset in Fig. 6.12), the procedure to determine Kb;t is identical to that for a circularcross-section pipe bend (see Section 6.6.3.1) except that K*b is determined from Fig.
6.12; and no correction is applied for the outlet pipe ow development (i.e., Cdev 1:
The loss coecients for composite miter bends and other bends are given by Miller
(1990) and Idelchik (1994).
410
FIGURE 6.9 Bend pressure loss coecient Kb* at Re 106 for a rectangular cross section bend
with the aspect ratio
* a=b 0:5. (From Miller, 1990.)
FIGURE 6.10 Bend pressure loss coecient Kb* at Re 106 for a square cross section bend.
(From Miller, 1990.)
411
FIGURE 6.11 Bend pressure loss coecient Kb* at Re 106 for a rectangular cross section bend
with the aspect ratio
* a=b 2. (From Miller, 1990.)
FIGURE 6.12 Bend pressure loss coecient Kb* at Re 106 for a miter bend. (From Miller,
1990.)
412
Example 6.6 Determine the pressure drop in a 908 miter bend using all the data given in
Example 6.5 for a 908 circular bend.
SOLUTION
Problem Data and Schematic: The schematic of the 908 miter bend is shown in Fig. E6.6
together with the input data.
b = 90
Flow
L
di
di = 0.025 m
b = 90
e = 0.000025 m
L = 0.25 m
= 997 kg/m2
= 0.000855 Pa s
FIGURE E6.6
Determine: Miter bend and downstream pipe pressure drop for the specied pipe.
Assumptions: The ow entering the bend is fully developed turbulent ow.
Analysis: Most of the information generated for the 908 circular bend is also applicable
here. For the miter bend, as mentioned in Section 6.6.3.3, Cdev 1 and K*b 1:2 from
Fig. 6.12. From Example 6.4, CRe 1:48 and Crough 1:30: Hence,
Kb K*b CRe Cdev Crough 1:2 1:48 1 1:30 2:31
and
Kb;t Kb Kf Kb f
4L
4 0:25 m
2:31 0:0056
2:534
Dh
0:025 m
The pressure drop associated with the 908 miter bend is then
pb Kb;t
u2m
997 kg=m3 4:086 m=s2
21,090 Pa 21:09 kPa
2:534
2gc
21
Ans:
Discussion and Comments: The pressure drop associated with the miter bend is 21.09 kPa;
for the same uid and ow rates, the pressure drop for the circular bend is 4.62 kPa. Thus
the miter bend resulted in an increase in the pressure drop by about 4.5 times for this
example. Hence, whenever there is an option, a circular bend is generally preferred over a
miter bend for a lower pressure drop through the bend.
6.7
For most heat exchanger surfaces, the core pressure drop [the core friction term in Eq.
(6.28)] for a given heat transfer surface is determined experimentally (using a small size
413
heat exchanger) as a function of the uid ow rate. Details on the experimental method
are presented in Section 7.3.4. These data may then be used for the design and analysis of
heat exchangers of dierent physical size, operating at dierent temperatures and/or
pressures, and operating with dierent uids compared to those for the test exchanger,
but using the same heat transfer surface. Hence, we need to present the test exchanger
core p vs. m_ results in a universal form so that they can be used for operating conditions, physical sizes, and uids beyond those for the test exchanger.
If the detailed geometry (such as the hydraulic diameter Dh , minimum free ow area
Ao , etc.) of the heat exchanger surface is known, the best approach is to present the core
pressure drop versus mass ow rate results in nondimensional form, such as an f vs. Re
curve. If the detailed geometry information is not available, the measured core p vs. m_
results are presented in dimensional form. Both approaches are presented next.
6.7.1
The core pressure drop in a dimensionless form is generally presented in two alternative
forms,y in terms of the pressure loss coecient K or Euler number Eu and the Fanning
friction factor f dened by
p K
p f
u2m
u2
Eu m
2gc
2gc
4L u2m
Dh 2gc
6:53
6:54
Thus, the Euler number is the same as the pressure loss coecient K. If the pressure loss
coecient K is constant along the ow length, such as in tube banks, manifolds, bends,
valves, and so on, as a result of turbulent ow, usually K or Eu is used to present the
pressure drop. For a tube bank with Nr rows, the Euler number is generally dened as an
Nr row average as in Eq. (6.37) for Nr;cc Nr .
The Fanning friction factor generally represents primarily the frictional component of
the pressure drop and is used when the given heat transfer surface has approximately the
same frictional pressure drop per unit length along the ow direction. Thus, use of the
friction factor allows pressure drop prediction of dierent ow lengths of the heat
exchanger surface. In uid dynamics books, generally the Darcy friction factor fD is
used and is related to the Fanning friction factor f as
fD 4f
p
Dh
u2m =2gc L
6:55
A comparison of Eqs. (6.54) and (6.55) indicates that one needs to know the hydraulic
diameter and ow length of the exchanger surface if the pressure drop is presented in
terms of the friction factor. No such information is needed for p to be presented in
terms of K or Eu.
See also Eq. (7.115) for presenting pressure drop in terms of the Hagen number.
414
The uid ow rate is presented in dimensionless form as the Reynolds number dened
as
Re
Ao
Ao
6:56
6.7.2
m_ std Dh m_ act Dh
std Ao
act Ao
6:57
Thus,
m_ std m_ act
std
act
6:58
Since m_ V_ , we get the following relationship between V_ std and V_ act using Eq. (6.58):
V_ std V_ act act std
std act
6:59
y
Note that the pressure drop is identical for two exchangers having dierent frontal areas but identical heat
transfer surfaces and ow length when the mass velocities G are identical in both exchangers. Thus, dierent
frontal areas are taken into account if p is plotted against G rather than m_ . However, in this case, the information
on the minimum free-ow area is required to determine G m_ =Ao .
415
std
act
6:60
2gc Dh
2gc A2o Dh
p
u2m 4L
m_ 2 4L
6:61
Now let us match the standard and actual friction factors for a given geometry using
Eq. (6.61).
fstd fact )
m_ 2std
m_ 2act
6:62
act m_ std 2
std m_ act
6:63
Thus
pstd pact
act std 2
std act
6:64
This is the most general relationship between pstd and pact that requires both density
and viscosity corrections, and should be used in all cases.
To illustrate the aforementioned eects of density and viscosity corrections, the
experimental pressure drop p with air as a function of the measured mass velocity G
is shown in Fig. 6.13a for a heat exchanger. The pressure drop p vs. mass velocity G
tests were conducted with air at six dierent inlet temperatures under isothermal conditions. There is about a 17% spread in the experimental pressure drop over the temperature range covered. The same results are replotted in Fig. 6.13b with a traditional density
correction to a standard temperature of 158C (598F) [i.e., the pressure drop is corrected
using Eq. (6.66), and no correction is applied to the measured mass velocity G]. The
results collapse to about a 4% spread with the test temperature. Figure 6.13c shows the
recommended correction of density and viscosity [i.e., the pressure drop is corrected
using Eq. (6.64) and the mass velocity is corrected using Eq. (6.60)]. The results collapse
to a single curve, as expected from the foregoing theoretical basis.
However, in industry the viscosity correction in Eq. (6.64) is usually neglected
and only the density correction is applied. It can be rationalized as follows. When the
friction factor is constant (independent of the Reynolds number for turbulent ow, such
as in a rough pipe; see Fig. 6.4) or when f is not strongly variable for turbulent ow in a
smooth pipe, we do not need to match Restd and React as is done in Eq. (6.57). Hence, in
this case,
m_ std m_ act
6:65
416
FIGURE 6.13 Pressure drop p as a function of the mass velocity G for a heat exchanger:
(a) measured data; (b) density correction applied to p; (c) density and viscosity correction
applied to p and the density correction applied to G.
417
FIGURE 6.14 Pressure drop p as a function of the mass velocity G for a perforated plate:
(a) measured data; (b) density correction applied to p; (c) density and viscosity correction
applied to p and the density correction applied to G.
418
act
std
6:66
To demonstrate that the viscosity correction is not signicant for a turbulent ow (or
where the pressure drop is approximately proportional to the velocity square), isothermal tests similar to those of Fig. 6.13a are shown in Fig. 6.14a with air at three temperatures on a perforated plate. The results are shown in Fig. 6.14b with the density
correction alone, and in Fig. 6.14c with the density and viscosity corrections together.
As it is found, whether only the density correction is applied (Fig. 6.14b) or both the
density and viscosity corrections are applied (Fig. 6.14c), the results of these tests collapse
to a single curve, as anticipated from the theoretical basis outlined above.
In most situations, whether or not f is constant is not known a priori, and therefore
Eq. (6.64) is recommended for correcting pact to pstd and Eqs. (6.58), (6.59), and
(6.60) for correcting m_ ; V_ , and G, respectively, for a plot of p vs. m_ ; V_ , or G at standard
conditions.
f
Re
>
>
< D3h 2gc A
p
"
#
>
>
>
1 0:046 0:2 4L2:8 1:8
>
>
_
m
>
: D3 2gc A1:8
h
6:67a
6:67b
p /
8
1
1
>
>
>
< D3 to D4
h
h
6:68a
>
1
1
>
>
: 3 to 4:8
Dh
Dh
6:68b
SUMMARY
p /
419
8
1
>
>
L;
>
<
Ao
6:69a
>
1
>
>
: L; 1:8
Ao
6:69b
p /
8
1
>
>
>
<;
6:70a
>
>
1
>
: ; 0:2
6:70b
L
A2o Dh
6:71
for a specied uid and its mass ow rate. Since the friction factors and Colburn factors
(or Nu) for enhanced heat transfer surfaces are higher than those for a plain surface, both
heat transfer rate and pressure drop for an enhanced surface will be higher than those for
a plain surface for a given uid mass ow rate m_ and the same heat exchanger dimensions. However, it is possible to maintain high-heat-transfer performance with the same
pressure drop (particularly in laminar ows) by properly choosing the heat exchanger
dimensions and surface geometry [i.e., the proper choice of L, Ao , and Dh in Eq. (6.71) so
that p remains the same]. Refer to Example 10.3 to gain some insight into this concept.
However, if the exchanger frontal area cannot be changed, one will end up with a larger
pressure drop with enhanced surface compared to that for the plain surface for a given
uid ow rate. In that case (i.e., for a xed frontal area), a plain unenhanced surface will
have a lower pressure drop but longer exchanger ow length (and hence larger volume
and mass of the exchanger) to meet the heat transfer requirement specied.
SUMMARY
In this chapter, important issues related to the pressure drop analysis and data presentation for a heat exchanger as a component are discussed with the following highlights.
. In a heat exchanger, the pressure drop (in addition to heat transfer) becomes an
important design consideration for gases in laminar, transition, and turbulent ow
and for highly viscous liquids in laminar ow. For other liquids, and particularly in
turbulent ow, the pressure drop is not as critical a design consideration as heat
transfer.
. Pressure drop associated with a heat exchanger is made up of two contributions:
(1) pressure drop associated with a core/matrix/heat transfer surface where heat
transfer takes place, and (2) pressure drop associated with ow distribution devices
420
that contribute to the pressure drop without any eective heat transfer. Ideally,
most of the design pressure drop allowed should be associated with the rst
contributionthe heat transfer surface.
Core pressure drop may consist of the following contributions: (1) core friction,
(2) entrance eect, (3) exit eect, (4) momentum eect, and (5) elevation change
eect. The rst contribution should be ideally above 80 to 90% because that
portion of the pressure drop is utilized where heat transfer takes place. The rst
four contributions are included in Eq. (6.28) and the fth one is given by Eq. (6.5).
Pressure drop associated with the following ow distribution devices is presented in
Section 6.6: (1) pipes, (2) sudden expansion and contraction at the inlet/outlet, and
(3) bends. For other ow distribution devices, refer to Miller (1990) and Idelchik
(1994).
For most heat exchanger surfaces, pressure drop data are obtained experimentally.
They are presented in dimensional or dimensionless form as outlined in Section 6.7.
When presented in dimensional form, they should be corrected for both density and
viscosity changes, as shown in Eq. (6.64).
Pressure drop is a strong function of the passage hydraulic diameter, and it is also
dependent on the exchanger ow length, free-ow area, and heat transfer surface
area. The specic functional dependency is shown in Eqs. (6.68) and (6.69). The
pressure drop is also dependent on the type of uid used, particularly its density
and viscosity, as shown in Eq. (6.70).
REFERENCES
Bell, K. J., 1988, Delaware Method for shell-side design, in Heat Transfer Equipment Design, R. K.
Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere Publishing, Washington, DC,
pp. 145166.
Crane Co., 1976, Flow of Fluids through Valves, Fittings, and Pipes, Technical Paper 410, Crane Co.,
Chicago.
Idelchik, I. E., 1994, Handbook of Hydraulic Resistance, 3rd ed., CRC Press, Boca Raton, FL.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger
Publishing, Malabar, FL.
Miller, D. S., 1990, Internal Flow Systems, 2nd ed., BHRA (Information Services), Craneld, UK.
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
6.1
6.2
Heat exchangers are ideally designed to have a signicantly higher pressure drop
in inlet and outlet headers than in the core.
(a) depends on the application
(b) true
REVIEW QUESTIONS
(c) false
(e) plate-n exchangers
421
6.3
Entrance and exit losses in a given uid side of an exchanger are important:
(a) for high core p
(b) at high Reynolds numbers
(c) for a high ratio of frontal to
(d) for long cores
free-ow area
6.4
Entrance and exit loss terms are usually treated as zero for:
(a) crossow to tube banks
(b) rotary regenerator with cylindrical ow passages
(c) plate heat exchangers
(d) normal ns on individual tubes
6.5
In total core pressure drop evaluation, the entrance and exit pressure losses are
generally:
(a) additive
(b) compensating
(c) cant tell in general
6.6
Two designs meet heat transfer performance requirement for a uid stream
(uid 1) in a plate heat exchanger: (1) a 1-pass 1-pass design (see Fig. 1.65a)
with 10 ow channels for uid 1 and 11 uid channels for uid 2, and (2) a series
ow arrangement (see Fig. 1.65f ) with seven ow channels for uid 1 and eight
ow channels for uid 2. If the ow is fully developed turbulent in both designs
(i.e., f is about the same), the pressure drop ratio for uid 1 for design 2 to design
1 is approximately:
(a) 0.7
(b) 0.49
(c) 100
(d) 700
(e) 343
(f ) 1.43
6.7
6.8
6.9
6.10
The pressure drop in the uid distribution system is generally presented in terms
of:
(a) Euler number
(b) Reynolds number
(c) velocity head
(d) momentum ux correction factor
6.11
422
6.12
Arrange the following exchangers in the order of most allowed design pressure
drop p to least allowed p.
(a) noncompact liquid-to-liquid exchanger
(b) highly compact gas-to-gas exchanger
(c) moderately compact gas-to-gas exchanger
(d) compact water-to-water exchanger
6.13
1
2
1
f
1
K
c
e
2gc i pi 6
o
7
pi
rh i m
4|{z} |{z}
|
{z} |{z}o 5
1
2
Identify which terms of this equation are treated as zero for the following
exchangers:
(a) gas-to-gas waste heat recovery exchanger:
(b) n side of a circular nned tube:
(c) shell side of a shell-and-tube exchanger for sulfuric acid cooling:
(d) plate-n exchanger during isothermal testing:
6.14
Air at a given mass ow rate goes in series through two passes of the same length
in a heat exchanger. Now consider an alternative arrangement where those two
passes are in parallel coupling and the total airow rate is the same (i.e., these
cases correspond to Fig. 3.21 vs. Fig. 3.22 for uid 2). The air-side core pressure
drop for the second case will be the following factor of the core pressure drop for
the rst case, assuming fully developed laminar ow through the exchanger. Note
that except for dierent coupling of two passes, both passes of both exchangers are
identical.
(a) 1
(b) 12
(c) 2
1
(e) 4
(d) 4
PROBLEMS
6.1 Determine the pressure drop on the hot and cold sides of the regenerator of
Problem 5.1 having " 95%. Compute p=pi on each side in percent and determine
the total p=p for this regenerator. Generally, design p=ptotal is kept below 4
to 5%. Discuss your results. Treat gas as air. Use 0:369 104 and
0.360 104 Pa s for the hot- and cold-uid sides, respectively, and f 14=Re:
6.2 Determine the pressure drops on the hot and cold sides of the regenerator of
Example 5.3. The following are additional data for the surface: f 17=Re,
Dh 0:44 mm, 17 105 Pa s, pa;i 442 kPa and pg;i 150 kPa.
6.3 Explain why Ke can be negative for some values of and Re for all geometries of
Fig. 6.3. Explain why Kc is not negative for any value of and Re.
PROBLEMS
423
6.4 The objective is to evaluate the pressure drop for air and water at (1) equal velocity
and (2) equal ow rate through a 25-mm-inside-diameter 3-m-long circular tube.
Air and water are at 258C. Assume f 16=Re for laminar ow and
f 0:00128 0:1143Re0:311 for turbulent ow. The following uid properties
are specied:
Thermophysical Properties
Specic heat cp (J/kg K
Dynamic viscosity (Pa s)
Thermal conductivity k (W/m K)
Prandtl number Pr
Density (kg/m3 )
Air
Water
1009
0:184 104
0.0261
0.71
1.183
4187
8:853 104
0.602
6.16
996.4
(a) Consider air and water velocities at 6 m/s. Determine the following:
(i) Air ow and water ow rates.
(ii) Pressure drop for each uid, neglecting entrance and exit losses.
(iii) Fluid pumping power for each uid.
(b) Now assume that air and water ow rates are the same as the airow rate in
part (a)(i). Determine the following.
(i) Pressure drop for each uid, neglecting entrance and exit losses.
(ii) Fluid pumping power for each uid.
(c) Discuss the results of parts (a) and (b) from an engineering viewpoint.
6.5 To select an appropriate shell type for a given application (with single-phase uids
on both sides), we want to evaluate all major shell types: E, F, G, H, J, and X.
Because of the thermal stress considerations, use U tubes (two tube passes) in a
single-shell exchanger for all shell types specied.
(a) For the identical tube uid ow rate, will the pressure drop on the tube side be
the same or dierent in the types of shells above for the same eective shell
length? Why? Assume all uid properties to be constant.
(b) Estimate the shell-side pressure drop for each shell type as a function of um and
L, where um is the mean shell-side velocity in the E shell and L is the shell length.
Note that the shell-side ow rate is the same for all shell types and each has
single-segmental baes/support plates at the same spacing. The shell and tube
diameters and number of tubes are the same for all shell types. The shell length
is much greater than the shell diameter. Select a shell type for the lowest shellside pressure drop. Hints: Dont forget to add ps qualitatively due to 1808
bends. Dont use Eqs. (6.37) and (6.39).
6.6 Estimate whether or not the pressure drops through the core are going to be acceptable for the design given below. The inlet temperature, pressure, and mass ow rate
of an airstream entering a gas-to-gas two-uid heat exchanger are: 2148C, 490 kPa,
and 21 kg/s, respectively. The outlet temperature of that uid is 3468C. The mass
ow rate of the other airstream is the same as for the rst, while the inlet temperature and pressure are 4178C and 103 kPa, respectively. The pressure drops are
limited to 4.9 kPa for the rst uid stream and to 2.57 kPa for the second uid
stream. Fanning and Colburn friction factors can be calculated using the following
424
correlations for a plain plate-n surface with the designation 19.86 of Kays and
London (1998):
j or f exp a0 ra1 rfa2 ra3 ra4 a5 rg
where aj ; j 1; . . . ; 5 are given below, and r lnRe.
f
j
a0
a1
a2
a3
0:81952 103
0:26449 103
0:54562 103
0:16720 103
0:14437 103
0:41495 102
0:18897 102
0:50051 101
a4
a5
For both heat transfer surfaces, the hydraulic radius is 0.001875 m and the extendedsurface eciency is 0.9. The number of transfer units for the exchanger is 4.9. The
heat transfer surface wall temperature may be assumed to be 3178C. Thermal conductivity of the n material and n thickness are 200 W/m K, and 0.152 mm. The
plate spacing is 6.35 mm. The ratio of minimum free-ow area to frontal area for
both heat transfer surfaces is 0.3728, while the minimum free-ow areas on both
sides are 0.4258 m2 and 1.5306 m2 . Elaborate all decisions on the assumptions
required if any of the input data are missing.
In exchanger design, the dimensional heat transfer and pressure drop vs. uid ow test
data obtained for one exchanger size cannot be used to size or rate accurately an exchanger of a dierent size (having the same basic surface geometries) because of the nonlinear
relationships among the geometrical and operating parameters of the exchanger. To rate
or size an exchanger of dierent size or dierent performance requirements, one must
have dimensionless heat transfer and uid friction characteristics{ of heat transfer
surfaces. That is why among the most important inputs to the heat exchanger thermal
and hydraulic design and analysis are accurate dimensionless heat transfer and uid
friction characteristics, as shown in Fig. 2.1. These heat transfer characteristics are
generally presented in terms of the Nusselt number Nu vs. the Reynolds number Re,
the dimensionless axial distance x*, or the Graetz number Gz. The experimental characteristics are usually presented in terms of the Stanton number St or the Colburn factor j
vs. the Reynolds number Re. Flow friction characteristics are generally presented in
terms of the Fanning friction factor f vs. the Reynolds number Re or the dimensionless
axial distance x ; alternatively, they are presented for ow over tube banks in terms of
the Euler number Eu or Hagen number Hg vs. the Reynolds number Re. Depending on
ow and heat transfer conditions, some additional important dimensionless numbers
may be involved, as discussed later. The surface characteristics are obtained analytically
for simple geometries. The surface characteristics are primarily obtained experimentally
for most exchanger surfaces because the ow phenomena are complex due to the geometry features of ow area and/or heat transfer surface. Now more emphasis is placed on
numerical (CFD) analysis to obtain the surface characteristics of complex heat exchanger
surfaces, but full three-dimensional numerical analyses are yet not practical at a reasonable cost covering a wide range of Re values, due to computing limitations.
In this chapter we start with basic concepts in Section 7.1 for understanding the
behavior of the surface characteristics. These will include the concepts of boundary
layers, types of ows usually encountered in industrial heat exchangers, convection
mechanisms, and the basic denitions of the mean velocity, mean temperature and
heat transfer coecient. In Section 7.2, dimensionless groups used for uid ow and
heat transfer characteristics of heat exchanger surfaces are presented along with an
{
We will not use the terminology surface performance data since performance in some industries is related to a
dimensional plot of the heat transfer rate and one uid-side pressure drop as a function of its uid ow rate for an
exchanger. Note that we need to distinguish between the performance of a surface geometry (of one side of an
exchanger) and the performance of a heat exchanger.
425
426
illustrative example for airow through a circular tube. Experimental techniques used for
measurement of the heat transfer and ow friction characteristics of heat exchanger
surfaces are presented in Section 7.3. These include the steady state, transient, and
Wilson plot techniques. In Section 7.4, analytical and semiempirical correlations are
presented for heat transfer and ow friction for simple geometries. In Section 7.5, experimental correlations are presented for complex geometries that include tubular surfaces,
plate heat exchanger surfaces, plate-n surfaces, tube-n surfaces, and regenerator
surfaces. The inuence of temperature-dependent uid properties can be signicant in
a heat exchanger. This is discussed in Section 7.6, particularly the property ratio method
for gases and liquids. Although forced convection is the major mode of heat transfer in
many heat exchangers, free convection and radiation heat transfer can be important in
some applications. These issues are summarized briey in Section 7.7. Thus, the major
motivation of this chapter is to outline most important issues related to the understanding and utilization of available literature information for the determination and minor
extrapolation of accurate heat transfer and ow friction characteristics of heat exchanger
surfaces.
7.1
BASIC CONCEPTS
Some of the basic concepts needed to understand heat transfer and ow friction characteristics of heat exchanger surfaces are described in this section. These include the
concepts of a boundary layer, ow types and convection mechanism, and denitions
of the mean velocity, temperature, and heat transfer coecient. Although these concepts
have been introduced in the rst courses in uid mechanics and heat transfer, our
emphasis here is to review them from the heat exchanger application point of view for
understanding and minor extrapolation of design correlations.
7.1.1
Boundary Layers
The concepts of velocity and thermal boundary layers are rst discussed below.
7.1.1.1 Velocity Boundary Layer. The ow eld around a body may be divided into
two regions for the purpose of analysis. The thin region close to the body surface, where
the inuence of uid viscosity becomes increasingly predominant when approaching the
surface, is referred to as the velocity or momentum boundary layer, as shown in Fig. 7.1
for ow over a at plate. The remainder of the ow eld can to a good approximation
BASIC CONCEPTS
427
1
um
m1
ri
FIGURE 7.3 Flow regimes in developing and developed turbulent ow in a tube at a high
Reynolds number Re 104 . (From Shah, 1983.)
428
Note that for a Newtonian uid (m 1), the velocity prole in Eq. (7.1) is governed by
u umax 1 r=ri 2 and umax 2um , where um represents mean velocity and umax the
maximum, centerline velocity.
The empirical, power-law velocity prole for fully developed turbulent ow in a
circular tube is given by
u
umax
r 1=n
1
ri
um
2n2
umax n 12n 1
7:2
where the exponent n varies with Re and the values of n are between 6 and 10 for the
range of Re between 104 and 106 (Hinze, 1975; Bhatti and Shah, 1987). In Eq. (7.2), u, um ,
and umax represent the time-averaged turbulent local, cross-sectional mean, and crosssectional maximum velocities. Thus as Re increases, n increases, and the velocity prole
becomes atter over most of the tube cross section. An analysis of Eqs. (7.1) and (7.2)
reveals that all turbulent velocity proles are signicantly steeper near the wall than the
laminar velocity prole, contributing signicantly higher pressure drops in turbulent
ow, as discussed later.
7.1.1.2 Temperature Boundary Layer. When heat transfer takes place between the
uid and the solid surface, the major temperature change occurs in a region very
close to the surface in most cases. This region is referred to as the temperature or
thermal boundary layer, shown in Fig. 7.4 for laminar internal ow. The temperature
prole across the cross section can be determined integrating the corresponding energy
equation of the boundary layer problem, and it is beyond the scope of this discussion.
The solution approach to determining these proles can be found in any advanced heat
convection textbook (see, e.g., Kays and Crawford, 1993). Specic temperature proles
can be found in Shah and Bhatti (1987) for laminar ow and in Bhatti and Shah (1987)
for turbulent ow. The temperature boundary layer, also sometimes referred to as uid
lm near the surface, may be interpreted in terms of the thermal resistance to heat
transfer from the uid to the solid surface (or vice versa). The heat transfer rate per unit
of surface area across the boundary layer in Fig. 7.4 Tm > Tw is
qw00 k
@T
T Tw
k m
t
@y y0
7:3
BASIC CONCEPTS
429
Instead of expressing the heat ux in terms of the thermal boundary layer thickness t ,
generally it is expressed in terms of the lm coecient{ or the heat transfer coecient
h k=t , where k is the uid thermal conductivity. The thinner the thermal boundary
layer (the smaller t ), the larger will be the heat transfer coecient h and the smaller the
lm resistance. The bulk temperature Tm (for the pertinent denition, see Section 7.1.4.2)
for internal ow is replaced by T1 for external ow; however, in some published correlations, some other temperatures, such as Ti , replaces Tm , and hence the user should check
the denition of the temperature dierence used in the Nusselt number evaluation.
The thermal boundary layer is not necessarily of the same thickness as the velocity
boundary layer. The boundary layer thickness and the growth rates are dependent on the
uid Prandtl number, discussed in Section 7.1.2.4 (Laminar Flow).
7.1.2
Types of Flows
In heat exchangers, many dierent types of ows are encountered because a wide variety
of heat transfer surfaces and a wide range of mass ow rates are being used. Our objective
for describing these ows is (1) to understand the uid ow and heat transfer behavior of
heat transfer surfaces, (2) to develop/apply analytical heat transfer and pressure drop
correlations for these surfaces, and (3) to facilitate further enhancement of heat transfer
for improved heat exchanger performance. With these objectives in mind, we will classify
major types of ows with their specic behavior. We restrict this classication to low
Mach number ows as found in most industrial heat exchangers. Most of the information presented in this section is condensed from Jacobi and Shah (1998).
7.1.2.1 Steady and Unsteady Flows. If the velocity and temperature throughout the
ow eld do not change with time, the ow is said to be steady. Such ows are easier to
model and analyze. Although in most heat exchanger applications, true steady ow is
not encountered, the steady ow approximation is often useful. On the other hand,
unsteady eects can be important.
The most obvious case of an unsteady ow occurs when the boundary conditions,
either velocity (pressure) or temperature, change with time. Such imposed unsteadiness
can be caused by changes in the fan or pump speed or uid and/or wall temperature.
However, even if the boundary conditions are steady, the ow may exhibit unsteadiness.
A simple example is ow normal to a circular cylinder; with a steady approach ow,
unsteadiness and vortex shedding occur downstream. Another example is the ow in an
oset strip n geometry where unsteadiness and vortex shedding (such as von Karman
vortices) occur downstream of the strips. In these cases, unsteadiness is self-sustained in
the ow. For oset strip ns, the self-sustained unsteadiness is periodic in nature along the
ow length after a few strips. A rigorous numerical modeling of such a self-sustained ow
and its experimental verication for a generic ow geometry of communicating channels
indicates a great potential for heat transfer enhancement of such a geometry (Amon et al.,
1992). Imposed or self-sustained unsteadiness can be laminar or turbulent and generally
causes an early transition to turbulent ow. Laminar unsteady ows are preferred over
turbulent ows because the pressure drops in the turbulent ows are higher than those
associated with unsteady laminar ows.
{
It is referred to as the lm coecient since the entire thermal resistance is assumed to be in the boundary layer or
lm.
430
7.1.2.2 Laminar, Transition, and Turbulent Flows. Flow is considered laminar when
the velocities are free of random uctuations at every point in the ow eld; thus, the
uid motion is highly ordered. Fluid particles ow along nearly parallel streamlines
with no rotation. However, in a developing laminar boundary layer (see Figs. 7.1 and
7.2), uid particles along a streamline have u and v velocity components for a 2D
geometry, with the v component being responsible for the momentum or enthalpy
transfer across the boundary layer. In laminar ow, viscous forces are dominant over
inertia forces. For steady laminar ow, all velocities at a stationary point in the ow
eld remain constant with respect to time, but the velocities may be dierent at dierent
points in the ow eld. For unsteady laminar ow, all velocities at a stationary point in
the ow eld will also be time dependent. Laminar ow, also referred to as viscous or
streamline ow, is associated with viscous uids, low ow velocities, and/or small ow
passages.
Flow is considered turbulent when the uid particles do not travel in a well-ordered
manner; however, it is dicult to dene turbulence in simple terms. Some important
characteristics are:
. Turbulent ows have self-sustained, irregular velocity uctuations (u 0 ; v 0 ; w 0 ) in all
directions. In turbulent ows, uid particles travel in randomly moving uid masses
of varying sizes called eddies. The irregularity of these uctuations distinguishes
turbulence from laminar self-sustained oscillations, which are usually periodic.
. Turbulent ows have eddies with a broad distribution of sizes. In a turbulent pipe
ow, the eddy length scale can range from a few millimeters to the pipe diameter.
The range of eddy sizes is the result of the generation of large eddies, their breakup
into smaller and smaller eddies, and their eventual dissipation at small scales. This
broad eddy-size distribution distinguishes turbulence from laminar self-sustained
unsteadinesslaminar unsteadiness usually produces larger eddies (a few length
scales or a few pipe diameters) with a small eddy-size distribution.
. A turbulent ow is accompanied by large-scale mixing due to the advective eect of
turbulent eddies. Mixing increases heat transfer and wall friction when the nearwall region of the ow is aected (refer to the discussion related to Table 7.1).
If a streak of colored dye is injected in turbulent water ow through a transparent
pipe, one will be able to see that the color will immediately disperse throughout the ow
eld. Thus, the transport of momentum and energy transverse to the main ow direction
is greatly enhanced in turbulent ow.
The turbulent boundary layer has a multilayer character. The simplest model includes
two layers, as shown in Fig. 7.3; the near-wall region is referred to as the viscous sublayer
(where molecular diusion plays a major role), and the outer region is called a fully
turbulent region or core (where turbulent mixing is dominant). However, molecular
diusion and turbulent mixing phenomena are dependent on the uid Prandtl number,
as summarized in Table 7.1 and discussed as follows. For liquid metals, Pr < 0:03,
molecular diusion or heat transfer by conduction within the uid in the fully turbulent
region is important in addition to turbulent mixing. Hence, the thermal resistance is
distributed throughout the ow cross section, and the heat transfer coecient is less
dependent on the Reynolds number than for higher Prandtl numbers (Pr > 0:5). For
all uids with Pr > 0:5, in the fully turbulent region, molecular diusion plays a minor
role and the turbulent mixing is dominant. For uids with 0 < Pr < 5, the viscous sub-
BASIC CONCEPTS
431
TABLE 7.1 Dominance of Molecular Diusion and Turbulent Mixing in a Two-Layer Turbulent
Boundary Layer as a Function of the Fluid Prandtl Number
Sublayer
Pr < 0:03
0:5 Pr 5
Pr > 5
Molecular diusion and
turbulent mixing
Turbulent mixing
Viscous sublayer
Molecular diusion
Molecular diusion
Turbulent core
Turbulent mixing
432
become turbulent, since wakes, which include a separated shear layer, become turbulent
at very low Reynolds numbers (50 or less; see Bejan, 1995). As the ow approaches the
next tube row, a favorable pressure gradient will accelerate the ow around the next row
of tubes. In this region of favorable pressure gradient, the ow could relaminarize, to
take on a turbulent character again in the wake. Thus, in this nned tube heat exchanger,
it may be possible to have a conned region of the ow that looks turbulent. A similar
situation occurs in oset-strip n and multilouver n designs at locations downstream
from the ns. How is one to classify such a ow? There is no doubt that a ow with
regions of turbulence should be classied as turbulent. Accurate predictions of the local
ow and heat transfer cannot be obtained without attention to the relevant turbulent
processes. Nevertheless, turbulent mixing is unimportant throughout much of such a
owlocally, the ow appears laminar. This type of a combined-ow situation is common in heat exchangers. Many such ows may be characterized as low-Reynolds-number
turbulent ows Re < 3000.
7.1.2.3 Internal, External, and Periodic Flows. The ows occurring in conning
passages of various regular or irregular, singly or doubly connected, constant or variable cross sections such as circular, rectangular, triangular, annular, and the like are
referred to as internal ows. Associated with these ows is a pressure gradient. The
pressure gradient either decreases along the ow length (favorable pressure gradient) or
increases in the ow direction (adverse pressure gradient), depending on whether the
cross-section size remains constant or increases. In the internal duct ow, the boundary
layers are eventually constrained by the surface, and after the development length, the
entire cross section represents a boundary layer.
The unconned ows occurring over surfaces such as at plates, circular cylinders,
turbine blades, and the like are referred to as external ows. In the external ow, the
boundary layers may continue to grow along the ow length, depending on the surface
geometry. Unlike the internal ows, these ows can occur without a sizable pressure
gradient in the ow direction as in ow past a at plate. Frequently, however, these ows
occur with pressure gradients that may be positive or negative in the ow direction. For
example, in ow over a circular cylinder both negative and positive pressure gradients
occur along the main ow direction.
Another distinction between the internal and external ows is that in the latter the
eect of viscosity is dominant near the solid wall only, with a potential inviscid ow being
away from the wall on the unbounded side. In internal ows, in general, the eect of
viscosity is present across the ow cross section. The only exception is the ow near the
passage inlet where a potential inviscid core develops near the center of the cross section
as shown in Fig. 7.2.
It should be noted that in spite of the enlisted distinctions between internal and
external ows, it is very dicult to dene the ow either as internal or as external in
many practical situations. Heat exchangers are considered to have internal owsa
somewhat articial view. The classication of ow in complex heat exchanger passages
depends on how the boundaries aect the ow. Consider the sketch of at plates shown
in Fig. 7.5. In the rst case, (Fig. 7.5a), the ow path appears long compared to the plate
spacing, and an internal ow is suggested. In Fig. 7.5b, the ow path appears very small
compared to the plate spacing, and an external ow is suggested. The simplied situation
reected in Fig. 7.5c may be dicult to decide since the length scales do not rmly suggest
internal or external ow. Spacing and length are important, but only inasmuch as they
BASIC CONCEPTS
433
FIGURE 7.5 Length scales important in continuous and interrupted passages: (a) ow length
much longer than the plate spacing; (b) ow length much shorter than the plate spacing; (c) ow
length and plate spacing comparable; (d) periodic array (oset strip ns) that provides an interrupted
passage. (From Jacobi and Shah, 1996.)
reect the eect of the boundary layer on the ow. It would be more precise to consider
the boundary layer thickness as a length scale.
When ow develops between neighboring plates (ns), velocity and temperature
boundary layers grow. If the neighboring boundary layers do not interact, the ow is
considered external; however, if the boundary layers interact, the ow should be considered internal. In cases where the ow is aected by boundary layers on neighboring
ns, a blockage eect occurs. That is, as the boundary layers grow (along with the
accompanying displacement boundary layers), the eective ow area for the out-ofthe-boundary-layer ow is reduced. By continuity, the core ow must accelerate and a
favorable pressure gradient (i.e., the static pressure decreasing with increasing ow length
x) must be established. The favorable pressure gradient thins the boundary layers, with a
commensurate increase in skin friction and heat transfer. As with turbulence, it is useful
to admit a paradigm with combined internal and external ows in dierent regions. Such
an approach suggests that ow development is important in classifying an exchanger ow
as internal or external; however, before addressing ow development, it is useful to
discuss periodic ows.
The last situation depicted in Fig. 7.5d shows an array of at plates in the ow. Such
arrays are common in contemporary heat exchanger designs (e.g., in oset strip n
designs). For these extended arrays, it is useful to drop the internal and external classication (view) and adopt the periodic classication, as it reects the boundary condition.
Periodic ows can be developing or fully developed (as discussed in Section 7.1.2.4). In
periodic ows, the surface pattern is spatially repeated, and the boundary conditions may
be normalized in such a way as to take a periodic structure. For such geometries, it is
doubtful that the ow will be steadyinstabilities in the wake regions almost ensure
unsteadiness. Modeling approaches which assume that the ow to be steady will
probably be of little value except at very low Reynolds numbers (Re < 200). It is very
434
7.1.2.4
Laminar Flow. Four types of laminar duct (internal) ows are: fully developed, hydrodynamically developing, thermally developing (abbreviated for thermally developing
and hydrodynamically developed ow), and simultaneously developing (abbreviated
for thermally and hydrodynamically developing ow). A further description of these
ows is now provided with the aid of Fig. 7.6.
Referring to Fig. 7.6a, suppose that the temperature of the duct wall is held at the
entering uid temperature (Tw Te ) and there is no generation or dissipation of heat
within the uid. In this case, the uid experiences no gain or loss of heat. In such an
isothermal ow, the eect of viscosity gradually spreads across the duct cross section
FIGURE 7.6 Types of laminar ows for constant wall temperature boundary condition: (a)
hydrodynamically developing ow followed by thermally developing and hydrodynamically
developed ow; (b) simultaneously developing ow, Pr > 1; (c) simultaneously developing ow,
Pr < 1. Solid lines denote the velocity proles, dashed lines the temperature proles. (From Shah
and Bhatti, 1987.)
BASIC CONCEPTS
435
beginning at x 0. The extent to which the viscous eects diuse normally from the duct
wall is represented by the hydrodynamic boundary layer thickness v , which varies with
the axial coordinate x. In accordance with Prandtls boundary layer theory, the hydrodynamic boundary layer thickness divides the ow eld into two regions: a viscous region
near the duct wall and an essentially inviscid region around the duct axis.
At x Lhy ,{ the viscous eects have spread completely across the duct cross section.
The region 0 x Lhy is called the hydrodynamic entrance region, and the uid ow in
this region is called the hydrodynamically developing ow. As shown in Fig. 7.6a, the axial
velocity prole in the hydrodynamic entrance region varies with all three space coordinates [i.e., u ux; y; z]. For hydrodynamically developed ow, the velocity prole at a
given cross section becomes independent of the axial coordinate and varies with the
transverse coordinates alone [i.e., u uy; z or ur; ]. The fully developed velocity
prole is essentially independent of type of the velocity prole at the duct inlet.
After the ow becomes hydrodynamically developed (x > Lhy ; Fig. 7.6a), suppose
that the duct wall temperature Tw is raised above the uid entrance temperature Te (i.e.,
Tw > Te ). In this case, the local temperature changes diuse gradually from the duct wall,
beginning at x Lhy . The extent to which the thermal eects diuse normally from the
duct wall is denoted by the thermal boundary layer thickness t , which also varies with
the axial coordinate x. The t is dened as the value of y 0 (the coordinate measured from
the duct wall in Fig. 7.6) for which the ratio Tw T=Tw Te 0:99. According to
Prandtls boundary layer theory, the thermal boundary layer thickness divides the ow
eld into two regions: a heat-aected region near the duct wall and an essentially unaected region around the duct axis. At x Lth ,{ the thermal eects have spread completely across the duct cross section. The region Lhy x Lth is termed the thermal
entrance region, and the uid ow in this region is called the thermally developing ow.
It may be emphasized that the thermally developing ow is already hydrodynamically
developed in Fig. 7.6a. As shown in this gure, in the thermally developing ow region,
the local dimensionless uid temperature # Tw T=Tw Te , varies with all three
spatial coordinates [i.e., # #x; y; z].
For Lth x < 1 in Fig. 7.6a, the viscous and thermal eects have completely diffused across the duct cross section. This region is referred to as the fully developed region.
The uid ow in this region is termed the fully developed ow. In this region, the dimensionless temperature # varies with the transverse coordinates alone, although the local
uid temperature T varies with all three-space coordinates, and uid bulk temperature
Tm varies with the axial coordinate alone. In the fully developed region, the boundary
conditions at the wall govern the convective heat transfer process so that the temperature
distribution is essentially independent of the duct inlet temperature and velocity distributions.
{
The hydrodynamic entrance length Lhy is dened as the duct length required to achieve a maximum duct crosssection velocity as 99% of that for a fully developed ow when the entering uid velocity prole is uniform. The
maximum velocity occurs at the centroid for the ducts symmetrical about two axes (e.g., circular tube and
rectangular ducts). The maximum velocity occurs away from the centroid on the axis of symmetry for isosceles
triangular, trapezoidal, and sine ducts (Shah and London, 1978). For nonsymmetrical ducts, no general statement
can be made for the location of umax : There are a number of other denitions also used in the literature for Lhy .
{
The thermal entrance length Lth is dened, somewhat arbitrarily, as the duct length required to achieve a value of
local Nusselt number equal to 1.05 times the Nusselt number for fully developed ow, when the entering uid
temperature prole is uniform. As discussed in Section 7.4.3.1, theoretically the Nusselt number is innity at x 0
and reduces asymptotically to a constant value in fully developed ow.
436
The fourth type of ow, called simultaneously developing ow, is illustrated in Fig.
7.6b and c. In this case, the viscous and thermal eects diuse simultaneously from the
duct wall, beginning at x 0. Depending on the value of the Prandtl number Pr, the two
eects diuse at dierent rates. In Fig. 7.6b, Pr > 1 and v > t , whereas in Fig. 7.6c,
Pr < 1 and v < t . This relationship among Pr, v and t is easy to infer from the
denition of the Prandtl number, which for this purpose can be expressed as
Pr =, a ratio of kinematic viscosity to thermal diusivity. The kinematic viscosity
is the diusion rate for momentum or for velocity in the same sense that the thermal
diusivity is the diusion rate for heat or for temperature change. For Pr > 1, the
velocity prole (hydraulic) development is faster than the temperature prole (thermal)
development (so that Lhy < Lth ). For Pr < 1, the velocity prole development is slower
than the temperature prole development (so that Lhy > Lth ). When Pr 1, the viscous
and thermal eects diuse through the uid at the same rate. This equality of diusion
rates does not guarantee that the hydrodynamic and thermal boundary layers in internal
duct ows will be of the same thickness at a given axial location. The reason for this
apparent paradox lies in the fact that with Pr 1, the applicable momentum and energy
dierential equations do not become analogous. In external laminar ow over a at plate,
on the other hand, the energy and momentum equations do become analogous when
Pr 1. When the boundary conditions for the momentum and thermal problems are also
analogous, we get v t for all values of x.
The region 0 x Lc in Fig. 7.6b and c is referred to as the combined entrance
region. It is apparent that the combined entrance length Lc is dependent on Pr.
For Pr > 1, Lc Lth , and for Pr < 1, Lc Lhy . It may also be noted that in the combined
entrance region, both the axial velocity and the dimensionless temperature vary with all
three space coordinates [i.e., u ux; y; z and # #x; y; z where # Tw;m T=
Tw;m Tm ]. The region Lc x < 1 is the fully developed region, similar to the one
depicted in Fig. 7.6a with axially invariant uy; z and #y; z, satisfying
@u
0
@x x>Lhy
@#
0
@x x>Lth
7:4
In a fully developed laminar ow, the uid appears moving by sliding laminae of innitesimal thickness relative to adjacent layers. If a dye were injected at the centerline of a
tube in laminar ow of water, the colored streak will continue to ow without being
mixed at the tube centerline. While fully developed laminar ow is obtained for
Re 2300 for a smooth circular tube with smooth tube inlet, the actual value of the
Reynolds number to achieve/maintain fully developed or developing laminar ow
depends on the tube cross-sectional geometry, straightness, smoothness and constancy
of the cross section along the ow length, and the ow inlet geometry. We will discuss this
after we dene the types of turbulent duct ows.
Turbulent Flow. The turbulent duct ows can also be divided into four categories: fully
developed, hydrodynamically developing, thermally developing, and simultaneously
developing. This division is identical to the one adopted for laminar duct ows in the
preceding subsection. However, there are some important dierences:
. The hydrodynamic entrance length and the thermal entrance length for turbulent
duct ow are characteristically much shorter than the corresponding lengths in
BASIC CONCEPTS
437
laminar duct ow. Typical values of Lth =D are 8 to 15 for air and less than 3 for
liquids. Consequently, results for fully developed turbulent uid ow and heat
transfer are frequently used in design calculations without reference to the hydrodynamic and thermal entrance regions. However, caution must be exercised in
using the fully developed results for the low Prandtl number liquid metals, since
the entrance region eects are quite pronounced for such uids even in turbulent
duct ows (Lth =Dh 5 to 30, depending on Re and Pr).
. The turbulent ow development is preceded by the laminar boundary layer development and through the transition region as shown in Fig. 7.3.
. Fully developed turbulent ow generally exists for Re > 10,000 in a circular pipe.
For a sharp corner ow passage geometry such as a triangular passage, while the
ow is turbulent in the core region for Re 10,000, it is laminar in the sharp corner
region.
. The ow is fully developed turbulent for Re < 10,000 for those ow geometries in
which turbulence is frequently generated within the ow passage geometry, such as
in a perforated n exchanger or a corrugated plate heat exchanger. Typically, ow
is fully developed turbulent for Re 200 for plate heat exchanger geometries (see
Fig. 1.16) and for Re > 100 on the shell side of a plate baed shell-and-tube
exchanger.
Thus the Reynolds number at which fully developed turbulent ow can be achieved
depends on tube cross-section geometry, its variation along the ow length, the surface
roughness, and the ow inlet geometry.
Periodic Flow. In the regularly interrupted heat exchanger surfaces, such as a strip
n surface, the conventional fully developed laminar or turbulent ow does not exit
[i.e., the velocity gradient of Eq. (7.4) is dependent on x]. Suciently downstream, the
velocity prole and the dimensionless temperature prole # Tw;m T=Tw;m Tm
at a given cross section of each strip n are identical:
ux1 ; y; z ux2 ; y; z
#x1 ; y; z #x2 ; y; z
7:5
where x2 x1 , where is the length of one spatial period. This condition means that
Nux x1 ; y; z Nux x2 ; y; z
7:6
Usually, two to eight spatial repetitions are required to attain fully developed conditions
in a periodic geometry, with the exact number depending on the geometry and the
Reynolds number.
7.1.2.5 Flows with Separation, Recirculation, and Reattachment. The purpose of the
surface interruptions, such as in oset strip ns and louver ns, is to break the growth
of the boundary layer. This results in thin boundary layers and enhances the heat
transfer coecient (usually accompanied by an increase in the friction factor).
However, an undesirable consequence of the surface interruption is often the separation, recirculation, and reattachment features in the ow, which results in more increase
in the f factor than in j or Nu. Consider, for example, the ow at the leading edge of a
n of nite thickness. The ow typically encounters such a leading edge at the heat
438
FIGURE 7.7 (a) Flow passing between two neighboring ns. Boundary layer separation,
reattachment, and growth are shown. Free shear layer and wake regions are also identied. (b)
Region A. (From Jacobi and Shah, 1996.)
exchanger inlet or at the start of new ns, oset strips, or louvers. For most Reynolds
numbers, ow separation will occur at the leading edge because the ow cannot turn the
sharp corner of the n, as shown in Fig. 7.7. Downstream from the leading edge, the
ow reattaches to the n. The streamline emanating from the leading edge and terminating at the reattachment point is called the separating streamline (see Fig. 7.7). Fluid
between the separating streamline and the n surface is recirculating; this region is
sometimes called a separation bubble or recirculation zone. Within the recirculation
zone, relatively slow-moving uid ows in a large eddy pattern. The boundary between
the separation bubble and the separated ow (along the separation streamline) consists
of a free shear layer. Free shear layers are highly unstable, and it can be expected that
velocity uctuations will develop in the free shear layer downstream from the separation point. These perturbations will be advected downstream to the reattachment
region, and there they will result in increased heat transfer. The n surface in contact
with the recirculation zone is subject to lower heat transfer because of the lower uid
velocities and the thermal isolation associated with the recirculation eddy. Boundary
layer separation occurs only in regions where there is an adverse pressure gradient. As
increasing pressure slows the ow, there may be a point at which the velocity gradient
at the surface vanishes; thus, along a curve on the n surface is the separation line.
Sharp corners and bends, such as those commonly found in interrupted-surface designs,
are subject to separation. If the ow does not reattach to the surface from which it
separated, a wake results. The separation bubble increases form drag and thus usually
represents an increase in uid pumping power with no heat transfer benet. Therefore,
as a general rule, ow separation should be avoided in surface design, where extremely
low uid pressure drop/pumping power is required.
7.1.3
BASIC CONCEPTS
439
conduction (molecular diusion) at the wall and within the uid and the movement of
uid is referred to as convection heat transfer. Energy transport due solely to bulk uid
motion (and having no heat conduction) is referred to as advection:
convection conduction advection
7:7
Thus, knowledge of both heat conduction and uid mechanics is essential for convection
heat transfer. If the motion of uid arises solely due to external force elds such as
gravity, centrifugal, magnetic, electrical, or Coriolis body forces, the process is referred
to as natural or free convection. If the uid motion is induced by some external means
such as pump, fan (blower or exhauster), wind, or vehicle motion, the process is referred
to as forced convection.
7.1.4
Basic Denitions
Now let us introduce some basic, but important denitions that are used in heat exchanger design and analysis. These are mean velocity, mean temperatures, and heat transfer
coecient.
7.1.4.1 Mean Velocity. The uid mean axial velocity um is dened as the integrated
average axial velocity with respect to the free-ow area Ao :
um
1
Ao
u dA
Ao
7:8
where u is the local velocity distribution across the ow cross section. In the heat
exchanger analysis ("-NTU, MTD, etc.), not only is the ow assumed to be uniform
for all ow channels but also uniform within a channel at velocity um . The true velocity
distribution in a ow passage is considered only for evaluation of the theoretical/
analytical value of the heat transfer coecient and friction factor for that ow
passage.
7.1.4.2 Mean Temperatures. The peripheral mean wall temperature Tw;m and the uid
bulk temperature Tm at an arbitrary duct cross section x are dened as
Tw;m
Tm
1
P
Tw ds
1
Ao um
7:9
uT dA
Ao
7:10
where P is the duct perimeter and s is the spatial coordinate at a point on the duct wall
(of nite thickness) along the inside perimeter. For a duct with uniform curvature, such
as a circular tube, Tw Tw;m . However, for a noncircular tube, Tw may not be uniform
but will be dependent on the boundary condition.
The uid bulk temperature Tm in Eq. (7.10) is an enthalpy average over the ow cross
section for constant cp . However, if we idealize uniform ow in a cross section with
u um , the Tm in Eq. (7.10) is the integrated average value over the cross section. The
uid bulk temperature Tm is also referred to as the mixed mean uid temperature,
or mixing cup, mass average, or ow average temperature. Conceptually, Tm is the
440
temperature one would measure if the duct were cut o at a section and escaping uid
were collected and thoroughly mixed in an adiabatic container.
7:11
7:12
Hence, the heat transfer coecient h represents quantitatively the convective heat ux q 00
per unit temperature dierence Tw Tm or Tw T1 between a surface (wall) and a
uid. Thus, the complex ow and heat transfer phenomena for a heat transfer surface are
all lumped into the denition of h, making it dependent on many variables or operating
conditions. Some of these variables/conditions are: phase condition (single-phase, condensation, boiling/evaporation, multiphase), ow regime (laminar, transition, turbulent), ow passage geometry, uid physical properties (i.e., the type of the uid), ow
and thermal boundary conditions, convection type (free and forced), heat transfer rate,
nonuniformity of wall temperature at individual ow cross sections, viscous dissipation,
and other parameters/variables, depending on the ow type. Thus, in general, the concept of h to simplify the convection phenomenon is useful only in a limited number of
applications and may not provide an easy solution to a wide range of convective heat
transfer problems. In external forced convection ow, q 00 is often directly proportional to
the temperature dierence/potential T Tw T1 , so h is found to be independent
of this temperature dierence. Moreover, the heat transfer coecient may be considered
as nearly constant in some ow regimes. Only in these situations (linear problem),
may Eqs. (7.11) and (7.12) be considered to provide a linear relationship between the
driving potential for heat transfer and convective heat ux. For nonlinear problems, h
may also be dependent on T, such as in natural convection ows (where convection
heat ux is proportional to, say, T 5=4 for laminar ow), or it may be dependent on q 00
and T, as in boiling. In such situations, the operational convenience of the conductance
concept is diminished (since now h is dependent on T and/or q 00 ; such as T 1=4
for laminar free convection), although the denitions of Eqs. (7.11) and (7.12) are still
valid.
Depending on the value of the temperature potential T Tw Tm and the
heat ux q 00 at a local point on the wall, averaged over a cross section or averaged
over the ow length, the magnitude of h will be local at a point, peripherally average
but axially local, or averaged over the ow length. The h signies a conductance in
the thermal circuit representation of Fig. 7.4 where q 00 qw =A signies the current
and T is the potential. This h is the characteristic of the heat transfer process, the
property of a uidheat transfer surface thermal interaction, not a uid property.
It is also referred to as the convection conductance, lm coecient, surface coecient
of heat transfer, unit conduction for thermal-convection heat transfer, and unit thermal
convective conductance. It is also represented by the symbol mostly in European
literature.
DIMENSIONLESS GROUPS
441
7:13
where the adiabatic temperature is the temperature that the heat transfer surface would
achieve if there were no heat transfer to or from it by radiation and conduction; thus
it represents the eective average uid temperature near the heat transfer surface.
Although this concept may be quite useful for simple geometries used in electronic cooling (Moat, 1998), where Tadiabatic can be determined readily, it is dicult to determine in
complex heat exchanger geometries having nonuniform temperature over the n cross
section.
In the convection heat transfer process using h [Eq. (7.11)], only the wall temperature
and uid bulk temperature (or the free stream temperature) are involved for the determination of convective heat ux. However, this heat ux is actually dependent on the
temperature gradient at the wall. Namely, combining Eqs. (7.3) and (7.11), the heat
transfer rate per unit surface area can be represented either as conduction or convection
heat ux as follows (heat transfer taking place from the wall to the uid):
qw00 k
@T
q 00 hTw Tm
@y y0
7:14
k@T=@yy0
Tw Tm
7:15
i.e., the heat transfer coecient is, indeed, dependent on the temperature gradient at the
wall. If viscous dissipation is signicant, a straightforward application of Eq. (7.14) may
yield a negative heat transfer coecient. This is a consequence of an incorrectly assumed
driving potential dierence for heat transfer (i.e., the imposed temperature dierence
would not be proportional to the temperature gradient at the wall). In such cases, the
heat transfer coecient must be based on the dierence between the given and the
adiabatic wall temperatures (Burmeister, 1993).
7.2
DIMENSIONLESS GROUPS
Heat transfer characteristics of an exchanger surface are presented in terms of the Nusselt
number, Stanton number, or Colburn factor vs. the Reynolds number, x*, or Graetz
number. Flow friction characteristics are presented in terms of the Fanning friction
factor vs. Re or x . These and other important dimensionless groups used in heat
exchanger design and internal ow forced convection are described next and are also
summarized in Table 7.2 with their denitions and physical meanings.
It should be emphasized that in all dimensionless groups, wherever applicable, for
consistency the hydraulic diameter Dh is used as a characteristic length. However, the
hydraulic diameter or any other characteristic length does not represent a universal
characteristics dimension. This is because three-dimensional boundary layer and wake
442
TABLE 7.2 Important Dimensionless Groups for Internal Flow Forced Convection Heat Transfer
and Flow Friction, Useful in Heat Exchanger Design
Dimensionless
Group
Reynolds number
Re
Fanning friction
factor
um Dh GDh
w
u2m =2gc
r
p rh
f p* h 2
L um =2gc L
f
r
Apparent Fanning fapp p* h
L
friction factor
Incremental
pressure drop
number
Darcy friction
factor
Euler number
Hagen number
Dimensionless
axial distance
for uid ow
problem
Nusselt number
Stanton number
Colburn factor
Prandtl number
L
Represents the excess dimensionless pressure
rh
drop in the entrance region over that for
K1 constant for x ! 1
fully developed ow
D
Four times the Fanning friction factor;
fD 4f p* h
L
commonly used in uid mechanics
literature
p
Pressure drop normalized with respect to the
Eu p* 2
um =2gc
dynamic velocity head; see an alternative
denition for a tube bank in Eq. (7.22)
g
p
Alternative to a friction factor to represent
Hg 2c D3h
x
the dimensionless pressure drop. It does
not have any velocity explicitly in its
denition, so it avoids the ambiguity of
velocity denitions
x
Ratio of the dimensionless axial distance
x
Dh Re
x=Dh to the Reynolds number; axial
coordinate in the hydrodynamic entrance
region
h
q 00 Dh
Ratio of the convective conductance h to pure
Nu
k=Dh kTw Tm
molecular thermal conductance k=Dh over
the hydraulic diameter
h
Nu
Nu
443
DIMENSIONLESS GROUPS
cp um Dh um Dh
k
Re Pr
Peclet number Pe
x
x
Dimensionless x*
Dh Pe Dh Re Pr
axial
distance
for heat
transfer
problem
m_ cp Pe P
P 1
Graetz
Gz
kL
4L
4Dh x*
number
Gz
for a circular
4x* tube
Leveque
number
1
D
x HgPr h
L
2 f
1
f
Re
xf
x*
Lq
Fluid Flow
Reynolds Number. The Reynolds number is dened for internal ow as
Re
um Dh GDh um Dh
7:16
It is usually interpreted as a ow characteristic proportional to the ratio of ow momentum rate u2m or inertia force to viscous force um =Dh for a specied duct geometry,
where the mathematical expressions in parentheses are provided as an example for ow
in a constant-cross-sectional duct. Note that the inertia force is zero for fully developed
internal ow, while the momentum rate is still nite. Hence, the Reynolds number Re is a
ratio of ow momentum rate to viscous force, and thus it is a ow modulus. It should be
added, though, that the physical interpretation of the Reynolds number as the ratio of
inertia force to viscous force usually referred to in textbooks is not necessarily correct.
444
This is due to the fact that in a boundary layer there is actually always a balance between
inertia and friction, which can easily be demonstrated by studying the order of magnitude
of respective terms in an energy equation. See Bejan (1995) for further discussion. If Re is
the same for two systems that are geometrically and kinematically similar, a dynamic
similarity is also realized, irrespective of the uid.
7.2.1.2 Friction Factor and Related Groups. The ratio of wall shear stress w to the
ow kinetic energy per unit volume u2m =2gc is dened as the Fanning friction factor:
f
w
u2m =2gc
7:17
If f is based strictly on the true wall shear stress, it truly represents the skin friction{
component only. It is the skin friction that relates to the convective heat transfer over a
surface in the Reynolds analogy discussed in Section 7.4.5. In a heat exchanger core,
depending on the geometry of the heat exchanger surface, there could be form drag and
internal contraction/expansion (as in a tube bank or a perforated plate core) included
in the experimental value of the f factor; in that case, w in Eq. (7.17) represents the
eective wall shear stress.
The friction factor is strongly dependent on the ow passage geometry in laminar
ow, and weakly dependent in turbulent ow. The friction factor is inversely proportional to the Reynolds number in fully developed laminar ow and is dependent on
x x=Dh Re in developing laminar ow. The friction factor is proportional to
Ren (where n 0:20 to 0.25 in turbulent ow) for smooth tubes. It is dependent on
the surface roughness in turbulent ow. In addition to the ow passage geometry and
ow regimes, the friction factor could also be dependent on uid physical properties
( and ), phase condition (single-phase, condensation, and vaporization), and other
parameters, depending on the ow type.
In a steady-state isothermal fully developed ow in a constant-cross-sectional
geometry, the momentum rate at any cross section is constant. The pressure drop is
then a result of the wall friction. In the absence of core entrance and exit losses, it can
be shown [see Eq. (6.12)] from the application of Newtons second law of motion that
p*
p
L
4L
f
f
rh
Dh
u2m =2gc
7:18
p
u2m =2gc
fapp
L
L
ffd Kx
rh
rh
7:19
{
Skin friction is the ow friction associated with scrubbing the surface (through or over which the uid ows) by
the uid.
DIMENSIONLESS GROUPS
445
Here ffd is the friction factor for fully developed ow and Kx is the incremental pressure
drop number (Shah and London, 1978). In the case of fully developed ow in a duct, there
is always an entrance region (see Fig. 7.2) in the beginning. In this case, Kx of Eq. (7.19)
takes its maximum value K1 reported for many duct geometries by Shah and London
(1978) and Shah and Bhatti (1987). However, the incremental contribution of K1 to
the pressure drop may be negligible and can be ignored in long ducts.
In the literature, several other denitions of the friction factor are also used. Hence
one should be careful to distinguish before its use. The other more common denition
used in uid mechanics literature for the friction factor is the large or Darcy friction
factor fD , also sometimes referred to as DarcyWeisbach friction factor:
fD 4f
7:20
L
Dh
or p fD
L u2m
Dh 2gc
7:21
For ow over a tube bank, the skin friction contribution to the pressure drop may not
be major, and no unique ow length can be dened for p to be proportional to L. For
such geometries, the pressure drop is presented in terms of an average Euler number Eu
per tube row instead of the friction factor f:
Eu
p
1
p*
Nr
u2m =2gc Nr
7:22
For other external ow geometries, the Euler number is reported in the literature for a
complete tube bank by eliminating Nr from Eq. (7.22) or per tube row as in Eq. (7.22), or
some other denitions, and hence care should be exercised to note the specic denition
of Eu used.
An alternative way of representing the driving force p=x of internal uid ow in
a dimensionless form is in terms of the Hagen number Hg, dened by the rst equality in
the following equation (Martin, 2002). The second equality relates the Hagen number to
the Fanning friction factor and the Euler number.
8
gc 3 p
2
>
>
< 2 Dh x 2f Re
Hg
> gc 2 p 1
>
: 2 do
Eu Re2d
Nr 2
7:23a
7:23b
where Red um do =. Thus the Hagen number is an alternative to a friction factor or
an Euler number to represent the dimensionless pressure drop, and it is an average value
per tube row for ow normal to a tube bundle. Note that the Hagen number does not
have any velocity explicitly in its denition, which may be advantageous in ow normal
to the tube bank and other external ow geometries having some ambiguity in dening
the maximum velocity, as needed in the Fanning friction factor denition. Hence when
relating to the Hagen number and friction factor or Euler number of Eq. (7.23), any
reference velocity may be used as long it is the same in both the f or Eu and Re number
denitions.
446
fD Re 64
Hg 32Re
7:24
x
Dh Re
7:25
The apparent friction factor fapp decreases with increasing value of x and asymptotically
approaches the fully developed value f as x ! 1.
7.2.2
Heat Transfer
7.2.2.1 Nusselt Number. The Nusselt number is one of the dimensionless representations of the heat transfer coecient. It is dened for an internal ow as the ratio of the
convective conductance h to the pure molecular thermal conductance k=Dh :
Nu
h
hDh
q 00 Dh
k
kTw Tm
k=Dh
7:26
The Nusselt number has a physical signicance in the sense that the heat transfer
coecient h in Nu represents the convective conductance in a thermal circuit representation (Fig. 7.4) with the heat ux q 00 as the current and (Tw Tm ) as the potential.
Alternatively, the Nusselt number may be interpreted as the ratio of convection heat
transfer to conduction heat transfer.
For external ow or in the thermal entrance region, the Nusselt number is dened by
either of the rst two equalities in the following equation:
Nu
hL hDh
@#*
k
k
@y* y 0
7:27
DIMENSIONLESS GROUPS
447
phase, condensation, and vaporization); uid physical properties, including Pr; and other
variables/parameters, depending on the ow type and convection type (free and forced).
7.2.2.2 Stanton Number. The Stanton number is another dimensionless representation
of the heat transfer coecient. It is dened as
St
h
h
Gcp um cp
7:28
By multiplying both the numerator and denominator by Tw Tm , the physical meaning of the Stanton number becomes apparent: namely, St is the ratio of convected heat
transfer (per unit duct surface area) to the enthalpy rate change of the uid reaching the
wall temperature (per unit of ow cross-sectional area).
Single-phase heat transfer from the wall to the uid (or vice versa) is related to its
enthalpy rate change:
hATw Tm Ao Gcp To Ti Gcp Ao T
7:29
7:30
Hence, the Stanton number can also be interpreted as being proportional to the temperature change in the uid divided by the convective heat transfer driving potential.
When axial uid heat conduction is negligible (large Pe), St is frequently preferred to
Nu as a dimensionless modulus for the convective heat transfer correlation. This is
because it relates more directly to the designers task of establishing the number of
exchanger transfer units NTU. Moreover, the behavior of St with Re parallels that of
the Fanning friction factor f vs. Re [see Fig. 7.8, in terms of modied St or the
j St Pr2=3 factor and f factor vs. Reynolds number]. For turbulent ow, since
0.2
h / u0:2
m or Re , it is not a strong function of Re; it is nearly constant.
The Stanton number is also directly related to the number of heat transfer units on
one uid side of the exchanger, as follows:
St
h
hA Ao
r
D
ntu h ntu h
L
4L
Gcp m_ cp A
7:31
The Nusselt number is related to the Stanton, Prandtl, and Reynolds numbers as
follows by denition:
Nu St Re Pr
7:32
448
As the Stanton number is dependent on the uid Prandtl number, the Colburn factor j is
nearly independent of the owing uid for 0:5 Pr 10 from laminar to turbulent ow
conditions. Thus the j vs. Re data obtained for a given heat exchanger surface for air can
be used for water under certain ow conditions, as discussed in Section 7.4.6. Note that
using Eqs. (7.31) and (7.33), it can be shown that j is related to ntu as follows:
j ntu
rh 2=3
D
Pr ntu h Pr2=3
L
4L
7:34
7.2.2.4 Prandtl Number. The Prandtl number is dened as the ratio of momentum
diusivity to the thermal diusivity of the uid:
Pr
cp
k
7:35
The Prandtl number is solely a uid property modulus. Its range for several uids is as
follows: 0.001 to 0.03 for liquid metals, 0.2 to 1 for gases, 1 to 13 for water, 5 to 50 for
light organic liquids, 50 to 105 for oils, and 2000 to 105 for glycerin.
7.2.2.5
cp um Dh um Dh
k
7:36
On multiplying the numerator and denominator of the rst equality of Eq. (7.36) by the
axial uid bulk temperature gradient dTm =dx, it can be shown that
Pe Dh
m_ cp dTm =dx
kAo dTm =dx
7:37
Thus the Peclet number represents the relative magnitude of the thermal energy transported to the uid (uid enthalpy change) to the thermal energy axially conducted within
the uid. The inverse of the Peclet number is representative of the relative importance of
uid axial heat conduction. Therefore, Pe is important for liquid metal heat exchangers
because of its low Pr values (Pr 0:03). Longitudinal heat conduction within the uid
for all other uids is negligible for Pe > 10 and x* > 0:005 (Shah and London, 1978).
From Eqs. (7.16) and (7.35), Eq. (7.36) becomes
Pe Re Pr
7:38
7.2.2.6 Dimensionless Axial Distance, Graetz Number, and Leveque Number. The
dimensionless distance x* in the ow direction for heat transfer in the thermal entrance
region and the Graetz number Gz are dened as
x*
x
x
Dh Pe Dh Re Pr
Gz
m_ cp Pe P Re Pr P
kL
4L
4L
7:39
449
DIMENSIONLESS GROUPS
This denition of the Graetz number is used conventionally in the chemical engineering
literature. If the ow length L in Gz is treated as a length variable, x* is related to Gz as
x*
P 1
4Dh Gz
7:40
which reduces to x* =4Gz for the circular tube. Thus, the inverse of the Graetz
number is proportional to the dimensionless axial distance for thermal entrance region
eects in laminar ow.
Leveque (1928) obtained a developing thermal boundary layer solution for the fully
developed laminar velocity prole in a circular tube of length L for the mean Nusselt
1=3
T boundary condition as Nu 1:615x*
1:615Re Pr Dh =L1=3 .
number for the *
Since f Re 16 for a circular tube, this Leveque solution can be generalized for other
ow passage approximately as follows by multiplying f Re=16 by the term in parentheses to get Nu 0:641 f Re2 Pr Dh =L1=3 0:404 fD Re2 Pr Dh =L1=3
0:4042Hg Pr Dh =L1=3 . Since many of the heat exchanger surfaces have developing
velocity and temperature proles, Martin (2002) proposed this generalized Nusselt
number expression for chevron plates, tube banks, cross rod matrices, and packed
beds in laminar and turbulent ow regimes using only the contribution of the skin
friction factor for f or fD in the equation above. Martin refers to this generalized
Nusselt number expression as the generalized Leveque equation and the parenthetic
group as the Leveque number, Lq, which is slightly modied as follows:
1
D
D
Lq xf Hg Pr h xf f Re2 Pr h
L
L
2
7:41
where xf is the fraction of the total or apparent friction factor that corresponds to the
skin friction. As noted near the end of Section 7.4.5, j=f 0:25; hence, xf 0:5 for many
interrupted ow geometries.
7.2.3
Since the majority of basic data for heat exchanger (particularly compact heat exchanger)
surfaces are obtained experimentally (because computational uid dynamics techniques
and modeling cannot analyze accurately three-dimensional real surfaces at present), the
dimensionless heat transfer and pressure drop characteristics of these surfaces are presented in terms of j and f vs. Re. As an example, basic heat transfer and ow friction
characteristics for airow in a long circular tube are presented in Fig. 7.8. This gure
shows three ow regimes: laminar, transition, and turbulent. This is characteristic of fully
developed ow in continuous ow passage geometries such as a long circular tube,
triangular tube, or other noncircular cross-sectional ducts. Generally, the compact interrupted surfaces do not have a sharp dip in the transition region (Re 1500 to 10,000), as
shown for the circular tube. Notice that there is a parallel behavior of j vs. Re and f vs. Re
curves, but no such parallelism exists between Nu and f vs. Re curves. The parallel
behavior of j and f vs. Re is useful for (1) identifying erroneous test data for some specic
surfaces for which a parallel behavior is expected but indicated otherwise by test results
(see Fig. 7.11 and related discussion); (2) identifying specic ow phenomena in which
the friction behavior is dierent from the heat transfer behavior (such as rough surface
ow for friction and smooth surface ow for heat transfer for highly interrupted n
450
FIGURE 7.8 Basic heat transfer and ow friction characteristics for airow through a long
circular tube. (From Shah, 1981.)
geometries in turbulent ow); and (3) predicting the f factors for an interrupted surface
(that has considerable form drag contribution) when the j factors are known by some
predictive method. It should be remembered that j vs. Re can readily be converted to Nu
vs. Re curves, or vice versa, for given uids because j Nu Pr1=3 =Re by denition.
Because the values of j, f, and Re are dimensionless, the test data are applicable to
surfaces of any hydraulic diameter, provided that a complete geometric similarity is
maintained.
451
periodic, or arbitrary rise/drop. The determination of the Fanning friction factor is made
under steady uid ow rates with or without heat transfer, regardless of the core construction and the method of heat transfer testing. A detailed review of thermal performance test methods for industrial heat exchangers is provided by Lestina and Bell (2001).
7.3.1
Generally, a crossow heat exchanger is employed as a test section. On one uid side
(the known side), a surface for which the j vs. Re characteristic is known is employed;
a uid with a high heat capacity rate ows on this side. On the other uid side (the
unknown side) of the exchanger, a surface is employed for which the j vs. Re characteristic is to be determined. The uid that ows over this unknown-side surface is
preferably the one that is used in a particular application of the unknown-side surface.
Often, air is used on the unknown side; while steam, hot water, chilled water, or oils
(resulting in a high value of hA) are used on the known side. In the following subsections,
we describe the test setup, experimental method, data reduction method, and test core
design. This method could become quite inaccurate, due to inaccuracies in temperature
measurements, and hence it is generally not used for high (> 3) or low (< 0:5) core NTU.
7.3.1.1 Test Setup and Experimental Procedure. In general, the test setup consists of
the following basic elements on the unknown side: (1) a test section; (2) a uid metering
device, such as a nozzle, orice, or rotameter; (3) a uid pumping device, such as a fan,
pump or high-pressure uid supply; (4) temperature measurement devices, such as
thermocouples or resistance thermometers; and (5) pressure measurement devices,
such a manometers or pressure transducers. Similar devices are also used on the
known side. As an example, Fig. 7.9 shows the air-side schematic of the test rig used
by Kays and London (1950) at Stanford University.
In the experiments, the ow rates on both uid sides of the exchanger are set at
constant predetermined values. Once the steady-state conditions are achieved, uid temperatures upstream and downstream of the test section on both uid sides are measured,
as well as all pertinent measurements for the determination of uid ow rates. The
upstream pressure and pressure drop across the core on the unknown side are also
recorded to determine the hot friction factors.{ The tests are repeated with dierent
FIGURE 7.9 Schematic of a steam-to-air steady-state heat transfer test rig. (From Shah, 1985.)
{
The friction factor determined from the p measurement taken during heat transfer testing is referred to as the
hot friction factor.
452
ow rates on the unknown side to cover the desired range of Reynolds number for the j
and f vs. Re characteristics. To assure high-accuracy data, the dierence between the
enthalpy rate drop of the hot uid (e.g., steam or water) and the enthalpy rate rise of
the cold uid (e.g., air) should be maintained below 3%; also, the heat capacity rate
ratio C * 0 for condensing steam on one uid side; C * should be maintained below 0.2
if a liquid is used on the Cmax side to maintain the Cmin side as the controlling thermal
resistance side.
7.3.1.2 Theoretical Analysis and Test Data Reduction Methods. The determination of
the airside lm coecient h, Colburn factor j, and Reynolds number Re is now shown
for condensing steam-to-air tests as an illustration. If water or other uid is used on the
known side, the following calculations need to be modied, as discussed below.
The uid temperature distributions in the core on both uid sides are shown in Fig.
7.10a. Generally, a slightly superheated steam is entered at the core inlet, to ensure dry
steam. However, the limited inuence of the degree of superheat on the data reduction is
neglected. The steam-side temperature is taken as uniform, corresponding to the average
core saturation pressure.
An energy balance equation for a control volume between x and x dx of an air
channel in the heat exchanger core (Fig. 7.10b), can be written as follows:
dTa;x
Ca Ta;x UP dxTs Ta;x Ca Ta;x
dx 0
7:42
dx
Therefore,
UP dxTs Ta;x Ca dTa;x )
dTa;x
U dA
Ts Ta;x
Ca
7:43
where Ca m_ cp a is the heat capacity rate for air (cold uid), P dx dA, and subscripts
a and s denote air and steam, respectively. Idealizing that the local overall heat transfer
coecient U is uniform throughout the channel (i.e., U Um ) and integrating Eq. (7.43),
we get
Ts Ta;o
eNTU 1 "
Ts Ta;i
7:44
where NTU UA=Ca . The second equality on the right-hand side of Eq. (7.44) denotes
the exchanger ineectiveness (1 "), which is obtained directly from the denition of
Ts
Ta,o
Ta,i
Condensing
steam at Ts
Air
UPdx(Ts Ta,x)
CaTa,x
Ca[Ta,x + (dTa,x/dx)dx]
dx
x=0 x=x
x=L
(a)
(b)
FIGURE 7.10 (a) Fluid temperature distributions in the test core for steam-to-air tests; (b) control
volume for energy balance.
453
" Ta;o Ta;i =Ts Ta;i or from Eq. (3.84) for C* Cmin =Cmax 0, since condensing steam is used on one uid side. Subsequently, NTU is computed from Eq. (7.44).
If hot water, chilled water, or some oil is used on the known side instead of steam, the
crossow test core would have a nite value of C* (C * 0:2). In that case, the hot uid
temperature in Fig. 7.10a will not be constant and integration of the energy balance
equation(s) would lead to the appropriate "-NTU relationship, dierent from Eq. (7.44).
These relationships are given in Table 3.6, such as Eq. (II.1) for an unmixedunmixed
crossow exchanger, or Eq. (II.2) or (II.3) for mixedunmixed crossow exchanger. In
such a case, NTU must be obtained from the corresponding "-NTU relationship iteratively (often) or directly (rarely), depending on the pertinent expression. When C* 6 0,
the heat capacity rate is determined from the measured mass ow rates on each uid side
and the specic heats of the uids at their average temperatures. On the known side, the
uid properties are evaluated at the arithmetic average temperature Ts . On the unknown
side, the uid properties (cp , , Pr, ) are evaluated at the log-mean average temperature:
Ta;lm Ts Tlm
where
Tlm
Ts Ta;i Ts Ta;o
7:45
7:46
The overall heat transfer coecient Ua based on the total airside surface area Aa is
then evaluated from NTU as Ua NTU Ca =Aa . Now, the reciprocal of Ua ; an overall
thermal resistance, is considered as having three components in series: (1) air-side
thermal resistance, including the extended surface eciency on the air side; (2) wall
thermal resistance; and (3) steam-side thermal resistance, including the extended surface
eciency on the steam side:
1
1
A
Aa
a w
Ua o;a ha Aw kw o;s As hs
7:47
Then
ha
1
1
1
A
Aa
a w
o;a Ua Aw kw o;s As hs
7:48
The test cores are generally new, and no fouling or scale resistance is on either side, so the
corresponding resistance is not included in Eq. (7.47). The wall thermal resistance is
constant and is evaluated separately for each test core. The steam side (or liquid side,
if liquid is employed) heat transfer coecient hs must also be evaluated separately for
each core and should be known a priori. However, the steam-side (or liquid-side)
resistance is generally a very small percentage of the total resistance, and a reasonable
estimate will suce, even though considerable uncertainty may be involved in its
determination.
The term o;a in Eq. (7.48) is the extended surface eciency (see Section 4.3.4) of the
air-side surface and is related to the n eciency f of the extended surface as follows
[see Eq. (4.160)]:
o;a 1
Af
1 f
A
7:49
454
where the n eciency f for various geometries has been presented in Table 4.5. For
many plate-n surfaces, the relation for the straight n with constant conduction cross
section may be used to a good approximation [see Eq. (4.146)]. In that case,
f
tanh m
m
7:50
where m2 2h=kf and is the n length from the root to the center of the n (see Fig.
4.15 as an example). Once the surface area and the geometry are known for the extended
surface, h and o are computed iteratively from Eqs. (7.48)(7.50). For example, calculate
the value of ha from Eq. (7.48) for assumed o;a 1, and subsequently determine f from
Eq. (7.50), o;a from Eq. (7.49), and ha from Eq. (7.48). With this new value of ha ,
compute f from Eq. (7.50) and o;a from Eq. (7.49) and subsequently the next value
of ha from Eq. (7.48). Continue such iterations until the desired degree of convergence is
achieved. Alternatively, a graph, tabular values, or a curve t to o h vs. h is determined
for the given extended surface using Eqs. (7.50) and (7.49), and ha is determined from it
once o;a ha is found from Eq. (7.48) or (7.47).
The Stanton number St and the Colburn factor j are then evaluated from their
denitions, knowing the heat transfer coecient h, the core mass velocity G, the air
water vapor mixture specic heat cp , and the Prandtl number Pr as follows:
St
h
Gcp
j St Pr2=3
7:51
The Reynolds number on the unknown side for the test point is determined from its
denition:
GDh
Re
7:52
where G m_ =Ao is the core mass velocity, Dh 4Ao =P is the hydraulic diameter, and
is the dynamic viscosity for the unknown-side uid evaluated at the temperature given by
Eq. (7.45).
Example 7.1 The objective of this example is to determine Colburn factor j and
Reynolds number Re for the air-side surface tested by a steady-state technique.{ The
test unit is a crossow heat exchanger with steam condensing on the known side and air
owing on the unknown side. The air-side surface has the oset strip n surface shown in
Fig. E7.1.
Geometrical properties on the air side of this core for the data reduction are:
Dh 0:00121 m
rh
0:005688
L
Af ;a 1:799 m2
Aa 2:830 m2
The data for this problem are for a test point of core 105 tested by London and Shah (1968).
455
FIGURE E7.1
Ao;a 0:0161 m2
w 0:406 mm
b 1:905 mm
Aa w
5:68
105 m2 K=W
A w kw
Show that the log-mean temperature on the air side is 77.718C. At this temperature,
air properties are cp 1:0166 kJ/kg K for humid air, Pr 0:697, and 2:0822
105 Pa s. Determine Re and j for this test point. Consider the steam-side thermal
resistance as zero. Fins are made from aluminum for which kf 192:1 W/m K.
SOLUTION
Problem Data and Schematic: The test core geometry information, test point inlet pressure, air-side pressure drop, and temperatures are given for the crossow test unit with
air as the test uid and steam as the known uid on the other side. A small section of the
test core is shown in Fig. E7.1.
Determine: The Reynolds number and j factor for this test point.
Assumptions: Constant uid properties apply, the steam-side thermal resistance is zero
and there is no fouling on either uid side.
Analysis: First, we calculate the log-mean average temperature of the air using Eqs. (7.46)
and (7.45):
Tlm
Ts Ta;i Ts Ta;o
lnTs Ta;i =Ts Ta;o
107:87 23:058C 107:87 101:608C
30:168C
ln107:87 23:058C=107:87 101:608C
The air temperatures at the inlet and outlet of the test core represent an average of 9 and 27 thermocouple
readings, respectively; the thermocouple wire spool was calibrated to 0:058F accuracy, traceable to the United
States National Bureau of Standards.
456
Hence,
Ta;lm Ts Tlm 107:87 30:168C 77:718C
Let us calculate the air-side Reynolds number from its denition after calculating
the mass velocity G:
G
Re
0:274 kg=s
m_ a
17:019 kg=m2 s
Ao;a 0:0161 m2
GDh 17:019 kg=m2 s
0:00121 m
989
2:0822
105 Pa s
To determine the j factor, rst compute ". Since the air side is the Cmin side,
"
0:9261
Ts Ta;i
107:87 23:058C
therefore;
1 " 0:0739
Aa
Aa
2:830 m2
256:38 W=m2 K
Since the steam-side thermal resistance is zero, h on the air side from Eq. (7.48) is then
ha
1
o;a 1=Ua Aa w =Aw kw
7:53
Here we need to know o;a to determine ha . The o;a is calculated from Eq. (7.49) for
known f , which in turn is dependent on h. Thus, we need to calculate ha iteratively. Let
us assume that o;a 1:00; then, from Eq. (7.53),
ha
1
260:16 W=m2 K
1
1=256:38 W=m2 K 5:68
105 m2 K=W
7:54
The n eciency of the oset strip n, from Eq. (4.146) with m from Eq. (4.147), is
"
#0:5
2ha
1=2
2
260:16 W=m2 K
0:102 mm
1
1
m
kf
s
2:819 mm
192:1 W=mK
0:102
103 m
165:881 m1
m 165:88 m1
0:851
103 m 0:1412
tanh m tanh0:1412
0:993
m
0:1412
1 f Af ;a
1 0:993
1:799 m2
1
1
0:996
Aa
2:830 m2
f
o;a
457
Hence, the rened value of ha using Eq. (7.53) with o;a 0:996 and the rest of the terms
from Eq. (7.54) is
ha
261:21 W=m2 K
o;a
0:996
With this value of ha , the new values of f and o;a are found to be
f 0:993
o;a 0:996
Ans:
Thus, the values of o;a and ha are converged in two iterations for this problem. The
Colburn factor is nally calculated from its denition:
j St Pr2=3
h
261:21 W=m2 K
0:697 2=3
Pr2=3
Gcp
17:019 kg=m2 s
1:0166
103 J=kg K
0:01187
Ans:
Discussion and Comments: This example demonstrates how to obtain the j factor for a
heat transfer surface when its thermal resistance is dominant in a two-uid heat exchanger. The methodology is straightforward. If hot or chilled water had been used on the
steam side, we would need to make a few changes in the procedure. Instead of using
Eq. (7.44) for the "-NTU relationship for C * 0, we would have used the appropriate
"-NTU formula from Table 3.6 [such as Eq. (II.1)]. In that case, NTU being implicit in
the formula, it would have been computed iteratively for known " and C*. The water-side
thermal resistance would have been nite and Eq. (7.48) should have been used for the
airside h. If the test core were not a new core, fouling on the water side should have been
included.
7.3.1.3 Test Core Design. The test core is designed with two basic considerations
in mind to reduce the experimental uncertainty in the j factors: (1) the appropriate
magnitudes of thermal resistances on each uid side as well as of the wall, and (2) the
proper range of NTU.
The thermal resistances in a heat exchanger are related by Eq. (7.47) by multiplying
1/Aa on both sides. To reduce the uncertainty in the determination of the thermal
resistance of the unknown side (with known overall thermal resistance, 1/UA), the
thermal resistances of the exchanger wall and the known side should be kept as small
as possible by design. The wall thermal resistance value is usually negligible when one of
the uids in the exchanger is air. This may be minimized further through the use of a thin
material with high thermal conductivity. On the known side, the thermal resistance is
minimized by the use of a liquid (hot or cold water) at high ow rates, or a condensing
steam, to achieve a high h, and also by using an extended surface. Therefore, the thermal
boundary condition achieved during steady-state testing is generally a close approach to
a uniform wall temperature condition.
The NTU range for testing is generally restricted between 0.5 and 3 or between 40 and
90% in terms of the exchanger eectiveness. To understand this restriction and point out
the precise problem areas, consider the test uid on the unknown side to be cold air being
heated in the test section and the uid on the known side to be hot water (steam replaced
by hot water and its ow direction reversed in Fig. 7.9 to avoid air bubbles). The high
458
NTU occurs at low airow values for a given test core. Both temperature and mass
ow rate measurements become more inaccurate at low airows, and the resultant
heat unbalances qw qa =qa increase sharply at low airows with decreasing air mass
ow rates. In this section, the subscripts w and a denote water and air sides, respectively.
Now, the exchanger eectiveness can be computed in two dierent ways:
"
qa
Ca Tw;i Ta;i
or
"
qw
Ca Tw;i Ta;i
7:55
In an ideal case, qa qw and the two relationships of Eq. (7.55) must give identical
results. However, in reality, qa 6 qw , as emphasized above. Thus, a large variation in "
will result at low airows, depending on whether it is based on qa or qw . Since "-NTU
curves are very at at high " (high NTU), there is a very large error in the resulting NTU,
and hence in h and j. The j vs. Re curve drops o consistently with decreasing Re, as
shown by a dashed line in Fig. 7.11. This phenomenon is referred to as rollover or drop-o
in j. Some of the problems causing the rollover in j are errors in temperature and air
mass ow rate measurements, as follows:
1. A thermocouple measures the junction temperature, not the ambient temperature.
Hence, the measured air temperature downstream of the test core, Ta;o , may be too
low, due to heat conduction along the thermocouple wire. This and other heat
losses at low airows result in quite low heat transfer coefcients associated with a
thermocouple junction or a resistance thermometer. This heat conduction error is
not so pronounced for the upstream temperature measurement since air is at a
lower temperature. However, the measured air temperature Ta;i upstream of the
test core may be too high, due to the radiation effect from the hot core and hot
walls of the wind tunnel because of heat conduction in the duct wall from the hot
test core. This error is negligible for the core downstream, since the duct walls are
at about the same temperature as the outlet air. Both the aforementioned errors in
Ta;i and Ta;o will decrease the qa calculated.
2. At low airows, temperature stratication in the vertical direction would be a
problem both upstream and downstream of the test core. Thus, it becomes difcult
to obtain true bulk temperatures Ta;i and Ta;o .
FIGURE 7.11 Rollover phenomenon for j vs. Re characteristic of a heat exchanger surface at
low airows. The dashed curve indicates the rollover phenomenon; the solid curve represents the
accurate characteristic. (From Shah, 1985.)
459
3. On the water side, the temperature drop is generally very small, and hence it will
require very accurate instrumentation for Tw measurements; the water ow rate
should be adjusted downward from a high value to ensure the magnitude of Tw
large enough for an accurate measurement. Also, care must be exercised to ensure
good mixing of water at the core outlet before Tw is measured.
4. There are generally some small leaks in the wind tunnel between the test core and
the point of air mass ow rate measurement. These leaks, although small, are
approximately independent of the air mass ow rate, and they represent an increasing fraction of the measured ow rate m_ a at low airows. A primary leak test is
essential to ensure negligible air leakage at the lowest encountered test airow
before any testing is conducted.
5. Heat losses to ambient are generally small for a well-insulated test section.
However, they could represent a good fraction of the heat transfer rate in the
test section at low airows. A proper calibration is essential to determine these
heat losses.
6. For some test core surfaces, longitudinal heat conduction in the test core
surface wall may be important and should be taken into account in the data
reduction.
The rst ve factors cause heat imbalances qw qa =qa to increase sharply at
decreasing low airow rates. To minimize or eliminate the rollover in j factors, the
data should be reduced based on qave qw qa =2, and whenever possible, by reducing
the core ow length by half (i.e., reducing NTU) and then retesting the core.
The uncertainty in the j factors obtained from the steady-state tests (C * 0 case)
for a given uncertainty in To Ts Ta;o or Tw;o Ta;o is given as
d j
dTo ntua eNTU
j
Tmax NTU NTU
7:56y
Here Tmax Tw;i Ta;i . In general, ntua =NTU 1:1 for most testing for j data. Thus,
a measurement error in the outlet temperature dierence [i.e., dTo magnies the
error in j by the foregoing relationship both at high NTU (NTU > 3) and low NTU
(NTU < 0:5). The error at high NTU is due to the errors in To and other factors
discussed above as well as due to eNTU in the numerator. The error in the j factor at
low NTU can also be signicant, as can be found from Eq. (7.56) (NTU2 in the denominator), and hence a careful design of the test core is essential for obtaining accurate j
factors.
In addition to the foregoing measurement errors, inaccurate j data are obtained for a
given surface if the test core is not constructed properly. The problem areas are poor
thermal bond between the ns and the primary surface, gross blockage (gross ow
maldistribution) on the water (steam) or air side, and passage-to-passage nonuniformity
(or maldistribution) on the air side. These factors inuence the measured j and f factors
{
This equation is obtained as follows: After multiplying Eq. (7.47) on both sides by Ca =Aa , the resulting equation
has NTU in terms of ntua and the rest of the terms, which are treated as constant; dierentiating this equation, we
get d(NTU) in terms of dntua . Substituting the last relationship in Eq. (7.44) after it is dierentiated [i.e., it has
the d(NTU) term], we eventually can get a formula for dntua =ntua . Using the denition of j from Eq. (7.34), we
nally get Eq. (7.56).
460
FIGURE 7.12 Inuence on measured j data of (a) poor thermal bond between ns and primary
surface, (b) water (steam)-side gross blockage, (c) air-side gross blockage, and (d) air-side passage-topassage nonuniformity. The solid lines are for the perfect core, the dashed lines for the imperfect
core specied. (From Shah, 1985.)
dierently in dierent Reynolds number ranges. Qualitative eects of these factors are
presented in Fig. 7.12 to show the trends. The solid lines in these gures represent the j
data of an ideal core having a perfect thermal bond, no gross blockage, and perfect ow
uniformity. The dashed lines represent what happens to j factors when the specied
imperfections exist. It is imperative that a detailed air temperature distribution be
measured at the core outlet during testing to ensure that none of the foregoing problems
are associated with the core.
The experimental uncertainty in the j factor for the foregoing steady-state method is
usually within 5% when the temperatures are measured accurately to within 0:18C
(0.28F) and none of the aforementioned problems exist in the test core. The uncertainty
in the Reynolds number is usually within 2% when the mass ow rate is measured
accurately within 0.7%.
7.3.2
To obtain highly accurate j factors, one of the requirements for the design of a test core
in the preceding method was to have the thermal resistance on the test uid (gas) side
dominant [i.e., the test uid-side thermal conductance o hA is signicantly lower than
that on the other (known) h side]. This is achieved by either steam or hot or cold water at
high mass ow rates on the known h side. However, if the uid on the unknown h side is
water or another liquid and it has a high heat transfer coecient, it may not represent a
461
dominant thermal resistance, even if condensing steam is used on the other side. This is
because the test uid thermal resistance may be of the same order of magnitude as the
wall thermal resistance. Hence, for liquids, Wilson (1915) proposed a technique to obtain
heat transfer coecients h for turbulent ow in a circular tube. This technique has been
used extensively over the years and its modications are often used today as well.
In this method, liquid (test uid, unknown side, uid 1) ows on one uid side for
which j vs. Re characteristics are being determined; condensing steam, liquid, or air
ows on the other uid side (uid 2), for which we may or may not know the j vs. Re
characteristics. The uid ow rate on the uid 2 side and the log-mean average temperature must be kept constant (through iterative experimentation) so that its thermal
resistance and C2 of Eq. (7.60) are truly constant. The ow rate on the unknown (uid
1) side is varied systematically. The uid ow rates and temperatures upstream and
downstream of the test core on each uid side are measured for each test point. Thus
when " and C* are known, NTU and UA for the test core are computed. For discussion
purposes, consider the test uid side to be cold and the other uid side to be hot. UA is
given by
1
1
1
Rc; f Rw Rh; f
UA o hAc
o hAh
7:57
Note that o 1 on the uid side, which does not have ns. For fully developed turbulent
ow through constant-cross-sectional ducts, the Nusselt number correlation is of
the form
w 0:14
Nu Co Rea Pr0:4
7:58
m
where Co is a constant and a 0:8 for the DittusBoelter correlation (see Table 7.6).
However, note that a is a function of Pr, Re, and the geometry; it is shown in Fig. 7.19 for
a circular tube. Theoretically, a will vary depending on the tube cross-sectional geometry,
particularly for augmented tubes or with turbulators, and it is not known a priori. Wilson
(1915) used a 0:82. The term w =m 0:14 takes into account the variable property
eects for liquids; for gases, it should be replaced by an absolute temperature ratio
function [see Eq. (7.157)]. By substituting the denitions of Re, Pr, and Nu in Eq.
(7.58) and considering uid properties as constant we obtain
0:42 0:18
0 0:82
hc Ac Ac Co k0:6 0:82 c0:4
Dh c u0:82
p
m C1 um
C1 u0:82
m
o;c
7:59
The test conditions are maintained such that the fouling resistances Rc; f and Rh; f remain
approximately constant, although not necessarily zero; Wilson (1915) had neglected
them. Since h is maintained constant on the uid 2 side, the last four terms on the
right-hand side of Eq. (7.57) are constant, let us say equal to C2 . A requirement for
this condition (as mentioned above) is to have the log-mean average temperature constant so that the thermal resistance and C2 remain constant. Now substituting Eq. (7.59)
into Eq. (7.57), we get:
1
1
C2
UA C1 u0:82
m
7:60
462
FIGURE 7.13 Original Wilson plot of Eq. (7.60). (From Shah, 1985.)
1. The uid ow rate and its log-mean average temperature on the uid 2 side must be
kept constant, so that C2 is a constant.
2. The Re exponent in Eq. (7.58) is presumed to be known (such as 0.82 or 0.8).
However, in reality it is a function of Re, Pr, and the geometry itself. Since the Re
exponent is not known a priori, the classical Wilson plot technique cannot be
utilized to determine the constant Co of Eq. (7.58) for most noncircular or
enhanced heat transfer surfaces.
3. All the test data must be in one ow region (e.g., turbulent ow) on uid 1 side, or
the Nu correlation of Eq. (7.58) must be replaced by an explicit equation with two
(or more) unknown constants, such as Eq. (7.58) or any other variant.
4. Variations in uid thermophysical properties and the n thermal resistance are not
taken into consideration on the unknown uid 1 side.
5. The fouling resistance on both uid sides of the exchanger must be kept constant so
that C2 remains constant in Eq. (7.60).
463
Shah (1990) discusses how to relax all the foregoing limitations of the Wilson plot
technique except for the third limitation (one ow region for complete testing), this is
discussed later.
The original Wilson plot technique has two unknowns, C1 and C2 in Eq. (7.60). In
general, the Wilson plot technique determines two unknowns of the overall thermal
resistance equation by a linear (or log-linear) plot or by a linear regression analysis. In
the problem above, if Rc; f , Rw , and Rh; f are known a priori, we can determine the Nu
correlation of the type of Eq. (7.58) for unknown constant Co when its exponent a is
known for the uid 2 side. Thus, the heat transfer correlation on the uid 2 side can also
be evaluated using the Wilson plot technique if the exponents on Re in Eq. (7.58) are
known on both uid sides. Alternatively, we could determine Co and a of Eq. (7.58) by
the Wilson plot technique if C2 is known. The Wilson plot technique thus represents a
problem with two unknowns. Briggs and Young (1963) extended this technique with
three unknowns.
For a more general problem (e.g., a shell-and-tube exchanger), consider the Nu
correlation on the tube side as Eq. (7.58) with Co Ct0 and the exponent on Re as a.
On the shell side, consider Nu correlation as given by Eq. (7.58) with Co Cs0 and the
Re exponent as d. We can rewrite Eq. (7.57) as follows after neglecting Rt; f and Rs; f for a
new/clean exchanger:
1
1
1
Rw
d
0:4
UA Ct Rea Pr0:4 Ak=Dh t w =m 0:14
Cs Re Pr Ak=Dh s w =m 0:14
t
s
7:61
where Ct o;t Ct0 and Cs o;s Cs0 . Thus, the more general Wilson plot technique has ve
unknowns (Ct , Cs , a, d, and Rw ); no veried solution procedure is reported in the
literature for this problem.
Example 7.2 An airwater round tube and at louvered n heat exchanger test core
with turbulators in the tubes is tested to determine the Colburn factor vs. Reynolds
number correlation on the turbulators (water) side. Since generally the air side has larger
(controlling) thermal resistance, the conventional Kays and London method cannot be
used for getting the j or Nu vs. Re characteristics for the water side with a turbulator.
Hence, the Wilson plot technique has been used to determine the j vs. Re characteristics.
Testing is performed for three dierent air mass ow rates and for a range of mass ow
rates for water. The unknown heat transfer coecient side is the water side. The air-side
thermal resistance and its log-mean average temperature are kept constant during testing
for a specied nominal airow rate. The air- and water-side heat transfer surface areas
are 2.211 and 0.1522 m2, respectively. The hydraulic diameter Dh;w of the water-side ow
channels is 0.005282 m. From the temperature and ow measurements, the partially
reduced test data shown in Table E7.2 are determined for three constant values of airow
rates.{ The calculated values in the sixth column were obtained assuming
Nu Co Re0:85 Pr0:4 correlation for heat transfer on the water side. Determine this
correlation based on the data provided using the Wilson plot technique, assuming
negligible viscosity variation on the water side. How will this correlation change if the
{
If the test setup and measurements are correct, the water-side heat transfer correlation should be identical for all
three airow rates. From the temperatures measured, the exchanger eectiveness is evaluated and NTU is determined for the known exchanger ow arrangement. Subsequently, 1=U is computed for known A and Cmin .
464
TABLE E7.2 Partially Reduced Test Data to Determine Water Side j vs. Re Characteristics.
Air Mass
Flow Rate
(kg/s)
Water Mass
Flow Rate
(kg/s)
Pr for
Water
Re for
Water Side
1=U
m2 K=W
Aa Dh;w =
Ak Re0:85 Pr0:4 w
m2 K=W
0.757
0.756
0.756
0.755
0.758
0.758
0.757
Test Data for Nominal Air Mass Flow Rate of 0.76 kg/s
0.399
2.32
8,960
5:4979
103
0.452
2.30
10,191
5:3426
103
0.624
2.29
14,178
5:1246
103
0.757
2.27
17,327
5:0402
103
1.145
2.25
26,404
4:8165
103
1.517
2.24
35,168
4:7137
103
2.256
2.23
52,610
4:5513
103
3:6306
105
3:2578
105
2:4676
105
2:0858
105
1:4617
105
1:1480
105
8:1658
106
1.133
1.134
1.135
1.134
1.135
1.135
1.134
Test Data for Nominal Air Mass Flow Rate of 1.13 kg/s
2.253
1.88
52,354
3:8866
103
1.537
1.90
35,505
3:9903
103
1.170
1.92
26,935
4:1135
103
0.756
1.94
17,209
4:3221
103
0.651
1.96
14,759
4:3957
103
0.460
1.98
10,319
4:6706
103
0.382
1.97
8,501
4:8089
103
8:1922
106
1:1371
105
1:4362
105
2:0940
105
2:3822
105
3:2173
105
3:7829
105
1.885
1.882
1.883
1.888
1.887
1.890
1.888
Test Data for Nominal Air Mass Flow Rate of 1.89 kg/s
3.864
2.27
8,532
4:0781
103
0.460
2.25
10,243
3:9381
103
0.617
2.22
13,884
3:7570
103
0.777
2.20
17,637
3:6245
103
1.149
2.18
26,328
3:4303
103
1.524
2.17
35,088
3:3350
103
2.209
2.15
51,124
3:2518
103
3:7609
105
3:2293
105
2:5026
105
2:0482
105
1:4620
105
1:1471
105
8:3478
106
465
Analysis: The following Nusselt number correlation on the water side is given:
Nu
hDh
Co Rea Pr0:4
k
Nu
Co Rea1 Pr0:07
Re Pr1=3
The overall thermal resistance is related to air, water, and wall thermal resistances as
follows [see Eq. (7.57)]:
1
1
1
Rw
UAa o ha Aa
Co Re0:85 Pr0:4 k=Dh Aw
Therefore,
1
Aa
1
R
A
w a
U Co Re0:85 Pr0:4 k=Dh Aw
o ha
or
y
1
mx b
U
1.89
1.13
0.76
28.58
31.36
31.99
b
Number of
(m2 K=W) Data Points
0.003019
0.003646
0.004333
7
7
7
2:81
1010
3:73
1010
9:91
1010
Since m 1=Co , Co can then be computed for each airow testing from the results above.
The desired correlation for the circular tube with a turbulator is then
j Co Re0:15 Pr0:07 0:0350Re0:15 Pr0:07
where in the expression after the second equality, Co 1=28:58 0:0350 is based on the
tests for 1.89 kg/s airow rates. The Colburn factors from the correlation above are
presented graphically in Fig. E7.2 for three airow rates. We nd that the Colburn
factors for the largest airow rate are slightly larger than the corresponding values for
two smaller airow rates. The symbols used in the plot correspond to the same air
mass ow rates as in the Wilson plot. To determine the cause of this deviation, the
466
0.006
1/U (m 2. K/W)
0.005
0.004
0.003
m = 0.76 kg/s
m = 1.13 kg/s
m = 1.89 kg/s
0.002
10-5
2 10-5
3 10 -5
4 10-5
10
-2
9 10-3
8 10-3
710-3
m = 0.76 kg/s
m = 1.13 kg/s
m = 1.89 kg/s
610 -3
4
8 10 4
8 10 5
Re
(b)
FIGURE E7.2
experimental procedure and equipment operation should be checked and ensure that the
log-mean average temperature on the air side is kept constant. A detailed study in this
case showed that during the experiments for the latter two set of data, a water valve leak
was present but not corrected. The correction of the water mass ow rates would lead to a
very good agreement of all three sets of data, as one would expect.
467
Transient test techniques are used to establish the j vs. Re characteristics for a matrixtype regenerator surface or a high-ntu surface. The most important advantage of most of
these techniques is that only one uid (air) is employed during the testing. Therefore,
these techniques are also referred to as single-blow techniques. The test section is a singleuid exchanger (matrix or core) built up from the heat transfer surface for which the j vs.
Re characteristics are to be determined. Generally, air is used as a working uid. Initially,
the matrix wall and uid temperatures are both constant and uniform at the uid temperature at time equals zero. A known monotonic increase or decrease is imposed on the
inlet temperature of the uid. The resultant temperaturetime history of the uid at the
core outlet is recorded. This outlet uid temperature response is related directly to heat
transfer and the average heat transfer coecient in the matrix. A comparison of theoretical and experimental outlet temperature responses then permits evaluation of the average heat transfer coecient. The uid temperature at the core inlet could be a step,
exponential, periodic, or ramp function, or any arbitrary variation. The step inlet temperature was utilized most commonly for the single-blow method before the automatic
data acquisition by a computer became common in the early 1980s. Heggs and Burns
(1988) provided a comparison of four commonly used methods for the single-blow
technique (i.e., the direct matching, maximum slope, shape factor, and dierential
uid enthalpy methods). They reported the direct matching and dierential uid
enthalpy methods to be most accurate. For details of various transient techniques and
data reduction methods, refer to Shah and Zhou (1997). The direct curve-matching data
reduction method is most accurate for the complete range of ntu and is now the most
commonly used and recommended method.
468
FIGURE 7.14 Schematic of the transient single-blow heat transfer test rig. (From Wheeler, 1968.)
469
The test section is a single-uid exchanger in which the heat transfer surface is stacked
up with or without plain sheets in between. The test uid is air. The heating device is
generally a heating screen of ne wires (wire mesh) or made up of thin strips. The latter
approach is preferred, due to the large surface area of thin strips and hence lower
temperature, resulting in less radiation error in the upstream temperature measurements
(see Shah and Zhou, 1997). Also, the air temperatures upstream and downstream are
measured by point measurements (using a thermocouple grid) or line measurements
(such as 10 platinum wires 50 mm in diameter across the test section height arranged
at equal distance in the test cross section). The line measurements provide more accurate
temperature measurement across the test section with less ow disturbance. There are a
number of choices for accurate ow and pressure measurement devices, and the selection
is made by the individual investigators.
The following is the experimental procedure used to obtain the air temperaturetime
history at core outlet during the core heating and cooling. The airow rate is set at a
constant predetermined value. The air is heated with the resistance-heating device to
about 118C (208F) above the ambient temperature; the heated air in turn heats the
matrix. The temperaturetime history of the air at the core inlet and outlet is recorded
continuously during matrix heating, as shown in Fig. 7.15a. Then heating continues until
the core reaches a uniform temperature exhibited by a negligible dierence between the
air temperature at the inlet and exit of the matrix. Once the stable condition is reached,
the power to the heating device is turned o. The temperaturetime history of the air
leaving the matrix is recorded continuously during the matrix cooling period as shown in
Fig. 7.15b. During each heating and cooling period of the matrix, measurements are
taken for airow rates, core upstream and downstream pressures, and core upstream
temperature before the heating device. Usually, two heating and two cooling curves are
recorded for each ow rate, and the average heat transfer coecient of four curves is
used to determine the j factor. Similar tests are repeated with dierent airow rates to
cover the desired range of the Reynolds number.
7.3.3.2 Theoretical Model. A brief background on the theoretical analysis of the test
point may be given as follows. Pertinent dierential equations for the matrix heating
case of Fig. 7.15a are identical to those of Eq. (5.71) for the hot-gas ow period in a
regenerator presented in Section 5.3. Similarly, the dierential equations for the matrix
Tf,o, =
Tf,i , = 0
Tf,o , =
Tf,i , =
Tf,i
Tf,o
Tf,i
Tf,o, = 0
Tf,o
Tf,i , = 0
Tf,i , = 0
Tf,i , =
Core inlet
(a)
Core outlet
Core inlet
Core outlet
(b)
FIGURE 7.15 Exponential temperature change imposed at the inlet and its response at the core
outlet: (a) matrix heating; (b) matrix cooling.
470
cooling case of Fig. 7.15b are those of Eq. (5.72). Generally, longitudinal heat conduction in the matrix wall is included in the analysis. In this case, the second equations of
Eqs. (5.71) and (5.72) are replaced by Eqs. (5.111) and (5.113). Hence, the pertinent
dierential equations, using the same nomenclature, for the single-blow analysis
problem are
@Tf
Tw Tf
@
7:62
@Tw
@ 2 Tw
Tf Tw ntu
@
@2
7:63
where
hA x
C L
hA
Mc cw
kw A k
LC
ntu
hA
C
7:64
Here Tf and Tw are the uid (air) and wall temperatures. The boundary and initial
conditions for this problem are
At 0:
7:65
At 0:
Tw ; 0 0
7:66
At 0 and 1:
@Tw 0;
0
@
Tf ; 0 0
@Tw 1;
0
@
7:67
where the time constant of heater H Mc=hAH and the subscript H denotes the values
for the heater (in contrast to the test core). Note that longitudinal heat conduction in the
air is negligible and hence has not been incorporated in Eq. (7.62).
A variety of methods have been used to solve Eqs. (7.62), (7.63), and (7.65)(7.67) as
summarized by Shah and Zhou (1997). One of the most accurate and rapid methods is
the numerical analysis of these equations for the measured inlet and outlet air temperatures from the core. In this case, the inlet and outlet temperatures are measured using an
online data acquisition system; the temperatures are then digitized as nely as desired
and are fed into the data reduction program, which employs a direct curve-matching
method. Also input to the data reduction program are the measured airow rate, core
geometrical properties Dh ; Ao ; A;
, and core and air thermophysical properties. For an
assumed value of the heat transfer coecient, Eqs. (7.62)(7.67) are solved numerically
and the outlet air temperature distribution is determined as a function of time. This is
then compared with the measured temperaturetime distribution. If the measured and
numerical outlet temperature distributions do not match within a desired degree of
accuracy, iterations are continued on the heat transfer coecient until the temperature
distributions do match. Mullisen and Loehrke (1986) used the Regula-falsi method for
iterative adjustment on h. Typical curve matching is shown in Fig. 7.16. Once h is known,
the j factor is determined from its denition. The Reynolds number is also calculated
from the measured ow rate, core geometrical properties, and air thermophysical properties. The j factors determined by this direct curve-matching method are accurate within
2% for the complete range of NTU. For further details on the data reduction method,
refer to Mullisen and Loehrke (1986).
471
FIGURE 7.16 Typical curve matching of experimental temperaturetime history at the core
outlet with the numerically predicted curve. (From Mullisen, 1983.)
The experimental determination of ow friction characteristics of compact heat exchanger surfaces is relatively straightforward. Regardless of the core construction and the
method of heat transfer testing (steady state or transient), determination of the f factor
is made under steady uid ow rates with or without heat transfer. For a given uid ow
rate on the unknown f side, the following measurements are made: core pressure drop,
core inlet pressure and temperature, core outlet temperature for hot friction data, uid
mass ow rate, and core geometrical properties. The Fanning friction factor f is then
determined:
f
rh
1
2gc p 1
1
1
1
2
2
1
K
1
K
2
c
e
L 1=m
i
o i
o
G2
7:68
This equation is an inverted form of the core pressure drop equation (6.28). For the
isothermal pressure drop data, i o 1=1=m . Here Kc and Ke are sudden contraction and expansion pressure loss coecients presented in Fig. 6.3. The friction factor thus
determined includes the eects of skin friction, form drag, and local ow contraction and
472
expansion losses, if any, within the core. Tests are repeated with dierent ow rates on
the unknown side to cover the desired range of the Reynolds number. The experimental
uncertainty in the f factor is usually within 5% when p is measured accurately within
1%.
The Reynolds number is determined in the same way as described in Section 7.3.1.2
for heat transfer tests. The uncertainty in Reynolds numbers for both j and f factor
testing is 2% when the uid ow rate is measured accurately within 0.7%.
Generally, the Fanning friction factor f is determined from isothermal pressure
drop data (no heat transfer across the core). The hot friction factor f vs. Re curve
should be close to the isothermal f v. Re curve, particularly when variations in the
uid properties are small (i.e., the average uid temperature for the hot f factors is
not signicantly dierent from the wall temperature). Otherwise, the hot f factors
must be corrected to take into account the temperature-dependent uid properties (see
Section 7.6.1).
Example 7.3 Calculate the friction factor for the heat exchanger operating point of
Example 7.1.
SOLUTION
Problem Data and Schematic: Test core, geometry, operating conditions, and the
schematic are the same as those in Example 7.1; the following are specic data for the
friction factor evaluation:
pi 101:60 kPa
G 17:019 kg=m s
2
p 1:493 kPa
Ti 23:058C
To 101:608C
rh =L 0:005688
Kc 0:37
Ke 0:48
0:3067
Determine: The friction factor for the given heat exchanger operating point.
Assumptions: Steady-state ow and constant uid properties apply.
Analysis: We use Eq. (7.68) to determine the friction factor. Let us rst calculate the air
density i , o , and 1=m . For air as a perfect gas, the gas constant R~ 287:04 J=kg K
or (N m=kg K) for air.
i
pi
101:60
103 Pa
1:1946 kg=m3
~
RTi 287:04 J=kg K
273:158C 23:058C
0:9307 kg=m3
R~To 287:04 J=kg K
273:158C 101:608C
1
1 1
1
1
1
1
0:9558 m3 =kg
m 2 i o
2 1:1946 kg=m3 0:9307 kg=m3
o
473
K
1
K
c
e
L 1=m
i
o i
o
G2
"
3
2
0:005688
2
1
1:493
10 Pa 1 0:3067 0:37
0:9558 m3 =kg
1:1946 kg=m3
17:019 kg=m2 s2
1
1
1
1 0:3067 2 0:48
2
0:9307 1:1946 kg=m3
0:9307 kg=m3
Ans:
The Reynolds number for this point was calculated in Example 7.1 as Re 989.
Discussion and Comments: This example shows that the friction factor determination is
again straightforward if the geometry is known and all required measurements are made.
For this particular test point, notice that the core pressure drop contribution is the
largest; the entrance and exit loss contributions compensate each other, with the net
result as a small eect (6.6%).
The closed-form or approximate (such as numerical) solutions of governing equations in this book are referred to
as theoretical solutions (not as analytical correlations); for example, the solutions are in terms of the temperature
distribution for the energy equation, or the velocity/pressure distribution for the momentum equation. When
secondary parameters are computed (such as the heat transfer coecient or the Nusselt number) from exact or
approximate theoretical solutions, we refer to them as analytical (or numerical, if appropriate) correlations. In this
case, analytical/numerical results are generally expressed in a closed-form (preferably simplied) equation (with or
without curve tting) without any experimental input. When experimental results are expressed in a theory-based
equation (with constants modied), we refer to them as semiempirical correlations; when the experimental results
are expressed in an equation form without much theory base (such as regression), we refer to such an equation as
an empirical correlation.
474
475
H1 and *
H2 boundary conditions
axial direction as well as in the peripheral direction. The *
may be realized in gas turbine regenerators, counterow exchangers with C* 1, and
H1 boundary condition
nuclear and electric resistance heating. In these applications, the *
could be realized for highly conductive materials (such as copper, aluminum, etc.) where
H2 boundary
the temperature gradients in the peripheral direction are at a minimum; the *
condition is realized for very poorly conducting materials (such as ceramics, plastics,
Teon, etc.) for which temperature gradients exist in the peripheral direction. For intermediate thermal conductivity values, the boundary condition will be in between that of
H1 and *
H2 . It may be noted that the *
H1 and *
H2 boundary conditions for the symmetrically
*
heated passages with no sharp corners (e.g., circular, at, and concentric annular ducts)
are identical. In general, NuH1 > NuT , NuH1 NuH2 , and NuH2 NuT .
7.4.1
7.4.1.1 Laminar Flow. Nusselt numbers are constant for fully developed laminar
ow in ducts of constant cross-sectional area;{ but they depend on the ow passage
geometry and thermal boundary conditions. The product of the Fanning friction factor
and the Reynolds number is also constant, but dependent on the ow passage geometry. The fully developed laminar ow problem has been analyzed extensively for many
duct geometries. Analytical correlations for some technically important ow passages
are presented in Table 7.3. More detailed analytical correlations for some important
geometries are also presented in closed-form equations in Table 7.4. The following
observations may be made from Table 7.3:
. There is a strong inuence of ow passage geometry on Nu and fRe. For micro- or
meso-scale ow passages, the surface roughness height prole can be nonnegligible
compared to the passage size (Dh ) and hence can change the ow passage geometry
shape regardless of the original shape, and thus aect Nu and f Re even in
laminar ow. Rectangular passages with a small aspect ratio exhibit the highest
Nu and f Re.
H1 , *
H2 , *
T have a strong inuence on the Nusselt
. The thermal boundary conditions *
H1 boundary condition
numbers. Depending on the ow geometry, j factors for *
H2 boundary condition, and about
may be roughly 50% greater than that for the *
T boundary condition.
20% greater than that for the *
. As Nu hDh =k, a constant Nu implies the convective heat transfer coecient h
independent of the ow velocity (Reynolds number) and uid type (Prandtl
number).
. An increase in h can best be achieved either by reducing Dh or by selecting geometry
with a low-aspect-ratio rectangular ow passage. Reducing the hydraulic diameter
is an obvious way to increase exchanger compactness and heat transfer (a direction
adopted for the development of meso and micro heat exchangers, the ultracompact
heat exchangers), or Dh can be optimized using well-known heat transfer correlations based on design problem specications.
Note that for micro- and meso-scale passage geometries, the cross-sectional area of the duct may not be uniform
along the ow length due to manufacturing processes. In this case, the analytical correlations presented in Table
7.3 may not be accurate and truly representative for performance of such ducts.
476
TABLE 7.3 Solutions for Heat Transfer and Friction for Fully Developed Laminar Flow through
Specied Ducts
Geometry
L=Dh > 100
Aspect
Ratio
NuH1
NuH2
NuT
f Re
jH1 a
f
K1b
c
L
hy
3.014
1.474
2.39
12.630
0.269
1.739
0.040
3.111
1.892
2.47
13.333
0.263
1.818
0.040
3.608
3.091
2.976
14.227
0.286
1.433
0.090
4.002
3.862
3.34
15.054
0.299
1.335
0.086
4.123
3.017
3.391
15.548
0.299
1.281
0.085
4.364
4.364
3.657
16.000
0.307
1.250
0.056
2b 1
2a 4
5.331
2.94
4.439
18.233
0.329
1.001
0.078
2b 1
2a 6
6.049
2.93
5.137
19.702
0.346
0.885
0.070
2b 1
2a 8
6.490
2.94
5.597
20.585
0.355
0.825
0.063
2b
0
2a
8.235
8.235
7.541
24.000
0.386
0.674
0.011
p
3
2b
2
2a
p
3
2b
2
2a
2b
1
2a
2b 1
2a 2
477
* 2b=2a
Isosceles triangle:
4ab
p
Dh
a a2 4b2
Rectangular:
4ab
4b
Dh
a b 1 *
* 2b=2a
5
NuH1 2:059* 1:2489* 1:0559* 0:2515* 0:1520* 0:0901=*
3
2
3
for 1 * < 8
0:912* 13:3739* 78:9211* 46:6239=*
NuH2
*
0:312=
for 8 * 1
For 1 * 1:
2
3
4
5
6
NuH1 2:0591 0:7139* 2:9540* 7:8785* 5:6450* 0:2144* 1:1387*
for * 0:125
1:088*
NuH2
2
3
0:21131 10:9962* 15:1301* 16:5921* for 0:125 < * 1
For 0 * 1:
2
3
4
5
NuT 7:5411 2:610* 4:970* 5:119* 2:702* 0:548*
2
3
4
5
f Re 241 1:3553* 1:9467* 1:7012* 0:9564* 0:2537*
For 0 * 1:
f Re, NuT ,
TABLE 7.4 Laminar Fully Developed Analytical Correlations for Friction Factors and Nusselt Numbers for Some Duct Geometries
Continued
478
* 2b=2a
b
Dh
Em
Elliptical:
* 2b=2a
Right triangle:
4ab
p
Dh
a b a2 b2
2
2
4
2
* 6* 1
4
2
17* 98* 17
Em
2
For 0 * 2:
2
3
NuH2 0:3258*1 15:6397* 29:5117* 16:2250* =Em2
2
NuH1 91 *
f Re 21 * =Em2
For 0 * 1:
2
3
4
5
f Re 121 0:27956* 0:2756* 0:0591* 0:0622* 0:0290*
For 0 * 1:
f Re, NuT ,
479
2b
161 r*2
; rm
2
2
1 r* 2rm*
1 r*
2 ln1=r*
1=2
2
3
4
5
6
NuH 6:20661 2:3108r* 7:7553r* 13:2851r* 10:5987r* 2:6178r* 0:4680r*
For 0:02 r* 1:
For 0 r* 0:02:
Dh 2ro ri
r
r* i
ro
Concentric annular:
2a
* 2b=2a
z
Dh is a function of *
Sine:
f Re
For 0 r* 1:
480
TABLE 7.5 Inuence of Increasing Specic Variables on Theoretical Fully Developed Laminar
Friction Factors and Nusselt Numbers
Variable
Nu
Entrance eect
Passage-to-passage nonuniformity
Gross ow maldistribution
Free convection in a horizontal passage
Free convection with vertical aiding ow
Free convection with vertical opposing ow
Property variation due to uid heating
Increases
Decreases slightly
Increases sharply
Increases
Increases
Decreases
Decreases for liquids
and increases for
gases
Increases for liquids
and decreases for
gases
Increases sharply
Aects only if the surface
roughness height prole
is signicant compared
to the passage size (Dh )
Increases
Decreases signicantly
Decreases
Increases
Increases
Decreases
Increases for liquids
and decreases for
gases
Decreases for liquids
and increases for
gases
Increases slightly
Aects only if the
surface roughness
height prole is
signicant compared
to the passage size
Dh
Fouling
Surface roughness
481
2
4ffd L
G
p
K1
Dh
2gc
7:69
482
f =2Re Pr
(7.77)
(7.78)
(7.79)
Nu
Nulam 1
Nuturb
1:33 Re=6000
Nu 0:023Re0:8 Pr0:4
Turbulent ow in
a smooth duct
Transition and
turbulent ow,
circular or noncircular duct
(7.76)
(7.75)
(7.74)
(7.71)
(7.72)
(7.73)
(7.70)
Eq.
No.
Nu
f =2Re 1000Pr
900
0:63
Re 1 10Pr
C 1:07
Nu
Blasius:
f 0:0791Re0:25
McAdams:
f 0:046Re0:2
Bhatti and Shah: f 0:00128 0:1143Re0:311
f A B Re1=m
1. A 0:0054; B 2:3
108 ; m 2=3
2. A 0:00128; B 0:1143; m 3:2154
Correlations
Transition and
turbulent ow,
circular or noncircular duct
Turbulent ow in
a smooth duct
Turbulent ow in
a smooth duct
Transition and
turbulent ow,
circular or noncircular duct
Type of Flow
and Geometry
Gnielinski correlation
Accuracy 10%
2300 Re 5
106 , 0:5 Pr 2,000
Not a good correlation in the transition
regime
Obtain f from Eq. (7.70) or (7.73)
Use Nuturb from Eq. (7.74) or (7.76) and
Nulam from Table 7.3
Applicable for 2000 < Re < 8000
PetukhovPopov correlation
Accuracy 5%
4000 Re 5
106
0:5 Pr 106
Obtain f from Eq. (7.70) or (7.73)
BhattiShah correlation
1. 2100 Re 4000
2. 4000 Re 107
Accuracy 2%
Remarks
Winterton (1998)
Taborek (1990)
Gnielinski (1976)
Reference
TABLE 7.6 Some Important Correlations for f and Nu for Transition and Turbulent Flows in Circular and Noncircular Smooth Tubes
483
Turbulent ow in
a smooth duct
Nu 0:023Ren Pr0:4
NuT 4:5
Pr0:86
w
(7.84)
7:83
(7.82)
0:0156Re0:85
f
Pr0:86
NuH 5:6 0:0165Re0:85
w
f
Liquid metal
turbulent ow
in a smooth duct
(7.80)
(7.81)
for heating
for cooling
Nu 0:023Re0:8 Pr1=3
0:024Re0:8 Pr0:4
0:026Re0:8 Pr0:3
Turbulent ow in
a smooth duct
Nu
Turbulent ow in
a smooth duct
10 < Re < 5
10 and 0 Pr 0:1
4
DittusBoelter correlation
2500 < Re < 1:24
105 and
0:7 Pr 120
See text for accuracy
Colburn correlation
104 < Re < 105 and 0:5 Pr 3
Accuracy within 27:6% and 19:8%
Chen and Chiou correlation
Present authors
484
485
FIGURE 7.18 NuH for a circular tube for fully developed turbulent ow. (From Bhatti and Shah,
1987.)
486
FIGURE 7.19 Dependence of the exponent n of Eq. (7.84) on Re and Pr when comparing Eq.
(7.84) with Eq. (7.74).
Most of the thermal or ow resistance is concentrated in the viscous sublayer near the
wall for turbulent ow, and the temperature and velocity proles are relatively at over
most of the channel cross section. Hence, provided that there are no sharp corners, the
inuence of the channel shape in turbulent ow is not as great as that in laminar ow. A
common practice is to employ hydraulic diameter as the characteristic length in the
circular tube correlations to predict Nu and f for noncircular channels. Hence, it is
generally an accepted fact that the hydraulic diameter correlates Nu and f for fully
developed turbulent ow in circular and noncircular ducts. This is true for the results
to be accurate to within 15%{ for most noncircular ducts except for those having sharp
or acute-angled corners in the ow passage or concentric annuli with the inner wall
heated. In these cases, the values of Nu and f factors could be more than 15% lower
than the values for the circular tube. Also, the corners and noncircularity of ow
passages would aect the ow phenomena, and Nu and f will be somewhat dierent
from those for the circular tube.
A careful observation of accurate experimental friction factors for all noncircular
smooth ducts (and the correlations in Table 7.7) reveals that ducts with laminar
f Re < 16 have turbulent f factors lower than those for the circular tube; whereas
ducts with laminar f Re > 16 have turbulent f factors higher than those for the
circular tube. Similar trends are observed for the Nusselt numbers. If one is satised
{
This order of accuracy is adequate for most engineering calculations for overall heat transfer and pressure drop,
although it may not be adequate for detailed ow distribution and local temperature distribution analyses as
required, for example, in a nuclear reactor or when mass production of heat exchangers requires accuracy within
5% or lower for cost savings.
487
TABLE 7.7 Fully Developed Turbulent Flow Friction Factors and Nusselt Numbers (Pr > 0.5) for
Some Technically Important Smooth-Walled Ducts
Duct Geometry
Recommended Correlations
Rectangular:
4ab
2a
; *
ab
2b
D 2 11
*2 *
Dh 3 24
Dh
Isosceles triangular:
4ab
p
Dh
a a2 4b2
Dg
1
Concentric annular:
Dh 2ro ri ; r*
2
ri
ro
2
D 1 r* 1 r* = ln r*
Dh
1 r*2
488
within 15% accuracy, Eqs. (7.70) and (7.76) in Table 7.6 for f and Nu can be used for
noncircular passages with the hydraulic diameter as the characteristic length in f, Nu,
and Re; otherwise, refer to Table 7.7 for more accurate results for turbulent ow in those
duct geometries.
Example 7.4 A parallelow exchanger is operating at the following conditions:
m_ h 3000 kg=h, m_ c 6000 kg=h, Th;i 508C, Th;o 308C, Tc;i 108C. The properties
for both streams (water) are 1000 kg=m3 , cp 4180 J=kg K, k 0:59 W=m K, and
0:001 Pa s. The heat transfer coecients on both sides are 5000 W/m2 K. Assume
fouling resistances and the wall thermal resistance to be negligible. Determine how much
surface area is necessary for this exchanger. Assume water properties constant with
temperature and Dh 15 mm on both uid sides.
The heat transfer coecients vary with the velocity (or ow rate) according to the
DittusBoelter correlation:
hDh
GDh 0:8 0:4
Pr
0:023
k
What are the exit temperatures if (1) the hot uid rate is doubled; (2) ow rates of both
uids are doubled? Also compute and compare the heat transfer rates for each case.
SOLUTION
Problem Data and Schematic: Flow rates, heat transfer coecients, and inlet and outlet
temperatures as well as uid properties are given in Fig. E7.4 for a parallelow exchanger. Also given is the variation of heat transfer coecients with water velocity in terms of
the DittusBoelter correlation.
50C
30C
20C
10C
m
h = 3000 kg/h
m
c = 6000 kg/h
hh = hc = 5000 W/m2 K
Dh = 15 mm
x
FIGURE E7.4
Determine: The surface areas of on both uid sides, and the exit temperatures for various
specied ow rates.
Assumptions: Fluid properties are constant, wall thermal resistance and fouling
resistances are negligible, and both uids ow through smooth surfaces.
Analysis: In the rst portion of the problem, " and C* are specied. We determine NTU
and subsequently, the surface area. Note that the hot uid is the Cmin uid since its ow
489
0:50
Th;i Tc;i 50 108C
0:5
Cmax Cc 6:967 kW=K
Using Eq. I.2.2 of Table 3.6 for the parallelow exchanger for the "-NTU relationship,
we have
NTU
1
1
1
1
ln
ln
0:9242
1 0:5
1 0:51 0:5
1 C*
1 "1 C *
Uh hh hc 5000 W=m2 K 5000 W=m2 K 2500 W=m2 K
) Uh 2500 W=m2 K
Since we have determined UA and Uh separately, the surface area is then computed as
Ah
UA
3219 W=K
1:288 m2
Uh
2500 W=m2 K
Ans:
Similarly,
Ac 1:288 m2
Ans:
490
or
Tc;o Tc;i
Ch
T Th;o 10 0:5
50 30 208C
Cc h;i
m_ new
m_ old
0:8
5000 W=m2 K
20:8 8705:5 W=m2 K
0:0003149 m2 K=W
Unew hh;new hc 8705:5 W=m2 K 5000 W=m2 K
Therefore, Unew 3175:9 W=m2 K. So the new value of NTU will be
NTU
3175:9 W=m2 K
1:288 m2
0:5871
6:967
103 W=K
Since Ch Cc 6:967 kW=K, C* 1. Hence, from Eq. I.2.1 of Table 3.6 or for parallelow in Table 3.3,
1 eNTU1C 1
1 e2NTU 0:51 e2
0:5871 0:3455
"
2
1 C*
The new outlet temperatures for this C* 1 case are
Th;o Th;i "Th;i Tc;i 508C 0:345550 108C 36:28C
Ans:
Ans:
491
Cc 13:933 kW=K
6000 kg=h
4180 J=kg K
0:5
12000 kg=h
4180 J=kg K
1
1
hh hc
1
8705:5 W=m2 K
4352:8 W=m2 K
2
UA
4352:8 W=m2 K
1:288 m2
0:8047
6967 W=K
Cmin
1 eNTU1C 1 e0:875
1:5
"
0:4673
1 0:5
1 C*
Finally, the new outlet temperatures are
Th;o Th;i "Th;i Tc;i 508C 0:467350 108C 31:18C
Ans:
Ans:
Ans:
3000 6000
6000 6000
6000 12000
5.0
8.7
8.7
5.0
5.0
8.7
C*
"
Th;o Tc;o
(8C) (8C)
Discussion and Comments: The objective of this example is to demonstrate the eect of
increasing ow rates of one or both uids on exchanger performance. Since the "-NTU
relationship is not linear in general [for this problem, it is given for parallelow in Table
3.3], one cannot expect that doubling the ow rates on one or both uid sides will double
the heat transfer rates. Also, the heat transfer coecient will vary nonlinearly with
492
the ow rate; in this example, it will vary according to the DittusBoelter correlation
mentioned in the problem statement.
Doubling the ow rate on the Cmin side as in case 1 decreases NTU, increases C*, and
decreases " as expected. However, the decrease in " is much less than linear, and Cmin
increases linearly with m_ h . Hence, the net heat transfer rate in the exchanger increases.
Notice that despite increasing q, the Th;o increases (and not decreases) due to the
increased ow rate. However, since we have not changed the Cmax Cc ow rate,
the cold uid outlet temperature increases due to the increased heat transfer in the
exchanger. Thus, this problem clearly demonstrates the peculiar behavior of arrows
going up or down for Th;o and Tc;o in Table 3.5. However, notice that we cannot use
Table 3.5 directly for this problem since Cmin Ch for this problem and Table 3.5 is for
Cmin Cc :
Next, when we double both uid ow rates, the value of NTU is in between the base
case and case 1, due to increased overall U. This, along with decreased C* compared to
case 1, produces higher " and q. However, notice that due to the nonlinear nature of the
problem, by doubling the ow rates on both sides, q is still not doubled compared to the
base case. As a result, we get Tc;o even lower than that for the base case. However, as
expected, Th;o will be in between that for the base case and case 1.
Thus, we nd that due to the nonlinear relationship between h and m_ , and between "
and NTU, we obtain outlet temperatures that are higher or lower than the base case from
the exchanger with increasing ow rates. Thus the use of Table 3.5 for Cmin Cc or its
counterpart for Cmin Ch can provide qualitative guidelines on the variations in Th;o and
Tc;o , but actual computation as shown in this example is necessary for quantitative
results.
Example 7.5 You as a designer need to decide whether to select a rectangular (aspect
ratio * 18) or a square duct (* 1) of 5000 mm2 cross-sectional area for air owing at
0.05 kg/s at 278C for maximum heat transfer and for minimum pressure drop. The duct
length is 5 m. The duct wall temperature is 1008C, due to the steam on the other side.
For fully developed turbulent ow for rectangular and square ducts, use the following
correlations for friction factors and Nusselt numbers:
f 0:0791 Re0:25 1:0875 0:1125*
Nu 0:024 Re0:8 Pr0:4
where * is the aspect ratio of the rectangular ow passage. Use the following properties
for air:
1:0463 kg=m3
cp 1008 J=kg K
202:5
107 Pa s;
k 29:1
103 W=m K
Pr 0:702
SOLUTION
Problem Data and Schematic: Two pipes for this problem are shown in Fig. E7.5. The
uid properties are specied in the problem statement.
493
m = 0.05 kg/s
Ti = 27C
L=5m
70.71 mm
25 mm
200 mm
70.71 mm
FIGURE E7.5
Determine: The pressure drop and heat transfer for each duct, and decide which geometry will yield a lower pressure drop and higher heat transfer.
Assumpions: Fully developed turbulent ow exists with constant uid properties.
Analysis: Let us rst compute the hydraulic diameters and Reynolds numbers for the
rectangular and square ducts.
* 1=8
Dh
* 1
4Ao 4
200 mm
25 mm
P
2200 25 mm
Dh
44:44 mm
G
Re
4Ao 4
70:71 mm
70:71 mm
P
270:71 70:71 mm
70:71 mm
m_
0:05 kg=s
10 kg=m2 s
Ao 5000
106 m2
GDh 10 kg=m2 s
44:44
103 m
202:5
107 Pa s
Re
21,946
m_
0:05 kg=s
10 kg=m2 s
Ao 5000
106 m2
GDh 10 kg=m2 s
70:71
103 m
202:5
107 Pa s
34,919
Heat Transfer. To calculate the air outlet temperature and heat transfer rate, we deterT
mine the exchanger eectiveness for computed values of NTU and C*. For the *
boundary condition, the temperature distributions of the wall and air are similar to
those of Fig. 3.1c. They correspond to the C * 0 case. Hence the "-NTU formula of
Eq. (3.84) applies. Since there is no thermal resistance specied for the wall, the other
uid side and fouling, and no ns, we have from Eq. (3.20),
UA hA
where h is computed from the Nusselt number correlation given. We now compute q and
Ta;o separately for * 18 and 1.
* 18 Case. Let us rst compute the air-side heat transfer coecient. From the correlation given,
Nu 0:024 Re0:8 Pr0:4 0:024219460:8 0:7020:4 61:92
Hence,
h Nu
k
29:1
103 W=m K
40:55 W=m2 K
61:92
Dh
44:44
103 m
494
To compute NTU UA=Cmin hA=Cmin for this problem, let us compute A and Cmin .
The surface area A of 5 m long rectangular pipe (* 18) is
A PL 2200 25
103 m
5 m 2:25 m2
Cmin m_ cp 0:05 kg=s
1008 J=kg K 50:40 W=K
Therefore,
NTU
hA
40:55 W=m2 K
2:25 m2
1:810
50:40 W=K
Cmin
Now
" 1 eNTU 1 e1:810 0:8364
q "Cmin Tw Ta;i 0:8364
50:40 W=K
100 278C 3077 W
Ans:
q
3077 W
278C
88:18C
Cmin
50:40 W=K
k
29:1
103 W=m K
36:95 W=m2 K
89:78
Dh
70:71
103 m
A PL 270:71 70:71
103 m
5 m 1:414 m2
Cmin m_ cp 0:05 kg=s
1008 J=kg K 50:40 W=K
NTU
hA
36:95 W=m2 K
1:414 m2
1:037
50:40 W=K
Cmin
Ans:
q
2375 W
278C
74:18C
Cmin
50:40 W=K
Ans:
495
4fLG2
4
0:006976
5 m
10 kg=m2 s2
150:0 Pa
2gc Dh 2
1
1:0463 kg=m3
44:44
103 m
Ans:
Pressure Drop for * 1. The friction factor for this case, from the correlation given, is
f 0:0791Re0:25 1:0875 0:1125*
0:0791349190:25 1:0875 0:1125
1
0:005642
Hence, the frictional pressure drop for this pipe is
p
4fLG2
4
0:005642
5 m
10 kg=m2 s2
76:3 Pa
2gc Dh 2
1
1:0463 kg=m3
70:71
103 m
Ans:
q (W)
p (Pa)
q=q1
p=p1
1
8
3077
2375
150.0
76.3
1.30
1.00
1.97
1.00
From the results above and intermediate aspect ratios, we nd that the lower the aspect
ratio of a rectangular duct, the higher is the heat transfer rate and pressure drop in
turbulent ows. The increase in p is higher than the increase in q. Hence, if the pressure
drop is the constraint, choose a square duct in the rectangular duct family. If the higher
heat transfer rate is the requirement, choose a rectangular duct with as low an aspect
ratio as possible compared to a square duct.
Discussion and Comments: This example demonstrates that the pressure drop increases
considerably as the aspect ratio of a rectangular duct decreases in turbulent ow. It can
also be shown that this pressure drop increase with decreasing aspect ratio is even more
pronounced (>1.97) for laminar ow. Hence, if one is interested in a rectangular duct
design for uid ow only, the square duct will provide the minimum pressure drop. The
heat transfer rate increases with decreasing aspect ratio as shown above, but the increase
is lower compared to the increase in the pressure drop. Although this problem is meant
for the duct design connecting various components of a thermal system, if one is interested in heat exchanger surface design, the exchanger surface selection criteria are presented in Section 10.3.
Circular Tube with Surface Roughness. A roughness element has no eect on laminar
ow unless the height of the roughness element is not negligible compared to the ow
cross-sectional size{. However, it exerts a strong inuence in transition and turbulent
{
This means that surface roughness does inuence j (or Nu) and f factors, even in laminar ow if the surface
roughness height is not negligible compared to the ow passage size. This is because the ow cross-section shape is
changed due to the surface roughness and hence Nu and f change, as discussed in Section 7.4.1.1.
496
p
eu* e w gc 1=2 e um di w gc 1=2 e
Re f =2
2
di
di
um
7:85
497
regime (0 e < 5), transition regime (5 e 70), and fully rough regime (e > 70).
Note that the currently accepted ranges of e for the three regimes dened above are
slightly dierent from those originally proposed by Nikuradse. For the fully rough
regime, there will not be a viscous sublayer. The friction factor correlations for these
three regimes are presented in Table 7.8.
The roughness obtained by Nikuradse with closely packed sand-grains can be
regarded as roughness of the maximum density. The roughness is adequately characterized by the height e or the ratio e=Dh . However, other types of roughness involving
uniformly distributed elements of nite sizes such as spheres, spherical segments,
cones, and the like, cannot be characterized by e or e=Dh alone. In such cases, it is
convenient to determine an equivalent sand-grain roughness es so that Nikuradses
measurements of fully rough regime friction factors and the velocity distribution with
sand-grain roughness can be utilized.
When two-dimensional rib roughness with ribs at right angles to the ow direction are
tted to the equivalent sand-grain roughness scale, a very large value of es =Dh results
compared to the e=Dh value of the ribs. For example, with e=Dh 0:001, a value of
es =Dh 0:025 is obtained in the fully rough ow regime. This means that a two-dimensional riblike roughness element perpendicular to the ow direction is appreciably more
eective in increasing f than is a sand-grain type of element of the same height.
The natural roughness occurring in commercial pipes is three-dimensional (similar to
the sand-grain roughness) and has a random distribution and arbitrary shape. The major
dierence in f vs. Re curves of sand-grain vs. natural roughness is in the transition
region of rough ow. For the sand-grain roughness, f reaches a minimum at an intermediate Re and then gradually rises to an asymptotic constant value with increasing Re.
For natural roughness, f monotonically decreases with Re and reaches an asymptotic
value at high Re. The Fanning friction factors for commercial rough pipes are shown in
Fig. 6.4.
Based on Eq. (7.87) in Table 7.8 or Fig. 6.4, the friction factor for the fully rough
region is constant, independent of Re. Hence, whenever experimental f factors for a
heat exchanger surface are almost constant, independent of Re (usually at high Re), we
characterize that surface as behaving as a rough surface in that Re range.
Turbulent ow heat transfer studies with sand-grain (two- and three-dimensional)
and repeated rib roughnesses have been conducted in detail as summarized by Dipprey
and Sabersky (1963) and Webb (1994). Dipprey and Sabersky recommended the follow-
TABLE 7.8 Circular Tube Rough Surface Flow Regimes and Friction Factor Correlations
Flow Regime
Range of e
Hydraulically
smooth
Transition
0e <5
Fully rough
e > 70
5 e 70
Correlation
See Eq. (7.70) in Table 7.6
1
2e 16:2426
p 3:48 1:7372 ln
ln A2
di
Re
f
1:1098
2e=di
7:149 0:8981
where A2
6:0983
Re
1
2e
e
p 3:48 1:737 ln 2:28 1:737 ln
di
di
f
Eq. No.
(7.70)
(7.86)
(7.87)
498
ing correlation based on heat transfer measurements for closed-packed sand-grain type
roughness in a circular tube:
Nu
f =2Re Pr
p
f =2
ge Pr0:44 Be
7:88
where and f is given by Eqs. (7.70), (7.86), and (7.87) (in Table 7.8); Be is given in
Table 7.9 for the appropriate ranges of e [obtained based on the approximate straightline curves given by Dipprey and Sabersky (1963)]; and ge Pr0:44 is presented in Fig.
7.20 and correlated by{
ge Pr
0:44
8
9:684 0:0658e 0:0003e 2
>
>
>
>
>
>
24:5 0:24e
>
>
>
>
>
>
> 2:5 0:2e
<
for e 70
7:90
Note that the last expression in Eq. (7.89) covers the complete range of Prandtl numbers
from 1.20 to 5.94, but the values for Pr 1:20 for e 20 have an error of up to 25%
compared to the data of Dipprey and Sebarsky (1963). The other individual expressions
are accurate within 8% of a one-decimal-point reading of the data from the gure of
Dipprey and Sabersky (1963). Equation (7.88) is valid for 1 < Pr < 6. A more general
equation applicable to the fully rough ow regime (e > 70) in the range 0:5 < Pr < 5000
is given by Shah and Bhatti (1988) as{
Nu
f =2Re 1000Pr
p
0:2 0:5 8:48
f =2f17:42 13:77Pr0:8
t ge Pr
7:91
0 e 3
3 e 7
7 e 14
14 e 70
e 70
5:5 2:5 ln e
7:30 0:318e
9.52
9:78 0:0186e
8.48
{
These correlations were developed by Ms. Zeng Deng, a Ph.D. student at Oklahoma State University, Stillwater,
OK (2002) from the graphical results of Dipperey and Sebarsky (1963).
{
In previous publications, there were some errors in this equation; this is the correct version.
499
FIGURE 7.20 Heat transfer correlation of sand grain roughness data. (From Dipprey and
Sabersky, 1963.)
where the turbulent Prandtl number Prt is a function of Pr being given by the following
correlation, presented by Malhotra and Kang (1984) for a circular tube in the range
104 < Re < 106 :
8
0:36
>
< 1:01 0:09Pr
Prt 1:01 0:11 ln Pr
>
:
0:99 0:29ln Pr1=2
for 1 Pr 145
for 145 Pr 1800
for 1800 < Pr 12,500
7:92
The prediction of Eq. (7.91) in conjunction with Eq. (7.92) agrees within 15%, with the
reported experimental data in the fully rough ow regime for 0:7 Pr 4600.
In the transition regime of ow 5 e 70 over sand-grain surface roughness, the
St number can increase by a factor of 2 to 3 with a corresponding increase in the friction
factor with St=f approaching 0.25 from lower values. This is the region where the surface
roughness can reduce the surface area requirement for a given pressure drop. However,
beyond a certain combination of Re and Pr, the surface roughness increases f in fully
rough region (see Fig. 6.4) with no signicant increase in St. In that region, it is found
that the j factor continues to decrease as Re increases for all e=di values, while the f
factor remains constant. The reason for this behavior is that the surface roughness
induces form drag that becomes dominant compared to skin friction at high Re. This
results in the attened (nondecreasing) f vs. Re curves shown in Fig. 6.4. This form drag
does not have a heat transfer counterpart, and hence j decreases as Re increases, similar
to a smooth surface, thus reducing the relative heat transfer performance of a rough
surface.
7.4.2
500
For a circular tube, this distance is given by (Shah and Bhatti, 1987; Bhatti and Shah,
1987)
(
0:056Re
for laminar flow Re 2100
Lhy
7:93
Dh
1:359Re1=4
for turbulent flow Re > 104
Equation (7.93) shows that Lhy =Dh 137 for laminar ow at Re 2100, whereas
Lhy =Dh 13:6 for turbulent ow at Re 104 . Thus, the hydrodynamic entrance eect
persists over a longer distance in laminar ow than in turbulent ow. For practical
purposes, the hydrodynamic entrance eect manifests itself as an increase in both the
friction factor and heat transfer coecient.
7.4.2.1 Laminar Flow. The friction factors in the hydrodynamic entrance region are
higher than those for the fully developed case. The eects of skin friction and momentum rate change in the entrance region are included in the apparent friction factor [see,
Eq. (7.19)]. The fapp Re factors for four passage geometries are presented as a function
of x in Fig. 7.21. For other geometries, refer to Shah and London (1978) and Shah
and Bhatti (1987). It can be seen from Fig. 7.21 that for very low values of x , the
values of fapp Re are about the same for all four geometries. This is because the eect
of the boundary layer development is the same and is not inuenced by the boundary
layer of the neighboring surface in the very early entrance region.
Based on the solutions for laminar boundary layer development over a at plate and
the fully developed ow in circular and some noncircular ducts, fapp Re are correlated
by the following equation (Shah and London, 1978):
fapp Re 3:44x 0:5
7:94
FIGURE 7.21 fapp Re factors for parallel plates, and circular, square, and equilateral triangular
ducts for developing ow. (From Shah and London, 1978.)
501
TABLE 7.10 K1
1, fERe, and C 0 for Use in Eq. (7.94)
K1
f Re
C0
Rectangular Ducts
____
*
1.00
0.50
0.20
0.00
1.43
1.28
0.931
0.674
14.227
15.548
19.071
24.000
0.00029
0.00021
0.000076
0.000029
1.69
13.333
0.00053
1.25
0.830
0.784
0.688
0.678
0.674
16.000
21.567
22.343
23.813
23.967
24.000
0.000212
0.000050
0.000043
0.000032
0.000030
0.000029
where K1, f Re, and C 0 are given in Table 7.10 for rectangular, equilateral triangular, and concentric annular ducts. Here fapp is dened the same way as f [compare
Eqs. (7.18) and (7.19)], but it includes the contributions of additional pressure drop due
to the momentum rate change and excess wall shear between developing and developed
ows. Considering only the rst term on the right-hand of Eq. (7.94), it can be shown that
f / Re0:5 .
It should be emphasized that while Eq. (7.94) will predict the apparent friction factors
accurately for a single duct, it generally does not predict the apparent friction factors
even fairly approximately for interrupted n geometries, such as strip ns and louver
ns used in compact heat exchangers. This is because it includes only the eect of skin
friction. The form drag associated with the smooth and burred edges of surface interruptions and the wake eect may contribute signicantly to the pressure drop. Hence,
analytical apparent friction factors are generally not used in designing exchangers. But
as a rule of thumb, f 4j or a similar relationship may be used to predict f factors
for interrupted surfaces for which j factors are already known either from the theory
or from experiments.
7.4.2.2 Turbulent Flow. In the entrance region, the turbulent ow friction factors are
higher than those in fully developed ow. However, the entrance length is very short,
generally less than 10 tube diameters, and other sources of pressure drop at the tube
entrance are far more important in most applications. Hence, the inuence of the
developing turbulent ow region is generally neglected in pressure drop evaluation.
Zhi-qing (1982) has developed a closed-form solution for fapp Re for developing
turbulent ow in a circular tube and has shown that fapp Re in this case is dependent
on both Re and x (Shah and Bhatti, 1988).
502
7.4.3
* 0:0335
Lth;T
7:95
This translates into Lth =Dh for air (Pr 0:7) at Re 2000 as 60 and 47 and signicantly
higher (over 100) for water and other liquids with high Prandtl numbers. The reason for
the longer thermal entrance length with high-Pr uids (i.e., uids with relatively low
thermal diusivity compared to the momentum diusivity) is that the thermal boundary
layer develops gradually (see Fig. 7.6b to get the basic concept).
The turbulent ow thermal entry length for gases and liquids is almost independent of
Re and thermal boundary condition. Lth =Dh varies from about 8 to 15 for air, and
Lth =Dh < 3 for liquids. For noncircular ducts, Lth =Dh may be as high as 30 to 40, due
to the coexistence of laminar ow in corner regions.
7.4.3.1 Laminar Flow. The thermal entrance Nusselt numbers are higher than those
for the fully developed case. The local Nusselt numbers for four ow geometries are
T and *
H1 boundary conditions. Theoretically, the
shown in Figs. 7.22 and 7.23 for the *
Nusselt numbers are innity at x 0 and asymptotically approach the fully developed
FIGURE 7.22 Comparison of Nux:T for parallel plates, and circular, rectangular, and isosceles
triangular ducts for developed velocity and developing temperature proles. (From Shah and
London, 1978.)
503
FIGURE 7.23 Comparison of Nux:H1 for parallel plates, and circular, rectangular, equilateral
triangular, and semicircular ducts for developed velocity and developing temperature proles.
(From Shah and London, 1978.)
values as x* increases. The thermal entrance solutions for Nu for many geometries are
presented in Shah and London (1978) and Shah and Bhatti (1987).
H1 boundary
T and *
The thermal entrance local and mean Nusselt numbers for the *
conditions for circular and noncircular ducts having laminar developed velocity proles
and developing temperature proles are correlated as (Shah and London, 1978)
Nux;T 0:427 f Re1=3 x*1=3
7:96
7:97
where f is the Fanning friction factor for fully developed ow, Re is the Reynolds
number, and x* x=Dh Re Pr. For interrupted surfaces, x s . The equations
above are recommended for x* < 0:001. The following observations may be made
from Eqs. (7.96) and (7.97) and other solutions for laminar ow surfaces having
developing laminar ows.
. The inuence of thermal boundary conditions on thermally developing ow
appears to be of the same order as that for thermally fully developed ow; we
have considered hydrodynamically fully developed ow in both cases.
. Since Nu / x*1=3 x=Dh Re Pr1=3 , Nu / Re1=3 / u1=3
m . Therefore, h
varies as u1=3
m .
. Since the velocity prole is considered fully developed, p / um , as noted earlier.
504
. The inuence of the duct shape on thermally developing Nu is not as great as that
for the fully developed Nu.
Equations (7.96) and (7.97) may not yield accurate values of the Nusselt numbers for
interrupted surfaces such as strip n, and louver n, due to ow separation, vortex ow,
burred edges, and other factors. In experimental data of many interrupted surface geometries, it is found that the exponent on x* varies from 0.3 to 0.6, depending on
the ow type, ow development, and other factors. Using an experimental value of the
exponent (which may be dierent from 0.33), Eqs. (7.96) and (7.97) do provide valuable
guidelines for predicting the performance of a surface of a family for which no experimental j and f data are available. For example, consider that we know j1 and f1 at some
Re (in laminar ow) for an interrupted surface having the interruption length 1 .
However, in the design, we would like to employ an interrupted surface of the same
type but having the interruption length 2 . As soon as we change the interruption length
from 1 to 2 , the original j and f data for the surface with the interruption length 1 are
no longer valid. In this case, the new j2 and f2 at the same Re for the same uid can be
obtained heuristicly as follows.
Since j Nu Pr1=3 /Re, from Eqs. (7.96) and (7.97), we get
j / x*1=3 /
Dh
1=3
7:98
Therefore,
j2
j1
2 =Dh;2 1=3
1 =Dh;1
7:99
Since all other quantities are known, j2 can readily be calculated from Eq. (7.99). If no
other information is available, consider
f2
j
2
f1
j1
7:100
to evaluate f2 . This is because the contribution of the form drag is of the same order of
magnitude as that of the skin friction. Hence, we cannot use the theory-based developing
ow friction solutions for interrupted nned surfaces. It should be emphasized that
Eq. (7.100) implies that the relationship between f2 and j2 is the same as it is between
f1 and j1 . Instead of changing the interruption length from 1 to 2 in the foregoing
example, if we had changed the n density, it would have resulted in the hydraulic
diameter changing from Dh;1 to Dh;2 . In that case, the same foregoing procedure could
be used to determine j2 and f2 .
The theoretical ratio Num =Nufd is shown in Fig. 7.24 as a function of x* for several
passage geometries having the constant wall temperature boundary condition. The
following observations may be made from this gure.
. The entrance region Nusselt numbers and hence the heat transfer coecients could
be two to three times higher than the fully developed values, depending on the
interruption length * x*.
505
FIGURE 7.24 Ratio of laminar developing to developed ow NuT for dierent ducts; the velocity
prole is developed for both Nus. (From Shah and Webb, 1983.)
. At x* 0:1, although the local Nusselt number approaches the fully developed
value, the value of the mean Nusselt number can be signicantly higher than that
for fully developed ow for a channel of length * x* 0:1:
. The ratio Num =Nufd increases from the lowest values for parallel plates to the
highest values for the equilateral triangular duct at a given x*, and thus this
order is a function of the channel shape. Notice that it is just the opposite nature
of that for Nufd for fully developed ow in Table 7.3. This is because the duct crosssectional shape aects Nux;T and Num;T marginally until the thermal boundary
layers of the neighboring walls start interacting. For a highly interrupted surface,
a basic inferior passage geometry for fully developed ow (such as triangular) does
not penalize signicantly in terms of low values of Nu or h in thermally developing
ow.
. A higher value of Num =Nufd at x* 1 means that the ow channel has a longer
entrance region.
7.4.3.2 Turbulent Flow. The Nusselt numbers (Nux ) in the thermal entrance region are
higher than those for the fully developed turbulent ow. This trend is similar to that for
the laminar ow case. However, unlike laminar ow, those Nusselt numbers are not
dependent on the thermal boundary conditions for Pr 0:7.
T and
The following correlations are for the local and mean Nusselt numbers for the *
H thermally developing ow in a circular passage (Bhatti and Shah, 1987):
*
Nux
C6
1
Nu1
10x=Dh
Num
C
1 6
Nu1
x=Dh
7:101
506
where Nu1 stands for the fully developed NuT or NuH derived from the formulas
recommended in Table 7.7 and
C6
x=Dh 0:1
3000
0:68
Re0:81
Pr1=6
7:102
These correlations are valid for x=Dh > 3, 3500 < Re < 105 , and 0:7 < Pr < 75. The
value of Num agrees within 12% with the experimental measurements for Pr 0:7.
Notice that Nux and Num of Eq. (7.101) are functions of x=Dh , Re, and Pr; in
contrast, the Nusselt numbers for developing laminar ow depend only on
x* x=Dh Re Pr. As an illustration, the turbulent entrance region local Nusselt
H boundary condition.
numbers are presented in Fig. 7.25 for a circular tube having the *
H turbulent ow thermal
*
7.4.4
507
7.4.4.1 Laminar Flow. In simultaneously developing ow, the velocity and temperature proles both develop in the entrance region. If they are uniform at the duct inlet,
the uid velocity, velocity gradients, and temperature gradients near the wall in the
entrance region will be higher than those for velocity proles already developed. The
higher velocities near the wall convect more thermal energy in the ow direction, and
heat transfer in the thermal entrance region is higher for the case of developing velocity
proles. As an example, Fig. 7.26 compares the circular duct Nux;H for developed and
developing velocity proles. The curve for Pr 1 in this gure represents the correlation for any uid (Pr < 1) having its velocity prole fully developed before the temperature prole starts developing. It is clear from Fig. 7.26 that the Nusselt number is
higher for the simultaneously developing ow situation (Pr 1).
Additionally, in simultaneously developing ow, the rate of development of the temperature boundary layer relative to the velocity boundary layer does depend on the uid
Prandtl number (refer to Fig. 7.6b and c). If the velocity and temperature proles are
uniform at the duct entrance, the lower the uid Prandtl number, the faster the development of the temperature boundary layer will be in comparison to the velocity boundary layer in the entrance region of the duct. This would result in lower temperature
gradients at the wall and in turn decrease the Nusselt number and heat transfer at a
given x x=Dh Re). Thus, the lower the Prandtl number, the lower the Nusselt
number will be at a given x for specied duct geometry. However, if the axial coordinate
is stretched or compressed by considering x* x =Pr, it is found that the lower the
508
FIGURE 7.27 Simultaneously developing laminar ow Nux;H1 for Pr 0:7 for several constantcross-section duct geometries. (From Shah and London, 1978.)
Prandtl number, the higher the Nusselt number will be at a given x*. The eect of the uid
Prandtl number on the entrance region Nusselt numbers is shown in Fig. 7.26 for the
circular tube. The Nux;H1 for simultaneously developing ow for the circular duct,
parallel plates, rectangular ducts, and equilateral triangular ducts are compared in
Fig. 7.27 for Pr 0:7.
The theoretical entrance-region Nusselt numbers for simultaneously developing ow
are higher than those for thermally developing and hydrodynamically developed ow.
These theoretical solutions do not take into account the wake eect or secondary
ow eect present in ow over interrupted heat transfer surfaces. Experimental data
indicate that the compact interrupted surfaces do not achieve the higher heat transfer
coecients predicted for simultaneously developing ows. The results for thermally
developing ows (and developed velocity proles), such as those in Fig. 7.24 or Eqs.
(7.96) and (7.97), are in better agreement with the experimental data for interrupted
surfaces, and hence those are recommended for design guidelines; the correct exponent
on x*, instead of 13, should be used in these equations if available based on experimental
data.
7.4.4.2 Turbulent Flow. The Nusselt numbers for simultaneously developing turbulent
ow are practically the same as the Nusselt numbers for thermally developing turbulent
ow. However, the Nusselt numbers for simultaneously developing ows are quite
sensitive to the passage inlet conguration (Bhatti and Shah, 1987).
Table 7.11 summarizes the dependence of p and h on um for developed and developing laminar and turbulent ows based on the solutions/correlations presented for the
specic ow type. Although these results are for the circular tube, the general functional
relationship should be valid for noncircular ducts as a rst approximation.
7.4.5
For convective transport of mass, momentum, and heat transfer across a boundary layer,
the associated coecients (mass transfer, skin friction, and heat transfer) in dimension-
509
TABLE 7.11 Dependence of Pressure Drop and Heat Transfer Coecient on the Flow Mean
Velocity for Internal Flow in a Constant Cross-Sectional Duct
h / uqm
p / upm
Laminar
Turbulent
Laminar
Turbulent
Fully developed
um
u1:8
m
u1:8
m
u1:8
m
u1:8
m
u0m
u0:8
m
Flow Type
Hydrodynamically developing
u1:5
m
Thermally developing
um
Simultaneously developing
u1:5
m
u1=3
m
u0:8
m
u1=2
m
u0:8
m
less form are related to each other for attached boundary layers over a heat/mass transfer
surface. This relationship in its simplest form was discovered by Reynolds in 1874 (see
Bejan, 1995). The analysis infers that the dierential momentum and energy equations
T
are similar for fully developed turbulent ow over a sharp-edge at plate; and for the *
boundary condition, the resulting dimensionless velocity and temperature proles are
identical for Pr 1 and the heat transfer coecient (i.e., St number) and friction factor
are related as
St
f
2
7:103
It has also been shown that the boundary layer equations for convective transport of
mass, momentum, and energy have the identical form when the pressure gradient dp=dx
along the ow direction is zero and Pr Sc 1 for laminar ow (Incropera and DeWitt,
2002). Here Sc, the Schmidt number, denotes a ratio of momentum diusivity to the mass
diusivity. Thus, the Reynolds analogy is now referred to as
f
Re Nu Sc
2
7:104
This is valid for laminar and turbulent boundary layer ows with zero axial pressure
gradient and the constant wall temperature boundary condition. By converting the Nu or
Sc in the corresponding Stanton numbers and extending the Prandtl number range based
on the experimental results, the modied Reynolds or ChiltonColburn analogies are
f
St Pr2=3 j
7:105
Stm Pr
7:106
2=3
jm
where Stm and jm are the Stanton number and Colburn factor for mass transfer.
However, for most heat exchanger surfaces, the analogy above is not quite correct.
This is because there is always some nite pressure gradient along the ow length (nite
pressure drop in the heat exchanger), and the boundary conditions are not necessarily
the constant wall temperature. Despite this fact, we know from the empirical evidence
that skin friction and convective heat transfer at the wall are related to each other.
Whenever there is convective heat transfer at the heat exchanger surface, there will be,
510
correspondingly, associated skin friction. We cannot get the convective heat transfer
without the expenditure of skin friction in the boundary layer ows. We refer to it as
an extended Reynolds analogy. It is formulated as (Shah and Bhatti, 1988)
j St Pr2=3
Nu
f
Pr1=3
w
Re
2
7:107
where
w
w duct geometry; flow type; boundary condition; Pr
7:108
Here,
w is the function that modies the Reynolds analogy relationship between j and f
factors; the ow type in
w indicates developing or developed laminar or turbulent ow.
Based on a variety of turbulent boundary layer models and experimental data, many
correlations have been developed for turbulent ow Nu, and some of them are summarized in Table 7.6. One of the most accurate correlations for fully developed turbulent ow
(and also for transition ow) in a circular tube is due to Gnielinski and is given in Eq.
(7.76) in Table 7.6. A comparison of the Gnielinski correlation with Eq. (7.107) will then
yield the most accurate value of
w , which is dependent on Pr and Re.
In general, based on the extended Reynolds analogy, skin friction and heat transfer at
the wall (heat transfer surface) for boundary layer ow are related to each other by a
parameter
w . It may or may not be possible to have an explicit expression/value of
w
for complex heat transfer surface geometries. However, this analogy clearly indicates
that an increase in heat transfer will be accompanied by an increase in the skin friction
factor! Then for an enhanced heat transfer surface, it is the heat exchanger designers
responsibility to maintain the same pressure drop by adjusting the geometry of the
exchanger [i.e., L, Ao , and Dh , see Eq. (6.29)] to take advantage of heat transfer enhancement (and associated increase in the f factor) without increasing the pressure drop and
uid pumping power for the exchanger. It should also be mentioned that the Reynolds
analogy in general would not be valid for nonboundary layer ows. For example, form
drag associated with the ow normal to a tube bank or form drag associated with the
ow in the wake region of an interrupted heat transfer surface does not enhance heat
transfer in general. Hence, for such geometries, j=f for gases and other uids is closer to 14
rather to 12, as indicated by Eq. (7.103), if the contribution of the form drag is of the same
order of magnitude as that of the skin friction. Due to the form drag associated with
surface roughness in turbulent ow, the friction factor takes on an approximate constant
value (see Fig. 6.4). However, there is no counterpart of the form drag for heat transfer,
hence the j factor continues to decrease with an increase in Re for a rough surface in the
fully rough region as discussed with the results of Fig. 7.20, a behavior similar to that for
a smooth surface!
7.4.6
The limitations of the j vs. Re plot, commonly used in presenting compact heat exchanger surface basic data, should be understood for various basic ow types as follows.
. In fully developed laminar ow, as discussed in Section 7.4.1.1, the Nusselt number
is theoretically constant, independent of Pr (and also Re). Since j St Pr2=3
Nu Pr1=3 /Re by denition, the j factor will be dependent on Pr in the fully
511
developed laminar region. Hence, the j factors presented in Chapter 7 of Kays and
London (1998) for gas ows in the fully developed laminar region should rst be
converted to a Nusselt number (using Pr 0:70), which can then be used directly
for liquid ows as constant-property results.
. Based on analytical correlations for thermally developing laminar ow in Section
7.4.3.1, Nu / x*1=3 . This means that Nu Pr1=3 is independent of Pr and hence
j is independent of Pr for thermally developing laminar ows.{ As a result, the j vs.
Re data of Kays and London (1998) could be used for air, water, and other uids as
constant-property data.
. For fully developed turbulent ow, Nu / Pr0:4 , and hence j / Pr0:07 . Thus, j is
again dependent on Pr in the fully developed turbulent region.{ In this case, if
the experimental data/correlations are available for specic geometries in terms
of the j vs. Re characteristics, convert those data into Nusselt numbers using the
denition j Nu Pr1=3 =Re with the properties of the test uids (i.e., using the
same Pr values as reported in the original work). Then use these Nusselt numbers
for the uids of interest for a given application. If the original experimental data/
correlations are available in terms of the Nusselt numbers, there is no need to
change anything.
All of the foregoing comments apply to either constant-property theoretical solutions
or almost-constant-property (low-temperature dierence) experimental data. The inuence of property variations, discussed in Section 7.6, must be taken into account by
correcting the aforementioned constant-property j or Nu when designing a heat
exchanger.
512
To determine the uid properties for the correlations of this section, refer to Section
9.1 for the appropriate mean temperature on each uid side.
7.5.1
Tube Bundles
Hglam 140Red
X* 0:62 0:75
1:6
Xt* 4Xt*X*= 1
7:110
Equation (7.110) is valid for all inline tube bundles. It is also valid for staggered tube
1:6
1:6
bundles except that the term Xt* needs to be changed to Xd* when the minimum freeow area occurs in the diagonal planes of the staggered tube bundle [i.e.,
X* < 0:52Xt* 11=2 .
("
Hgturb;i
#
)
0:61 0:94=X*0:6
0:47X=Xt1:5
*
*
10
0:11
0:015Xt 1X 1
Xt* 0:851:3
20:1X=Xt
Hgturb;s
Red
("
t;n Re2d
!3
!3 )
#
0:6
X*
Xt*
1
1
0:2
1:25
0:005
Xt*
X*
Xt* 0:851:08
7:111
Re1:75
t;n Re2d
d
7:112
Equation (7.112) is valid for Red 250,000. For higher Red , correct Hgturb;s of Eq.
(7.112) as follows:
Re 250,000
7:113
Hgturb;s;corr Hgturb;s 1 d
for Red > 250,000
325,000
8
1
1
1
>
>
for 5 Nr 10 and
>
>
2
>
2X * Nr 10
>
>
< t
X* 0:52Xt* 11=2
!
#
"
t;n
2
>
Xd* 1
1
1
>
>
>
for 5 Nr 10 and
2
>
>
*
*
N
10
X
X
1
>
r
t
t
:
X* < 0:52Xt* 11=2
(7.114)
513
and
t;n 0 for Nr > 10. It should be emphasized that
t;n Re2d takes into account the
inuence of tube bundle inlet and outlet pressure drops while the rst bracketed term on
the right side of Eqs. (7.111) and (7.112) takes into account the frictional pressure drop
in the tube bundle. For the total pressure loss of a tube bundle, the complete Eq. (7.111)
or (7.112) should be used (i.e., both terms on the right-hand sides of these equations
should be included).
The foregoing correlation of Eq. (7.109) is valid for 1 < Red < 300,000 and Nr 5
for both inline and staggered tube bundles: 1:25 Xt* 3:0 and 1:2 X* 3:0 for
inline tube bundles and 1:25 Xt* 3:0, 0:6 X* 3:0, and Xd* 1:25 for staggered
tube bundles. The experimental data for this correlation had 7:9 do 73 mm.
The pressure drop for ow normal to the tube bundle is then computed from
p
2 Nr
Hg
gc do2
7:115
where Nr is the number of tube rows in the ow direction. Note that only when the
minimum ow area occurs in the diagonals for a staggered tube bundle should the term
Nr be replaced by Nr 1, considering the number of ow resistances in the diagonal ow
area.
In all correlations for heat transfer and pressure drop for tube bundles,
Red um do =, where
8
Xt*
>
>
> u1 *
>
Xt 1
>
>
>
<
Xt*
um u1
*
>
X
t 1
>
>
>
>
>
Xt*
>
: u1
2Xd* 1
7:116
Martin (2002) developed comprehensive correlations for heat transfer with ow normal
to inline and staggered plain tube bundles as follows:
8
0:1
>
Red 1
< 0:404Lq1=3
Red 1000
Nu
>
:
1=3
0:404Lq
7:117
where
8
4Xt*= 1
>
>
>
1:18 Hg Pr
>
>
X*
>
>
>
<
*
4Xt = 1
Lq 0:92 Hg Pr
>
Xd*
>
>
>
>
>
4Xt*X*= 1
>
>
0:92
Hg
Pr
:
X*Xd*
514
where Hg is obtained from Eq. (7.109). Note that the Leveque number Lq is dened in
Eq. (7.41).
The foregoing heat transfer correlation of Eq. (7.117) is valid for
1 < Red < 2,000,000, 0:7 Pr 700 (validity expected also for Pr > 700, but not for
Pr < 0:6), and 2 Nr 15 for inline tube bundles (908 tube bundles) and 4 Nr 80
for staggered tube bundles (30, 45, 608 tube bundles and other staggered tube bundles
within the correlation tube pitch ranges); the tube pitch ratio range is 1:02 Xt* 3:0
and 0:6 X* 3:0 for both inline and staggered tube bundles. The experimental data
for this correlation had 7:9 do 73 mm. The Nusselt numbers are predicted accurately
within 20% for inline tube bundles and 14% for staggered tube bundles using Eq.
(7.117), where the Hagen number of Eq. (7.118) is determined from the Gaddis and
Gnielinski correlation of Eq. (7.109). The Nusselt number prediction may be better if
the experimental friction factors are used. Note that when the Gaddis and Gnielinski
correlation is extrapolated outside their ranges of Red and Nr for Nu calculations, it has
predicted Nu within the accuracy mentioned.
7.5.2
One of the most commonly used high-performance plate heat exchanger surfaces has
chevron plates with the important geometrical parameters identied in Fig. 7.28. A
considerable amount of research has been conducted to determine heat transfer and
ow friction characteristics of this geometry. Martin (1996) provides comprehensive
correlations for friction factors and Nusselt numbers for this geometry. His correlation
for the Fanning friction factors is
1
cos
1 cos
p
p
f 0:045 tan 0:09 sin f0 = cos 1=2
3:8f1
7:119
where
8
>
< 16
for Re < 2000
f0 Re
>
:
for Re 2000
1:56 ln Re 3:02
8
149:25
>
>
0:9625
for Re < 2000
<
Re
f1
>
>
: 9:75
for Re 2000
Re0:289
GDh
Re
q
p
4a
1
m_
G
1 1 X 2 4 1 X 2 =2
Dh
6
2aW
7:120a
7:120b
7:121
X
2a
7:122
As shown in Fig. 7.28, a is the amplitude and is the wavelength of chevron corrugations, and W is the plate width between gaskets. An exact formula for is given by
Eq. (8.131). The friction factor correlation of Eq. (7.119) is valid for the corrugation
angle within 0 to 808 and is accurate within 50% and 100%. If the model plate data
(those having only the central portion of the plates without inlet and outlet ports and
515
distribution regions) are eliminated, the correlation of Eq. (7.119) is based on industrial
plates of PHEs and is within 40% accuracy. Of course, this correlation can be improved
further if the actual detailed geometrical information would be available.
Martin (1996) also obtained the Nusselt number correlation as follows, using the
momentum and heat transfer analogy from a generalized Leveque solution in thermal
entrance turbulent ow in a circular pipe (Schlunder, 1998):
Nu
1=6
hDh
f Re2 sin 20:374
0:205Pr1=3 m
k
w
7:123
This correlation is valid for the corrugation angle within 10 to 808, and is accurate
within 30%, and within 13% for industrial plates. Note that if Eq. (7.123) is used
for gases, the viscosity correction term m =w 1=6 should be omitted. Note that a 100%
error in f will translate to 30% error in Nu, due to the exponent 0.374 on f in Eq. (7.123).
7.5.3
In this section we provide correlations for the oset strip n and louver n geometries,
and briey mention other geometries.
A careful examination of all accurate published data has revealed the ratio j=f 0:25
for strip n, louver n, and other similar interrupted surfaces. This can be justied
approximately as follows. The ow develops along each interruption in such a surface.
Based on the Reynolds analogy for fully developed turbulent ow over a at plate, in the
absence of form drag, j=f should be 0.5 for Pr 1 (see Section 7.4.5). Since the contribution of form drag is of the same order of magnitude as the skin friction in developing
laminar ows for such an interrupted surface, j=f will be about 0.25. Published data for
strip and louver ns may be questionable if j=f > 0:3, since they would indicate that the
contribution of form drag in interrupted ns is small; however, the form drag contribution will not be small in actual ns because of the nite n thickness and possible burrs at
the edges due to the manufacturing process. All pressure and temperature measurements
and possible sources of ow leaks and heat losses must be checked thoroughly for all
basic surface characteristics having j=f > 0:3 for strip and louver ns (refer to some
problems identied in Section 7.3.1.3 for the test core design).
516
7.5.3.1 Oset Strip Fins. This is one of the most widely used enhanced n geometries
(Fig. 1.29d) in aircraft, cryogenics, and many other industries that do not require mass
production. This surface has one of the highest heat transfer performances relative to
the friction factor. Extensive analytical, numerical and experimental investigations have
been conducted over the last 50 years. The most comprehensive correlations for j and f
factors for the oset strip n geometry are provided by Manglik and Bergles (1995) as
follows.
0:1541 0:1499 0:0678
s
h0
s
s
0:504 0:456 1:055 0:1
s
1 5:269
105 Re1:340 0
h
s
s
j 0:6522 Re0:5403
1 7:669
108 Re4:429 0
h
s
s
7:124
f 9:6243 Re0:7422
7:125
where
Dh
4Ao;cell
4sh 0 s
Acell =s 2ss h 0 s h 0 s
7:126
517
FIGURE 7.29 Denition of geometrical parameters of corrugated louver ns. (From Chang and
Wang, 1997.)
corrugated louver ns (see Fig. 7.29a or 1.29e), based on an extensive database for
airow over louver ns, is obtained by Chang and Wang (1997) and Wang (2000) as
follows:
j Re0:49
p
pt
7:127
p
p
p
p
90
p
p
where Rep Gp = represents the Reynolds number based on the louver pitch p . Also,
is the louver angle (deg), pf the n pitch (mm), b the vertical n height (mm), Wt the
tube outside width ( total n length in the airow direction if there are no overhangs;
mm), the louver cut length (mm), pt the tube pitch (mm), and the n thickness (mm).
These geometrical parameters are shown in Fig. 7.29. Equation (7.127) is valid for
the following ranges of the parameters: 0:82 Dh 5:02 mm, 0:51 pf 3:33 mm,
0:5 p 3, 2:84 b 20 mm, 15:6 Wt 57:4 mm, 2:13 18:5 mm, 7:51
pt 25 mm, 0:0254 0:16 mm, 1 Nr 2, and 8:4 358. This correlation
predicts 89% of experimental j factors of 91 test cores within 15% for
30 < Rep < 5000 with a mean deviation of 8%.
Chang and Wang (1997) also presented a simplied correlation for Eq. (7.127) as
j 0:425 Re0:496
p
7:128
518
They report that this correlation predicts 88% of data points within 25% with the mean
deviation of 13%.
The correlation for the Fanning friction factor based on the same database by Chang
et al. (2000) is
f f1 f2 f3
7:129
where
f1
f2
8
0:805pf =b
< 14:39 Rep
fln1:0 pf =p g3:04
:
0:2
0:60491:064=
4:97 Rep
8
< fln=pf 0:48 0:9g1:435 Dh =p 3:01 ln0:5 Rep 3:01
8
< pf = 0:308 Lf = 0:308 e0:1167pt =Ht 0:35
f3
: p =H 0:0446 fln1:2 =p 1:4
3:553 0:477
t
t
p
f
7:131
Rep < 150
150 < Rep < 5000
7:132
Additional parameters for the friction factor correlations are: Dh is the hydraulic diameter of the n geometry (mm), Ht the tube outside height (mm), and Lf the n length in
the airow direction (mm). Note that in Eqs. (7.130) and (7.132) is in degrees. The
conventional denition of the hydraulic diameter [the rst denition in Eq. (3.65)] is used
in Eq. (7.131) assuming that the louver ns were plain ns (without cuts) for the calculation of Ao and A; the eect of the braze llets has also been neglected since no such
information is available in the open literature. The correlation of Eq. (7.129) predicts
83% of the experimental friction factor data points within 15%, with a mean deviation
of 9% for the parameter ranges the same as those for Eq. (7.127).
Note that the authors included the tube width Wt in the heat transfer correlation (to
take into account the correct surface area). However, they used the n length Lf { for the
friction factor correlation since the friction factor f of Eq. (7.129) is for the n friction
component only (excluding entrance and exit pressure losses from the pressure drops
measured).
7.5.3.3 Other Plate-Fin Surfaces. Perforated and pin n geometries have been investigated and it is found that they do not have superior performance then that for oset
strip and louver n geometries (Shah, 1985). Perforated ns are now used only in a
limited number of applications. They are used as turbulators in oil coolers and in
cryogenic air separation exchangers as a replacement for the existing perforated n
exchangers; modern cryogenic air separation exchangers use oset strip n geometries.
Wt < Lf for 15 cores of Achaichia and Cowell (1988); for all other cores in the correlations, Wt Lf :
519
Considerable research work has been reported on vortex generators using winglets
(Jacobi and Shah, 1999), and research continues for heat exchanger applications.
7.5.4
Two major types of tube-n extended surfaces are (1) individually nned tubes and (2)
at ns (sometimes referred to as plate ns) with or without enhancements/interruptions
on an array of tubes as shown in Fig. 1.31. An extensive coverage of the published
literature and correlations for these extended surfaces is provided by Webb (1994),
Bemisderfer (1998), Kays and London (1998), and Wang (2000). Empirical correlations
for some important geometries are summarized below.
7.5.4.1 Individually Finned Tubes. This n geometry, helically wrapped (or extruded)
circular ns on a circular tube as shown in Fig. 1.31a or 8.5, is commonly used in
process and waste heat recovery industries. The following correlation for j factors is
recommended by Briggs and Young (1963) for individually nned tubes in a staggered
tube bank:
j 0:134Re0:319
d
0:2 0:11
s
s
7:133
where Red um do = (where um occurs at the minimum free-ow area dened in Section
8.2.1.2), de do =2 is the radial height of the n, is the n thickness, s pf is
the distance between adjacent ns, and pf is the n pitch; see the circular n gure in
Table 4.5 for the denitions of f , and pf . Equation (7.133) is valid for the
following ranges: 1100 Red 18,000, 0:13 s= 0:63, 1:01 s= 7:62,
0:09 =do 0:69, 0:011 =do 0:15, 1:54 Xt =do 8:23, n root diameter do
between 11.1 and 40.9 mm, and n density Nf 1=pf between 246 and 768 ns/m.
All data have been obtained on equilateral triangular pitch tube bundles (308 in Table
8.1). The standard deviation of Eq. (7.133) with experimental results is 5.1%.
For friction factors, Robinson and Briggs (1966) recommended the following correlation:
ftb
9:465Re0:316
d
0:927 0:515
Xt
Xt
do
Xd
7:134
1=2
Here Xd Xt2 X2
is the diagonal pitch, and Xt and X are the transverse and
longitudinal tube pitches, respectively. The correlation is valid for the following ranges:
2,000 Red 50,000, 0:15 s= 0:19, 3:75 s= 6:03, 0:35 =do 0:56,
0:011 =do 0:025, 1:86 Xt =do 4:60, 18:6 do 40:9 mm, and 311 Nf 431
ns/m. The standard deviation of Eq. (7.134) with correlated data is 7.8%.
For crossow over low-height nned tubes, Ganguli and Yilmaz (1987) correlated all
data of Rabas and Taborek (1987) as well as some additional data published in the
literature for air. They correlated the heat transfer result data as follows:
j 0:255Re0:3
d
0:3
de
s
7:135
520
where de is the n tip diameter of the radial (low height) ns. This correlation is valid
for the following ranges: 6:35 mm, 800 Red 800,000, 208 408, Nr 4,
0:6 Pr 0:7, and 5 de =s 60, where is the tube layout angle for a tube bundle
(see Table 8.1). All the heat transfer data correlated within 20%, and 95% of the data
correlated within 15%. Note that the low-nned tube j factors for a given Red are lower
than those for a plain tube and the j factors approach the plain tube value asymptotically
at high Red .
For low-nned tubes, Ganguli and Yilmaz (1987) correlated the plain tube bank
friction factor data as follows:
7:136
f Fs ftb; p
where Fs is the surface factor (a ratio of the friction factor for the nned tube to that for
the plain tube) and ftb; p is the friction factor for a plain tube bank. They are given by
3
A
Fs 2:5 tan1 0:5
5
Ap
7:137
2
d do2 de Nf
4 e
7:138y
where
A do 1 Nf 2Nf
Ap do
7:139
3
3
X*
X*
Kp 2:5 1:2Xt* 0:851:06 0:4 1 0:01 t 1
Xt*
X*
7:140
This friction factor correlation is valid for the same parameter ranges mentioned for the
heat transfer correlation. All of the friction data correlated within 20%, and 95% of
the data correlated within 15%.
7.5.4.2 Plain Flat Fins on a Tube Array. This geometry, plain at ns (Fig. 1.31b) on a
staggered tube bank, is used in the air-conditioning/refrigeration industry as well as
where the pressure drop on the n side prohibits the use of enhanced/interrupted at
ns. An inline tube bank is generally not used unless very low n side pressure drop is
the essential requirement. A heat transfer correlation for plain at ns on staggered
tube banks (Fig. 1.31b) is provided by Wang and Chi (2000) and summarized by Wang
(2000) as follows:
8
c1 1:084 0:786 c2
>
pf
pf
pf
>
0:29 Xt
>
0:108Re
>
dc
<
X
dc
Dh
Xt
j
c5 c6 0:93
>
>
p
pf
pf
>
> 0:086Recdc3 Nrc4 f
:
dc
Dh
Xt
for N r 1
7:141
for Nr 2
{
Here A is the total outside surface area of the nned tube and Ap is the plain tube outside surface area, and both A
and Ap are per unit tube length.
521
where
c1 1:9 0:23 ln Redc
0:058Nr
ln Redc
c6 5:735 1:21 ln
Redc
Nr
7:142c
where pf is the n pitch, dc is the collar diameter of the n and Redc um dc =. This j
factor correlation predicts 89% of the test points of 74 cores within 15%, with a mean
deviation of 8%. Wang and Chi (2000) also provided the following correlation for the
friction factors:
f 0:0267Recdc7
Xt
X
c8 c9
pf
dc
7:143
where
pf
Xt
0:00758
0:177
c7 0:764 0:739
X
dc
Nr
c8 15:689
64:021
ln Redc
c9 1:696
15:695
ln Redc
7:144a
7:144b
Equations (7.141) and (7.143) are valid for the following ranges of the parameters:
300 Redc 20,000,
6:9 dc 13:6 mm,
1:30 Dh 9:37 mm,
20:4 Xt
31:8 mm, 12:7 X 32 mm, 1:0 pf 8:7 mm, and 1 Nr 6. This friction factor
correlation of Eq. (7.143) predicts 85% of experimental friction factors of 74 test cores
within 15%, with a mean deviation of 8%.
7.5.4.3 Corrugated Flat Fins on a Tube Array. There are a number of variations
available for at ns with a sharp vs. smooth wave. The specic at n geometry
shown in Fig. 7.30 is designated as a corrugated (herringbone or sharp-wave) n.
The heat transfer and ow friction correlations are developed by Wang (2000) and
presented separately for large- and small-diameter tubes as follows. For larger tube
diameters (do 12:7 and 15.88 mm, before tube expansion), the following are the correlations:
j 1:7910Recdc1
0:456
1:343 0:317
pf
X
pd
Nr0:27
dc
xf
f 0:05273Recdc2
c3 c4
pf
pd
A 2:726 Dh 0:1325 0:02305
ln
Nr
xf
Xt
dc
Ap;t
7:145
7:146
522
where
0:493 0:886
0:0296
pf
X
p
c1 0:1707 1:374
Nr0:143 d
dc
xf
7:147a
0:25
0:2
pf
A
pd
c2 0:1714 0:07372
ln
X
xf
Ap;t
7:147b
0:3
pf
A
c3 0:426
ln
Xt
Ap;t
7:147c
c4
10:2192
ln Redc
Here xf is the projected n pattern length for one-half wavelength, pd the n pattern
depth (peak-to-valley distance, excluding n thickness, as shown in Fig. 7.30), and
Ap;t the tube outside surface area when there are no ns. Equations (7.145) and
(7.146) are valid for the following ranges of the parameters: 500 Redc 10,000,
3:63 Dh 7:23 mm, 13:6 dc 16:85 mm, 31:75 Xt 38:1, 27:5 X 33 mm,
2:98 pf 6:43 mm, 1 Nr 6, 12:3 14:78, 6:87 xf 8:25 mm, and
pd 1:8 mm. The correlation of Eq. (7.145) predicts 93% of experimental Colburn
factors for 18 test cores within 10%, with a mean deviation of 4%. Similarly, the
correlation of Eq. (7.146) predicts 92% of experimental friction factors for 18 test
cores within 10%, with a mean deviation of 5%.
For smaller-diameter tubes (do 7:94 and 9.53 mm before tube expansion), following
are the correlations for the j and f factors:
j 0:324Recdc1
pf
X
c2
tan c3
pf
X
c4
X
Nr0:428
Xt
c7
A 5:35 Dh 1:3796 0:0916
ln
Nr
dc
Ap;t
7:148
7:149
523
where
0:6 0:54
pf
X
c1 0:229 0:115
Nr0:284 ln0:5 tan
Dc
Dh
c2 0:251
0:09 1:75
pf
X
c3 0:439
Nr0:93
Dh
Xt
0:232Nr1:37
lnRedc 2:303
c4 0:502lnRedc 2:54
1:4
pf
A
ln
c6 3:247
Xt
Ap;t
7:150a
c5 0:4604 0:01336
c7
pf
X
20:113
ln Redc
0:58
ln
7:150b
A
tan 1:5
Ap;t
7:150c
7:150d
Equations (7.148) and (7.149) are valid for the following ranges of the parameters:
300 Redc 8000, 1:53 Dh 4:52 mm, 8:58 dc 10:38 mm, Xt 25:4 mm,
19:05 X 25:04 mm, 1:21 pf 3:66 mm, 1 Nr 6, 14:5 18:58, 4:76
xf 6:35 mm, and 1:18 pd 1:68 mm. The correlation of Eq. (7.148) predicts 95%
of experimental Colburn factors of 27 test cores within 15%, with a mean deviation of
6%. Similarly, the correlation of Eq. (7.149) predicts 97% of experimental friction
factors for data points of 27 test cores within 15%, with a mean deviation of 5%.
7.5.5
Regenerator Surfaces
The two most common types of regenerator surfaces are (1) continuous cylindrical
passages for rotary regenerators and some compact xed-matrix regenerators, and
(2) randomly packed woven screens, crossed rods, and packed beds using a variety of
materials.
For compact regenerators, the continuous cylindrical ow passages have simple geometries, such as triangular, rectangular, and hexagonal passages. The Nu and f factors
of Tables 7.3 and 7.4 are a valuable baseline for such passages. Due to the entrance length
eect, actual Nusselt numbers should be higher than that for fully developed ow.
However, the actual ow passages are never ideal and uniform, due to manufacturing
processes and tolerances. The passage-to-passage nonuniformity, discussed in Section
12.1.2, reduces heat transfer more than the gain by the thermal entrance eect. The
friction factors are generally higher than those for fully developed ow because of the
signicant eect of the hydrodynamic entrance length. The passage-to-passage nonuniformity reduces the friction factor and p only slightly. Thus, generally j or Nu is
lower and f is higher than those for fully developed ow. Also, the thermal boundary
condition for heat transfer may not exactly correspond to any of the boundary conditions described previously. Hence, accurate j and f versus Re characteristics are generally
determined experimentally even for simple geometries in addition to those for the
complex geometries. As an illustration, London et al. (1970) presented the following
correlations for airow through triangular passages (40 < Re < 800):
f
14:0
Re
3:0
Re
7:151
524
FIGURE 7.31 Cross rod geometries. (From Kays and London, 1998.)
London and Shah (1973) presented the following correlations for airow through
hexagonal passages (80 < Re < 800):
f
17:0
Re
4:0
Re
7:152
Crossed rod geometries of Fig. 7.31 have alternate layers of parallel solid rods touching each other and stacked 908 to each other. Martin (2002) has correlated the j factors as
follows for these geometries based on the generalized Leveque solution in thermal
entrance turbulent ow in a circular pipe (Schlunder, 1998):
Nu 0:535 p1 p0:25 f Re2 Pr1=3
7:153
where the friction factors of Kays and London (1998) data are correlated by Das (2001)
as
8
0:136
0:104
Xt*
>
< 0:603Re
f 0:728Re0:188 Xt*0:913
>
:
0:475Re0:108 Xt*0:458
and dw is the wire diameter, Xt* Xt =dw , and Xt is the transverse pitch of the rods. Note
that the hydraulic diameter Dh , the ratio of free-ow area to frontal area
, the porosity
(void fraction) p, and the ratio of heat transfer surface area to volume of the crossed
rod matrices (of square mesh using circular cylinder rods) are related to dw and Xt as
follows:
Dh
4Xt
pdw
dw
1p
Xt dw 2
Xt2
p1
dw
4Xt
Xt
7:155
The correlations of Eqs. (7.153) and (7.154) are valid for dw 9:53 mm, and
1:571 Xt =dw 4:675 or 0:500 p 0:832. The correlation for Nu (the j factor) is
accurate within 4.8% and that for the f factors is accurate within 7.8, 6.7, and
7.1% for inline, staggered, and random arrangements of Kays and London data (1998).
Example 7.6 The purpose of this example is to determine the change in performance of
a single-pass unmixedunmixed crossow exchanger (Fig. E7.6a) when it is made a twopass conguration either in series coupling (over and under) Fig. E7.6b), or in parallel
525
FIGURE E7.6 (a) Single-pass unmixedunmixed crossow exchanger; (b) series coupled (overand-under) two-pass cross-counterow exchanger; (c) parallel coupled (side-by-side) two-pass
exchanger.
coupling (side by side, Fig. E7.6c), with water mixed between passes for two cases: 1:1
and 10:1 thermal resistances on air and water sides. Note that the total heat transfer
surface area is the same in all three exchangers. The performance specications for the
single-pass exchanger of Fig. E7.6a are: number of heat transfer units, 1.2; ratio of heat
capacity rates, 0.8; eectiveness, 0.547; and pressure drop on the air side, 0.25 kPa. All
areas, geometrical properties, ow rates, heat transfer coecients, friction factors, and
uid properties are known (i.e., the exchangers have been designed).
Consider ja / Re0:4
and fa / Re0:2
on the air side and assume uid properties to be
a
a
constant. Determine the exchanger eectivenesses and the pressure drops for the exchangers for Fig. E7.6b and c. Evaluate " and p for the following two cases of thermal
resistances for the single-pass exchanger of Fig. E7.6a: (a) Ra Rw and (b) Ra 10Rw .
Neglect the wall resistances and fouling resistances in all three exchangers.
SOLUTION
Problem Data and Schematic: An exchanger with known thermal capacities is given as in
Fig. E7.6a. Also given are two exchangers that vary in overall conguration from the rst
exchanger as shown in Figs. E7.6b and c. For the exchanger of Fig. E7.6a,
NTU 1:2
C*
Cair
0:8
Cwater
" 0:547
Determine: The exchanger eectivenesses and pressure drops for the exchangers of Fig.
E7.6b and c.
Assumptions: Constant uid properties apply; pressure drops due to the entrance, exit,
and momentum eects on the air side are neglected, as is the eect of ow turning from
pass 1 to pass 2 on heat transfer and pressure drop.
Analysis: Let us use subscripts 1, 2, and 3, respectively, for all pertinent quantities of
exchangers of Fig. E7.6a, b, and c. Careful observations of the geometry and ow
arrangement will yield the following results:
526
Variable
Exchanger 1
Exchanger 2
Exchanger 3
Airow length
Air-side mass velocity
Air-side surface area per pass
Water-side surface area per pass
Air-side heat capacity rate per pass
Heat capacity rate ratio air/water for each
pass
Air-side heat transfer coecient
Water-side heat transfer coecient
U based on air-side surface area per pass
L1
G1
A1
Aw;1
C1
C*
L2 2L1
G2 2G1
A2 0:5A1
Aw;2 0:5Aw;1
C 2 C1
Cp* C *
L3 L1
G 3 G1
A3 0:5A1
Aw;3 0:5Aw;1
C3 0:5C1
Cp* 0:5C*
h1
hw
U1
h3 h1
hw
U3 U1
hAw 1
hA1
h2
hw
U2 will depend
on Ra and Rw
0:5hAw 1
h2 0:5A1
0:5hAw 1
0:5hA1
We rst calculate the change in the heat transfer coecient on air side for exchanger 2.
By denition, h ja Gcp Pr2=3 , and ja Re0:4
is specied where Rea GDh =. Hence,
a
G / G0:4 G G0:6
h ha / ja G / Re0:4
a
Therefore,
h2
h1
G2
G1
0:6
20:6 1:516
Now, for the single-pass exchanger (exchanger 1), we designate o hAa hA1 . Then
we get the following:
R a Rw
Ra 10Rw
Case
Case
1
1
1
10
1
1
1
2
UA 1 hA 1 hA 1 hA 1
1
1
1
1:1
UA 1 hA 1 10hA 1 hA 1
Now let us calculate the exchanger eectiveness for exchanger 2, assuming a two-pass
exchanger, for the two cases of thermal resistance distribution. The appropriate relationships derived in Chapter 3 are used here.
Ra Rw
Case
UA 2; p 0:758hA 1 0:5hA 1 hA 1
Ra 10Rw
Case
UA 2; p 0:758hA 1 5hA 1 hA 1
527
UA2; p 0:602UA1
UA2; p 0:724UA1
"2
1 0:8
0:422=1 0:422 2 1
1 0:8
0:422=1 0:422 2 0:8
0:610
"2
1 0:8
0:468=1 0:468 2 1
1 0:8
0:468=1 0:468 2 0:8
0:657
"2 0:610
1:12 or 12%
"1 0:547
"2 0:657
1:20 or 20%
"1 0:547
Thus, the increase in heat transfer rate for a two-pass exchanger is 12% and 20%,
respectively, over similar single-pass balanced (the Ra Rw case) and highly unbalanced
(the Ra 10Rw case) exchangers.
Now we calculate the eectiveness of exchanger 3, which is a parallel coupled (sideby-side) two-pass exchanger with the uid (water) mixed between passes. The unfolding of this exchanger will not result in exchanger 1 because the water side is not unmixed
throughout. As noted earlier, for each pass of exchanger 3, Cair is one-half that of the
exchanger 1. Hence,
Cp*
Ca;3 0:5Ca;1
0:5C*1 0:4
Cw
Cw
U1 0:5A1
NTU1 1:2
0:5C1
Note that the overall U will be the same for exchangers 1 and 3 as noted earlier. Hence, "p
from Fig 3.9 (or from the "-NTU formula for an unmixedunmixed crossow exchanger
in Table 3.3) is 0.6175 for NTUp 1:2 and Cp* 0:4. In this case, the parallel airstream is
the uid stream 2 of Fig. 3.22. Hence, the exchanger eectiveness "3 for the airside is
calculated from modied Eq. (3.162) in terms of " "p Cp* formula as follows with
"p 0:6175, Cp* 0:4, and n 2.
"3
n
1
1
1 1 0:6175
0:4 2 0:5412
1 1 "p Cp*
*
2
0:4
nCp
Since the dierence between exchanger 1 and exchanger 3 on the water side is only
unmixed (exchanger 1) vs. mixed (exchanger 3) between passes (here we considered two-
528
pass on the water side for exchanger 3), we expect the eectiveness of exchanger 3 to be
slightly lower,{ and this is what we found "3 0:541 vs. "1 0:547.
Now let us rst calculate the air-side pressure drop for exchanger 2. Since fa / Re0:2
a
and Rea GDh =,
0:2
Re2 0:2
G2
Re1
G1
f2
f1
7:156
because Dh on the air side is the same for both exchangers, and we assume that the uid
properties do not vary with relatively small changes in the air temperature. If we consider
the core frictional pressure drop as the main component of the total pressure drop, then,
from Eq. (6.29),
p
4fLG2 1
2gc Dh m
p1
G2
G1
0:2
L2
L1
G2
G1
2
1:8
G
2 2
22 1:8 6:96
G1
or
p2 6:96
0:25 1:74 kPa
Ans:
Thus, the pressure drop on the air side for the two-pass exchanger 2 is about seven times
that for a similar single-pass exchanger 1. Thus, the penalty in the air-side pressure drop
is substantial compared to the gain in heat transfer (12 to 20%) for a two-pass crosscounterow exchanger.
For exchanger 3, the pressure drop on the air side will not change and the pressure
drop on the water side will be slightly higher, due to the return tank (a 1808 bend) after
the rst pass. The only advantage of exchanger 3 could be dierent packaging.
Discussion and Comments: The motivation of this problem is to demonstrate the calculation procedure and dierences in heat transfer and pressure drop performance of a heat
exchanger when it is packaged in dierent ways, ideally having the same surface areas
and the same individual heat transfer surface geometries. Since exchanger 3 is theoretically almost identical to exchanger 1, there is no signicant dierence in either heat
transfer or pressure drop. However, exchanger 2 is two-pass exchanger, and as a result,
its eectiveness improves over that for the single-pass exchanger as expected. However,
{
Refer to Chapter 11 for detailed discussion and thermodynamics arguments for the reasons of this performance
reduction.
529
there is a signicant penalty in the pressure drop. Hence the design decision on which
arrangement to select (exchanger 1 or 2) will depend on the allowable pressure drop on
the air side.
Note that if we had made exchanger 2 a two-pass cross-counterow exchanger with
the water side as two-pass and airow as a straight throughow, the resulting increase in
the water-side pressure drop (about 7- to 8-fold) due to the increased velocity (twice),
double the ow length, and a 1808 turn may be tolerable due to low uid pumping power
(refer to the end of Section 6.1.1). Such a ow arrangement is used in liquid-to-gas
exchangers when the liquid-side heat transfer coecient is very low (or hA on the liquid
side is equal or lower than the gas side); a substantial gain in heat transfer performance
may be obtained in that case, due to the increased velocity and hence h and U.
7.6
One of the basic assumptions made in the theoretical correlations for Nu and f of Section
7.4 is that the uid properties remain constant throughout the ow eld. Most of the
original experimental j and f data presented in Section 7.5 involve small temperature
dierences so that the uid properties generally do not vary signicantly. In certain heat
exchanger applications, uid temperatures vary signicantly. At least two questions
arise: (1) Can we use the j and f data obtained for air at 50 to 1008C (100 to 2008F)
for air at 500 to 6008C (900 to 11008F)? (2) Can we use the j and f data obtained with air
(such as all data in Kays and London, 1998) for water, oil, and viscous liquids? The
answer is yes, by modifying the constant-property j and f data for the variations in the
uid properties within a heat exchanger, and we discuss this method in this section.
Let us rst illustrate how the variation in uid properties aect the pressure drop and
heat transfer in a heat exchanger. Consider a specic case of inuence of the variable
liquid viscosity on pressure drop and heat transfer for fully developed laminar ow
through a circular tube. In absence of heat transfer (isothermal case), the velocity prole
will be parabolic (Fig. 7.32). Now consider heat being transferred from the tube wall to
the liquid. This will set up a temperature gradient in the radial direction in the liquid at a
tube cross section. Liquid will be hotter near the wall and cooler near the tube centerline.
Since the liquid viscosity decreases with increasing temperature, the uid near the wall
will have lower viscosity than that of the isothermal case. This will result in increased
velocity near the tube wall and a decreased velocity near the tube center to satisfy the
continuity equation for the steady-ow case. The resulting velocity prole is atter, as
shown in Fig. 7.32. The decrease in liquid viscosity near the wall yields lower f and lower
p. The increased velocity near the wall may mean a more ecient convection heat
530
7.6.1
531
TABLE 7.12 Property Ratio Method Exponents of Eqs. (7.157) and (7.158) for Laminar Flow
Fluid
Heating
Cooling
Gas
Liquid
Nucp
Tw
Tm
n
f cp
Tw
Tm
m
7:157
Nucp
w
m
n
f cp
w
m
m
7:158
Here the subscript cp in Eqs. (7.157) and (7.158) refers to constant properties (i.e., Nu
and f for constant uid properties), and all properties in the dimensionless groups of
Eqs. (7.157) and (7.158) are evaluated at the bulk temperature. The values of the exponents n and m for fully developed laminar and turbulent ows in a circular tube are
summarized in Tables 7.12 and 7.13 for heating and cooling situations.
TABLE 7.13 Property Ratio Method Correlations or Exponents of Eqs. (7.157) and (7.158) for
Turbulent Flow
Fluid
Gas
Liquid
Heating
0:83
Cooling
n
Pr 0:29Tw =Tm
Nu 5 0:012Re
n log10 Tw =Tm 1=4 0:3
for 1 < Tw =Tm < 5, 0:6 < Pr < 0:9, 104 < Re < 106 ,
and L=Dh > 40
m 0:1 for 1 < Tw =Tm < 2:4
a
n0
m 0:1 (tentative)
n 0:25a for 1 < w =m < 40
m 0:24b for 1 < w =m < 2
532
These correlations [Eqs. (7.157) and (7.158)] with exponents from Tables 7.12 and
7.13, are derived for the constant-heat ux boundary condition. The variable-property
eects are generally not important for fully developed ow having a constant wall
temperature boundary condition, since Tm approaches Tw for fully developed ow.
Therefore, to take into account the minor inuence of property variations for the constant wall temperature boundary condition, the correlations of Eqs. (7.157) and (7.158)
are adequate. The values of the exponents n and m may also depend on the duct crosssectional shape, surface interruptions, developing ows and thermal boundary conditions dierent from the constant heat ux or constant wall temperature. However, no
such information is available in the literature. Hence, Eqs. (7.157) and (7.158) may not be
able to fully correct the variable-property eects for many diverse situations in heat
exchangers, but they still represent the best primary corrections in open literature.
7.7
When the velocities of secondary ows due to free convection eects are not negligible in
comparison to the mean velocity of the forced convection ow, the free convection may
have a signicant inuence on heat transfer and pressure drop calculated by a pure forced
convection correlation. The inuence of superimposed free convection over pure forced
convection ow is important when the uid velocity is low, a high temperature dierence
Tw Tm is employed, or the passage geometry has a large hydraulic diameter Dh . The
eect of the superimposed free convection is generally important in the laminar ow of
many shell-and-tube heat exchangers; it is quite negligible for compact heat exchangers.
The eect of free convection is correlated by combinations of the following dimensionless numbers: Grashof number Gr, Rayleigh number Ra, Prandtl number Pr, and
L/D along with the Reynolds number Re. The Grashof and Rayleigh numbers are
dened as
Gr
Ra Gr Pr
7:159
7:160
533
For horizontal tubes, free convection sets up secondary ows at a cross section that aids
the convection process. Hence, the heat transfer coecient and Nusselt number for the
combined convection are higher than those for the pure forced convection for ow in a
horizontal tube. The maximum heat transfer occurs at the bottom of the tube. When the
free convection eect is signicant in laminar ow, large temperature gradients exist near
the wall, and the temperature variations in the horizontal and vertical directions at a
given ow cross section are also markedly dierent from the parabolic velocity distribution for Poiseuille ow.
Metais and Eckert (1964) recommend a free, mixed, and forced convection regime map,
as shown in Fig. 7.33, for a horizontal circular tube with axially constant wall temperature
boundary condition. The limits of the forced and free convection regimes are dened in
such a manner that free convection eects contribute only about 10% to the heat ux.
Figure 7.33 may therefore be used as a guide to determine whether or not free convection is
important. Recently, Ghajar and Tam (1995) have presented a ow regime map for a
horizontal circular tube with uniform wall heat ux boundary condition and three inlet
congurations to the tube. They have developed this chart for a limited range of the
Rayleigh number Ra and provided correlations for the Nusselt number.
A number of correlations for developed and developing mixed convection ows are
presented by Aung (1987) for a horizontal circular tube with various boundary conditions. As an illustration, Morcos and Bergles (1975) presented the following combined
convection correlation in a horizontal circular tube:
NuH4 f4:362 0:145Gr* Pr1:35 Kp0:25 0:265 2 g1=2
7:161
FIGURE 7.33 Horizontal circular tube free, forced, and mixed convection ow regimes for
102 < PrDh =L < 1. (From Metais and Eckert, 1964.)
534
where
Gr* Gr Nu
g*D4h q 00
2 k
Kp
kw w
kDh
7:162
Here Gr* is the modied Grashof number and Kp is the peripheral heat conduction
parameter. The subscript H4 denotes the axially constant wall heat ux boundary condition with nite heat conduction in the peripheral direction (Shah and London, 1978).
This correlation is valid for 3
104 < Ra < 106 , 4 < Pr < 175, and 0:015 < Kp < 0:5,
where Ra is dened in Eq. (7.160) with Tw replaced by the mean wall temperature Tw;m .
All the uid properties in NuH4 , Gr*, Pr, and Kp should be evaluated at the lm temperature [i.e., at Tw;m Tm =2].
The thermal entrance length is reduced signicantly in the presence of the free convection eects. Thus, for most cases when free convection eects are superimposed, the
combined convection ow is fully developed. Hence, Eq. (7.161) is sucient to determine
the Nusselt numbers for combined convection.
The friction factors are also higher for the combined convection case. Based on the
ethylene glycol data, Morcos and Bergles (1975) presented the following correlation:
f
1 0:195Ra0:15 15 1=15
ffc
7:163
where ffc 16=Re is the isothermal forced convection friction factor. Both f and Ra are
based on the uid properties evaluated at the lm temperature.
It should be emphasized that the free convection eects are unimportant for compact
heat exchanger surfaces due to the small Dh , illustrated in the following example.
Example 7.7 Investigate the inuence of superimposed free convection on the air side
of the automobile radiator of Problem 3.8. Use the following additional information:
Tlm 49:98C, Dh 0:003 m for air-side surface,
0:20
104 Pa s,
3
1:09 kg=m , Pr 0:7, ef =Dh 0:5, and the operating Reynolds number Re 1500.
SOLUTION
Problem Data and Schematic: Data are provided for the radiator shown in Fig. E7.7.
Determine: The inuence of superimposed free convection.
Airflow out
Coolant in
Coolant out
Airflow in
Radiator
Airflow
FIGURE E7.7
Tm = 49.9
Dh = 0.003 m
ef /Dh = 0.5
Re =1500
= 0.20 104Pa s
= 1.09 kg/m3
Pr = 0.7
535
Assumptions: The ow pattern map of Fig. 7.33 is applicable for this problem.
Analysis: Using Fig. 7.33, we determine whether or not the free convection eects are
important for this problem. To determine the Grashof number, let us determine
Tw Tm and *. For this problem,
Tw Tm Tlm 49:98C 49:9 K
The mean air temperature is the arithmetic mean water temperature minus the log-mean
temperature dierence.
Tm;c Tm;h Tlm
98:9 93:38C
49:98C 46:28C
2
1
1
1
0:0031
Tm;c 273:15 46:2 K
K
Hence
Gr Pr Dh
0:7
121:7
170:4
ef
0:5
Note that we consider the interrupted length (louver length in the ow direction) ef
for ef =Dh in the correlation of Fig. 7.33. We nd from Fig. 7.33 that for
Gr Pr Dh =ef 170:4 and Re 1500, the ow regime is forced convection laminar
ow, and hence the inuence of superimposed free convection is negligible.
Discussion and Comments: For the radiator, the mean temperature dierence is relatively
small (49.98C for this case), the hydraulic diameter is small (Dh 0:003 m), and the
velocity is relatively high compared to the free convection velocities of 1 m/s or less.
Hence, we would not expect the free convection eects to be important. And this is
what is found from this example.
7.7.2
Unlike horizontal tubes, the eect of superimposed free convection for vertical tubes is
dependent on the ow direction and whether or not the uid is being heated or cooled.
For uid heating with upward ow or uid cooling with downward ow, free convection
aids forced convection since the velocities due to free and forced convection are in the
536
FIGURE 7.34 Vertical circular tube free, forced, and mixed convection ow regimes for
102 < PrDh =L < 1. (From Metais and Eckert, 1964.)
same direction. The resultant friction factor and heat transfer coecient are higher than
the pure forced convection coecient. However, for uid cooling with upward ow or
uid heating with downward ow, free convection counters forced convection, and lower
friction factor and heat transfer coecient result. The ow regime chart of Metais and
Eckert (1964) for vertical tubes, as shown in Fig. 7.34, provides guidelines to determine
the signicance of the superimposed free convection. The results of Fig. 7.34 are applicable for upow and downow and both constant heat ux and constant wall temperature
boundary conditions.
Chato (1969) summarizes the literature on mixed convection in vertical channels and
suggests the following correlations. For upward liquid ow (water and oil) with liquid
heating in a circular tube with decreasing density (and decreasing viscosity) along the ow
direction (@=@x < 0), the Nusselt numbers and friction factors are given by the following
correlations devised by Worse-Schmidt as presented by Chato (1969) for fully developed ow:
8
>
>
>
< 4:73
Nu
0:25
>
Gr
>
>
: 2:5
Re
8
>
>
>
< 16
f Re
0:475
>
Gr
>
>
: 5:8
Re
for
Gr
< 12:8
Re
for
Gr
> 12:8
Re
for
Gr
< 8:4
Re
for
Gr
> 8:4
Re
7:164
7:165
where Gr is dened by Eq. (7.159). The measured temperature proles and Nusselt
numbers agree quite well with Eq. (7.164) for moderate heat input. Equation (7.165)
lies just above the upper range of the friction factor data obtained with water and oil.
For upward liquid ow with liquid cooling in a circular tube with increasing density
(and increasing viscosity) along the ow direction (@=@x > 0), the Nusselt numbers and
537
friction factors both decrease rapidly beyond Gr=Re 12:8 and 8.4, respectively, and the
friction factors even become negative, due to reverse ows occurring at the wall.
For upward gas ow and constant heat ux, the local Nusselt numbers and friction
factors are given by the following correlations:
Nux
8
1=2 Gzx 3Gzx 20
>
>
> Nucp 0:025q*
3=2
>
>
Gzx
<
>
Gz 8
>
>
Nucp 0:07q*1=2 x 1=2
>
>
:
Gzx
1:58 Gz0:3
x
4:2
7:167
and q* qdi =2ke Te Aw , with the subscript e denoting the condition at the channel
entrance. In Eqs. (7.166) and (7.167), the Graetz number Gzx denotes a local Graetz
number Gzx m_ cp =kx, where x is the length along the tube. Similarly, Nux represents
a local Nu number based on the hydraulic diameter of the channel. For upward-owing
gases with heating, the friction factors are given by
8
16 Tw
>
>
>
>
Re Tm
>
>
>
>
>
< 15:5 T 1:10
w
f
> Re Tm
>
>
>
>
>
>
15:5 Tw 1:25
>
>
:
Re Tm
for
Tw
< 1:2 to 1:5
Tm
for gases
Tw
<3
Tm
Tw
<2
Tm
7:168
where Re is the mean Reynolds number with respect to the ow length. For upwardowing gases with cooling, the friction factors are given by
f
7.8
16 Tw 0:81
Re Tm
Tw
<1
Tm
7:169
The theory and correlations presented in this chapter so far are applicable if the radiation
eect is negligible (which is the case in many heat exchanger applications). However,
when the heat exchanger wall is at a temperature signicantly higher than that of the
uid, or vice versa, heat transfer by radiation may also take place in parallel with heat
transfer by convection. Some examples of high-temperature uids in heat exchangers are
(1) the gas turbine exhaust gas owing over tube bundles in cogeneration applications,
538
(2) exhaust gases from metal industries used in a xed-matrix regenerator, (3) liquid
metals (e.g., sodium) in a fast breeder reactor heat exchanger, and (4) high-temperature
uids used in some heat exchanger applications. However, liquid metals are opaque and
the radiation eect is negligible.
7.8.1
Radiative heat transfer between the heat exchanger surface and the liquid is generally
negligible for water and many liquids having considerably high density compared to
gases such as air. In other words, the eect of radiation would be suppressed due to a
too high uid participation through absorption of any thermal radiation that may take
place. However, there may be some wavelength ranges where the transmissivity may not
be negligible; this wavelength range is dependent on the particular liquid and the mean
beam length for radiation. Most liquids, including water used in heat exchanger applications, are opaque and the radiation heat transfer to and from liquids is negligible,
although some of the liquids may have slight absorptivity in some wavelength range.
7.8.2
n
X
7:170
j1
where Fkj is the view factor or conguration factor for thermal radiation from the surface
k to surfaces j,
is the StefanBoltzmann constant,{ and Tk and Tj must be absolute
temperatures (K or 8R). Equation (7.170) implies that all n black surfaces in the enclosure
are at. Howell (1982) has provided the view factors for many congurations, and the
latest catalog is available in a CD with the book by Siegel and Howell (2002). Some view
factors applicable to heat exchanger applications are summarized in Table 7.14.
In the case of absobing gas ows, the radiation eect is important only for those
components of gas that have signicant concentration and also absorb/emit infrared
radiation. Water vapor and carbon dioxide often fall in this category. Sulfur and nitrogen oxides, on the other hand, are usually not present in high-enough concentration to
warrant their consideration.
{
5:6696
108 W=m2 K4 0:1714
108 Btu/hr-ft2-8R4.
539
TABLE 7.14 View Factors and Radiation Heat Transfer between Some Heat Exchanger Surfaces
for Transparent Flowing Gas
A1, T1, 1
q12
F12 1
A2, T2, 2
A
T 41 T 42
1="1 1="2 1
r1
A1 r1
A2 r2
r2
q12
A1 T 41 T 42
1="1 1 "2 ="2 r1 =r2
F12 1
Concentric Spheres
A1 r21
A2 r22
r1
q12
A1 T 41 T 42
1="1 1 "2 ="2 r1 =r2 2
F12 1
r2
A1, T1, 1
q12
A1 "1 T 41 T 42
F12 1
A2, T2, 2
Consider a gas component having emissivity "g and absorptivity g , with a bounding
heat transfer surface area A to which the gas will interchange radiation. The wall surface
temperature is Tw , and the gas owing through the heat exchanger is at a mean temperature Tg (which is the same as Tm in the rest of this book). Then the net radiation heat
transfer by the gas to a black wall surface is given by
qrad
A"g Tg4 g Tw4
7:171
Here Tg and Tw must be absolute temperatures (in K or 8R). Since the radiation band
structure (e.g., absorption coecient as a function of wavelength) of the gas is dependent
on its temperature, the energy emitted by gas depends on Tg , whereas the energy
absorbed by the gas depends as well on the radiation emitted by the wall at Tw in addition
to its dependence on Tg . Thus "g Tg and g Tg ; Tw depend on the temperatures indicated. Additionally, both "g and g depend on the partial pressure of the gas component,
which participates in radiation heat transfer. Note that if the wall radiates heat to the gas,
540
qrad in Eq. (7.171) will be negative. We explain how to evaluate "g and g by the mean
beam length treatment of gas properties after we develop the magnitude of the combined
heat transfer coecient in a heat exchanger when radiation cannot be neglected in
relation to convection.
In the case of gray walls with an emissivity of "w , the evaluation of qrad is more
complicated, due to multiple reections. It can be shown that the net heat transfer
from the gas to the wall is given by
qrad
"w
A
"g Tg4 g Tw4
1 1 "w 1 g
7:172
4
A
T Tw4
1="g 1="w 1 g
7:173
Since the convection heat transfer equation is in terms of the temperature dierence
Tw Tg , the radiation heat transfer coecient hrad can be presented similarly using
Eq. (7.172):
hrad
ATg Tw 1 1 "w 1 g Tg Tw
7:174
In this equation, Tw and Tg must be given in K or 8R. Note that hconv for forced
convection is generally not a strong function of the temperatures Tw and Tg , but hrad
is a strong function of Tw and Tg , as shown in Eq. (7.174).
In heat exchangers, the combined convection and radiation eects are then taken care
of approximately by considering the convection and radiation heat transfer phenomena
in parallel. Hence,
hcombined hconv hrad
7:175
The rest of the heat exchanger analysis for a combined convection and radiation problem
is performed the same as before by replacing h (or hconv ) with hcombined , assuming that
hcombined is dened using jTw Tg j jTw Tm j.
Now we consider the determination of "g and g of Eq. (7.174) from the experimental
results/correlations of Hottel and Sarom (1967), and some of them reproduced by
Incropera and DeWitt (2002), among others. The gas emissivity "g has been correlated
in terms of the temperature Tg , the total pressure p of the gas, the partial pressure pg of
the gas species (such as water vapor, carbon dioxide, etc.), and the radius L of an
equivalent hemispherical gas mass. More precisely, Le , the mean beam length, is the
required radius of a gas hemisphere such that it radiates a heat ux to the center of its
base equal to the average ux radiated to the area of interest by the actual volume of gas
(Siegel and Howell, 2002). In most exhaust gases, water vapor and carbon dioxide are the
most important components from the radiation eect viewpoint, and hence we only
consider them here. For other gas components, refer to Hottel and Sarom (1967).
The water vapor emissivity "H2 O is presented in Fig. 7.35 as a function of the gas
temperature Tg and pH2 O Le . Here pH2 O is the partial pressure of water vapor in the gas
mixture at a total pressure of 1 atm; and Le is the mean beam length to take into account
541
0.8
0.6
20
10
5
3
2
0.4
0.3
0.2
1
Emissivity, H O
2
0.6
0.4
0.1
0.08
0.2
0.06
0.1
0.04
0.06
0.04
0.03
0.02
0.02
0.015
0.01
0.007
0.01
0.008
pH OL e = 0.005 ft atm
2
0.006
300
600
900
1200
1500
Gas temperature, Tg (K)
2100
1800
FIGURE 7.35 Emissivity of water vapor in a mixture with nonradiating gases at 1 atm total
pressure and of hemispherical shape. (From Hottel, 1954.) (1 ft atm 0.305 m atm).
pH OLe =
2
1.8
0 0.05 ft atm
0.25
0.55
1.0
2.5
5.0
10.0
Pressure correction, CH
2O
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0
0.2
0.4
(pH
+
2O
0.6
0.8
p)/ 2 (atm)
1.0
1.2
FIGURE 7.36 Correction factor for obtaining water vapor emissivity at pressures other than
1 atm; here, "H2 O; p61 atm CH2 O "H2 O; p1 atm . (From Hottel, 1954.) (1 ft atm 0.305 m atm).
542
Characteristic Length
Le
Diameter D
Diameter D
Diameter D
Diameter D
0:65D
0:95D
0:65D
0:60D
the size and shape of the gas geometry, which is dierent from a hemispherical geometry.
If the total pressure of the gas is dierent from 1 atm, a correction factor CH2 O (to be
multiplied to "H2 O for 1 atm pressure) is obtained from Fig. 7.36. The mean beam length
Le is provided in Table 7.15 for various gas geometries (Incropera and DeWitt, 2002).
Similarly, the carbon dioxide emissivity "CO2 and corresponding correction factor CCO2
are provided in Figs. 7.37 and 7.38, respectively.
The gas absorptivity g can be evaluated using the approximate procedure from the
following expressions (Siegel and Howell, 2002), where the emissivity and the correction
for total pressure dierent from 1 atm for a specic component are obtained from Figs.
7.35 through 7.38.
Tm 0:5
T
"H2 O Tw ; pH2 O Le w
Tw
Tm
0:5
Tm
T
CCO2
"CO2 Tw ; pCO2 Le w
Tw
Tm
H2 O CH2 O
7:116a
CO2
7:116b
The aforementioned emissivity and absorptivity values for water vapor and carbon
dioxide apply only when they are alone with other nonradiating components in the
gas mixture. However, when both are present in the gas with other nonradiating components, the total gas emissivity and absorptivity are given by
"gas CH2 O "H2 O CCO2 "CO2 "
7:177
Now evaluate " where the correction factor " is given in Fig. 7.39 (see Hottel
and Sarom, 1967).
SUMMARY
The most important inputs for the thermal and hydraulic design of a heat exchanger are
heat transfer coecients and friction factors for the heat transfer surfaces used. The
thermal design of the exchanger depends on the accuracy of this information. Except
for the major heat exchanger industries having extensive test facilities, such information
is not available on modern heat transfer surfaces to most small companies, consultants,
543
SUMMARY
0.3
0.2
0.06
2.0
1.0
0.8
0.4
0.2
0.1
0.06
0.04
0.04
0.02
0.1
Emissivity, CO2
0.08
0.03
0.01
0.02
0.006
0.005
0.004
0.01
0.003
0.008
0.006
0.004
0.003
0.002
0.001
0.002
0.001
300
600
900
1200
1500
Gas temperature, Tg (K)
1800
2100
FIGURE 7.37 Emissivity of carbon dioxide in a mixture with nonradiating gases at 1 atm total
pressure and of hemispherical shape. (From Hottel, 1954.) (1 ft atm 0.305 m atm).
2.0
1.5
1.0
0.8
0.6
0.5
0.4
0.3
0.05
0.08 0.1
0.2
0.3
0.5
p (atm)
0.8 1.0
2.0
3.0
5.0
FIGURE 7.38 Correction factor for obtaining carbon dioxide emissivity at pressures other than 1
atm; here, "CO2 ; p61 atm CCO2 "CO2 ; p1 atm . (From Hottel, 1954.) (1 ft atm 0.305 m atm).
544
Mixture correction,
0.07
Tg 540C
L e( p + p
Tg 540C
L e( p + p ) =
Tg 125C
0.06 Le( pH O + pCO ) =
2
2
5 ft atm
0.05
H2O
H2O
CO2
5 ft atm
5 ft atm
3
2
1.5
1.0
0.75
0.04
3
0.03 0.75
0.5
0.02 0.3
0.2
2
1.5
1.0
0.01
0
0.2
3
2
1.5
1.0
0.75
0.50
0.30
0.20
0.4 0.6
pH O
0.50
0.8 1.0 0
pCO + pH O
2
0.2
0.30
0.20
0.4 0.6
p
0.8 1.0
H2O
pCO + pH O
2
1.0 0
)=
CO2
pCO + pH O
2
FIGURE 7.39 Correction factors associated with mixtures of water vapor and carbon dioxide.
(From Hottel, 1954.) (1 ft atm 0.305 m atm).
and academia. Usually, this information may be obtained from the correlations in the
open literature, analytical solutions, and/or experimentation. Hence, in this chapter we
have provided sucient details on all these aspects with background information so that
one may have a broad understanding of what to look for, may be able to either obtain
accurate j (or Nu) and f data, or be able to use the analytical correlations for approximations and extrapolations with sucient accuracy. An approximate correlation or methodology is also presented in the chapter to cover situations where uid properties vary
signicantly, there is combined free and forced convection, or there is combined convection and radiation in a heat exchanger. With this broad understanding of the topics
covered in this chapter, the reader will be able to obtain approximate values of the
surface characteristics and be able to design exchangers with reasonably accurate
required performance level or the size. Hence, after learning the basic design theory of
Chapter 3, the information provided in this chapter is most important for the accurate
thermal design of heat exchangers.
REFERENCES
Achaichia, A., and T. A. Cowell, 1988, Heat transfer and pressure drop characteristics of at tube
louvered plate n surfaces. Exp. Thermal Fluid Sci., Vol. 1, pp. 147157.
Amon, C. H., D. Majumdar, C. V. Herman, F. Mayinger, B. B. Mikic, and D. P. Sekulic, 1992,
Numerical and experimental studies of self-sustained oscillatory ows in communicating
channels, Int. J. Heat Mass Transfer, Vol. 35, pp. 31153129.
Aung, W., 1987, Mixed convection in internal ow, in Handbook of Single-Phase Convective Heat
Transfer, S. Kakac, R. K. Shah, and W. Aung, eds., Wiley, New York, Chap. 15.
Baclic, B. S., P. J. Heggs, and H. Z. Z. A. Ziyan, 1986a, Dierential uid enthalpy method for
predicting heat transfer coecients in packed beds, Heat Transfer 1986, Proc. 8th Int. Heat
Transfer Conf., Vol. 5, pp. 26172622.
REFERENCES
545
Baclic, B. S., D. D. Gvozdenac, D. P. Sekulic, and E. J. Becic, 1986b, Laminar heat transfer
characteristics of a plate-louver n surface obtained by the dierential uid enthalpy method,
in Advances in Heat Exchanger Design, eds. R. K. Shah and J. T. Pearson, HTD-Vol. 66, ASME,
New York, pp. 2128.
Bejan, A., 1995, Convection Heat Transfer, 2nd ed., Wiley, New York.
Bemisderfer, C. H., 1998, Contemporary developments in the thermal design of nned-tube heat
exchangers, J. Enhanced Heat Transfer, Vol. 5, pp. 7190.
Bhatti, M. S., and R. K. Shah, 1987, Turbulent and transition convective heat transfer in ducts, in
Handbook of Single-Phase Convective Heat Transfer, S. Kakac, R. K. Shah, and W. Aung, eds.,
Wiley, New York, Chap. 4.
Briggs, D. E., and E. H. Young, 1963, Convection heat transfer and pressure drop of air owing
across triangular pitch banks of nned tubes, Chem. Eng. Prog. Symp. Ser. 41, Vol. 59, pp. 110.
Burmeister, L. C., 1993, Convective Heat Transfer, Wiley, New York.
Chang, Y. J., and C. C. Wang, 1997, A generalized heat transfer correlation for louver n geometry,
Int. J. Heat and Mass Transfer, Vol. 40, pp. 533544.
Chang, Y. J., K. C. Hsu, Y. T. Lin and C. C. Wang, 2000, A generalized friction correlation for
louver n geometry, Int. J. Heat Mass Transfer, Vol. 43, pp. 22372243.
Chato, J. C., 1969, Combined free and forced convection ows in channels, in Advanced Heat
Transfer, B. T. Chao, ed., University of Illinois Press, Urbana, IL, pp. 439453.
Chen, C. J., and J. S. Chiou, 1981, Laminar and turbulent heat transfer in the pipe entrance for
liquid metals, Int. J. Heat Mass Transfer, Vol. 24, pp. 11791190.
Churchill, S. W., and S. C. Zajic, 2002, The improved prediction of turbulent convection, Heat
Transfer 2002, Proc. 12th Int. Heat Transfer Conf., pp. 279284.
Colburn, A. P., 1933, A method of correlating forced convection heat transfer data and a comparison with uid friction, Trans. Am. Inst. Chem. Eng., Vol. 29, pp. 174-210; reprinted in Int. J.
Heat Mass Transfer, Vol. 7, pp. 13591384, 1964.
Cowell, T. A., M. R. Heikal, and A. Achaichia, 1995, Flow and heat transfer in compact louvered n
surfaces, Exp. Thermal Fluid Sci., Vol. 10, pp. 192199.
Das, S. K., 2001, Private Communication, Department of Mechanical Engineering, IIT Madras,
Chennai, India.
Dipprey, D. F., and R. H. Sabersky, 1963, Heat and momentum transfer in smooth and rough tubes
at various Prandtl numbers, Int. J. Heat Mass Transfer, Vol. 6, pp. 329353.
Dittus, F. W., and L. M. K. Boelter, 1930, Heat transfer in automobile radiators of tubular type,
University of Calif. Publications in Engineering, Vol. 2, pp. 443461.
Gaddis, E. S., and V. Gnielinski, 1985, Pressure drop in cross ow across tube bundles, Int. Chem.
Eng., Vol. 25, pp. 115.
Ganguli, A., and S. B. Yilmaz, 1987, New heat transfer and pressure drop correlations for crossow
over low-nned tube banks, AIChE Symp. Ser. 257, Vol. 83, pp. 914.
Ghajar, A. J., and L. M. Tam, 1994, Heat transfer measurements and correlations in the transition
region for a circular tube with three dierent inlet congurations, Exp. Thermal Fluid Sci., Vol. 8,
pp. 7990.
Ghajar, A. J., and L. M. Tam, 1995, Flow regime map for a horizontal pipe with uniform wall heat
ux and three inlet congurations, Exp. Thermal Fluid Sci., Vol. 10, pp. 287297.
Gnielinski, G., 1976, New equation for heat and mass transfer in turbulent pipe and channel ow,
Int. Chem. Eng., Vol. 16, pp. 359368.
Gvozdenac, D. D., 1994, Experimental prediction of heat transfer coecients by use of double-blow
method, Warme- und Stoubertragung, Vol. 29, pp. 361365.
Heggs, P. J., and D. Burns, 1988, Single blow experimental prediction of heat transfer coecients: a
comparison of commonly used techniques, Exp. Thermal Fluid Sci., Vol. 1, pp. 243251.
546
REFERENCES
547
Nikuradse, J., 1933, Stromungsgesetze in rauhen Rohren, Forsch. Arb. Ing.-Wes., No. 361; English
translation, NACA TM 1292.
Petukhov, B. S., and V. N. Popov, 1963, Theoretical calculation of heat exchange in turbulent ow
in tubes of an incompressible uid with variable physical properties, High Temp., Vol. 1, No. 1,
pp. 6983.
Rabas, T. J., and J. Taborek, 1987, Survey of turbulent forced-convection heat transfer and pressure
drop characteristics of low-nned tube banks in cross ow, Heat Transfer Eng., Vol. 8, No. 2, pp.
4962.
Robinson, K. K., and D. E. Briggs, 1966, Pressure drop of air owing across triangular pitch banks
of nned tubes, Chem. Eng. Prog. Symp. Series 64, Vol. 62, pp. 177184.
Schlunder, E.-U., 1998, Analogy between heat and momentum transfer, Chem. Eng. Process, Vol.
37, pp. 103107.
Shah, R. K., 1981, Compact heat exchangers, in Heat Exchangers: Thermal-Hydraulic Fundamentals
and Design, S. Kakac, A. E. Bergles, and F. Mayinger, eds., Hemisphere Publishing Corp.,
Washington, DC, pp. 111151.
Shah, R. K., 1983, Fully developed laminar ow through channels, in Low Reynolds Number Flow
Heat Exchangers, S. Kakac, R. K. Shah and A. E. Bergles, eds., Hemisphere Publishing Corp.,
Washington, DC, pp. 75108.
Shah, R. K., 1985, Compact heat exchangers, in Handbook of Heat Transfer Applications, 2nd ed.,
W. M. Rohsenow, J. P. Hartnett, and E. N. Ganic, eds., McGraw-Hill, New York, pp. 4174 to
4311.
Shah, R. K., 1990, Assessment of modied Wilson plot techniques for obtaining heat exchanger
design data, Heat Transfer 1990, Proc. 9th Int. Heat Transfer Conf., Vol. 5, pp. 5156.
Shah, R. K., and M. S. Bhatti, 1987, Laminar convective heat transfer in ducts, in Handbook of
Single-Phase Convective Heat Transfer, S. Kakac, R. K. Shah, and W. Aung, eds., Wiley, New
York, Chap. 3.
Shah, R. K., and M. S. Bhatti, 1988, Assessment of correlations for single-phase heat exchangers, in
Two-Phase Flow Heat Exchangers: Thermal-Hydraulic Fundamentals and Design, S. Kakac, A. E.
Bergles, and E. O. Fernandes, eds., Kluwer Academic Publishers, Dordrecht, The Netherlands,
pp. 81122.
Shah, R. K., and A. L. London, 1978, Laminar Flow Forced Convection in Ducts, Supplement 1 to
Advances in Heat Transfer, Academic Press, New York.
Shah, R. K., and A. C. Mueller, 1988, Heat exchange, in Ullmanns Encyclopedia of Industrial
Chemistry, VCH Publishers, Weinheim, Germany, Unit Operations II, Vol. B3, Chapter 2.
Shah, R. K., and Webb, R. L., 1983, Compact and enhanced heat exchangers, in Heat Exchangers:
Theory and Practice, J. Taborek, G. F. Hewitt and N. Afgan, eds., Hemisphere/McGraw-Hill,
Washington, DC, pp. 425468.
Shah, R. K., and S. Q. Zhou, 1997, Experimental techniques for obtaining design data for compact
heat exchanger surfaces, in Compact Heat Exchangers for the Process Industries, R. K. Shah, ed.,
Begell House, New York, pp. 365379.
Siegel, R., and J. R. Howell, 2002, Thermal Radiation Heat Transfer, Taylor & Francis, New
York.
Slanciauskas, A., 2001, Two friendly rules for the turbulent heat transfer enhancement, Int. J. Heat
Mass Transfer, Vol. 44, pp. 21552161.
Sleicher, C. A., and M. Tribus, 1957, Heat transfer in a pipe with turbulent ow and arbitrary walltemperature distribution, Trans. ASME, Vol. 97, pp. 789796.
Taborek, J., 1990, Design method for tube-side laminar and transition ow regime with eects of
natural convection, paper presented in the Open Forum at the 9th Int. Heat Transfer Conf.,
Jerusalem, Israel.
548
Tam, L. M., and A. J. Ghajar, 1997, Eect of inlet geometry and heating on the fully developed
friction factor in the transition region of a horizontal tube, Exp. Thermal Fluid Sci., Vol. 15,
pp. 5264.
Wang, C. C., 2000, Recent progress on the air-side performance of n-and-tube heat exchangers,
Int. J. Heat Exchangers, Vol. 1, pp. 4976.
Wang, C. C., and K. U. Chi, 2000, Heat transfer and friction characteristics of plain n-and-tube
heat exchangers; Part 2; Correlation, Int. J. Heat Mass Transfer, Vol. 43, pp. 26922700.
Webb, R. L., 1994, Principles of Enhanced Heat Transfer, Wiley, New York.
Wheeler, A. J., 1968, Single-blow transient testing of matrix-type heat exchanger surfaces at low
values of Ntu, TR. 68, Department of Mechanical Engineering, Stanford University, Stanford,
California.
Wilson, E. E., 1915, A basis for rational design of heat transfer apparatus, Trans. ASME, Vol. 37,
pp. 4782.
Winterton, R. H. S., 1998, Where did the DittusBoelter equation come from? Int. J. Heat Mass
Transfer, Vol. 41, pp. 809810.
Yu, B., H. Ozoe, and S. C. Churchill, 2001, The characteristics of fully developed turbulent convection in a round tube, Chem. Eng. Sci., Vol. 56, pp. 17811800.
Zhi-qing, W., 1982, Study on correction coecients of laminar and turbulent entrance region eect
in round pipe, Appl. Math. Mech., Vol. 3, No. 3, pp. 433446.
Zukauskas, A., 1987, Convective heat transfer in cross ow, Handbook of Single-Phase Convective
Heat Transfer, S. Kakac, R. K. Shah, and W. Aung, eds., Wiley, New York, Chap. 6.
REVIEW QUESTIONS
For each question, circle one or more correct answers. Explain your answers briey.
7.1
7.2
7.3
7.4
REVIEW QUESTIONS
549
7.5
In a simultaneously developing ow in a circular tube, consider the velocity boundary layer thickness as 10 mm. The approximate thickness in millimeters of the
temperature boundary layer for air, water and sodium ow will be (circle one
number from each group):
(a) air: 5, 10, 15 mm
(b) water: 5, 10, 15 mm
(c) liquid sodium: 5, 10, 15 mm
7.6
7.7
The heat transfer coecient for forced convection could be dependent on the:
(a) thermal boundary condition
(b) uid Prandtl number
(c) ow velocity
(d) wall-to-uid bulk temperature
(e) ow passage shape
dierence
(f) ow passage size
7.8
7.9
The groups that designate dimensionless pressure drop characteristics are the:
(a) Fanning friction factor
(b) Reynolds number
(c) axial distance x
(d) incremental pressure drop number
(e) Euler number
(f ) Hagen number
7.10
7.11
7.12
(b) ow modulus
(d) pressure drop modulus
(f ) none of these
In a good design, NTU of the test core for the steady-state technique for determining heat transfer coecient is usually:
(a) above 3
(b) below 0.4
(c) 0:4 NTU 3
(d) cant tell
H1 boundary condition, the following are kept uniform and constant for
For the *
developing ow in a tube:
(a) heat ux along the periphery
(b) wall temperature along the axial direction
(c) heat transfer coecient
(d) nite wall thermal conductivity (kw < 1)
(e) none of these
(f ) all of these
550
7.13
Consider a circular tube and a sharp cornered quadrilateral duct both having the
same ow area. Compared to the circular tube value, the Nusselt number for
the quadrilateral duct for fully developed laminar ow will be:
(a) higher
(b) lower
(c) equal
(d) cant tell
7.14
7.15
In fully developed ow of the following regimes, the Nusselt number is independent of the Prandtl number:
(a) laminar
(b) transition
(c) turbulent
(d) none of these
7.16
7.17
7.18
7.19
7.20
If we change from two to four passes on the tube side of a given shell-and-tube
exchanger (the same shell diameter and number of tubes) with the same tube ow
rate, the tube-side velocity in turbulent ow will:
(a) remain the same
(b) increase by 40%
(c) increase by 74%
(d) increase by a factor of 2
(e) increase by a factor of 4
(f ) increase by a factor of 8
7.21
For the case of Question 7.20, the tube-side pressure drop will:
(a) remain the same
(b) increase by 40%
(c) increase by 74%
(d) increase by a factor of 2
(e) increase by a factor of 4
(f ) increase by a factor of 8
7.22
For the case of Question 7.20, the tube-side heat transfer coecient will:
(a) remain the same
(b) increase by 40%
(c) increase by 74%
(d) increase by a factor of 2
(e) increase by a factor of 4
(f ) increase by a factor of 8
Hint: Use DittusBoelter correlation of Eq. (7.79) for Nu.
7.23
The heat transfer coecient for fully developed laminar ow increases with:
(a) increase in surface area
(b) decrease in hydraulic diameter
(c) decrease in uid velocity
(d) increase in uid thermal conductivity
REVIEW QUESTIONS
551
7.24
For a fully developed laminar ow through a circular tube, identify how the heat
transfer coecient will vary with increasing values of the following variables or
uid properties (circle one for each item):
(a) ow velocity um :
(i) increase
(ii) decrease
(iii) doesnt change
(b) thermal conductivity k: (i) increase
(ii) decrease
(iii) doesnt change
(c) density :
(i) increase
(ii) decrease
(iii) doesnt change
(i) increase
(ii) decrease
(iii) doesnt change
(d) specic heat cp :
(e) dynamic viscosity :
(i) increase
(ii) decrease
(iii) doesnt change
(i) increase
(ii) decrease
(iii) doesnt change
(f ) tube diameter di :
(g) tube length L:
(i) increase
(ii) decrease
(iii) doesnt change
(i) increase
(ii) decrease
(iii) doesnt change
(h) heat ux q 00 :
7.25
7.26
Pressure drop for airow through a long circular tube at Re 2000 is measured as
7 kPa and the heat transfer coecient as 100 W/m2 K. Neglect the entrance and
exit losses and the momentum eect in the pressure drop evaluation. Consider
p 16m_ f ReL=2gc D4h 4fL Re2 2 =2gc D3h . Assume that Nu 4:36
for laminar ow, Pr 0:7, and constant uid properties.
(a) If the tube diameter is doubled, the resulting pressure drop will be the following factor of the original pressure drop (assume that the ow rate remains
constant):
1
(ii) 12
(iii) 2
(iv) 16
(v) 1
(vi) cant tell
(i) 14
(b) If the tube diameter is doubled, the resulting heat transfer coecient will be the
following factor of the original heat transfer coecient:
1
(i) 14
(ii) 12
(iii) 2
(iv) 16
(v) 1
(vi) cant tell
(c) If the tube length is doubled, the resulting pressure drop will be the following
factor of the original pressure drop:
1
(ii) 12
(iii) 2
(iv) 16
(v) 1
(vi) cant tell
(i) 14
(d) If the tube length is doubled, the resulting heat transfer coecient will be the
following factor of the original heat transfer coecient:
1
(ii) 12
(iii) 2
(iv) 16
(v) 1
(vi) cant tell
(i) 14
(e) If the ow rate is increased ve times, the resulting pressure drop will be the
following factor of the original pressure drop:
(i) 5
(ii) 25
(iii) 24.7
(iv) doesnt change
(v) cant tell
(f ) If the ow rate is increased ve times, the resulting heat transfer will be the
following factor of the original heat transfer coecient:
(i) 5
(ii) 3.62
(iii) 8.64
(d) 25
(e) doesnt change
Hint: Use Eqs. (7.72) and (7.79), for items (e) and (f ).
7.27
Given the physical location x in a circular tube, the Nusselt number in the
laminar thermal entrance region will:
(a) increase with increasing Pr
(b) decrease with increasing Pr
(c) cant tell
7.28
The Nusselt number for fully developed laminar duct ow depends on:
552
(a)
(c)
(e)
(g)
(i)
Reynolds number
cross-sectional geometry
Tw Tm
St
all of these
(b)
(d)
(f )
(h)
Prandtl number
thermal boundary condition
heat ux
none of these
7.29
7.30
7.31
The Nusselt number for fully developed turbulent duct ow Pr > 0:5 depends
on:
(a) Reynolds number
(b) Prandtl number
(c) cross-sectional geometry
(d) thermal boundary condition
(f ) heat ux
(e) Tw Tm
(g) St
(h) none of these
(i) all of these
7.32
7.33
7.34
7.35
PROBLEMS
553
7.36
Variable uid properties at a cross section yield higher heat transfer coecients
for:
(a) liquid heating
(b) liquid cooling
(c) gas heating
(d) gas cooling
7.37
7.38
Superimposed free convection over forced convection will yield higher heat transfer coecients for:
(a) gas heating in a horizontal tube
(b) liquid cooling in a horizontal tube
(c) liquid heating in a vertical tube with upow
(d) gas cooling in a vertical tube with downow
7.39
PROBLEMS
7.1
A crossow heat exchanger uses strip ns on the air side with the following
geometry: n pitch 549 m1 , plate spacing 9:53 mm, n length 3:18 mm,
ow passage hydraulic diameter 2:68 mm, n metal thickness 0:25 mm,
total heat transfer area/volume between plates 1250 m2 =m3 , and n area/total
area 0:840. Fins and parting sheets are made of stainless steel
kw 20:77 W=m K. Parting sheet thickness is 0.381 mm. Air ows at
u1 3:05 m/s with 1:1213 kg=m3 , 1:58
105 m2 =s, Pr 0:70, and
cp 1:00 kJ=kg K. For the air side,
Ao =Afr 0:402.
(a) Determine h for Re 1000.
(b) If the n thickness is reduced from 0.25 mm to 0.16 mm, how would h and
f be aected? Give qualitative reasons only.
(c) If every geometric dimension of the strip n geometry is scaled up by a factor
of 5, estimate j and f factors for operation at Re 1000. Use j 0:0192 and
f 0:0927 at Re 1000 for this surface.
7.2
A plate-n condenser is to be designed on the air side with 790 ns/m and a
0.025 mm n thickness. You are to choose between the triangular and rectangular
ow passage surfaces shown in Fig. P7.2. The mass ow rate m_ and the frontal
area Afr of the air side are xed, and the frontal velocity u1 3:05 m/s. Use the
following uid properties: 1:1213 kg/m3, 1:58
105 m2/s, Pr 0:70, and
554
2.5 mm
2.5 mm
60
2.17 mm
0.025
mm
0.025 mm
Dh = 1.57 mm
= Ao /Afr = 0.990
Dh = 1.45 mm
= Ao /Afr = 0.988
FIGURE P7.2
cp 1:00 kJ/kg K. For the air side, calculate the ratios of (a) heat transfer coecients hT =hR , (b) surface areas AT =AR for the same hA, and (c) uid pumping
power P T =P R . Which surface would you select? Why? Note that the subscripts T
and R here denote triangular and rectangular passages. Hint: Knowing Re for
each surface, calculate j (and h), and f from theoretical solutions provided in
the text.
7.3
7.4
A heat exchanger consists of 300 tubes 1.83 m long and 25.4 mm OD. The tubes
are arranged in 15 rows with Xt X 50:8 mm. The tube surface temperature is
maintained at 93.38C. Air at 1 atm and 48.98C ows normal to the tube bank at
6.1 m/s. Calculate the air-side heat transfer coecient, outlet air temperature,
total heat transfer rate, and air-side pressure drop considering (a) inline and
(b) staggered tube arrangements. Compare results and discuss.
7.5
Determine the heat transfer coecient for air, water, and liquid sodium ow
through a tube of circular, rectangular with * 0:125, and equilateral triangular
cross sections, each with Dh 25:4 mm. First consider Re 1000 and repeat the
H1
calculations at Re 10,000. Assume fully developed ow for each case and the *
boundary condition. Which uid and which cross-section geometry yield the highest heat transfer coecient? Why? Consider each uid at Tm of 365 K and at an
appropriate pressure having the following uid properties.
cp kJ=kg K
Pa s
k W=m K
Pr
Air
Water
Sodium
1.011
2:15
105
0.0311
0.699
4.209
3:06
104
0.677
1.90
1.38
6:98
104
86.2
0.011
555
PROBLEMS
7.6
The coolant passages in a reactor core are 0.127 m long and have a rectangular
cross section of 19 mm
25.4 mm. The walls of the passages are to be maintained
at a constant temperature of 3718C.
(a) If the coolant is nitrogen gas at 8 atm pressure and a temperature of 2608C,
what velocity in the passages will result in a nitrogen discharge temperature of
3168C?
(b) If the passage walls are assumed to have a roughness equivalent to that of
commercial steel pipes (e 0:046 mm), what velocity is required for nitrogen
discharge temperature of 3168C?
7.7
7.8
Re
0.00965
0.0103
0.0110
0.0117
0.0125
0.0362
0.0398
0.0432
0.0467
0.0500
0.0551
600
500
400
300
200
150
0.0138
0.0152
0.0170
0.0202
0.0639
0.0713
0.0828
0.102
0.137
0.171
FIGURE P7.8
556
mm, wall thickness 2.3 mm, and duct length as 3 m. The temperature of the air
inlet to the duct is 658C and the ambient temperature is 158C. The airow rate is
0.08 kg/s. The purpose of the problem is to minimize heat losses to the ambient
and hence to investigate the inuence of an insulation layer of 3.2 mm thickness
inside this duct. Determine without insulation; (a) the heat transfer coecient
inside the duct, (b) the heat exchange system eectiveness, (c) the outlet air
temperature, and (d) the heat loss. Repeat the calculations and determine the
same four quantities with insulation. Discuss the results. Use the following air
properties: 1:058 kg/m3, cp 1:008 kJ/kg K, 2:04
104 Pa s,
k 0:0288 W=m K, and Pr 0:701. The thermal conductivity of the wall and
insulation are 2.5 and 0.045 W=m K, respectively. Consider the natural convection coecient on the outside of the duct as 5 W=m2 K. Specify clearly any
assumptions that you may make.
7.9
7.10
Dh 3:475 mm,
0:48, 557:7 m2/m3, o 0:8, Pr 0:7,
cp 1005 J=kg K, 2:07
105 Pa s
PROBLEMS
557
Core depth L1
Water
Width L3
Air
Height L2
FIGURE P7.11
the water to air is 29.31 kW. Water and air inlet temperatures are 82.28 and
26.78C, respectively, and the specic heats are 4187 and 1009 J=kg K, respectively. The air-side pressure drop is 250 Pa, and the water-side pressure drop is
34.5 kPa. The objective of this problem is to determine the new heat transfer rate
and pressure drop on both uid sides (a) when the core width is doubled, (b) when
the core height is doubled, (c) when the core depth is doubled, and (d) when the air
inlet temperature of the original problem is reduced to 15.68C. In all cases, assume
that the core mass velocities on the air and water sides do not change due to the
ram eect on the air side and changing the water pump on the water side.
7.12
Consider a single-pass shell-and-tube air heater with condensing steam outside the
tube bundle and turbulent airow inside the tubes. The following equations are
applicable for the design with Nu, f, G, and p expressions for the tube/air side:
q m_ cp a Ta;o Ta;i m_ cp a Ta
q Udi LNt Tlm "m_ cp a Ts Ta;i
Tlm
"
Ts Ta;i Ts Ta;o
lnTs Ta;i =Ts Ta;o
Ta;o Ta;i
1 eNTU
Ts Ta;i
UA
Ta
m_ cp a Tlm
0:8
hDh
Gdi
0:023
Nu
Pr0:4
k
4m_
4fLG2
U hi
G 2a
p
2gc di
di Nt
NTU
Gdi 0:2
f 0:046
Below is a list of independent variables and some dependent variables. When one
independent variable is changed, all other independent variables are assumed to
558
remain constant. Assume constant uid properties and turbulent ow on the shell
side.
Independent Variables
1. Inlet temperature of air
2. Outlet temperature of air
3. Tube diameter
4. Tube wall thickness
5. Mass ow rate of air
6. Steam pressure
7. Total cross-sectional area of tubes
8. Heat transfer coecient on steam
side
9. Thermal conductivity of tube
material
10. Density of tube material
Dependent Variables
A.
B.
C.
D.
E.
F.
G.
Increases
Remains Same
Decreases
A
B
C
D
E
F
G
H
I
Change to the following conditions, and answer as in part (a):
(b) For constant values of independent variables 1, 2, 3, 4, 5, 7, 8, 9, and 10, and if
the independent variable 6 is increased.
(c) For constant values of independent variables 1, 2, 3, 4, 5, 6, 8, 9, and 10, and if
the independent variable 7 is increased.
(d) For constant values of independent variables 1, 2, 3, 4, 5, 6, 7, 9, and 10, and if
the independent variable 8 is increased.
(e) For constant values of independent variables 1, 2, 3, 4, 5, 6, 7, 8, and 9 and if
the independent variable 10 is increased. For most materials, the thermal
conductivity and density of tube material are related. However, for this
problem, we assume them to be independent variables.
7.13
PROBLEMS
559
and waterglycol mixture) is the same in the radiator. Determine the following at a
given vehicle speed.
(a) Increase or decrease in the heat transfer coecient for the waterglycol
mixture compared to that for pure water. Consider fully developed turbulent
ow in the water tubes and use the following DittusBoelter correlation for the
Nusselt number: Nu 0:023Re0:8 Pr0:4 . The physical properties of water and
the waterglycol mixture at 938C are as follows:
Property
Water
Density (kg/m3)
Specic heat cp (J/kg K)
Dynamic viscosity Pa s
Thermal conductivity k W=m K
Prandtl number
WaterGlycol Mixture
963.5
4212
3:05
104
0.675
1.90
1017.6
3592
7:50
104
0.410
6.57
VT NuL
P L NuT
P T NuL
XL*
XT*
di;T
di;L
2
2
di;L
di;T
2
f ReL
f ReT
ReL
ReT
2
Afr;L
Afr;T
XL*
XT*
2
ReT di;L
ReL di;T
Compute LL =LT from known ratios of VL =VT and Afr;L =Afr;T . Consider water
owing through the tubes at 310 K with the thermal conductivity
k 0:628 W=m K and Pr 4:62. Assume that ReT 5
104 and ReL 100.
Here the subscripts T and L denote turbulent ow and laminar ow, respectively.
(a) Determine the tube diameter required for a laminar ow exchanger to achieve
the same heat transfer coecient as found in the 20 mm tube. The following
correlations are given:
560
NuL 3:657
NuT 0:024Re
Pr
0:4
(b) Determine the ratios of exchanger volume, pumping power, frontal area, and
core length for laminar ow to those for turbulent ow.
(c) Discuss the results of part (b), including peculiarities, advantages, and disadvantages of the laminar ow exchanger.
7.15
A water pipe is embedded in an icewater bath such that the pipe wall temperature
is 08C. The water mass ow rate through the pipe is 1 kg/s. The water inlet
temperature is 508C and the specic heat of water is 4187 J/kg 8C.
(a) If the exchanger eectiveness is 80%, determine the outlet temperature of
water.
(b) What is the total heat transfer rate between water and the icewater bath?
(c) If the pipe diameter is doubled D2 2D1 , what would be the corresponding
new pipe length ratio L2 =L1 if the exchanger eectiveness remains unchanged?
Assume fully developed laminar ow.
7.16
During the performance testing of a compact heat exchanger, if the room air is not
conditioned, there is no control on the inlet temperature of air. Similarly, since the
fan delivers volumetric ow rate as a function of its speed, the mass ow rate will
depend on the air density (inlet temperature). Hence, in general, one cannot attain
the exact desired air inlet temperature and airow rate during testing. The objective of this problem is to correct the performance of a heat exchanger from the
measured values to the standard values.
A crossow heat exchanger has been tested for performance by the steady-state
technique using hot water and ambient air as the uids on the tube and n sides,
respectively. The following data have been measured:
Air inlet temperature 32:28C
Use cp for air and water as 1005 and 4187 J/kg K. Calculate the heat transfer rate
of this exchanger. For the simplicity of calculations, consider C* 0 for all sections of the problem below. Also note that the mass ow rate is given, so that it will
not depend on the inlet temperature.
(a) Now consider the desired inlet air temperature of 26.78C and no change in air
and water mass ow rates as well as water inlet temperature. Assume constant
uid properties. Determine the heat transfer rate of this exchanger. Also compute the percentage change in the heat transfer rate and the percentage change
in the inlet temperature dierence (ITD).
(b) Instead, now consider the desired air mass ow rate as 1.76 kg/s, and determine
the heat transfer rate of the exchanger. Assume no change in air and water inlet
temperatures, water mass ow rate, and air-side heat transfer coecient. For
PROBLEMS
561
this case, also compute the percentage changes in the heat transfer rate, air
mass ow rate, and the exchanger eectiveness.
(c) Next determine the heat transfer rate of this exchanger if the air mass ow rate
and air inlet temperature would have been 1.76 kg/s and 26.78C. Note that the
water mass ow rate and water inlet temperature remain the same. Continue
all assumptions parts (a) and (b).
(d) Discuss the results of parts (a), (b), and (c). Hint: First discuss individual
functional relationships for changes.
7.17
You are asked to design an oil cooler to cool the lubricating oil for a 168 kW diesel
engine. The oil inlet temperature is 828C, the oil ow rate is 7:57
104 m3 =s
( 0:643 kg/s), and the required heat rejection is 19.78 kW. The allowable pressure
drop on the oil side is 3:8
104 Pa. Water at 278C and 1:262
103 m3/s
( 1:285 kg/s) is available as a coolant. You have designed a single-pass counterow shell-and-tube heat exchanger with 100 tubes of 3.18 mm inside diameter and
0.3 m length to do the job. Calculate p on the oil side to check whether or not it is
within specications. Assume f Re 18 for the oil ow through the tubes. Since
there was a serious mistake in the design, the oil cooler is unable to deliver the
desired ow at design p. What is the mistake? For your reanalysis of p, consider the mean wall temperature as 358C since h on the water side is very high. Also
use the following uid properties: cp 4:187 kW/kg K for water; for oil
855:7 kg/m3, and cp 2081 J/kg K at the exchanger mean temperature.
The dynamic viscosity for the engine oil varies with temperature as
36:87
105 1:8T 323:59
where T is in 8C and is in Pa s. Neglect entrance and exit losses and the ow
momentum eect for the p calculations.
7.18
The purpose of this problem is to investigate the inuence of ow gross maldistribution on an air-cooled air-conditioning condenser with round tubes and
at ns (Fig. P7.18). Because of the specic packaging arrangement, the condensers face area is blocked by 50%, as shown in Fig. P7.18b. Assume that the total
airow rate over the partially blocked condenser of Fig. P7.18b is the same as the
airow rate over the unblocked condenser of Fig. P7.18a. The Nusselt number and
friction factor correlations for the airside surface are as follows:
Nu 2:0Re1=3
f 4:0Re0:5
Refrigerant
Blocked
w
rflo
Ai
Blocked
(a)
FIGURE P7.18
(b)
562
Determine the following, mentioning clearly any additional assumptions that you
may make for the solution during each step.
(a) Percentage increase or decrease in the air-side heat transfer coecient due to
blockage of the face area.
(b) Percentage increase or decrease in the air-side pressure drop.
(c) Percentage increase or decrease in o hA on the air side. Here assume that o is
suciently high that it changes negligibly with the change in h.
(d) The total thermal resistance for the unblocked condenser in terms of the airside thermal resistance assuming negligible fouling and wall resistances. For
this case, the ratio o hA on the air side to hA on the refrigerant side is 13. Note
that there are no ns on the refrigerant side. Hint: Use Eq. (3.24).
(e) The total thermal resistance of the partially blocked condenser in terms of its
air-side thermal resistance, again neglecting fouling and wall resistances. Note
that the refrigerant-side heat transfer coecient for unblocked and partially
blocked condensers remains the same (since the refrigerant passages are not
blocked); only its heat transfer surface area is aected.
(f ) The NTU of the partially blocked condenser, assuming the NTU of the
unblocked condenser to be 0.5 and UAs known from parts (d) and (e).
(g) Finally, determine the reduction in the heat transfer rate of the condenser
assuming that C* 0 and the same inlet temperature dierence Tmax .
The objective of this chapter is to outline surface geometrical characteristics that are
used in the determination of experimental j (or Nu) and f factors and in the design of
various heat exchangers. If the surface geometries required are determined dierently
than those for the original correlations, the heat transfer and pressure drop computed
can be signicantly dierent from the real (or measured) values regardless of how highly
accurate the original j and f data are. Important geometrical characteristics are: the heat
transfer area (both primary and secondary, if any), minimum free-ow area, frontal area,
hydraulic diameter, and ow length on each uid side of the exchanger [the ow lengths
could be dierent for heat transfer and pressure drop calculations e.g., see Eqs. (8.7) and
(8.8)]. The ratio of free-ow area to frontal area is needed for the determination of
entrance and exit pressure losses. Heat transfer surface area density is an important
parameter used in heat exchanger calculations. For a nned surface, an appropriate
length is needed for the n eciency determination. The foregoing geometrical characteristics are derived from basic geometric measurements of a heat exchanger and its
surfaces. In this chapter, we take one set of basic dimensions known for each geometry
and arrive at the geometrical characteristics for the following exchangers: tubular, tuben, plate-n, and simple cylindrical passage regenerators; these are presented in Sections
8.1 through 8.4. All important geometrical characteristics associated with ow bypass
and leakages for shell-and-tube heat exchangers are presented in Section 8.5 in terms of
known geometrical parameters for segmental baed exchangers.
8.1
Geometrical characteristics are derived separately for the inline and staggered tube
arrangements. Flow is idealized as being normal to the tube bank on the outside.
Tubes are considered bare (without ns) in this section.
8.1.1
Inline Arrangement
The basic core geometry for an idealized single-pass crossow tubular exchanger with an
inline tube arrangement is shown in Fig. 8.1. The header (tubesheet) dimensions for this
tube bank are considered as L2 L3 such that the overhangs of X =2 are idealized on
each end in the L2 dimension and of Xt =2 on each end in the L3 dimension. Thus the core
Fundamentals of Heat Exchanger Design. Ramesh K. Shah and Duan P. Sekulic
Copyright 2003 John Wiley & Sons, Inc.
563
564
length for ow normal to the tube bank is L2 and the noow dimension is L3 . Thus the
geometrical characteristics to be derived are for an innite tube array.y
8.1.1.1 Tube Inside. Tubes have inside diameter di ; length between headers L1 ; total
length, including header plates, L1 2h ; and a total number of tubes Nt , where
Nt
L2 L3
Xt Xl
8:1
The geometrical characteristics of interest for analyses for this geometry are straightforward:
total heat transfer area A d i L1 Nt
total minimum free-flow area Ao di2 Nt
4
core frontal area Afr L2 L3
ratio of free flow to frontal area
=4di2 Nt =4di2
L2 L3
Xt X
hydraulic diameter Dh di
tube length for heat transfer L1
8:2
8:3
8:4
8:5
8:6
8:7
8:8
A
d L N
i 1 t
L1 L2 L3
Vtotal
8:9
y
An innite tube array means an array having no end eects from the thermal hydraulic performance point of
view.
565
The heat transfer area associated with the header plates on the tube side, represented
by a plane through line BC (the leftmost plane) in Fig. 8.1b, is generally neglected. Note
that dierent tube lengths are considered for heat transfer and pressure drop calculations. If there is a tube overhang beyond header plates or tubesheets, that length should
be added for pressure drop calculations in Eq. (8.8), but that added length will not
contribute to heat transfer from one uid to the other uid.
8.1.1.2 Tube Outside. The geometrical characteristics of the inline tube bank of Fig.
8.1 are summarized now. The total heat transfer area consists of the area associated
with the tube outside surface and that with the two header plates:
d 2
A do L1 Nt 2 L2 L3 o Nt
4
8:10
L3
Xt
8:11
8:12
Afr L1 L3
8:13
Hence,
Ao Xt do Nt0 Xt do
Afr
L3
Xt
Dh
4Ao L2
A
8:14
8:15
8:16
8:17
A do
2
V
pt
8:18
where A and V are given by Eqs. (8.10) and (8.17), respectively. The second equality in
Eq. (8.18) is for a unit cell (shown by dashed lines in Fig. 8.1) for an inline arrangement.
It must be emphasized that the foregoing denition of the hydraulic diameter, Eq.
(8.15), is used by Kays and London (1998) for tube banks. However, Zukauskas (1987)
and other investigators use the tube outside diameter as the characteristic dimension in
heat transfer and pressure drop correlations. In such correlations, the length L2 may not
566
be required and the tube bundle p is computed from Eq. (6.37) or (6.38) if the Euler
number Eu or the Hagen number Hg is known, or other appropriate correlations. So it is
essential to know that dierent researchers use dierent denitions of the characteristic
length and other appropriate geometrical characteristics, and one must nd out rst the
specic denition of the geometrical characteristics before using a particular correlation.
8.1.2
Staggered Arrangement
The geometrical characteristics are derived for the staggered arrangement of Fig. 8.2. In
this case, any tube is halfway (in the Xt direction) between the two neighboring tubes in
the next tube row, and the pattern of two transverse tube rows is repeated along the X
direction. However, if the transverse tube row pattern repeats after three tube rows along
the X direction, with the tubes in successive tube rows oset by Xt =3 distance (and thus
tubes in every fourth tube rows arranged identical), it is referred to as a three-row echelon
arrangement. An n-row echelon tube arrangement with n 3 is possible as well. For
geometry calculations for staggered tube arrangement, similar to the inline tube arrangement, we consider the overhangs X =2 on both sides of L2 , the overhangs Xt =2 on both
sides of L3 , and half tubes at each end in the alternate tube row to simulate an innite
tube array. The total number of tubes in such a tube bundle is given by Eq. (8.l). If the
half tubes are eliminated from the alternate intermediate tube rows, the number of tubes
in the rst row becomes L3 =Xt and in the second row, L3 =Xt 1. The total number of
tube rows then is L2 =X , and the total number of tubes
Nt
L3 L2 =X 1
L3
L2 =X 1
1
Xt
Xt
2
2
8:19
567
8.1.2.1 Tube Inside. Tubes have inside diameter di , length between headers L1 , total
length including header plates as L1 2h and the total number of tubes as Nt , as
expressed by Eq. (8.1). The geometrical characteristics are identical to those for the
inline arrangement given in Eqs. (8.2)(8.9).
8.1.2.2 Tube Outside. The total heat transfer area consists of the area associated with
the tube outside surface and that with the two header plates.
d 2
A do L1 Nt 2 L2 L3 o Nt
4
8:20
The minimum free ow area occurs either at a plane through AA or at a plane through
diagonals such as BB and BC of a unit cell of Fig. 8.2a. A unit cell for the analysis is
shown in Fig. 8.3. From Fig. 8.3,
2a Xt do
b
Xt
2
8:21
1=2
2
X2
do pt do
8:22
where the bracketed term is pt for 308, 608, and 458 tube layouts of Table 8.1. Now
dene the minimum of 2a and 2b as c:
c
2a
2b
if 2a < 2b
if 2b < 2a
8:23
8:24
Here the last term, Xt do L1 , corresponds to the free-ow area between the last tube
(at each end in the rst row) and the exchanger wall.
568
Ratio of
minimum free
ow area
to frontal
area, Ao =Afr
Longitudinal
tube pitch X
Transverse
tube pitch Xt
p
3 pt do
p
p
for t 3:732
do
3 pt
pt do
pt
2pt do
p
p
for t 3:732
do
3 pt
pt
2
p
3
pt
2
pt
308 Triangular
Staggered
Array
pt do
pt
pt
pt
908 Square
Inline
Array
TABLE 8.1 Nomenclature and Geometrical Properties of Tube Banks Common in Shell-and-Tube Exchangers
2pt do
p
p
for t 1:707
do
2 pt
p
2 pt do
p
p
for t 1:707
do
2 pt
p
pt
2
569
8:25
Ao
Afr
8:26
4Ao L2
A
8:27
8:28
8:29
A
V
8:30
V X Xt L1 pt sin 608pt L1
p
3 2
p L
2 t 1
8:31
and
o
A
d
p o 2
V
3 pt
8:32
For shell-and-tube exchangers, tube bundles with three specic staggered arrangements are commonly used, and are referred to as 308, 458, and 608 tube layouts, while
the tube bundle with the inline arrangement is referred to as the 908 tube layout.
The relationship between the tube pitch pt and transverse and longitudinal pitches
(Xt and X ) for these tube layouts is outlined in Table 8.1. Also summarized in this
table is a ratio of the minimum free-ow area to the frontal area for these tube layouts.
Note that Eq. (8.24) for Ao is valid for any staggered arrangement; its value in terms of
in Table 8.1 is specic for those specic tube layouts.
8.2
Geometrical characteristics are derived for two specic tube-n exchangers: circular
tubes having individual circular ns and circular tubes having at plain ns.
8.2.1
The basic core geometry of an idealized single-pass crossow exchanger is shown in Fig.
8.4 for an inline tube arrangement. The nned tubes could also be in the staggered
y
If the tube-side dimensionless pressure drop correlation is based on an average Euler number per tube row, there
is no need to use or dene L2 for the p calculations.
570
arrangement similar to those in Fig. 8.2. The total number of tubes in this exchanger is
given by Eq. (8.1) for the inline or staggered arrangement.
8.2.1.1 Tube Inside. The core geometrical characteristics for the tube side applicable
for both inline and staggered arrangements are identical to those of Eqs. (8.2)(8.9).
8.2.1.2 Geometrical Characteristics for Tube Outside. The determination of geometrical characteristics for tube outside is somewhat complicated due to the presence of
circular ns. It is idealized that the root of the circular n has an eective diameter
do and the n tip has a diameter de . Depending on the manufacturing techniques, do
may be the tube outside diameter or tube outside diameter plus the thickness of two
collars made from the n hole material for spacing ns evenly.
The total heat transfer area A consists of the area associated with the exposed tubes
and header plates (primary surface area) Ap , and ns (secondary surface area) Af . The
primary surface area is the same as that given in Eq. (8.10) or (8.20) minus the area
blocked by the ns:
d 2
Ap do L1 Nf L1 Nt 2 L2 L3 o Nt
8:33
4
where is the n thickness and Nf is the number of ns per unit length. The n surface
area is given by
"
#
2de2 do2
Af
8:34
de Nf L1 Nt
4
The factor 2 in the rst term on the right-hand side is for two sides of a n. The total heat
transfer surface area is then
A Ap A f
8:35
571
The minimum free-ow area for the inline arrangement is that area for a tube bank
[Eq. (8.12)] minus the area blocked by the ns:
Ao Xt do L1 de do Nf L1
L3
Xt
8:36
For the staggered tube arrangement, the minimum free-ow area could occur either
through the front row or through diagonals similar to those of Fig. 8.3. A unit cell is
shown in Fig. 8.5. The dimensions 2a and b as calculated by Eqs. (8.21) and (8.22) must
be modied for the area blocked by the circular ns. We refer to these modied dimensions as 2a 0 and b 0 ; note, they cannot be depicted in Fig. 8.5. They are given by
b0
2a 0 Xt do de do Nf
8:37
1=2
2
Xt
X2
do de do Nf pt do de do Nf
2
8:38
where the bracketed term is pt for 308, 608 and 458 tube layouts of Table 8.1. Now
dene c 0 such that
(
0
2a 0
if
2a 0 < 2b 0
2b0
if
2b0 < 2a 0
8:39
L3
1 c 0 Xt do de do Nf L1
Xt
8:40
FIGURE 8.5 Unit cell of a staggered nned-tube arrangement. (From Shah, 1985.)
572
Conceptually, the fabrication of this exchanger is simple and amenable to mass production techniques. Proper holes are made into plain sheet metal (ns) of proper dimensions
L2 L3 . Tubes are then slipped into the properly stacked ns. Tubes are either
expanded mechanically or are brazed. The basic core geometry of an idealized singlepass crossow exchanger is shown in Fig. 8.6 for a staggered tube arrangement. The
tubes could also be in an inline arrangement. The total number of tubes in this exchanger
is given by Eq. (8.1) for the inline or staggered arrangement.
8.2.2.1 Tube Inside. The geometrical characteristics for the tube inside are the same as
those for the preceding geometries, such as those given by Eqs. (8.2)(8.9).
8.2.2.2 Tube Outside. The geometrical characteristics for the ow normal to the tubes
and ns are similar to those for the circular ns on circular tubes except for some
modications due to the at n geometry. The total heat transfer area consists of the
area associated with the exposed tubes and header plates (primary surface area) and
the ns (secondary surface area). The primary surface area is the same as that given by
Eq. (8.10) or (8.20) minus the area blocked by the ns:
d 2
Ap do L1 Nf L1 Nt 2 L2 L3 o Nt
4
8:41
573
8:42
edges area
8:43
The minimum free-ow area for an inline arrangement is the area for a tube bank
[Eq. (8.12)] minus the area blocked by the ns:
Ao Xt do L1 Xt do Nf L1
L3
Xt
8:44
For the staggered tube arrangement, the minimum free-ow area could occur either
through the front row or through the diagonals similar to those of Fig. 8.3 or 8.5. The
dimensions 2a and b as calculated by Eqs. (8.21) and (8.22) must be modied for the area
blocked by at plain ns. These modied dimensions, referred to as 2a 00 and b 00 , are given
by
b 00
2a 00 Xt do Xt do Nf
8:45
1=2
2
Xt
X2
do Xt do Nf pt do Xt do Nf
2
8:46
where the bracketed term is pt for 308, 608, and 458 tube layouts of Table 8.1. Now
dene c 00 such that
( 00
2a
if 2a 00 < 2b 00
8:47
00
c
00
00
00
if 2b < 2a
8:48
2b
The minimum free ow area is then given by
Ao
L3
1 c 00 Xt do Xt do Nf L1
Xt
8:49
Other geometrical characteristics of interest are identical to those given by Eqs. (8.25)
(8.30) with appropriate values of Ao and A from the equations above.
For the determination of entrance (sudden contraction) and exit (sudden expansion)
pressure losses, the area contraction and expansion ratio 0 is needed at the leading and
trailing n edges. It is given by
0
L3 L1 L3 Nf L1
L3 L1
8:50
574
8.2.3
In Sections 8.2.1 and 8.2.2 we showed how to evaluate A (Ap and Af ), Ao , Afr , and Dh for
the tube-n surfaces on both uid sides. Now we present how to evaluate and , and
the relationship between them, through denition of the hydraulic diameter for both
individually nned tubes and at ns on round or at tubes. For these cases, the surface
area density is the ratio of the total transfer surface area A on one side of the exchanger
to the total volume V of the exchanger as follows:
1
A1
V
2
A2
V
8:51
Similarly, the ratio of minimum free-ow area to frontal area for each uid side is given
by
1
Ao;1
Afr;1
2
Ao;2
Afr;2
8:52
41
1
Dh;2
42
2
8:53
Note that the surface area density has no meaning for tube-n exchangers since the
volumes occupied by each uid side are not mutually independent. Hence, we refer to s
only for tube-n exchangers.
8.3
A large number of n geometries are available for plate-n heat exchangers. Some of the
n geometries are shown in Fig. 1.29. It is beyond the scope of this section to present
derivations for geometrical characteristics for all these corrugated ns. As an illustration, geometrical characteristics are derived below for one side of a plate-n heat
exchanger having oset strip ns in Section 8.3.1 and corrugated louver ns in Section
8.3.2. In Section 8.3.3, a relationship is presented relating the s and s of the plate-n
surfaces.
8.3.1
575
FIGURE 8.7 (a) Plate-n exchanger; (b) oset strip n geometry; (c) small section of an idealized
oset strip n geometry; the unit cell is bounded by b1 and pf . (From Shah, 1985.)
(Fig. 8.7c). The primary surface area is then the sum of components 1, 3, and 4 minus
component 2.
These four components of the primary surface area are now derived:
total plate area Fig: 8:7a 2L1 L2 Np
8:54
8:55
8:56
576
8:57
Here, we have considered the total number of passages on the uid 2 side as one more
than that on the uid 1 side. The total primary surface area on the uid 1 side, from Eqs.
(8.54)(8.57), is then
Ap;1 2L1 L2 Np 2Lf nf 2b1 L1 Np 2b2 2w Np 1L2
8:58
The three components of the secondary (n) area are (refer to Fig. 8.7c)
fin height area 2b1 Lf nf
8:59
The oset strip ns have leading and trailing edges at each strip, which contribute to the
heat transfer area. The area associated with the edges is divided into two components:
the area associated with the edge height and the area associated with the edge width.
Because of the overlap of edge widths between two oset ns in the ow direction, only
half of the edge width area is available at the front and half at the back edge of each n,
except for the edges at the front and back face of the core (refer to Fig. 8.7c):
fin edge height area 2b1 noff nf
8:60
where noff is the total number of oset strips in the ow (L1 ) direction:
fin edge width area pf noff 1nf 2pf nf
8:61
where the last term of Eq. (8.61) represents the n edge width area at the front and back
face of the core. Thus, from Eqs. (8.59) through (8.61), the total secondary area on uid 1
side is
Af ;1 2b1 Lf nf 2b1 noff nf pf noff 1nf 2pf nf
8:62
8:63
where Ap;1 and Af ;1 are given by Eqs. (8.58) and (8.62), respectively.
The free-ow area on uid 1 side is given by the frontal area on uid 1 side minus the
area blocked by the ns at the entrance of the core on that side.
Ao;1 b1 L2 Np b1 pf nf
8:64
8:65
4Ao L
A
8:66
Ao
Afr
8:67
Dh
A
Vp
577
8:68
8:69
8:70
8:71
8:72
where s pf and h 0 b1 . The hydraulic diameter Dh 4Ao s =A for the unit cell
is then given by
Dh
4Ao; cell s
4sh 0 s
Acell
2s h 0 s h 0 s
8:73
b1
2
8:74
The perimeter of the n at a cross section is 2s 2 and the cross section area is s :
Thus the value of m for the oset strip n f of Eq. (4.146) is
2h
1=2 b1
1
2
kf
s
m
8:75
Example 8.1 Determine the air-side core geometrical characteristics needed for data
reduction of core 105 of Example 7.1. The following are the primary measurements, with
most of them representing an arithmetic average of at least ve individual measurements,
for each item.
Airow length L1 0:0532 m
Core width L2 0:2148 m
578
SOLUTION
Problem Data and Schematic: All pertinent geometrical data for the air side of a singlepass crossow heat exchanger are provided above. See Fig. 8.7 as a schematic for this
problem.
Determine: The air-side geometric characteristics, such as n pitch pf , total number of
ns nf , number of n osets noff , total primary area Ap , total secondary (n) area Af ,
total heat transfer area A, surface area density , minimum free-ow area Ao , frontal area
Afr , heat conduction area for wall thermal resistance Aw , and hydraulic diameter Dh .
Assumptions: The ideal n geometry for the air side is shown in Fig. 8.7c. It is assumed
that the total number of passages on the steam side is one more than that on the air side.
Analysis: Let us rst calculate the n pitch pf , total number of ns in the core nf , and
number of n osets noff before we use Eqs. (8.54)(8.70) to determine the necessary
geometrical characteristics.
fin pitch pf
core width L2
0:2148 m
0:001053 m 1:05 mm
fins
number of passages 204 46 9384 Ans:
total number of fins nf
passage
number of fin offsets noff
18
fin offset length s 0:00282 m
Ans:
Ans:
579
Ans:
Ans:
Af 1:8040 m2
0:636
A
2:8371 m2
A
2:8371 m2
2936:2 m2 =m3 Ans:
Dh
Ans:
Ao 0:0162 m2
0:309
Afr 0:0525 m2
A
2:8371 m2
Ans:
Ans:
We compute the hydraulic diameter based on the unit cell approach [i.e., using Eqs.
(8.71)(8.73)]:
580
1:21 mm 0:00121 m
A
Acell
16:023 mm2
cell
Discussion and Comments: Calculations for the geometrical characteristics are straightforward once the specic model for geometry is chosen. This example shows how the
geometric characteristics are computed for surfaces reported by Kays and London
(1998). Note that use of the cell approach for this test core provides the value of the
hydraulic diameter, which is in excellent agreement with that which includes all secondary eects; however, this may not always be the case.
8.3.2
8:76
8:77
8:78
p2f 1=2
8:79
8:80
Ao; cell
Afr; cell
8:81
4Ao; cell Lf
Acell
hydraulic diameter Dh
8:82
581
8:83
Acell
Vcell
8:84
8:85
For a full core, Dh and are the same as above. To calculate the free-ow area Ao , the
frontal area Afr , and the primary and n surface areas Ap and Af for the full core, rst
compute the number of ns in the core as follows for known core width W (n passage
width, corresponds to L2 in Fig. 8.7), the number of n pitches or ns in the core width
( W=pf ), and the number of nned passages Np . In this case, the total number of ns or
unit cells in the core are given by
nf N p
W
pf
8:86
Af nf Af ; cell
A nf Acell
Ap nf Ap; cell
Aw np 1Aw; cell
8:87
Heat exchanger volume Vp between plates for this geometry can also be calculated using
Eq. (8.69).
For the n eciency of the louver n geometry, like the oset strip n geometry, it is
assumed that the heat ow from both sides (of plates or tubes representing the primary
surface) is uniform and the adiabatic plane occurs at the middle point of the plate
spacing. Hence,
12 b2 p2f 1=2
8:88
The perimeter of the n at a cross section is 2Lf 2, and the cross section for
heat conduction through the n is Lf . Thus the value of m for the louver n f of
Eq. (4.146) is
1=2
2h
1 2
2 1=2
1
b pf
m
kf
Lf
2
8:89
In practical applications it is assumed that the contact resistance between the n and
the primary surface is zero. In a real design this assumption is never fullled, but the
inuence of this eect is usually small and lumped into the experimental j factor of a test
core (assuming that the same problem exists for the test core). Although the n eciency
reduction due to the poor thermal contact may be signicant in some local areas where
the brazing is not good, the overall eect on the heat exchanger performance may be
small (Zhao and Sakulic, 2001). This inuential factor has to be considered independently for each particular manufacturing process.
582
The geometry used by Chang and Wang (1997) for the j and f correlations of Eqs.
(7.127) and (7.129) neglected the inclination of the louver ns with respect to the primary
surface. Hence, Eqs. (8.77) and (8.79) were modied for one term as follows:
b2 p2f 1=2 b
8:90
SOLUTION
Problem Data and Schematic: All pertinent geometrical data for the air side of a at tube
and corrugated multilouver n crossow exchanger are provided above. The detailed
geometry is shown in Fig. 7.29.
Determine: The air-side geometrical characteristics: the hydraulic diameter Dh , the ratio
of free-ow area to frontal area , and the surface area density ; and for the full core, the
minimum free-ow area, frontal area, primary surface area, n surface area, total heat
transfer surface area, and heat conduction area for wall thermal resistance.
Assumptions: The ideal n geometry for the air side is shown in Fig. 7.29. The inuence of
braze llets is neglected; note that the at tube and corrugated multilouver n crossow
exchangers are brazed.
Analysis: We rst calculate Dh , , and for the unit cell and then compute the core
geometrical parameters using this information.
Primary tube surface area Ap; cell 2Wt pf 2pf Ht
2 28:0 mm 1:00 0:10 mm 2 1:00 mm
2:00 mm 54:40 mm2
Corrugated fin length b2 p2f 1=2 6:002 1:002 1=2 mm 0:10 mm
5:98 mm
Fin surface area Af ; cell 2Lf b2 p2f 1=2 2 30:0 mm 5:98 mm 358:80 mm2
Total heat transfer surface area Acell Ap; cell Af ; cell 54:40 mm2 358:80 mm2
413:20 mm2
583
0:675
Afr; cell
8:00 mm2
Ans:
1:569 mm
Acell
413:20 mm2
Ans:
Ans:
Wall conduction area per unit cell Aw; cell Wt W 28:0 mm 225:0 mm
6300 mm2
The total number of ns in the exchanger using Eq. (8.86) is:
nf Np
W
225:0 mm
20
4500
pf
1:00 mm
Now all areas related to full core are computed from Eq. (8.87) as follows:
Core primary surface area Ap nf Ap; cell 4500 54:40 mm2 244;800 mm2
0:2448 m2
Ans:
1:6146 m2
Ans:
2
Ans:
Total wall conduction area Aw Np 1Aw; cell 19 6300 mm2 119;700 mm2
0:1197 m2
Ans:
Although industry practice is to ignore the edge area of the cut louvers, let us compute
the approximate eect on the surface area. We will assume that the louver length is 80%
of the louver corrugated n length computed above as 5.98 mm, and the total number of
louvers in the two banks for a n is 12. In this case,
louver edge area 0:8 5:98 mm 0:1 mm 14 6:698 mm2
Since the total surface area for the cell is 413.20 mm2 , the edge area of 6.698 mm2 is 1.6%.
Hence our idealization of neglecting the edge area is good.
584
Discussion and Comments: The calculation procedure for computing geometrical characteristics of corrugated multilouver n is straightforward. Here we adopted the cell
approach as a starting point since it is accurate for the simplied geometry considered.
It becomes clear why industry neglects the surface area of cut louver edges.
8.3.3
In Sections 8.3.1 and 8.3.2 we showed how to evaluate A, Ao , Afr , Dh , , Vp , , and for
the oset strip ns and louver ns on one side of a plate-n exchanger. Now we show the
general relationship between , , and for a plate-n surface valid for any corrugated
n geometry.
If L1 and L2 are the ow lengths and Np and Np 1 are the number of ow passages
on the uid 1 and 2 sides, respectively, the volume between plates on each side is
Vp;1 L1 L2 b1 Np
Vp;2 L1 L2 b2 Np 1
8:91
After including the volume occupied by the plates or parting sheets, the total volume of
the exchanger is
V b1 Np b2 Np 1 2w Np 1L1 L2
8:92
A2 2 Vp;2
8:93
Here 1 and 2 are the surface area densities on each uid side based on unit volume
between plates. The ratio of minimum free-ow area to frontal area on the uid 1 side, 1 ,
is as follows after introducing the denitions of the hydraulic diameter and expressions
for Vp;1 and V from Eqs. (8.91) and (8.92):
1
Ao;1
Ao;1 L1 A1 Dh;1 =4 Vp;1 1 Dh;1 =4
Afr;1 Afr;1 L1
V
V
L1 L2 b1 Np 1 Dh;1 =4
b1 Np b2 Np 1 2w Np 1L1 L2
b1 Np 1 Dh;1 =4
b1 1 Dh;1 =4
b1 Np b2 Np 1 2w Np 1 b1 b2 2w
for Np 1
8:94
Here w is the thickness of the parting sheets. The last approximate equality is for the case
when Np 1 or the number of passages on each uid side are the same. Similarly,
2
b2 Np 12 Dh;2 =4
b2 2 Dh;2 =4
b1 Np b2 Np 1 2w Np 1 b1 b2 2w
8:95
The heat transfer surface area on one uid side divided by the total volume V of the
exchanger, designated as 1 , is obtained as follows by using the denitions of Dh;1
585
4Ao;1 =Dh;1
A1
A1
A =L
4
b1 1
1 1
1
V
L1 Afr;1
Afr;1
Afr;1
Dh;1 b1 b2 2w
8:96
Similarly,
2
A2
b2 2
V
b1 b2 2w
8:97
These relationships between and will be useful in sizing a plate-n heat exchanger, as
will be shown in Sections 9.2.2.2 and 9.2.2.3.
8.4
Some of the continuous cylindrical passage geometries for the rotary regenerator are
shown in Fig. 1.43. Geometrical characteristics of triangular passages (Fig. 1.43b) are
derived below as an illustration.
8.4.1
The ideal triangular passage model is shown in Fig. 8.8 along with the nomenclature to be
used in the derivation. Note that most of the symbols for this subsection are local. In
addition to the basic dimensions shown in Fig. 8.8, the total number of cells within the
unit core face area is determined from the enlarged photographs of the core face and is
designated as nc with units of m2 or ft2 . Hence
nc
number of cells
unit area
1
face area
nc
unit cell
8:98
Note that the face or frontal area of one cell then is the reciprocal of nc as designated in
the second equation above. For the ideal model of Fig. 8.8a,
face area for one cell
face area
1
dc
unit cell
nc
8:99
Cell bounds
x
y
c
a
(a)
(b)
FIGURE 8.8 (a) Triangular passage model; (b) triangular half-cell model. (From Young, 1969.)
586
8:100
;
cos
cos
c
d c2 1=2
2
8:101
so that
x
1=2
2
d
1
c
8:102
d
1
Acell 2L 1
c
8:103
The porosity is the ratio of free-ow to the frontal area for a cell:
y
2 tan
d 2
nc 2 2
1=nc
1=nc
c c
8:104
where the last equality comes from Eq. (8.99). The heat transfer surface area density is
given by
Acell 2 xL
2nc x
Vcell
1=nc L
8:105
Substituting the value of x from Eq. (8.102) and from Eq. (8.104) into Eq. (8.105) yields
2nc
=nc
d=c
1=2
1
2
1=2
d
1
c
8:106
4
Ao Afr
8:107
void volume
solid volume
M =V
1
1 w
total volume
total volume
w
8:108
587
TABLE 8.2 Surface Geometrical Properties for Some Idealized Flow Passages used in Compact
Regenerators
Geometry
Hydraulic
Diameter
Dh (m)
0.370.39
61
b
2b
31
1
b 2
b2
b 2
4b
b 2
2
p
3 b 2
b2
b 2
4b
b 2
2
p
3 b 2
b2
p
2 3 b 2
2b
p
3 b 2
1
b* b
b2 *
b* b
21 *b
b* b
2b*
1 *
p
4 3
2b 22
4b2
2b 32
24b
2b 32
2b
3
Cell Density
Nc (cells/m2 )
Porosity
The foregoing geometrical characteristics were used for ceramic regenerator cores of
London et al. (1970). For some simple regenerator surfaces, the geometrical characteristics are summarized in Table 8.2 for completeness.
8.5
In this section, geometrical characteristics required for the rating and sizing of shell-andtube exchangers with single segmental baes are derived.
8.5.1
Tube Count
588
b
B
c
ctl
Ds 2c
Dotl
wp
Ds
FIGURE 8.9 Nomenclature for basic bae geometry relations for a single segmental exchanger
(from Taborek, 1998).
tubes-in-window design or impingement plates, and the start of the drilling pattern
relative to the shell inside diameter and pass dividers. For a xed tubesheet design, the
outermost tubes can be close to the shell inside diameter, or the diameter of the outer tube
limit, Dotl (see Fig. 8.9), can be the largest followed by that for a split-ring oating head S
(see Fig. 10.7), and Dotl being the smallest for a pull-through head T (see Fig. 10.5). For a
U-tube bundle, some tubes are also lost near the centerline of the U-tube pattern because
of the manufacturing limitations on the tube bend radius. Because of many variables
involved, it is dicult to determine accurately the total number of tubes in an exchanger
except for a direct count. As an alternative, the tube count may be determined approximately using published tabular values, such as those of Bell (1988), among others. For a
specied diameter of the circle through the centers of the outermost tubes, Dctl , the eect
of the tube bundle type on the total number of tubes Nt is eliminated. Taborek (1998)
provides an approximate expression for the tube count as follows in terms of Dctl :
8
>
=4D2ctl
>
>
1
<
Ct p2t
Nt
2
>
=4Dctl
>
>
1
:
Ct p2t
8:109
where
8
0
>
>
>
>
< ctl sin ctl
2
2
>
>
sin ctl
>
ctl
>2
:
2
2
no impingement plate
impingement plate on one side
8:110
n is given in Fig. 8.10, Ct 0:866 for 308 and 608 tube layouts and Ct 1:00 for 458
and 908 tube layout. The angle ctl in Eq. (8.110) is in radians and is given by Eq. (8.114).
589
0.30
0.34 at Ntp = 4
0.28
0.26
25
200
0.20
0.18
30
0
Correction factor, n
0.22
ct =
0.24
0.16
35
0.14
0.12
0
0
40
600
800
1000
0.10
0.08
1500
2000
0.06
0.04
0.02
0
1
3
4
5
6
Number of tube passes, Ntp
FIGURE 8.10 The correction factor n for estimation of number of tubes for tube bundles with
the number of tube passses np 2 8: (From Taborek, 1998.)
The accuracy in predicting tube count using Eq. (8.109) is 5% for single-tube-pass
exchangers provided that large tubes are not used in relatively small shells. For multiple
tube passes, the accuracy is approximately 10% for Ds < 400 mm and 5% for larger shell
diameters.
8.5.2
The single segmental E shell exchanger is one of the most common exchangers used in the
process, petroleum, and power industries. The geometrical information needed for rating
such an exchanger by the BellDelaware method (discussed in Section 9.5.1) will be
derived here. The original geometry of Bell (1988) is modied based on the suggestions
made by Taborek (1998). As shown in Fig. 8.11a, b, and c, the shell side of an E shell
exchanger can be divided into three sections: internal crossow, window, and entrance
and exit sections, respectively. We calculate the necessary geometrical characteristics for
window and internal crossow sections next. The eect of larger bae spacings at the
entrance and exit sections will be included by a correction factor when we discuss the
thermal design of shell-and-tube heat exchangers in Section 9.5. In addition, in this
section we compute various bypass and leakage ow areas needed for the
thermal design of shell-and-tube heat exchangers.
8.5.2.1 Window Section. As shown in Fig. 8.9, the gross window area (i.e., without
tubes in the window) or the area of a segment ABC corresponding to the window
section is
sin b
D2
2
Afr;w D2s b
s b 1 c sin b
8:111
2
2
4 2
Ds
2
4
590
FIGURE 8.11 TEMA E shell exchanger: (a) internal cross ow sections, (b) window sections;
(c) entrance and exit sections. (From Taborek, 1998.)
where b is the angle in radians between two radii intersected at the inside shell wall with
the bae cut and is given by
2
b 2 cos1 1 c
Ds
8:112
To calculate the number of tubes in the window zone, we consider the tube eld
uniform within the shell cross section. This idealization is violated when there are tube
pass lanes (in a multipass exchanger) or when tubes are removed due to impingement
plates in the nozzle entry area. We will ignore this fact and assume that it causes
negligible second-order eects. Then the fraction Fw of the number of tubes in one
window section encircled by the centerline of the outer tube row (Fig. 8.9) is
Fw
8:113
591
where ctl is the angle in radians between the bae cut and two radii of a circle through
the centers of the outermost tubes (see Fig. 8.9) as follows:
Ds 2c
ctl 2 cos1
8:114
Dctl
where Dctl Dotl do . Consequently, the number of tubes in the window section is
obtained from
Nt;w Fw Nt
8:115
and the area occupied by tubes in the window section is
Afr;t do2 Nt;w do2 Fw Nt
4
4
8:116
8:117
where the right-hand-side terms are evaluated from Eqs. (8.111) and (8.116).
By application of the conventional denition, the hydraulic diameter of the window
section of a segmental bae is
Dh;w
4Ao;w
4Ao;w
P
do Nt;w Ds b =2
8:118
where b is given by Eq. (8.112) and P is the wetted perimeter (of all tubes and the shell
within the window region); the wetted perimeter of the bae edge is usually neglected.
This Dh;w is used for shell-side pressure drop calculations in laminar ow Re < 100.
The nal geometrical input required for the window section is the eective number of
tube rows in crossow needed for the heat transfer and pressure drop correlations. The
uid in the window section eectively makes a 1808 turn while owing from one internal
crossow section to another. In the window region, the uid has both crossow and
longitudinal ow components of varying magnitudes as a function of the position. Based
on the visual and experimental evidence (Bell, 1963), the eective distance of penetration
for crossow in the tube eld in the bae window is about 0.4c;eff in the region AB in
Fig. 8.11b while owing away from the internal crossow section, and about 0.4c;eff in
the region BC in Fig. 8.11b while owing toward the internal crossow section. Here c;eff
is the distance between the bae cut and Dctl (Fig. 8.9). Hence, the number of eective
tube rows in crossow in the window section is
Nr;cw
0:8c;eff 0:8
1 D Dctl
X
X c 2 s
8:119
8.5.2.2 Crossow Section. The fraction Fc of the total number of tubes in the crossow section is found from
sin ctl
8:120
Fc 1 2Fw 1 ctl
where the expression for Fw is obtained from Eq. (8.113). The number of tube rows Nr;cc
crossed during ow through one crossow section between bae tips may be obtained
from a drawing or direct count or may be estimated from
Nr;cc
Ds 2c
X
8:121
where X is the longitudinal tube pitch summarized in Table 8.1 for various tube layouts.
592
The crossow area at or near the shell centerline for one crossow section may be
estimated from
D
Ao;cr Ds Dotl ctl Xt do Lb;c
Xt
8:122
for 308 and 908 tube layout bundles. Here Dctl =Xt denotes the number of Xt do Lb;c
free-ow area in the given tube row. This equation is also valid for a 458 tube bundle
having pt =do 1:707 and for a 608 tube bundle having pt =do 3:732: For 458 and 608
tube bundles having pt =do lower than those indicated in the preceding line, the minimum
free-ow area occurs in the diagonal spaces, and hence the term Xt do in Eq. (8.122)
should be replaced by 2pt do ; or
D
Ao;cr Ds Dotl 2 ctl pt do Lb;c
Xt
8:123
for 458 and 608 tube bundles. If the tubes have circular ns, the area blocked by the ns
should be taken into account as in Eq. (8.36). Hence, Eq. (8.122) modies to
Ao;cr
Ds Dotl
Dctl
Xt do de do Nf Lb;c
Xt
8:124
which is valid for 308 and 908 tube bundles, 458 tube bundles having pt =do 1:707, and
608 tube bundles having pt =do 3:732. For circular nned tube bundles having 458 tube
layout and pt =do 1:707 or 608 tube layout and pt =do 3:732, Eq. (8.124) modies to
Ao;cr
D
Ds Dotl 2 ctl pt do de do Nf Lb;c
Xt
8:125
The number of baes Nb is required to compute the total number of crossow and
window sections. It should be determined from the drawings or a direct count.
Otherwise, compute it from the geometry of Fig. 8.11c as
Nb
L Lb;i Lb;o
1
Lb;c
8:126
where Lb;c is the central bae spacing, and Lb;i and Lb;o are the bae spacings in the inlet
and outlet regions.
8.5.3
Flow area available for bypass streams C and F (see Fig. 4.19) associated with one
crossow section, normalized with respect to the crossow open area at or near the
shell centerline, is
Fbp
Ao;cr
Ao;cr
8:127
593
where Np is the number of pass divider lanes through the tube eld that are parallel to the
crossow stream B, wp is the width of the pass divider lane (Fig. 8.9), and Ao;cr is given by
Eqs. (8.122)(8.125). Since the tube eld is on both sides of the pass divider bypass lane,
the F stream is more eective in terms of heat transfer than is the C stream. Hence, the
eective bypass lane width is considered as 0.5wp , as indicated in Eq. (8.127).
Now let us determine the tube-to-bae leakage area Ao;tb for one bae. The total
number of tubes associated with one bae is
Nt;b Nt 1 Fw Nt
1 Fc
2
8:128
where the value of Fw was substituted from the rst equality of Eq. (8.120). If the
diametral clearance (the dierence between the bae hole diameter d1 and the
tube outside diameter do ) is tb d1 do ), the total tube-to-bae leakage area for
one bae is
Ao;tb
d N 1 Fw
d tb 2 do2 Nt 1 Fw o tb t
4 o
2
8:129
Finally, the shell-to-bae leakage area for one bae is associated with the gap
between the shell inside diameter and the bae. Note that this gap exists only within
the sector ABC in Fig. 8.12. The shell-to-bae leakage area
Ao;sb Ds
sb
1 b
2
2
8:130
c
b
sb
C
Ds
FIGURE 8.12 Single segmental bae geometry showing shell-to-bae diametral clearance sb .
594
SOLUTION
Problem Data and Schematic: All pertinent geometrical variables for the shellside are
provided above. The detailed geometry is shown in Figs. 8.9, 8.11 and 8.12.
Determine: The shell-side geometrical characteristics: bae cut angle, fraction of total
number of tubes in the window section, the area for ow through the window section,
number of eective crossow rows in each window, fraction of total number of tubes in
crossow, the number of tube rows in one crossow section, crossow area at or near
centerline, number of baes, fraction of crossow area available for ow bypass, tubeto-bae leakage area for one bae, and shell-to-bae leakage area for one bae.
Assumptions: The shell-and-tube heat exchanger is assumed to have the ideal geometrical
characteristics summarized in Section 8.5.
Analysis: As outlined in the text, we compute geometrical characteristics for the window
section, crossow section, and bypass and leakage ow areas.
Window Section. Let us start the calculations with computing the angle b from Eq.
(8.112):
!
2c
2 86:7 103 m
1
2 cos
b 2 cos
1
1
2:131 rad 122o
Ds
0:336 m
1
0:01813 m2
1
sin
4
2
2
0:336 m
Afr;w
595
In order to calculate the fraction Fw of total tubes in the window section, rst compute
the bae cut angle, using Eq. (8.114), as
ctl 2 cos1
Ds 2c
0:336 m 2 86:7 103 m
2 cos1
2:004 rad 115o
Dctl
0:302 m
where Dctl Dotl do 0:321 m 19:0 103 m 0:302 m. Now the fraction Fw of
total tubes in the window section is given by Eq. (8.113) as
Fw
0:1747
2
2
2
2
Consequently, the number of tubes in the window section, from Eq. (8.115), is
Nt;w Fw Nt 0:1747 102 17:8
The area occupied by tubes in the window section, Eq. (8.116), is
Afr;t
2
d F N 0:0190 m2 0:1747 102 0:00505 m2
4 o w t 4
The net ow area in one window section is then, from Eq. (8.117),
Ao;w Afr;w Afr;t 0:01813 m2 0:00505 m2 0:01308 m2
The hydraulic diameter for the window section is given by Eq. (8.118) as
Dh;w
4Ao;w
do Nt;w Ds b =2
4 0:01308 m2
0:03683 m
0:0190 m 17:8 0:336 m 2:131=2
Finally, the number of eective tube rows in crossow in each window is computed using
Eq. (8.119) as
Nr;cw
0:8
1
c Ds Dctl
X
2
0:8
86:7 103 m 1=20:336 m 0:302 m 3:15 3
3
17:7 10 m
Crossow Section. The fraction Fc of the total number of tubes in the crossow section is
calculated from Eq. (8.120) as
Fc 1 2Fw 1 2 0:1747 0:6506
596
Next calculate the number of tube rows Nr;cc crossed during ow through one crossow
section between the bae tips [Eq. (8.121)] as
Nr;cc
X
17:7 103 m
The crossow area for the 458 tube layout bundle with plain tubes at or near the shell
centerline for one crossow section can be calculated, using Eq. (8.123), as
D
Ao;cr Lb;c Ds Dotl 2 ctl pt do
Xt
0:302 m
0:279 m 0:336 m 0:321 m 2
0:0250 m 0:0190 m
0:0354 m
0:03275 m2
Now, compute the number of baes from Eq. (8.126) as
Nb
L Lb;i Lb;o
4:3 m 0:318 m 0:318 m
1
1 14:13 14
0:279 m
Lb;c
Bypass and Leakage Flow Areas. To calculate the fraction of crossow area available for
ow bypass, Fbp [Eq. (8.127)], we rst have to calculate the magnitude of crossow area
for ow bypass:
Ao;bp Lb;c Ds Dotl 0:5Np wp 0:279 m 0:336 m 0:321 m 0:5 2 0:0190 m
0:00949 m2
Consequently,
Fbp
Ao;bp 0:00949 m2
0:2898
Ao;cr 0:03275 m2
2
2
0:001995 m2
Finally, the shell-to-bae leakage area for one bae [Eq. (8.130)] is
Ao;sb Ds
sb
0:002946 m
2:131
1 b 0:336 m
1
0:001027 m2
2
2
2
2
This concludes all geometrical characteristics needed for the thermal design/rating of a
shell-and-tube heat exchanger using the BellDelaware method.
597
Discussion and Comments: The calculation procedure for computing geometrical characteristics of a shell-and-tube heat exchanger is tedious but straightforward. The same
geometry has been utilized in the analysis of Example 9.4.
8.6
A large number of plate corrugation patterns have been developed worldwide. The
chevron plate geometry (see Fig. 1.18b and c) is the most common in use today. We
outline below the geometrical characteristics of the chevron plate gasketed PHE as an
example by evaluating the actual heat transfer surface area due to corrugations.
However, one of the common practices in industry is to ignore the eect of corrugations
altogether and treat the chevron plates as if they were plain (uncorrugated) plates.
A plate with nomenclature used in the following geometrical derivations is shown in
Fig. 7.28. The chevron corrugations increase surface area over the plain (uncorrugated)
plate of the same outer (overall) dimensions. The ratio of the developed (actual) surface
area of a chevron plate to the projected (for a plain or at plate) is given by
q
p
1=2
2
1
1 1 2 *2
E* 1 1 X 2 4 1 X 2 =2
8:131
6
where * 2a=, E* is the elliptical integral given in Table 7.4 with the formula for
the elliptical duct, [i.e., Em; and X 2a=. While the expression with the rst
equality is exact, Martin (1996) used the last approximate formula in Eq. (8.131) using
a three-point integration method. Thus the heat transfer surface area on one uid side of
a PHE is given by
A 2WLh 2aLh Np 2WLh Np
since
aW
8:132
where W (width of the plate between gaskets) and Lh (length of the plate for heat
transfer) are dened in Fig. 7.28 and Np is the number of channels (passages) on the
uid side considered.
The free-ow area on one uid side of a PHE is given by
Ao 2aWNp
8:133
4Ao Lh 8aWNp Lh 4a
A
2WLh Np
8:134
Another commonly used set of denitions for the heat transfer surface area, free-ow
area, and characteristic dimension is based on the projected surface area (i.e., as if there
were no corrugations). In that case,
A 2WLh Np
Ao 2aWNp
De 4a
8:135
8:136
598
The typical range of is from 1.15 to 1.25 with 1:22 for =a 2: This can
approach to 2 when =a is as small as 2.46.
The total number of plates Nt in a PHE is related to the number of passes np and the
number of channels per pass as Nc;p as follows.
Nt np Nc;p 1 np Nc;p 2 1
8:137
where the subscripts 1 and 2 refer to uids 1 and 2. The number of thermal plates in this
PHE is then Nt 2. The overall height of the corrugation, 2a p (where p is the plate
thickness) in Fig. 7.28, represents the thickness of a fully compressed gasket since
the plate corrugations are in metallic contact. It can be determined as the compressed
plate pack length Lpack (see Section 1.5.2.1 for the denition) divided by the total number
of plates Nt .
The plate length for heat transfer, Lh , and that for pressure drop, Lp , are related as
follows where Dp is the port diameter.
Lp Lh Dp
8:138
SUMMARY
Heat transfer and pressure drop correlations are highly dependent upon the geometrical
characteristics of a heat transfer surface. In this chapter, the geometrical characteristics
are derived for determination of heat transfer and pressure drop correlations and exchanger performance (q and p) of the following exchangers: tubular, tube-n, plate-n,
regenerative and plate heat exchangers. Also pertinent geometries are derived for computing the eects of leakage and bypass ows in segmental bae shell-and-tube exchangers. It should be emphasized that if any heat transfer or friction factor correlation from
the literature is used for design and analysis of any exchanger, the geometrical characteristics must be evaluated in the same manner as in the original source of the correlations.
REFERENCES
Bell, K. J., 1963, Final report of the cooperative research program on shell-and-tube heat
exchangers, Univ. Del. Eng. Exp. St. Bull., Vol. 5.
Bell, K. J., 1988, Delaware method for shell-side design, in Heat Transfer Equipment Design, R. K.
Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere Publishing, Washington, DC,
pp. 227254.
Chang, Y. J., and C. C. Wang, 1997, A generalized heat transfer correlation for louver n geometry,
Int. J. Heat Mass Transfer, Vol. 40, pp. 533544.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger
Publishing, Malabar, FL.
London, A. L., M. B. O. Young, and J. H. Stang, 1970, Glass ceramic surfaces, straight triangular
passages: Heat transfer and ow friction characteristics, ASME J. Eng. Power, Vol. 92, Ser. A,
pp. 381389.
Martin, H., 1996, A theoretical approach to predict the performance of chevron-type plate heat
exchangers, Chem. Eng. Process., Vol. 35, pp. 301310.
REVIEW QUESTIONS
599
Mondt, J. R., 1980, Regenerative heat exchangers: the elements of their design, selection and use,
Rep. No. GMR-3396, Research Laboratories, GM Technical Center, Warren, MI,
Sekulic, D. P., A. J. Salazar, F. Gao, J. S. Rosen, and H. S. Hatchins, 2003, Local transient behavior
of a compact heat exchanger core during brazingequivalent zonal (EZ) approach, Int. J. Heat
Exchangers, Vol. 4, No. 1, in print.
Shah, R. K., 1985, Compact heat exchangers, in Handbook of Heat Transfer Applications, 2nd ed.
W. M. Rohsenow, J. P. Hartnett, and E. N. Ganic, eds., McGraw-Hill, New York., pp. 4-174
to 4-311.
Taborek, J., 1998, Shell-and-tube heat exchangers: single phase ow, in Handbook of Heat Exchanger Design, G. F. Hewitt, ed., Begell House, New York, pp. 3.3.3-1 to 3.3.11-5.
Young, M. B. O., 1969, Glass-ceramic, triangular and hexagonal pasage surfacesheat transfer and
ow friction characteristics, TR HE-2, Department of Mechanical Engineering, Stanford
University, Stanford, California.
Zhao, H., and D. P. Sekulic, 2001, Brazed n-tube joint thermal integrity vs. joint formation, Proc.
2001 National Science Foundation Design, Service and Manufacturing Grantees and Research
Conference, Tampa, FL.; CD edition, University of Washington, Seattle, WA, DMII-998319.
Zukauskas, A. A., 1987, Convective heat transfer in cross ow, in Handbook of Single-Phase
Convective Heat Transfer, S. Kakac, R. K. Shah, and W. Aung, eds., Wiley, New York,
Chap. 6.
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
8.1
In a given tube bank, the actual total length of the tube used is 3 m, the tubesheets
are 6 mm thick, and the blanket insulation is 150 mm (Fig. RQ8.1). The eective
length of a tube for heat transfer is:
(a) 3 m
(b) 2.988 m
(c) 2.838 m
(d) 2.688
(e) cant tell
blanket
insulation
150 mm
6 mm tubesheet
3m
FIGURE RQ8.1
8.2
In Question 8.1, the eective tube length for pressure drop is:
(a) 3 m
(b) 2.988 m
(c) 2.838 m
(d) 2.688
(e) cant tell
8.3
8.4
The following surface area components are needed to evaluate primary surface
area in a plate-n heat exchanger:
600
8.6
For the same b and (see sketches in Table 8.2 for denitions), the cell density nc
of a regenerator of square passages compared to hexagonal passages is:
(a) higher
(b) the same
(c) lower
(d) cant tell
8.7
8.8
Which of the following arrangements have more tubes per pass for a specied shell
inside diameter, diameter of the outer tube limit, tube outside diameter, and tube
pitch?
(a) square
(b) rotated square
(c) triangular
(d) cant tell
8.9
8.10
As mentioned in Chapter 2 and tabulated in Table 3.11, there are a large number of heat
exchanger design problems, dened broadly as rating and sizing problems. In a rating
problem, we determine the heat transfer rate and/or outlet temperatures and pressure
drop performance of either an existing or an already designed heat exchanger. In a sizing
problem, we design a heat exchanger; this involves the determination/selection of
exchanger construction type, ow arrangement, tube/plate and n material, and/or the
physical size of an exchanger to meet the specied heat transfer and pressure drops within
specied constraints. We discuss the selection of heat exchanger types, ow arrangements, and so on, in Chapter 10 and thermodynamic analysis and operating problems
in Chapters 11 through 13. These aspects are equally or sometimes more important for
the design of heat exchangers. We concentrate here, then, in a narrow sense on the
determination of the physical size in a sizing problem for specic types of heat exchangers. Also, we determine heat transfer and pressure drop performance of a heat exchanger in a rating problem. Hence, in this chapter, stepwise solution procedures are
presented separately for the rating (problem 12 in Table 3.11) and sizing (problem 2 or
4 in Table 3.11) of plate-n, tube-n, plate, and shell-and-tube heat exchangers. To
present the solution procedures for rating and sizing, we use the theory, correlations,
and geometrical properties summarized in Chapters 3, 6, 7, and 8.
Before presenting the procedures for rating and sizing problems, we start with how to
evaluate the mean temperature of each uid in a two-uid heat exchanger. This mean
temperature is needed to calculate uid thermophysical properties for heat transfer and
pressure drop calculations on each uid side. Also note that an important assumption
made in the heat exchanger design theory (see Section 3.2.1) is the uniformity of uid/
solid thermophysical properties. Fluid properties should be determined at uid mean
temperatures.
9.1
The uid properties are determined at the mean (ow length average) temperature on
each uid side in a heat exchanger. The single-phase uid properties needed for heat
transfer and pressure drop calculations are density, specic heat, viscosity, thermal conductivity, and Prandtl number. These properties are available for a large number of uids
in the literature; some of them are summarized in Appendix A. Now let us determine the
appropriate uid mean temperatures for various heat exchanger congurations.
Fundamentals of Heat Exchanger Design. Ramesh K. Shah and Duan P. Sekulic
Copyright 2003 John Wiley & Sons, Inc.
601
602
For the "-NTU or MTD method, we need to obtain a single temperature value to
represent the temperature level of uid through each uid side of the exchanger. For
counterow and parallelow exchangers, the uid temperature varies in the ow direction as well as over the cross section of each ow passage (for example, as shown in Fig.
7.4). In a crossow heat exchanger, it also varies in the ow direction of the other uid. In
more complex arrangements, the uid temperature variation is generally both in the ow
direction and across a given ow cross section. The temperature changes in the ow
direction aect the uid bulk properties. A scheme to correct the friction factor and
convective heat transfer coecient due to the temperature changes across the ow passage cross section was discussed in Section 7.6.1. If the uid properties vary substantially
within an exchanger, a ow length average temperature may not be adequate to
determine the heat transfer rate and pressure drop accurately. In that case, refer to
Section 4.2.3.2 for the stepwise calculations where arithmetic average temperatures for
each uid are used in each segment.
In the determination of the heat transfer rate through the exchanger, the true mean
temperature dierence is either used directly (as in the MTD method,
q UA Tm UAF Tlm or indirectly (as in the "-NTU or P-NTU method). Note
that Tm Tlm for counterow and parallelow exchangers, or Tm F Tlm
Tlm for a well-designed exchanger of any other ow arrangement. Here F is the logmean temperature dierence correction factor discussed in Section 3.7.2. Therefore, we
use Tlm to evaluate the log-mean average temperature of the Cmin uid as described
below, when the temperature rise or drop of the Cmax uid is small. The average temperature of the Cmax uid in this case is the arithmetic average temperature. When the
temperature drop of the hot uid (Th ) and the temperature rise of the cold uid (Tc )
are of the same order of magnitude, the mean temperature on each uid side is considered
simply as the arithmetic average temperature. An approximate procedure is outlined next
for evaluation of this temperature for two-uid heat exchangers having some specic
ow conditions/arrangements. The results are also summarized in Table 9.1.
TABLE 9.1 Approximate Mean Temperatures on the Hot- and Cold-Fluid Sides of a Two-Fluid
Exchanger
Cmax Hot Fluid; Cmin Cold Fluid
Th;m
Th;i Th;o
Tc;m
Tc;i Tc;o
2
Tlm
C* 0:5 Case
Th;m
Th;i Th;o
2
Tc;m
Tc;i Tc;o
2
9.1.1
603
Typical temperature distributions for both uids are shown in Fig. 9.1. The ow length
average temperatures for the case shown in Fig. 9.1a are
Tc;m
Tc;i Tc;o
2
9:1
9:2
9:3
where
Tlm
The ow length average temperatures for the case of Fig. 9.1b are
Th;m
Th;i Th;o
2
9:4
9:5
9:6
where
Tlm
It can be shown theoretically that for C* 0, the foregoing method of computing the
ow length average temperature is exact (Kays and London, 1998). In this case, the
temperature of the Cmax uid remains truly constant along the ow length, in contrast
to the small variations shown in Fig. 9.1.
FIGURE 9.1 Temperature distributions in a heat exchanger with (a) Ch =Cc 0 and
(b) Cc =Ch 0: (From Shah, 1981.)
604
9.1.2
Typical temperature distributions for these exchangers are shown in Figs. 1.50 (in the
ow directions) and 1.54 (at inlet and outlet cross-sections). From Eqs. (3.192) and
(3.199), the true mean temperature dierence for these exchangers is
Tm
Tlm
FTlm
for counterflow
for crossflow
9:7
Since no simple relation is available for determining mean temperatures on both uid
sides for a counterow exchanger for C* > 0, we use the following approximation. When
C* is closer to zero (arbitrarily we limit C* < 0:5, we recommend to compute the mean
temperatures on each uid side by Eqs. (9.1)(9.6). For C* > 0:5, again arbitrarily, we
calculate the mean temperatures on both uid sides as arithmetic mean values:
Th;m 12 Th;i Th;o
9:8
For a good crossow exchanger design, F generally varies between 0.80 and 0.99 and
the resulting error in the average temperature, by considering F 1, does not generally
produce signicant variations in the uid properties, except possibly for oils. Hence, we
evaluate the average temperature on each uid side for a crossow exchanger by considering it as a counterow heat exchanger (i.e., using Tlm where appropriate).
It should be emphasized that if we assume a value for exchanger eectiveness, the
integral mean temperature could be calculated for those exchangers where a closed-form
solution for the temperature distribution is available. In that case, the afore mentioned
approximate procedure is not needed. However, in the lumped parameter approach (i.e.,
considering one value of the mean temperature along the complete ow length for a uid
as considered in the design theory of Chapter 3), the foregoing proposed approximate
values of mean temperatures should be adequate. If the uid properties vary signicantly
along the ow length, the "-NTU or other methods outlined in Chapter 3 are not
adequate. In that case, one should conduct numerical analysis as outlined in Section
4.2.3.2 by dividing the exchanger into suciently small elements so that arithmetic
average temperatures can represent the mean temperatures reasonably accurately.
9.1.3
Fluid temperatures are rst determined after every pass so that the inlet and outlet uid
temperatures are known for every pass. Then arithmetic averages of the terminal temperatures are taken as the ow length average temperature for each uid in each pass.
This would constitute the rst iteration. If needed, this procedure can be repeated with
modied thermophysical/process properties/parameters.
The eectiveness "p or Pp of each pass in terms of overall eectiveness " or P and C*
or R is given by Eq. (3.136) for an overall counterow arrangement, and that for an
overall parallelow arrangement by Eq. (3.144). Mean outlet temperatures Th;o and Tc;o
for each pass are then determined from the denition of individual pass eectiveness "p ;
use Eq. (3.44) with " replaced by "p , and the inlet and outlet temperatures as those of
individual passes. Note that both inlet temperatures of each pass are known in a parallelow exchanger starting with the rst pass. For an overall counterow arrangement, use
Eq. (3.44) modied as mentioned above to determine the mean outlet temperatures.
605
9.2
In this section we consider rating and sizing of plate-n exchangers (see Section 1.5.3.1
for a description) with corrugated ns on one or both sides. We use the "-NTU method
since it is used most commonly in industry for these types of exchangers.
9.2.1
Rating Problem
In this section we describe a step-by-step method for rating (problem 12 in Table 3.11) of
single-pass counterow and crossow exchangers. A similar step-by-step method for a
two-pass cross-counterow exchanger is presented by Shah (1988a). The basic steps
involved in the analysis of a rating problem are the determination of the following
information: surface geometrical properties, uid physical properties, Reynolds numbers, surface basic characteristics, corrections to the surface basic characteristics due to
temperature-dependent properties, heat transfer coecients and n eciencies, wall
thermal resistance and overall thermal conductance, NTU, C*, exchanger eectiveness,
heat transfer rate, outlet temperatures, and pressure drops on each uid side. These steps
are outlined in detail now.
1. Determine the surface geometrical properties on each uid side (see, e.g., Section
8.3). This includes the minimum free-ow area Ao , heat transfer surface area A
(both primary and secondary), ow length L, hydraulic diameter Dh , heat transfer
surface area density , ratio of minimum free-ow area to frontal area , n
geometry (l, , etc.) for n efciency determination, and any specialized dimensions
used for heat transfer and pressure drop correlations.
2. Compute the uid mean temperature and uid thermophysical properties on each
uid side. Since the outlet temperatures are not known for the rating problem, they
are guessed initially. Unless it is known from past experience, assume exchanger
effectiveness as 50 to 75% for most single-pass crossow exchangers, or 75 to 85%
for single-pass counterow and two-pass cross-counterow exchangers. For the
assumed effectiveness, calculate the uid outlet temperatures as follows:
606
Cmin
Th;i Tc;i
Ch
9:9
Cmin
Th;i Tc;i
Cc
9:10
Nucp
Nu
Nucp
Tw
Tm
w
m
n
fcp
n
fcp
m
Tw
Tm
w
m
for gases
9:11
for liquids
9:12
m
where the subscript cp denotes constant properties, and m and n are empirical
constants presented in Tables 7.12 and 7.13. Note that Tw and Tm in Eq. (9.11) and
Tables 7.12 and 7.13 are absolute temperatures and Tw is computed from Eq.
(3.33).
4. From Nu or j, compute the heat transfer coefcients for both uid streams from
the following equations.
h Nu
k
Dh
or
h jGcp Pr2=3
9:13
tanh m
m
m
hP
kf A k
o 1 1 f
1=2
Af
A
or
2h
kf
1=2
9:14
9:15
Here P, Ak and are the wetted perimeter, n cross-sectional area and n thickness, respectively. Refer to Table 4.5 for the formulas for some additional n
607
geometries for a plate-n exchanger. Also calculate the wall thermal resistance
Rw w =Aw kw . Finally, compute the overall thermal conductance UA from
1
Rh Rh; f Rw Rc; f Rc
UA
1
1
1
1
Rw
o hf A h
o hf A c o hAc
o hAh
9:16
9:17
m 2 i o
9:18
Next, the entrance and exit loss coecients, Kc and Ke , are obtained from Fig. 6.3
for known , Re, and the ow passage entrance geometry. The friction factor on
each uid side is corrected for variable uid properties using Eq. (9.11) or (9.12).
Here, the wall temperature Tw is computed from
Tw;h Tm;h Rh Rh; f q
9:19
where the various resistance terms are dened by Eq. (9.16). The core pressure
drops on each uid side are then calculated from
i
G2
i
L
1
2
2
1 Kc 2
1 f i
1 Ke
p
2gc i
o
o
rh
m
9:20
608
L 1 = L g = 0.3 m
AIR
L 3 = 1.0m
L 2 = 0.3m
FIGURE E9.1A Gas-to-air single-pass crossow heat exchanger for the rating problem. (From
Shah, 1981.)
Basic surface geometries on the gas and air sides along with their j and f data are given
in Fig. E9.1B and are summarized below along with the operating conditions. The subscripts g and a are used for the gas and air sides, respectively.
Af
bg 2:49 mm Dh;g 0:00154 m g 2254 m2 =m3 g 0:102 mm
0:785
A g
Af
ba 2:49 mm Dh;a 0:00154 m a 2254 m2 =m3 a 0:102 mm
0:785
A a
V_ g 3:494 m3 =s Tg;i 9008C
kf kw 18 W=m K
609
FIGURE E9.1B Surface basic characteristics for an oset strip n surface 1/8 19:86. (From
Kays and London, 1998.)
Determine: Heat duty and pressure drops (gas and air sides) for this exchanger.
Assumptions: The assumptions listed in Section 3.2.1 as applicable to a plate-n
exchanger are invoked.
Analysis: We follow the steps outlined in the preceding text, starting with the calculation
of surface geometrical properties.
610
Surface Geometrical Properties. We assume that there are Np passages for the gas and
Np 1 passages for the air to minimize heat loss to the ambient.y The noow height
(stack height) is given by
L3 Np bg Np 1ba 2Np 2w
Therefore
Np
166:6 167
bg ba 2w
2:49 mm 2:49 mm 2 0:5 mm
Here b is the n height (plate spacing) and w is the plate thickness. The frontal areas on
the gas and air sides are
Afr;g L2 L3 0:3 m 1:0 m 0:3 m2
Afr;a L1 L3 0:3 m 1:0 m 0:3 m2
The heat exchanger volume between plates, on each uid side, is
Vp;g L1 L2 bg Np 0:3 m 0:3 m 2:49 103 m 167 0:03742 m3
Vp;a L1 L2 ba Np 1 0:3 m 0:3 m 2:49 103 m 168 0:03765 m3
The heat transfer areas Ag and Aa are
Ag g Vp;g 2254 m2 =m3 0:03742 m3 84:345 m2
Aa a Vp;a 2254 m2 =m3 0:03765 m3 84:863 m2
The minimum free-ow area is then calculated from the denition of the hydraulic
diameter, Dh 4Ao L=A:
Ao;g
Dh Ag 0:00154 m 84:345 m2
0:1082 m2
4Lg
4 0:300 m
Ao;a
Dh Aa 0:00154 m 84:863 m2
0:1089 m2
4La
4 0:300 m
Ao;g 0:1082 m2
0:361
Afr;g
0:3 m2
a
Ao;a 0:1089 m2
0:363
Afr;a
0:3 m2
Mean Temperatures and Fluid Properties. To determine the mean temperatures on each
uid side, we need to calculate C*. Since the ow rates are specied as volumetric at inlet
temperatures, let us rst calculate the gas and air densities and then the mass ow rates.
y
This is a common practice, whenever practical, to minimize the heat losses to ambient.
g;i
pg;i
160 103 Pa
0:4751 kg=m3
R~Tg;i 287:04 J=kg K 273:15 900:0 K
a;i
pa;i
200 103 Pa
1:4726 kg=m3
~
RTa;i 287:04 J=kg K 273:15 200:0 K
611
where R~ 287:04 J=kg K is the gas constant for air. Note that all temperatures are in
kelvin. Hence the mass ow rates are
m_ g V_ g g 3:494 m3 =s 0:4751 kg=m3 1:66 kg=s
m_ a V_ a a 1:358 m3 =s 1:4726 kg=m3 2:00 kg=s
Hence, gas will be the Cmin side since the change in the specic heat is not a strong
function of temperature for air/gas. Now assume that " 0:75 for the crossow exchanger. Then using the denition of the exchanger eectiveness [Eqs. (9.9) and (9.10)], we
have
Tg;o Tg;i "Tg;i Ta;i 9008C 0:75900 2008C 375:08C
m_ g
1:66 kg=m3
Ta;o Ta;i "
Tg;i Ta;i 2008C 0:75
900 2008C
m_ a
2:0 kg=m3
635:88C
Note that we used cp;a cp;g as a rst approximation for determining Ta;o . Since
C* m_ g =m_ a 0:83, we will use the arithmetic average temperature from [Eq. (9.8)] as
the appropriate mean temperature on each uid side.
Tg;m
900:0 375:08C
637:58C 910:65 K
2
Ta;m
200:0 635:88C
417:98C 691:05 K
2
In the absence of information on the composition of the gas, we treat both the gas and
air as dry air. The properties of air are obtained from any source of thermophysical
properties (see, e.g., Appendix A) as
Gas at 637.58C
Air at 417.98C
Pa s
cp (kJ/g K
Pr
Pr2=3
40:1 106
33:6 106
1.122
1.073
0.731
0.694
0.811
0.784
612
Gg
m_ g
1:66 kg=s
Ao;g 0:1082 m2
Reg
15:342 kg=m s
GDh
Ga
2:0 kg=s
m_ a
Ao;a 0:1089 m2
Rea
GDh
a
18:365 kg=m s
We get the j and f values from the curve t of tabular values given in Kays and London
(1998) as follows; the other sources are the graphical values from Fig. E9.1B or generalized correlations from the literature such as those given in Section 7.5.3.1.
Gas
Air
Re
589
842
0.0170
0.0134
0.0669
0.0534
Since Reynolds numbers indicate the ow as laminar on both gas and air sides, the
correction to the j factor is unity because n 0 from Table 7.12. However, the correction
to the f factor will not be unity since m 6 0 from Table 7.12. We will determine this
correction after calculating the wall temperature Tw .
Heat Transfer Coecients and Fin Eciency. We compute the heat transfer coecient
from the denition of the j factors as follows:
hg
ha
jGcp
Pr2=3
jGcp
Pr2=3
Now let us calculate the n eciency for air and gas sides. Since the oset strip ns are
used on both gas and air sides, we will use Eq. (4.147) with Lf replaced by s to take into
account the strip edge exposed area.
1=2
2h
2 360:83 W=m2 K
0:102 mm 1=2
1
kf
s
3:175 mm
18 W=m K 0:102 103 m
mg
634:94 m1
1=2
2h
2 336:81 W=m2 K
0:102 mm 1=2
1
1
ma
kf
s
3:175 mm
18 W=m K 0:102 103 m
615:37 m1
a g b=2 2:49 mm=2 0:102 mm 1:143 mm 0:00114 m
613
Thus,
f ;g
mg
636:94 m1 0:00114 m
f ;a
ma
615:37 m1 0:00114 m
Af
o;g 1 1 f
1 1 0:8581 0:785 0:8886
A g
Af
o;a 1 1 f
1 1 0:8657 0:785 0:8946
A a
It should be pointed out that the n conduction length for the end passages on the
airside will be b and not (b=2 ). This will result in lower n eciency for the end
passages. However, its inuence will be smaller on the weighted average n eciency
considering all air passages. Hence, we have neglected it here. However, in a computer
program, it can easily be incorporated.
Wall Resistance and Overall Conductance. For the Rw determination, the wall conduction
area Aw is
Therefore,
Rw
w
0:5 103 m
Since the inuence of fouling is negligibly small for a gas-to-gas heat exchanger, we will
neglect it. Then 1=UA from Eq. (9.16) is
614
1
1
1
Rw
o hAc
UA o hAh
1
9:186 107 K=W
0:8886 360:83 W=m2 K 84:345 m2
1
0:8946 336:81 W=m2 K 84:863 m2
1
1
9:186 107 K=W
27043:8 W=K
25,570:1 W=K
0:868
Cmax Ca 2146 W=K
UA
12985 W=K
6:970
Cmin
1863 W=K
For NTU 6:970 and C* 0:868, the eectiveness for the crossow exchanger with
both uids unmixed from the expression of Table 3.3, is
" 0:8328
This eectiveness is higher than normally used for a crossow exchanger. The reason for
the selection of somewhat hypothetical example to yield high " is to demonstrate how to
take into account the eect of longitudinal conduction in the wall. We evaluate the
decrease " in " due to longitudinal heat conduction now. Let us rst calculate the
conduction cross-sectional area for longitudinal heat conduction in the wall.
Ak;g 2Np La w 2 167 0:3 m 0:5 103 m 0:0501 m2
Ak;a 2Np 2Lg w 2 167 2 0:3 m 0:5 103 m 0:0504 m2
Longitudinal conduction parameters on the gas and air sides are
kw A k
18 W=m K 0:0501 m2
0:0016
h g
LC g
0:3 m 1863 W=K
kw A k
18 W=m K 0:0504 m2
0:0014
c a
LC a
0:3 m 2146:4 W=K
615
1:06
o hAc 25,570:1 W=K
c 0:0016
1:14
h 0:0014
From the interpolation of tabular results of Chiou (1980) (see Table 4.1), it is found that
"=" 0:002. Thus " 0:0017, and actual exchanger eectiveness is
"actual 0:8328 0:0017 0:8311
The heat transfer rate q is then
q "Tg;i Ta;i Cmin 0:8311 900 2008C 1863 W=K 1083:8 103 W
The outlet temperatures are then
Tg;o Tg;i
q
1083:8 103 W
9008C
318:38C 591:5 K
Cg
1863 W=K
Ta;o Ta;i
q
1083:8 103 W
2008C
705:08C 978:2 K
Ca
2146 W=K
Since these outlet temperatures are dierent from those assumed for the initial determination of the uid properties, two more iterations were carried out with uid properties evaluated at the new average temperatures. The values of C*, NTU, ", Tg;o and Ta;o
were 0.857, 7.082, 0.8382, 314.48C and 701.98C, respectively, and after the third iteration
were 0.857, 7.079, 0.8381, 314.58C and 701.88C respectively.
Pressure Drops. We use Eq. (9.20) to compute the pressure drop on each uid side. The
densities are evaluated using the perfect gas equation of state:
Gas
Air
Ti K
To (K)
i (kg/m3 )
o (kg/m3 )
m (kg/m3 )
1173
473
591.5
978.2
0.4751
1.4726
0.9424
0.7123
0.6318
0.9602
Note that we have also considered the outlet pressures as 160 kPa and 200 kPa for gas
and air, respectively, since the pressure drop across the core is usually small and hence is
neglected in the rst iteration. The mean density in the last column is the harmonic mean
value from Eq. (9.18).
Now let us determine Kc and Ke . Oset strip ns are used on gas and air sides. In such
n geometries, because of the frequent boundary layer interruptions, the ow is well
mixed and is treated as having the Reynolds number very large (Re ! 1). The aspect
ratio of the rectangular passage, height/width 2:49=1=0:615 0:15 1:15. Since
Re 1 curves for parallel plate and square passage geometries of Fig. 6.3 are identical,
we could determine Kc and Ke from either geometry for a 0:36 as
Kc 0:36
Ke 0:42
616
Before we compute the pressure drop, we need to correct the values of the isothermal
friction factors by the method of Section 7.6.1 to take into account the temperaturedependent properties. A review of Eq. (9.11) indicates that we need to calculate the uid
bulk mean temperatures and the wall temperature. The mean temperatures on the gas
and air sides, based on the latest outlet temperatures, are
900 318:38C
609:28C 882:4 K
2
200 705:08C
452:58C 725:7 K
2
Tg;m
Ta;m
1
0:3698 104 K=W
o hAh
Ra
1
0:3911 104 K=W
o hAa
Rg
0:946
Ra
If we neglect the wall resistance,
q
Tg;m Tw Tw Ta;m
Rg
Ra
so that
Tw
533:08C 806:2 K
1 Rg =Ra
1 0:946
Since the gas is being cooled, using Eq. (9.11) and the exponent m 0:81 from Table
7.12,
m
Tw
806:2 K 0:81
0:0669
0:0622
f fcp
Tm
882:4 K
Since the air is being heated, using Eq. (9.11) and the exponent m 1:00 from Table 7.12,
m
Tw
806:2 K 1:00
0:0534
0:0593
f fcp
Tm
725:7 K
Now let us calculate the pressure drops using Eq. (9.20).
15:342 kg=m2 s2
0:4751 kg=m3
2
pg
1
1 0:361 0:36 2
2 1 0:4751 kg=m3
0:9424 kg=m3
0:0622 0:3 m 0:4751 kg=m3
0:4751 kg=m3
2
1
0:361
0:42
0:00154=4 m 0:6318 kg=m3
0:9424 kg=m3
2
2
7:0484 kg=m s
1:3827 kg=m3
2
pa
1 0:437 0:33 2
1
2 1 1:3827 kg=m3
0:8194 kg=m3
0:0683 0:6 m 1:3827 kg=m3
1:3827 kg=m3
2
1
0:437
0:31
617
18:365 kg=m2 s2
1:4726 kg=m3
2
1
0:363
0:36
2
1
2 1 1:4726 kg=m3
0:7123 kg=m3
0:0593 0:3 m 1:4726 kg=m3
1:4726 kg=m3
2
1
0:363
0:42
9.2.2
Sizing Problem
The sizing problem is more dicult. Many early decisions to choose the construction
type and basic geometries on each uid side are based on experience (including rules of
thumb and engineering judgments), operating conditions, maintenance, manufacturing
capability, and the expected life of the exchanger. Some of these issues were mentioned in
Chapter 2. We discuss the selection of exchanger construction type, ow arrangement,
surface geometries, and so on, in Chapter 10. With those inputs, the sizing problem then
reduces to the determination of the core or exchanger dimensions for the specied heat
transfer and pressure drop performance. One could, of course, reduce this problem to the
rating problem by tentatively specifying the dimensions, then calculate the performance
for comparison with the specied performance. This type of search for a solution is
usually performed in the case of shell-and-tube exchangers and regenerators where one
needs to take care of leakage and bypass ows in a very complex manner. However,
leakage and bypass ows are not signicant for plate-n and tube-n exchangers. The
solution method can be made more straightforward, with fast convergence for these
exchangers by reforming the surface characteristics input to include j=f vs. Re for
surfaces on each uid side, in addition to the separate j and f versus Re characteristics.
This coupling of heat transfer and ow friction is now made in the derivation of the core
mass velocity equation that has been proposed by Kays and London (1998). Once the
618
core mass velocity is determined, the solution to the sizing problem is carried out iteratively in a manner similar to the rating problem discussed in Section 9.2.1.
9.2.2.1 Core Mass Velocity Equation. The coupling of heat duty and pressure drops
is done by the core mass velocity equation as follows. From the required heat duty
(and hence the exchanger eectiveness ") and known heat capacity rates on each uid
side (known C*), the overall NTU is determined for the selected exchanger ow
arrangement. The overall conductance as a rst approximation is given by
1
1
1
UA o hAh o hAc
9:21
Here we have neglected the wall and fouling thermal resistances. The overall NTU is
related to individual side ntuh and ntuc as follows [see Eq. (3.67)]:
1
1
1
C*
1
9:22
9:23
Alternatively, if we have liquid on one side and gas on the other side, consider 10%
thermal resistance on the liquid side:
1
1
0:10
UA
o hAliq
9:24
Substituting Eq. (9.24) into Eq. (9.21) with Cc Cgas Cmin and Ch Cliq , we can
determine ntugas and with its subsequent substitution in Eq. (9.22), we get
ntugas 1:11 NTU
9:25
The ntu on each uid side is related to the Colburn factor j as follows by using Eq. (7.31)
and (7.33):
ntu
o hA
h A
A
L
o
o j Pr2=3
o j Pr2=3
m_ cp
Gcp Ao
Ao
rh
9:26
The pressure drop on each uid side is given by Eq. (9.20). Substituting L=rh from Eq.
(9.26) into Eq. (9.20) and simplifying, G reduces to
f ntu 2=3 1
1
1
1
1 2 Kc
Pr
2
j o
m
o i
i
1 1=2
1 2 Ke
o
619
G 2gc p1=2
9:27
Equation (9.27) is a more generalized core mass velocity equation than that provided
by Kays and London (1998), who considered only the rst term in brackets in Eq. (9.27).
Since the contribution of the last three terms in the brackets in Eq. (9.27) is generally very
small, they can be neglected in light of other approximations already mentioned. In this
case, Eq. (9.27) reduces to
G
2gc
o p
1=m Pr2=3 ntu
1=2
j
f
9:28
Equation (9.28) is referred to as the core mass velocity equation. The feature that
makes this equation so useful is that the ratio j=f is a relatively weak function of the
Reynolds number for most extended surfaces (see Fig. 2-41 of Kays and London, 1998,
and Fig. E9.2). Thus, one can readily estimate a fairly accurate value of j/f in the
operating range of Re. Also, for a good design, the n geometry is chosen such that
o is in the range 70 to 90% and higher. Hence, o 80% may be assumed for the rst
approximation in Eq (9.28), unless a better value is known from the past experience. All
other information in Eq. (9.28) is known or evaluated from the problem specication.
Thus, the rst approximate value of G can be computed from Eq. (9.28). As a result,
the iterative solution to the sizing problem converges relatively fast with this estimated
value of G.
We will use Eq. (9.28) for the rst iteration of a sizing problem, as described next for
single-pass counterow and crossow exchangers.
9.2.2.2 Sizing of a Single-Pass Counterow Exchanger. Now we outline a detailed
procedure for arriving at core dimensions for a counterow exchanger for specied
heat transfer and pressure drop. In a single-pass counterow heat exchanger of any
construction, if the core dimensions on one side are xed, the core dimensions for the
other side (except for the passage height) are also xed. Therefore, the design problem
for this case is solved for the side that has more stringent pressure drop specication.
This method is also applicable to the C* 0 exchanger, such as a gas-to-liquid or
phase-changing uid exchanger. In this case, the thermal resistance is primarily on
the gas side and the pressure drop is also more critical on the gas side. As a result,
the core dimensions obtained are based on the gas-side p and ntugas . The dimensions
on the other side are then chosen such that the calculated pressure drop is within the
specied p (i.e., pcalculated pspecified ). Thus either for counterow or for a C* 0
exchanger, the core dimensions are calculated for the side having the most stringent p.
Following is a step-by-step procedure for the solution.
1. To compute the uid mean temperature and the uid thermophysical properties
on each uid side, determine the uid outlet temperatures, for the specied heat
duty, from the following equation considering uid 1 as the hot uid.
q m_ cp 1 T1;i T1;o m_ cp 2 T2;o T2;i
9:29
620
2.
3.
4.
5.
6.
7.
If the exchanger eectiveness is specied, use Eqs. (9.9) and (9.10) to compute
outlet temperatures. For the rst iteration, estimate the cp values to determine the
outlet temperatures from Eq. (9.29). Subsequently, determine the mean temperatures on both uid sides using the procedure discussed in Section 9.1.2 or from
Table 9.1. With these mean temperatures, determine the cp s, and iterate one more
time for the outlet temperatures if warranted. Subsequently, determine , cp , k,
Pr, and on each uid side.
Calculate C* and " (if q is given), and determine NTU from the "-NTU expression, with tabular or graphical results for the ow arrangement selected [in this
case, use Eq. (3.86) for counterow]. The inuence of longitudinal heat conduction, if any, is ignored in the rst iteration since we dont yet know the exchanger
size.
Determine ntu on each uid side by the approximations discussed with Eqs. (9.23)
and (9.25) unless it can be estimated differently (i.e., instead of 50 : 50% or
90 : 10% thermal resistance distribution) from past experience.
For the surfaces selected on each uid side, plot j=f vs. Re curve from the given
surface characteristics, and obtain an approximate mean value of j=f over the
complete Reynolds number range; an accurate mean value of j=f is not necessary
since we are making a number of approximations to get the rst estimate of G. If
ns are employed, assume that o 0:80 unless a better value can be estimated
based on the experience.
Evaluate G from Eq. (9.28) on each uid side using the information from steps 1
through 4 and specied values of p.
Calculate Reynolds number Re, and determine j and f on each uid side from the
given design data for each surface. The design data may be in the form of graphs,
curve t to tabulated data or an empirical generalized equation.
Compute h, f , and o using Eqs. (9.13)(9.15). For the rst iteration, determine
U1 on the uid 1 side from the following equation derived from Eq. (3.24):
1
1
1
=
=
1 2 1 2
U1 o h1 o hf 1 o hf 2 o h2
9:30
where 1 =2 A1 =A2 , A=V, and V is the exchanger total volume, and subscripts 1 and 2 denote the uid 1 and 2 sides. For a plate-n exchanger, s are
related to s by Eqs. (8.96) and (8.97) with usually provided with the surface
basic characteristics (see, e.g., Fig. E9.1B)
1
b1 1
b1 b2 2w
2
b2 2
b1 b2 2w
9:31
Note that the wall thermal resistance in Eq. (9.30) is ignored in the rst iteration
since we do not yet know the size of the exchanger (i.e., Aw =A1 is unknown). In
second and subsequent iterations, compute U1 from
1
1
1
A
A =A
A =A
w 1 1 2 1 2
U1 o h1 o hf 1 kw Aw o hf 2 o h2
9:32
where the necessary area ratios A1 =A2 and A1 =Aw are determined from the geometry calculated in the preceding iteration.
621
8. Now calculate the core dimensions. In the rst iteration, use NTU as computed in
step 2. For subsequent iterations, calculate the longitudinal conduction parameter . With known ", C*, and , determine the correct value of NTU using
Eq. (4.15) or (4.20) . Determine A1 from NTU using U1 from step 7 and known
Cmin :
NTU Cmin
U1
9:33
A2
A 2A
A1 1 1 1
9:34
A1
and hence
A2
m_
G 1
Ao;2
m_
G 2
9:35
Ao;1
1
Afr;2
Ao;2
2
9:36
so that
Afr;1
where 1 and 2 are generally specied for the surface or can be computed from
Eqs. (8.94) and (8.95) as follows using given geometrical properties:
1
b1 1 Dh;1 =4
1 Dh;1
b1 b2 2w
4
2
b2 2 Dh;2 =4
2 Dh;2
b1 b2 2w
4
9:37
where the term after the second equality sign comes from the denition of s
from Eq. (9.31). In a single-pass counterow exchanger, Afr;1 and Afr;2 must be
identical, and those computed in Eq. (9.36) may not be identical. In this case, use
the greater of Afr;1 and Afr;2 . Finally, the core length L in the ow direction is
determined from the denition of the hydraulic diameter of the surface employed
on each uid side.
L
Dh A
Dh A
4Ao 1
4Ao 2
9:38
The value of L calculated from either of the equalities will be the same, as can be
shown using Eqs. (9.34), (9.36), and (9.37), and for Afr;1 Afr;2 .
Once the frontal area is determined, any choice of exchanger width and height
(product of which should be equal to the frontal area) will theoretically be a
correct solution. If there are any constraints imposed on the exchanger dimension, select the frontal area dimensions accordingly. Also, from the header design
viewpoint as well as from the ow distribution viewpoint, select the frontal area
dimensions to make it the least prone to maldistribution; see Chapter 12 for a
discussion of related issues.
622
9. Now compute the pressure drop on each uid side, after correcting f factors
for variable property effects, in a manner similar to step 8 of the rating problem
(section 9.2.1).
10. If the values calculated for p are within the input specications and close to
them, the solution to the sizing problem is completed; ner renements in the
core dimensions, such as integer numbers of ow passages, may be carried out at
this time. Otherwise, compute the new value of G on both uid sides using Eq.
(9.27), in which p is the input specied value, and f, Kc , Ke , and geometrical
dimensions are from the previous iteration.
11. Iterate steps 6 through 10 until both heat transfer and pressure drops are met as
specied. Probably, only one of the two pressure drops (whichever is the most
critical) will be matched, the other will be lower than specied for a gas-to-gas
exchanger. Only two or three iterations may be necessary to converge to the nal
size of the exchanger within 1% or the accuracy desired.
12. If the inuence of longitudinal heat conduction is important, the longitudinal
conduction parameter is computed from Eq. (4.13), and subsequently, NTU is
computed iteratively from the " formula/results of Section 4.1.2. This new value
of NTU is then used in step 8 in the second and subsequent iterations.
9.2.2.3 Sizing of a Single-Pass Crossow Exchanger. For a crossow exchanger,
determining the core dimensions on one uid side (Afr and L) does not x the dimensions on the other uid side. In such a case, the design problem is solved simultaneously
on both uid sides. The solution procedure follows closely that of Section 9.2.2.2 and is
outlined next through detailed steps.
1. Determine G on each uid side by following steps 1 through 5 of Section 9.2.2.2.
2. Follow steps 6 through 8 of Section 9.2.2.2 and compute A1 , A2 , Ao;1 ; Ao;2 , Afr;1 ;
and Afr;2 .
3. Now compute the uid ow lengths on each uid side (see Fig. E9.1A) from the
denition of the hydraulic diameter of the surface employed on each uid side as
follows:
Dh A
Dh A
L2
9:39
L1
4Ao 1
4Ao 2
Since Afr;1 L2 L3 and Afr;2 L1 L3 ; we obtain
L3
Afr;1
L2
L3
Afr;2
L1
9:40
Thus the noow (or stack) height L3 can be determined from the denition of
either Afr;1 or Afr;2 and known L2 or L1 and should be identical. In reality, they
may be slightly dierent because of the round-o error in calculations. In that case,
consider an average value for L3 .
4. Now follow steps 9 and 10 of Section 9.2.2.2 to compute p on each uid side. If
p on one of two uid sides does not match (i.e., too high compared to the
specication), calculate new values of G on both uid sides as mentioned in step
10 of Section 9.2.2.2.
623
5. Iterate steps 1 through 4 until both heat transfer and pressure drops are met as
specied within the accuracy desired.
6. If the inuence of longitudinal heat conduction is important, the longitudinal
conduction parameter h , c , and other appropriate dimensionless groups are
calculated based on the core geometry from the preceding iteration and input
operating conditions with the procedure outlined in Section 4.1.4. Subsequently,
NTU is computed iteratively from the " results of Section 4.1.4. This new value of
NTU is then used in step 8 of Section 9.2.2.2.
It should be emphasized that since we have not imposed any constraints on the
exchanger dimensions, the procedure above will yield unique values of L1 , L2 , and L3
for the surface selected such that theoretically the design will exactly meet the heat duty
and pressure drops on both uid sides.
Example 9.2 Consider the heat exchanger of the rating problem in Example 9.1. Design
a gas-to-air single-pass crossow heat exchanger operating at " 0:8381 having gas and
air inlet temperatures as 9008C and 2008C respectively, and gas and air mass ow rates as
1.66 kg/s and 2.00 kg/s, respectively. The gas side and air side pressure drops are limited
to 9.05 and 8.79 kPa, respectively. The gas and air inlet pressures are 160 kPa and 200 kPa
absolute. The oset strip n surface on the gas and air sides has the surface characteristics
as shown in Fig. E9.1B. Both ns and plates (parting sheets) are made from Inconel 625
alloy (its thermal conductivity as 18 W/m K). The plate thickness is 0.5 mm. Determine
the core dimensions of this exchanger.
SOLUTION
Problem Data and Schematic: The following information is provided for the sizing of a
crossow exchanger.
" 0:8381
Basic surface geometry parameters on the gas and air sides along with their j and f
data are given in Fig. E9.1B. Geometry and operating parameters are:
Af
0:785
bg 2:49 mm Dh;g 0:00154 m g 2254 m2 =m3 g 0:102 mm
A g
Af
0:785
ba 2:49 mm Dh;a 0:00154 m a 2254 m2 =m3 a 0:102 mm
A a
m_ g 1:66 kg=s Tg;i 9008C
m_ a 2:0 kg=s
Ta;i 1008C
kf kw 18 W=m K
Determine: The length, width and height of this exchanger to meet specied exchanger
eectiveness (heat duty) and pressure drops.
Assumptions: The assumptions listed in Section 3.2.1 applicable to a plate-n exchanger
are invoked. Neglect the eect of longitudinal heat conduction and treat the gas as air for
uid property evaluation.
Analysis: We will follow the steps outlined in Section 9.2.2.3 for the solution.
624
Outlet Temperatures. To determine outlet temperatures from the known ", we rst need
to know which uid side is the Cmin side. Since m_ a > m_ g , the Cmin side will be the gas side.
Assuming that the specic heats of air and gas in the rst iteration are the same
cp;a cp;g , using Eqs. (9.9) and (9.10), we get
Tg;o Tg;i "Tg;i Ta;i 9008C 0:8381 900 2008C 313:38C
Ta;o Ta;i "
m_ g
T Ta;i 2008C 0:8381 0:83 900 2008C 686:98C
m_ a g;i
This value of Ta;o will be rened after we determine the uid properties.
Fluid Properties. Since C* m_ g =m_ a 0:83, we will evaluate the uid properties at the
arithmetic mean temperatures.
9008C 313:38C
606:78C 879:8 K
2
2008C 686:98C
443:58C 716:6 K
2
Tg;m
Ta;m
The cp values of gas and air at these temperatures are 1.117 and 1.079 J/kg K, respectively (Raznjevic, 1976). Hence, using Eq. (9.9), the correct Ta;o will be
m_ g cp;g
Tg;i Ta;i
m_ a cp;a
1:66 kg=s 1:117 kJ=kg K
9008C 2008C 704:18C
2008C 0:8381
2:00 kg=s 1:079 kJ=kg K
704:18C 2008C
452:08C 725:2 K
2
The specic heat of air at 725.2 K is 1.081 kJ/kg K, which has negligible dierence from
the previous value of 1.079 kJ/kg K and hence there is no further need of iterations. The
air properties at Tg;m 879:8 K and Ta;m 725:2 K from Appendix 1 are as follows.
Gas at 879.8 K
Air at 725.2 K
Pa s
cp kJ=kg K
Pr
Pr2=3
39:3 106
34:7 106
1.117
1.081
0.721
0.692
0.804
0.782
The inlet and outlet gas densities are evaluated at 160 kPa and 150.95 ( 160 9:05) kPa,
respectively. The inlet and outlet air densities are evaluated at 200 kPa and
191.21 200 8:79 kPa, respectively. The mean densities are evaluated using Eq.
(9.18):
Gas
Air
625
Ti (K)
To (K)
i (kg/m3 )
o (kg/m3 )
m (kg/m3 )
1173.2
473.2
586.5
977.2
0.4751
1.4726
0.8966
0.6817
0.6212
0.9319
C* and NTU. From the foregoing values of cp and given ow rates, we evaluate
Cg m_ cp g 1:66 kg=s 1:117 103 J=kg K 1854 W=K
Ca m_ cp a 2:00 kg=s 1:081 103 J=kg K 2162 W=K
C*
0:858
Cmax 2162 W=K
Neglecting longitudinal heat conduction, NTU for a crossow exchanger with both
uids unmixed for " 0:8381 and C* 0:858, from the expression of Table 3.3, is
NTU 7:079
Now we need to estimate ntug and ntua from the overall NTU. The better the initial
estimate, the closer will be the value of G as a rst estimate. For a gas-to-gas heat
exchanger, a good estimate would be equal resistances on each uid side, considering
a thermally balanced design. This would correspond to Eq. (9.23). Hence,
ntua 2 NTU 2 7:079 14:16
Then neglecting the wall thermal resistance (since we do not know Aw yet), we get, from
Eq. (9.22),
ntua 2C* NTU 2 0:858 7:079 12:15
While ntua 12:15 is a somewhat rened value, we could have taken ntua 14:16 for
the rst iteration.
Core Mass Velocities. To determine G from Eq. (9.28), we need to estimate the values of
j=f and o . Since j and f versus Re characteristics are specied for the surfaces on the gas
and air sides, j=f versus Re curves are constructed as shown in Fig. E9.2. Since we do not
know Re yet, an approximate average (ballpark) value of j=f over the complete range
of Re is taken for each surface from this gure as
j=f g 0:25
j=f a 0:25
Again, a more precise value of j=f is not essential since we are getting a rst approximate
value of G with a number of other approximations. In absence of any specic values of o ,
we will assume o on both the gas the air sides to be 0.80. Now substituting all values on
the right-hand side of Eq. (9.28), we get
626
0.40
j/f
0.20
200
300
500
1000
Re
2000
5000
FIGURE E9.2 j=f vs. Re characteristics of surfaces of Fig. E9.1B. (From Shah, 1981).
Gg
2gc
o p j
1=m Pr2=3 ntu f
1=2
"
#1=2
2 1 0:8 9:05 103 Pa 0:25
1=0:6212 kg=m3 0:8039 14:16
14:06 kg=m s
2gc
o p j
Ga
1=m Pr2=3 ntu f
1=2
a
"
#1=2
2 1 0:8 8:79 103 Pa 0:25
18:56 kg=m s
2
Reynolds Numbers and j and f Factors. Compute the Reynolds number on each uid side
from its denition as
GDh
14:06 kg=m2 s 0:00154 m
Reg
551
g
0:0000393 Pa s
Rea
GDh
18:56 kg=m2 s 0:00154 m
824
a
0:0000347 Pa s
From Fig. E9.1B, (or the curve t of j and f data), determine the j and f factors for
these Reynolds numbers as follows:
Gas
Air
Re
551
824
0.0174
0.0135
0.0695
0.0539
Since Reynolds numbers indicate the ow as laminar on both gas and air sides, the
correction to the j factor for the temperature-dependent property eects is unity because
n 0 from Table 7.12.
627
Heat Transfer Coecients, Fin Eectivenesses, and Overall Heat Transfer Coecient. We
compute the heat transfer coecient from the denition of the j factor as follows:
hg
ha
jGcp
Pr2=3 g
jGcp
Pr2=3
Let us calculate mg and ma in order to calculate the n eciency on each uid side. Since
the oset ns are used on both gas and air sides, we use Eq. (4.147) with Lf replaced by s
to take into account strip edge exposed area.
mg
2h
1=2
2 339:88 W=m2 K
0:102 mm 1=2
1
1
kf
s g
3:175 mm
18 W=m K 0:102 103 m
618:17 m1
2h
1=2
2 346:36 W=m2 K
0:102 mm 1=2
1
1
ma
kf
s a
3:175 mm
18 W=m K 0:102 103 m
624:04 m1
a g
b
2:49 mm=2 0:102 mm 1:143 mm 0:00114 m
2
Thus
f ;g
mg
618:17 m1 0:00114 m
f ;a
ma
624:04 m1 0:00114 m
The overall surface eciencies with Af =A values from Fig. E9.1B or input are
Af
1 1 0:8609 0:785 0:8908
A
Af
1 1 f
1 1 0:8592 0:785 0:8895
A
o;g 1 1 f
o;a
To calculate Ua from Eq. (9.30), we need to rst calculate a and g using Eq. (9.31):
a
ba
2:54 mm 2254 m2 =m3
941:6 m2 =m3
g
bg
2:54 mm 2254 m2 =m3
941:6 m2 =m3
628
Hence,
Aa a 941:6 m2 =m3
1:0
Ag g 941:6 m2 =m3
Thus Ug from Eq. (9.30), with no fouling, is
g =a
1
1
1
1:0
1:66 kg=s
m_
0:1181 m2
G a 14:06 kg=m2 s
The air ow length is then computed from the denition of the hydraulic diameter:
Lg
Dh A
0:00154 m 85:95 m2
0:280 m
4Ao g
4 0:1181 m2
2:00 kg=s
m_
0:1078 m2
G a 18:56 kg=m2 s
and
La
Dh A
0:00154 m 85:95 m2
0:307 m
4Ao a
4 0:1078 m2
To calculate the core frontal area on each uid side, we rst need to determine Dh =4
as
a
0:363
4
4
g
0:363
4
4
629
Hence,
Afr;g
Ao;g 0:118 m2
0:3253 m2
g
0:363
Afr;a
Ao;a 0:1078 m2
0:2970 m2
a
0:363
Afr;a 0:2970 m2
Afr;g 0:3253 m2
1:061 m or L3
1:060 m
Lg
0:280 m
La
0:307 m
Kc 0:36
Ke 0:42
Air side:
Kc 0:36
Ke 0:42
To correct f factors for the temperature-dependent property eects, let us rst calculate
Tw . The thermal resistances on the hot and cold uid sides are
Rg
1
1
Ra
1
1
Therefore,
Rg 3:843 105 K=W
1:018
Ra 3:776 105 K=W
Now
Tw
528:78C 801:8 K
1 Rg =Ra
1 1:018
Since the gas is being cooled, using Eq. (9.11) and the exponent m 0:81 from Table
7.12, the corrected f factor is
m
T
801:8 K 0:81
0:0695
0:0645
fg fcp w
Tm
879:8 K
g
630
Since the air is being heated, using Eq. (9.11) and the exponent m 1:00 from Table 7.12,
we have
m
T
801:8 K 1:00
fa fcp w
0:0539
0:0596
Tm
725:2 K
a
The pressure drops, using Eq. (9.20), are
"
!
14:06 kg=m2 s2
0:4751 kg=m3
2
1 0:363 0:36 2
1
pg
2 1 0:4751 kg=m3
0:8966 kg=m3
#
0:0645 0:280 m 0:4751 kg=m3
0:4751 kg=m3
2
1 0:363 0:42
0:0596 0:307 m 1:4726 kg=m3
1:4726 kg=m3
2
1
0:363
0:42
0:00154=4 m 0:9319 kg=m3
0:6817 kg=m3
G2g
36:0522 ! Gg 15:44 kg=m2 s
2 1 0:4751 kg=m3
8:79 103 Pa
G2a
77:6803 ! Ga 18:26 kg=m2 s
2 1 1:4726 kg=m3
Knowing the rating problem solution, we can see that this new value of Gg and Ga have
almost converged to the true values. Repeating Steps 2 through 4 of section 9.2.2.3 with
the new values of G yields the following results:
Iterations
NTU
Lg
La
L3
Original
First
Second
Third
7.079
7.079
7.079
7.079
0.280
0.300
0.300
0.300
0.307
0.295
0.297
0.299
1.061
1.009
1.003
1.003
631
Thus with the iterations, the solution can be converged to the actual core dimensions
within any desired accuracy.
Discussion and Comments: The foregoing method clearly indicates how fast a solution to
the sizing problem will converge to the core dimensions that will meet the heat transfer
and pressure drops on both sides for a crossow exchanger when no constraints are
imposed on the dimensions. However, with the imposed constraints on the core dimensions, the design will not meet the heat transfer and pressure drops specied on both uid
sides, and a constraint on the geometric parameters or operating condition variables
must be relaxed. Also, in a sizing problem, one would like to nd the optimum set of
core/surface geometries, and/or operating conditions for the problem specication. In
that case, it becomes an optimization problem. We discuss it in Section 9.6.
9.3
Tube-n exchangers are mostly used as single-pass crossow or multipass cross-counterow exchangers. A heat pipe heat exchanger is eectively two tube-n single-pass crossow exchangers placed side by side separated by a splitter plate and connected to each
other by the same tubes, which are heat pipes. Fluids (usually air and exhaust gas) ow in
counterow (opposite) directions (crossow to nned tubes). The design theory for heat
pipe heat exchangers has been presented by Shah and Giovannelli (1988), and will not be
discussed here.
The solution procedures for the rating and sizing problems for tube-n exchangers
with individually nned tubes or at ns (see Fig. 1.31), either in single-pass crossow or
two-pass cross-counterow, are identical to those for plate-n exchangers described in
detail in preceding sections. Hence, rather than repeating the same steps, only the dierences are highlighted.
9.3.1
Surface Geometries
In this case, the surface area density , a ratio of total transfer surface area A on one side
of the exchanger to total volume V of the exchanger is used for heat transfer surfaces used
in tube-n exchangers. Hence, , , and Dh are computed from Eqs. (8.51)(8.53). Note
that Eq. (9.31) and the rst equality of Eq. (9.37) have no physical meaning for tube-n
exchangers. The heat transfer surface area density does not have signicance for the
tube-n exchangers.
9.3.2
All heat transfer equations, except for Rw and f , remain the same as those for plate-n
exchangers. The overall thermal resistance Eq. (9.16) should include a term for contact
resistance if ns are wrapped tension wound or mechanically expanded onto the tubes.
Also, the wall thermal resistance term should be for a tube. For a circular tube, it is as
given in Eq. (3.26).
The n eciency for circular ns of Fig. 1.31a or at ns of Fig. 1.31b is dierent from
that for the straight ns [Eq. (9.14)]. For circular ns of Fig. 1.31a, the n eciency is
given by Eq. (4.151), and an approximate formula, which does not involve Bessel functions, is given in Table 4.5. The n eciency of at ns is obtained by an approximate
632
method referred to as the sector method, discussed in Section 4.3.2.3. The details of how
to evaluate the n eciency for this case are also presented there and hence are not
repeated here.
9.3.3
The core pressure drop Eq. (9.20) for plate-n exchangers needs to be modied for tuben exchangers for the tube outside, as discussed in Section 6.2.2.2. For individually
nned tube exchangers, the entrance and exit pressure losses cannot readily be measured
and evaluated separately. Hence, they are lumped into experimentally determined
friction factors. In this case, the pressure drop is computed from Eq. (6.32) if f is the
Fanning friction factor based on the hydraulic diameter. If, instead, the Euler number or
Hagen number per tube row for the tube bank is used for the pressure drop evaluation,
Eq. (6.33) should be used for the pressure drop calculation.
For continuous at ns, the pressure drop components of Eq. (9.20) are all valid.
However, while the entrance and exit pressure losses are evaluated based on the ow area
at the leading and trailing edges of the ns, the core friction and momentum eect terms
are based on G computed from the minimum free-ow area within the core. Thus, Eq.
(6.36) should be used, instead of Eq. (9.20), for pressure drop evaluation of continuous
at ns on tubes.
9.3.4
For a tube-n exchanger with at ns, if the ow friction and heat transfer correlations
are based on the hydraulic diameter on the tube outside, the core mass velocity equation
of Eq. (9.28) is also valid for the tube outside. For nned tubes, there are a number of
dierent ways of correlating heat transfer and ow friction characteristics, such as that
Nu is based on the tube outside diameter and the pressure drop is based on the Euler
number or Hagen number. In that case, the core mass velocity equation should be
derived for the specic cases depending on the nature of the correlations, such as
using the Nu/Eu ratio instead of the j=f ratio.
9.4
Plate heat exchangers can be designed with m passes on the uid 1 side and n passes on
the uid 2 side, depending on ow rate imbalance, available pressure drop, and other
design criteria. One of the most common ow arrangements is 1-pass 1-pass counterow
design, selected for reasonably balanced ow rates on hot and cold uid sides. If the ow
maldistribution within the PHE is ignored and if all plates have the same geometry, the
rating of this exchanger is identical to that for the counterow plate-n exchanger
described in Section 9.2.1 except that o and f are unity since there are no ns in a
PHE. Unlike plate-n exchangers, sometimes it is not possible to meet the pressure drop
and heat transfer specied even on one side of a PHE. So let us rst discuss this condition
and the limiting cases for the design of a PHE; these limiting cases involve limitations
imposed on the specied heat transfer and/or pressure drops. For a PHE with mixed
plate design (i.e., having two dierent plate geometries), the channel-to-channel ow
maldistribution must be taken into account for rating and sizing. Detailed analysis of
633
Let us discuss how the design of a 1 pass1 pass counterow PHE diers from that of a
pure counterow plate-n exchanger. For a plate-n exchanger, the minimum free-ow
area Ao and the surface area A on each uid side are independent of each other. For
example, for a specied (or selected) Ao , the surface area A could be varied by changing
the n density (this change has a minor eect on Ao ). As a result, both heat transfer and
pressure drop on one of the two uid sides can be matched exactly. The design method
for plate-n exchangers then involves the coupling of specied NTU and p through a
core mass velocity equation as presented by Eq. (9.28). This approach cannot be used for
plate heat exchangers since Ao and A are not independent. Once Ao is xed (i.e., the
number of plates is selected), A is xed automatically (A 4Ao L=De ) for a specied plate
geometryy (the plate pattern, De , and L specied) because there is no secondary (or
extended) surface. Hence, in most cases, it is not possible (unless mixed channels are
used as discussed later) to match specied pressure drop and heat transfer identically
even on one uid side. The plate exchanger design in general is either pressure drop or
heat transferlimited. In the pressure droplimited design, the free-ow area is determined that satises the pressure drop limit; however, the corresponding surface area will
be higher than that required to meet the heat duty. In the heat transferlimited design, the
surface area specied transfers the required heat duty; however, the corresponding freeow area will be higher than that required to take advantage of the available pressure
drop on either uid side. Hence, the resulting pressure drops on both uid sides will be
lower than specied. This is explained further through Fig. 9.2 next.
A channel (a ow passage) in a PHE is made of two plates. Two given plates (with
two dierent chevron angles) can be used to obtain three dierent channel types. For
example, a plate type has two chevron angles low and high ; three channel (plate) combinations are possible: low and low , low and high , and high and high . For discussion
purposes, let us refer to them as channel types 1, 2, and 3, respectively. It is worth noting
that there exists an apparent chevron angle app for a mixed-plate channel, which is
approximately equal to low high =2. This approximate relationship can be unsatisfactory when a 908 plate is mixed with another plate having ( < 908 to form a channel. A
more precise value of app can be obtained experimentally by testing the mixed-plate
channels, determining their j (or Nu) and f vs. Re characteristics, comparing them with
the data for one type of chevron plate of various , and nding the value of the closest
match to the data. However, no such app values are available in the open literature.
For a given design, there exists an eective (or ideal) chevron angle eff that will meet
the design criteria: the heat transfer specied, the pressure drop on one uid side
matched, and the pressure drop on the other uid side lower than the value specied.
In general, it is not possible to match the pressure drop on both sides because of the
limited number of plate geometries available in a given size. In reality, the value of eff
(consider it less than app for discussion purposes) indicates whether the heat exchanger
should have channel types 1 and 2 or 2 and 3: In other words, if eff < app , channel types
1 and 2 should be used; and if eff > app , channel types 2 and 3 should be used in the
y
There are a limited number of plate patterns available, due to the very high cost of tooling and of the press
required to stamp the plates.
634
q2
p2
qr
2
2
pr
q1
i
1
p1
1
A1 A2
Aid
A1
A2
A
FIGURE 9.2 Heat-transfer-limited (channel type 1) and pressure-drop-limited (channel type 2)
designs. Curves labeled 1 have low low plates; curves labeled 2 have high high plates. (From
Shah and Focke, 1988.)
PHE. In both cases, a specic combination of one type of plates and mixed plates are
chosen to meet the required qr or pr on one uid side, depending on whether the design
is heat transfer or pressure droplimited.
We now explain the concept of heat transfer and pressure droplimited designs and
its relation to the concept of mixing the plates using Fig. 9.2. Consider two dierent
channel types as possible candidates to meet the required heat duty qr and pressure drops
pr for the two uid streams. Generally, in a heat exchanger, one uid stream has a more
severe pressure drop restriction than the other stream. We consider only that uid stream
having the more severe pressure drop constraint,y designated as pr (required or specied p). The heat transfer rate and pressure drop (on the more constrained side) as a
function of the surface area (or the number of thermal plates) for these two channels are
shown in Fig. 9.2 with solid line curves labeled as 1 and 2, with heat transfer rate and
pressure drop scales on the left- and right-hand y axes, respectively. Also the specied qr
and pr are shown in the same gure by horizontal long dashed lines. The following
discussion assumes that eff lies between those given by channel types 1 and 2.
y
Note that the uid side having a severe pressure drop constraint does not necessarily have a lower pressure drop
specied than that on the other uid side; in fact, it could have higher pressure drop. All it means is that we have to
ensure that the design pressure drop is equal or lower than the pressure drop specied.
635
Channel Type 1. As shown in Fig. 9.2, heat transfer of channel type 1 matches the
required heat transfer qr (see the intersection of dashed line qr and solid line q for surface
1) with surface area A1 and has pressure drop p10 ; thus, it does not utilize the available
pressure drop since p10 < pr . If the entire specied pressure drop would have been
utilized (consumed), the surface area required would be A10 , but the heat transfer rate of
the exchanger would be only q10 , which is signicantly lower than qr . Hence, the exchanger design with channel type 1 is limited by the surface area A1 to transfer the required qr .
Hence, it is designated as a heat transferlimited design.
Channel Type 2. Pressure drop of channel type 2 matches the pressure drop requirement
pr with surface area A2 , but in doing so, it utilizes much more free-ow area (and hence
surface area), which yields the heat transfer rate q20 . This is higher than the value required.
If the heat transfer would have been matched, the surface area required would have been
A20 , but the resulting pressure drop would have been p20 , which is signicantly higher
than pr . Hence, the exchanger design with channel type 2 is limited by the surface area
A2 to meet the specied pr constraint. Hence, it is designated as a pressure droplimited
design.
Mixed Channels. From Fig 9.2, it is clear that the ideal channel, designated as i, which
meets the heat transfer and pressure drop specications simultaneously, will require the
ideal amount of the surface area Aid . Thus, a proper mixing (combination) of channel
types 1 and 2 will yield the q and p curves as those indicated by i in Fig. 9.2. That design
would enable the designer to satisfy the design heat duty and pressure drop on one of the
two streams considered.
The use of two channel types in a given pass creates channel-to-channel ow maldistribution discussed in Section 9.4.2 and thereby a reduction in heat duty. Marriot (1977)
reports that the eect of maldistribution of this type on q is typically less than 7%. In
practical situations, a design based on mixed channels should be rated to quantify the
eects of ow maldistribution (refer to Sections 12.1.2 and 12.1.3 for detailed discussion
of ow maldistribution). If these eects are too severe, a pressure droplimited design
(using uniform channels giving a chevron angle higher than ideal) or a heat transfer
limited design (using uniform channels, giving a lower chevron angle) may be preferable
to a mixed-channel design.
9.4.2
Since 1 pass1 pass counterow PHE is the most common in application, its rating and
sizing can be accomplished by using the methods described in Sections 9.2.1 and 9.2.2
for plate-n exchangers, if the ow distribution is assumed uniform through all ow
channels. However, due to the nature of exchanger construction, it leads to several ow
maldistributions: within the channel, channel to channel, and manifold induced. These
maldistributions are described further in some detail in Sections 12.1.2 and 12.1.3. To
explain the rating procedure for a PHE with a mixed-plate design, we must consider, as a
minimum, channel-to-channel ow maldistribution. This type of ow maldistribution
occurs due to the presence of two dierent plate groups in a PHE. For example, consider
two types of plates used in a PHE: part of an exchanger made up with all 308 chevron plates
and the rest with alternating 308 and 608 chevron plates (i.e., having mixed-plate channels).
In such a heat exchanger, in addition to having manifold-induced maldistribution in
any plate group, the ow will be maldistributed among dierent plate groups, due to
636
their dierent ow resistance (such as f vs. Re) characteristics. This can be quantied
readily as shown below on a given uid side if we imagine that the pressure drop across
all channels (all plate groups) on a given uid side is the same. Hence, we rst summarize
the theory as to how to determine dierent ow rates through two groups of plates.
Subsequently, we show how to compute the heat transfer rate of this mixed-plate PHE.
Heat transfer rate (heat duty) for a PHE can be determined by idealizing the two
dierent plate groups in a PHE exchanger as two exchangers in parallel coupling (Fig.
9.3). Hence, one needs to determine the individual mass ow rates through these two
plate groups rst. Consider the same core (frictional) pressure drop for each plate group,
and neglect manifold and port pressure drops and momentum and elevation change
eects; using the core frictional term only [the second term on the right-hand side of
Eq. (6.44)], we get
fI G2I
f G2
II II
De;I
De;II
9:41
where subscripts I and II denote plate groups I and II and De is the equivalent diameter
[see Eq. (8.135) for the denition]. We consider that the friction factor can be represented
as
f a Ren
9:42
Combining Eqs. (9.41) and (9.42), and noting that m_ GAo , the ratio X of the mass ow
rates through plate groups I and II can be presented as follows:
X
m_ I
m_ II
1=2n n=2n
De;I 1n=2n Ao;I
aII
II
aI
I
De;II
Ao;II
9:43
where Ao;I and Ao;II are the total free-ow areas in plate groups I and II, respectively and
can readily be calculated with a known number of plates or channels in each plate group.
Then, from the mass balance, the total mass ow rate is
m_ m_ I m_ II
9:44
X m_
1X
m_ II m_ m_ I
m_
1X
low = 30
high = 60
Group I
Group II
FIGURE 9.3 Idealized counterow PHE with two plate groups in parallel.
9:45
637
Once the individual ow rates are determined, the pressure drop for each plate group can
be determined from the last three terms on the right-hand side of Eq. (6.44). The manifold and port pressure drops then should be added to get the total pressure drop on each
uid side. If the f Re correlation is about the same, ideally the total pressure drop on one
uid side of a PHE will be lower for the two dierent plate groups compared to that for
only one plate group. This can readily be understood with an electric analogy that an
electric circuit having two dierent resistances (such as 4 and 8 ) in parallel will have a
lower electric potential than that of an electric circuit having two identical electrical
resistances (6 and 6 , a mean value for the two individual resistances of the rst circuit)
in parallel with the same total electric current, despite the fact that the sum of individual
resistances is equal.
Once the total ow rates in each channel group are determined, the heat transfer
analysis is straightforward by considering two exchangers in parallel, as shown in Fig.
9.3, corresponding to two plate groups. The temperature eectiveness of each plate group
for a counterow exchanger is given by
P1;I
8
1 expNTUI 1 RI
>
>
>
< 1 R expNTU 1 R
I
I
I
>
> NTUI
>
:
1 NTUI
for RI 6 1
9:46
for RI 1
where
NTUI
UAI
UAI
C1;I
m_ I cp;I 1
RI
C1;I m_ I cp;I 1
C2;I m_ I cp;I 2
9:47
and
1
1 1
^ 1; f w R
^ 2; f 1
R
kw
UAI AI h1
h2 I
9:48
with AI A1;I A2;I Aw;I . Similarly, the temperature eectiveness P1;II of the second
plate group can be expressed in terms of NTUII and RII dened in the same manner. The
total exchanger heat duty is then given by
q qI qII P1;I C1;I P1;II C1;II Th;i Tc;i
9.4.3
9:49
Rating a PHE
We now present a rating procedure or determination of heat transfer and pressure drop
performance of a PHE that has two plate groups. In group I, all plates have the same
chevron angle (such as low ); in group II, two plate geometries (such as having low and
high are stacked alternately, thus having a mixed-plate pack (see Fig. 9.3). Since the
performance of a given unit is to be determined, the following quantities are specied:
. Exchanger geometry (i.e., plate width and length, channel gap, number of plates,
types of plates and how the mixing of plates is achieved in the given exchanger, etc.)
638
. Plate surface pattern with their heat transfer and pressure drop characteristics
. Flow arrangement of the two uids (i.e., the number of passes on each uid side and
overall uid ow direction)
. Mass ow rates, inlet temperatures, uid physical properties, and fouling resistances for each uid stream
With the foregoing known information, the following is a step-by-step rating procedure. This procedure is outlined for a PHE having two plate groups. If there is only one
plate group, use the same procedure with all quantities for the plate group II ignored.
1. Calculate uid properties , , k, and cp ) at the bulk mean temperature for each
uid side.
2. Compute m_ I and m_ II for both uids from Eq. (9.45).
3. Determine Re for both uids in each plate group.
4. Calculate hh and hc for both plate groups using the specied Nu or j vs. Re
correlations.
5. Compute (UAI using Eq. (9.48). Similarly, compute UAII .
6. Calculate NTUI and RI using Eq. (9.47). Similarly, calculate NTUII and RII .
7. Determine P1;I using Eq. (9.46). Similarly, determine P1;II .
8. Compute the heat duty q using Eq. (9.49).
9. Calculate f factors from Eq. (9.42).
10. Determine the combined channel pressure drops and other pressure drop components from Eq. (6.44) for both uid sides.
Next we illustrate this procedure with one rating example.
Example 9.3 A 1 pass1 pass counterow water-to-water plate heat exchanger has 47
thermal plates or 48 uid channels (24 channels for each uid). On each uid side,
chevron plates of 308 are used for 8 channels and 308 and 608 mixed chevron plates
are used for 16 channels. Assume that eff 39:88 and the following are empirical
correlations for the Nusselt and Reynolds numbers based on De .
0:646
Nu 0:724
Re0:583 Pr1=3
308
(
f
0:80Re0:25
3:44Re0:25
for 308
for 30 and 60 mixed plates
Hot Fluid
Cold Fluid
Water
18
40
Water
10
20
639
30
30
0.5
0.1
7 103
38
20
1.1
3.5
0.55
Fluid properties [use the same constant properties (for simplicity) on both hot- and cold-uid
sides]:
Dynamic viscosity Pa s
Thermal conductivity (W/m K)
8:1 104
0.619
Additional information:
Total fouling resistance 4 105 m2 K W
Density (kg/m3 )
Specic heat (J/kg K
995.4
4177 Pr = 5.47
SOLUTION
Problem Data and Schematic: The detailed process, geometry, and other data are provided in the problem statement for the PHE.
Determine: The heat transfer and pressure drop on each uid side for this exchanger.
Assumptions: The assumptions listed in Section 3.2.1 applicable to a PHE are invoked.
Analysis: We follow the steps outlined in Section 9.4.3 for the solution after we compute
ow and heat transfer areas as follows:
Ao;I Ao;c;I Ao;h;I W2aNc;I 0:5 m 3:5 103 m 8 0:014 m2
Ao;II Ao;c;II Ao;h;II W2aNc;II 0:5 m 3:5 103 m 16 0:028 m2
Ac;I Ah;I 0:55 m2 8 2 8:8 m2
Ac;II Ah;II 0:55 m2 16 2 17:6 m2
Since we have one type of chevron plates 308 in group I of the PHE and mixed
chevron plates 308 and 608) in group II of the PHE, let us rst evaluate the ow
distribution of each uid in these sections using Eq. (9.43). For this equation, the ratio of
dynamic viscosities and equivalent diameters will be unity from the problem statement.
Hence, from Eq. (9.43), we get
1=2n n=2n
De;I 1n=2n Ao;I
II
I
De;II
Ao;II
1=20:25
3:44
0:014 m2
1 1
1:151
0:80
0:028 m2
aII
aI
where the values for aj , j I or II, and n of Eq. (9.42) are given in the problem formulation. The mass ow rates from Eq. (9.45) are then
640
m_ c;I
X m_ c
1:151 10 kg=s
1X
1 1:151
5:351 kg=s
m_ h;I
X m_ h
1:151 18 kg=s
1X
1 1:151
9:632 kg=s
Gc;I
m_ c;I
5:351 kg=s
Ao;c;I
0:014 m2
382:21 kg=m2 s
Gh;I
m_ h;I
9:632 kg=s
Ao;h;I
0:014 m2
688:00 kg=m2 s
Gc;II
m_ c;II
4:649 kg=s
Ao;c;II
0:028 m2
166:04 kg=m2 s
Gh;II
m_ h;II
8:368 kg=s
Ao;h;II
0:028 m2
298:86 kg=m2 s
Rec;I
8:1 104 Pa s
Rec;II
8:1 104 Pa s
Reh;I
8:1 104 Pa s
Reh;II
8:1 104 Pa s
Now calculate the heat transfer coecients on the cold and hot sides for groups I and
II using the given correlation for Nu written in terms of heat transfer coecients.
0:646
k
I
hc;I 0:724
Re0:583
Pr1=3
c;I
308
De
0:619 W=m K 308 0:646
0:724
33030:583 5:471=3
308
7 103 m
12,701 W=m2 K
641
0:619 W=m K 39:88 0:646
hc;II 0:724
14350:583 5:471=3 9377 W=m2 K
308
7 103 m
0:619 W=m K 308 0:646
hh;I 0:724
59460:583 5:471=3 17,894 W=m2 K
308
7 103 m
0:619 W=m K 39:88 0:646
hh;II 0:724
25830:583 5:471=3 13,210 W=m2 K
308
7 103 m
The overall conductance (UA) for each group is computed from Eq. (9.48) as follows:
1
1
1
^ c; f w R
^ h; f 1
R
kw
UAI Ac;I hc;I
hh;I
1
1
4 105 m2 K=W
2
8:8 m 12,701 W=m2 K
1
3 106 m2 K=W
17,894 W=m2 K
2:0184 105 K=W
or
UAI 49,544 W=K
Similarly,
1
1
1
13,210 W=m2 K
1:2804 105 K=W
or
Next, determine NTU and R for groups I and II from their denitions.
NTUI
UAI
49,544 W=K
2:217
m_ cp c;I 5:351 kg=s 4177 J=kg K
NTUII
UAII
78,103 W=K
4:022
m_ cp c;II 4:649 kg=s 4177 J=kg K
RI
0:556
m_ h;I cp 9:632 kg=s 4177 J=kg K
RII
0:556
m_ h;II cp 8:368 kg=s 4177 J=kg K
642
The temperature eectiveness for the two groups of the counterow exchanger,
P1;I and P1;II are given by Eq. (I.1.1) of Table 3.6.
P1;I
1 expNTUI 1 RI
1 exp2:217 1 0:556
0:7906
1 RI expNTUI 1 RI 1 0:556 exp2:217 1 0:556
P1;II
1 expNTUII 1 RII
1 exp4:022 1 0:556
0:9179
1 RII expNTUII 1 RII 1 0:556 exp4:022 1 0:556
Finally, the heat transfer rate from the hot water to cold water in this exchanger is given
by
q P1;I m_ c;I cp Th;i Tc;i P1;II m_ c;II cp Th;i Tc;i
0:7906 5:351 kg=s 4177 J=kg K 40 20 K
0:9179 4:649 kg=s 4177 J=kg K 40 20 K
353:4 103 W 356:5 103 W 710 kW
Ans:
To compute the pressure drop, the friction factors for the Reynolds numbers above
can be computed from Eq. (9.42) as follows:
fc;I 0:8033030:25 0:1055
Now we compute the pressure drop associated within the plate pack on the cold and hot
sides using Eq. (6.29) with given m :
pc;I
4fc;I LG2c;I
4 0:1055 1:1 m 382:212
4866 Pa
2gc De
2 1 995:4 kg=m3 7 103 m
pc;II
ph;I
ph;II
Theoretically, pc;I pc;II and ph;I ph;II . As found above, this is true within the
round-o error margins. Thus we consider the following values for the pressure drop
associated with this plate pack.
pc 4866 Pa
ph 13,615 Pa
The other components of the pressure drop are the momentum eect, the elevation
change eect, and the inlet and outlet manifolds and ports pressure drops. The rst two
eects are negligible for this case (no change in the density and negligible elevation
643
change for 1.1-m-long plates) and are ignored. For the manifold and port pressure drop
component, the corresponding mass velocities are
Gc;p
10 kg=s
m_ c
1273 kg=m2 s
=4D2p =40:1 m2
Gh;p
18 kg=s
m_ h
2292 kg=m2 s
=4D2p =40:1 m2
The manifold and port pressure drops for the cold and hot uid sides are computed from
the rst term on the right-hand side of Eq. (6.44):
pc;p
1221 Pa
2gc i
2 1 995:4 kg=m3
ph;p
Ans:
Note that the pressure drops associated with the manifold and port on the cold and
hot uid sides are 20% [ 1221=6087 100 and 22.5% 3958=17,573 100;
respectively, of the total pressure drop on individual uid sides.
To compare the eect of mixed-plate performance to that for a single-plate geometry
PHE, let us recalculate the performance of a similar PHE of the same number of plates
(or channels), but all made from 308 chevron plates. The ow and surface areas on one
side of that PHE are
Ao;c Ao;h W2aNc 0:5 m 3:5 103 m 24 0:042 m2
Ac Ah 0:55 m2 24 2 26:4 m2
The mass velocities, Reynolds numbers, and heat transfer coecients on both uid sides
are as follows:
Gc
10 kg=s
m_ c
238:10 kg=m2 s
Ao;c 0:042 m2
Gh
18 kg=s
m_ h
428:57 kg=m2 s
Ao;h 0:042 m2
Rec
8:1 104 Pa s
Reh
8:1 104 Pa s
644
k
0:619 W=m K
20580:583 5:471=3
Re0:583
hc 0:724
Pr1=3 0:724
c
De
7 103 m
9640 W=m2 K
k
0:619 W=m K
1=3
hh 0:724
37040:583 5:471=3
Pr
0:724
Re0:583
h
De
7 103 m
13,579 W=m2 K
The overall thermal conductance UA is given by
1
1 1
^ c; f w R
^ h; f 1
R
kw
UA Ac hc
hh
1
1
4 105 m2 K=W
26:4 m2 9640 W=m2 K
1
8:3476 106 K=W
3 106 m2 K=W
13,579 W=m2 K
or
UA 119,790 W=K
Now we compute NTU, R, P, and q for this PHE.
NTU
UA
119,790 W=K
2:868
m_ cp c 10 kg=s 4177 J=kg K
0:556
m_ cp h 18 kg=s 4177 J=kg K
1 expNTU1 R
1 exp2:868 1 0:556
0:8528
1 R expNTU1 R 1 0:556 exp2:868 1 0:556
fh 0:8037040:25 0:1025
pc;core
2gc De
2 1 995:4 kg=m3 7 103 m
ph;core
645
Discussion and Comments: This example illustrates how to rate a PHE when it has a plate
pack consisting of one type of plates in one group and mixed plates in another group,
with the following results:
Group
m_ c (kg/s)
m_ h (kg/s)
q (kW)
pc;core (kPa)
ph;core (kPa)
5.351
4.649
9.632
8.368
353.4
356.5
4.866
4.865
13.615
13.607
I
II
We have also compared the performance of this exchanger with the one having all
chevron plates of 308. Following is the comparison of main results:
PHE
m_ c (kg/s)
m_ h (kg/s)
q (kW)
pc;core (kPa)
ph;core (kPa)
10
10
18
18
710
712
4.866
2.126
13.615
5.944
Mixed Plates
Single Plates
For the present problem, the friction factors for a given Re are over four times
larger for the mixed-plate group II. This, in turn, reduces the ow to only 46.5%
100 8:368 kg=s=18 kg=s of the total ow in group II despite the ow area being
double for group II than for group I. Hence, even though the heat transfer coecient for
the mixed-plate region is larger than that for the single-plate region at a given Re, the
reduction in the ow reduces h, and as a result, the overall heat transfer is about the same
(710 vs. 712 kW). Hence, it is important to keep in mind that the mixed-plate section
should not have excessive friction factors which would otherwise defeat the advantage of
having a mixed-plate section.
9.4.4
Sizing a PHE
When sizing a PHE, we have very little choice in the selection of plate dimensions, unlike
plate-n and tube-n heat exchanger designs, because we cannot arbitrarily select a plate
width W or plate length L. Instead, we should select from a relatively small pool of
available plate sizes from any manufacturer. As the dies used for forming the plates
are extremely expensive, each manufacturer oers only up to about 30 plate sizes. In
selecting an appropriate plate size, we may compute the uid velocity in the port and
limit this value to a maximum of 6 m/s (20 ft/sec), as a rough rule of thumb. Further,
the manifold and port pressure drops may not be allowed to exceed a certain percentage
(typically 10% but up to 30% in rare cases) of the total pressure drop. Most plate sizes
are generally available only in two chevron angles. However, by mixing plates of
dierent chevron angles in various proportions, the designer is able to obtain considerable exibility in eff for any PHE.
Two methods are published in the literature (Shah and Focke, 1988; Shah and
Wanniarachchi, 1991) for sizing a PHE. However, we do not describe them here because
(1) those procedures are quite involved, (2) engineers in the PHE industry use their own
proprietary computer programs with their own data for j and f factors, and (3) it is easy
to add or delete some plates if the designed PHE does not perform to the specications.
646
Since sophisticated proprietary computer programs are available for rating a PHE which
converge quickly, such programs have iterative rating schemes build into them to arrive
at a size to meet the specied heat transfer and/or pressure drop; the size will depend on
whether it is a heat transferlimited design or a pressure droplimited design, as discussed
in Section 9.4.1.
9.5
Accurate prediction of performance and design characteristics of conventional shell-andtube heat exchangers is more dicult than that for plate-n and tube-n exchangers. This
is due primarily to the complexity of shell-side ow conditions and the impact of that
complexity on heat transfer performance. There are many variables associated with the
geometry (i.e., baes, tubes, front- and rear-end heads, etc.) in a shell-and-tube heat
exchanger in addition to those for the operating conditions. So complete sizing (with a
unique design) of a shell-and-tube exchanger is not possible as for a plate-n exchanger,
described earlier. As a result, the common practice is to presume the complete geometry
of the exchanger and perform the rating of the exchanger to determine the tube (shell)
length if the heat duty is given, or outlet temperatures if the length is given. In both cases,
pressure drops are to be determined. Preliminary sizing (design) of a shell-and-tube heat
exchanger is possible based on a number of approximations and the experience of past
designs. Once the preliminary design is obtained, the design calculations are essentially a
series of iterative rating calculations made on the preliminary design until a satisfactory
design is achieved. In this section, we outline the basic steps of (1) a rating procedure with
an example, and (2) a preliminary design and subsequent iteration technique for sizing of
a shell-and-tube exchanger. It should be added that modern design practices are based
almost exclusively on sophisticated commercial or proprietary computer software that
takes into account many complex eects on the shell side that are beyond the simplied
methods presented here.
In this section, we start with how to compute the shell-side heat transfer and pressure
drop by taking into account various ow leakage, bypass, and other eects before
providing the rating procedure. These eects are taken into consideration by a widely
utilized method in the open literature referred to as the BellDelaware method. It was
originally reported by Bell (1963) for rating of fully tubed segmentally baed heat
exchangers with plain tubes based on the experimental data obtained for an exchanger
with geometrical parameters closely controlled. This method has been extended to rate
low-nned-tube E shell, no-tubes-in-window E shell, and F shell heat exchangers (Bell,
1988b; Taborek, 1998).
9.5.1
Heat transfer and pressure drop calculations constitute the key part of the rating or
design of an exchanger. Tube-side calculations are straightforward. The heat transfer
coecient is computed using available correlations for internal forced convection as
presented in Section 7.4, and Eq. (9.20) is used for pressure drop calculations. The
shell-side calculations, however, must take into consideration the eect of various leakage streams (A and E streams, Fig. 4.19) and bypass streams (C and F streams, Fig. 4.19)
in addition to the main crossow stream B through the tube bundle. Several methods
have been in use over the years, but the most accurate method in the open literature is the
647
BellDelaware method. The set of correlations for calculating shell-side heat transfer
coecients and pressure drops discussed next constitutes the core of the BellDelaware
method (Bell, 1988b).
9.5.1.1 Shell-Side Heat Transfer Coecient. In the BellDelaware method, the shellside heat transfer coecient hs is determined using Eq. (4.169) by correcting the ideal
heat transfer coecient hid for various leakage and bypass ow streams in a segmentally
baed shell-and-tube exchanger. The hid is determined for pure crossow in a rectangular tube bank assuming that the entire shell-side stream ow across the tube bank is
at or near the centerline of the shell. It is computed from the Nusselt number correlations of Eq. (7.117) or other appropriate Nu or j vs. Re correlations, modied for
property variation eects as outlined in Section 7.6.1. It is then corrected by ve
correction factors as follows:
hs hid Jc J Jb Js Jr
9:50
where
Jc correction factor for bae conguration (bae cut and spacing). It takes into
account heat transfer in the window and leads to the average for the entire heat
exchanger. It is dependent on the fraction of the total number of tubes in crossow
between bae tips. Its value is 1.0 for an exchanger with no tubes in the windows
and increases to 1.15 for small bae cuts and decreases to 0.65 for large bae cuts.
For a typical well-designed heat exchanger, its value is near 1.0.
J correction factor for bae leakage eects, including both tube-to-bae and baeto-shell leakages (A and E streams) with heavy weight given to the latter and credit
given to tighter constructions. It is a function of the ratio of the total leakage area
per bae to the crossow area between adjacent baes, and also of the ratio of the
shell-to-bae leakage area to tube-to-bae leakage area. If the baes are too
close, J will be lower, due to higher leakage streams. A typical value of J is in
the range 0.7 to 0.8.
Jb correction factor for bundle and pass partition bypass (C and F) streams. It varies
from 0.9 for a relatively small clearance between the outermost tubes and the shell
for xed tubesheet construction to 0.7 for large clearances in pull-though oating
head construction. It can be increased from about 0.7 to 0.9 by proper use of the
sealing strips in a pull-through bundle.
Js correction factor for larger bae spacing at the inlet and outlet sections compared
to the central bae spacing. The nozzle locations result in larger end bae spacing
and lower velocities and thus lower heat transfer coecients. Js usually varies from
0.85 to 1.0.
Jr correction factor for any adverse temperature gradient buildup in laminar ows.
This correction applies only for shell-side Reynolds numbers below 100 and fully
eective for Res < 20; otherwise, it is equal to 1.
The combined correction factor, made up of ve correction factors, in a well-designed
shell-and-tube exchanger is about 0.6 (i.e., a reduction of 40% in the ideal heat transfer
coecient). The combined correction factor can be as low as 0.4. Comparison with a
large amount of proprietary experimental data indicates that compared to measured
values, the shell-side hs predicted from Eq. (9.50) is from 50% too low to 200% too
high, with a mean error of 15% low (conservative) at all Reynolds numbers.
648
TABLE 9.2 Correction Factors for the Heat Transfer Coecient on the Shell Side by the Bell
Delaware Method
Correction Factors Cs
Jc 0:55 0:72Fc
2:2rlm
J 0:441 rs 1 0:441 rs e
Jb
1
Crb 12Nss 1=3
Js
Nb
1n
1 L
i
Nb 1 L
i
1n
L
o
L
o
L
i
Lb;i
Lb;o
; L
o
Lbc
Lbc
n
Jr
10=Nr;c 0:18
for Res 20
These correction factors were determined from well-controlled experiments, and the
results were presented graphically (Bell, 1963). Those correction factors have been curve
tted (Taborek, 1998) and are presented in Table 9.2.
The BellDelaware method can be used for a low-nned tube-bundle E shell, notubes-in-window E shell, and F shell exchangers. The modications for these extensions
are summarized briey next.
. External low-nned tubes are used when the shell-side heat transfer coecient is
low, such as with viscous liquids. In this case, the ideal heat transfer coecient for
low-nned tubes is computed from the appropriate correlations, such as Eq.
(7.135). Subsequently, the eective shell-side hs is calculated from Eq. (9.50).
. The no-tubes-in-window design is used to minimize/eliminate the ow-induced tube
vibration problem. In this case, the ow area Ao;w through one window is given by
Afr;w of Eq. (8.111) since Afr;t 0 in Eq. (8.117). The tube count for this exchanger
is given by Nt Fc , where Nt is the number of tubes for a fully tubed exchanger. The
fraction Fc of the total number of tubes in the crossow section is given by Eq.
(8.120). Also, Jc of Eq. (9.50) is unity and Nr;cw 0 for the calculation of Jr from
Table 9.2 for Eq. (9.50). The rest of the procedure remains the same.
. For the F shell exchanger, we have two-tube and two-shell passes by the use of a
longitudinal bae. If this bae is not welded on both sides to the shell, there will
be uid leakage from the upstream to the downstream pass on the shell side due to
649
the pressure dierence. Also, there will be heat leakage across the bae by heat
conduction from the hotter to colder side of the shell-side pass. These eects may
not be negligible in some cases. If we neglect these eects, the BellDelaware
method remains identical except that all ow and surface areas need to be reduced
by half compared to a single shell-side pass.
9.5.1.2 Shell-Side Pressure Drop. Similar to shell-side heat transfer, the shell-side
pressure drop is also aected by various leakage and bypass streams in a segmentally
baed exchanger. The shell-side pressure drop has three components: (1) pressure drop
in the central (crossow) section, pcr ; (2) pressure drop in the window area, pw ; and
(3) pressure drop in the shell-side inlet and outlet sections, pi-o (see Section 6.4.2.2). It
is assumed that each of the three components is based on the total ow rate, and that it
can be calculated correcting the corresponding ideal pressure drops.
The ideal pressure drop in the central section, pb;id , assumes pure crossow of the
uid across the ideal tube bundle. This pressure drop should be corrected for (1) the two
leakage streams A and E in Fig. 4.19 using the correction factor
, and (2) the bundle and
pass partition bypass ow streams C and F in Fig. 4.19 using the correction factor
b . The
ideal window pressure drop, pw , also has to be corrected for both bae leakage
streams. Finally, the ideal inlet and outlet section pressure drops, pio , are based on
an ideal crossow pressure drop in the central section. These pressure drops should be
corrected for bypass ow (correction factor
b ) and for uneven bae spacing in inlet
and outlet sections (correction factor
s ). Thus, the total shell-side pressure drop, from
Eq. (6.43), is given as
ps pcr pw pio Nb 1pb;id
b Nb pw;id
Nr;cw
2pb;id 1
Nr;cc b s
9:51
The formulas for pb;id and pw;id are given by Eqs. (6.37) and (6.39) respectively. The
Hagen number for Eq. (6.37) is obtained from Eq. (7.109). Various correction factors of
Eq. (9.51) are dened as follows:
0:4 to 0.5, although lower values are possible with small bae spacing.
s correction factor for inlet and outlet sections having dierent bae spacing from
that of the central section, in the range 0.5 to 2.
These correction factors, originally presented in graphical form (Bell, 1963, 1988b), are
given in Table 9.3 in equation form by Taborek (1998).
The combined eect of pressure drop corrections reduces the ideal total shell-side
pressure drop to 20 to 30% of the pressure drop that would be calculated for ow
through the corresponding exchanger without bae leakages and bundle bypass streams
(i.e., ps;actual 0:2 to 0.3 ps;id : Comparison with a large number of proprietary
experimental data indicate that compared to measured values, the shell-side ps
650
TABLE 9.3 Correction Factors for the Pressure Drop on the Shell Side by the BellDelaware
Method
Correction Factors,
s
exp1:331 rs rplm
Lb;c
Lb;o
2n 0
Lb;c
Lb;i
2n 0
computed from Eq. (9.51) is from about 5% low (unsafe) at Res > 1000 to 100% high at
Res < 10:
Despite the facts above, it should be emphasized that the window section contributes
high pressure drop [compared to the other components of Eq. (9.51)] with insignicant
contribution to heat transfer. This results in an overall lower heat transfer rate to
pressure drop ratio for the segmental bae exchanger than that for grid bae and
most newer shell-and-tube heat exchanger designs.
9.5.2
Rating Procedure
The following is a step-by-step rating procedure using the BellDelaware method (Bell,
1988b). For the rating problem, the detailed exchanger geometry is specied and we
determine the heat duty, outlet temperatures as well as pressure drops on both uid
sides. We then describe the changes in the solution method if the exchanger length is
to be determined.
1. Compute the surface geometrical characteristics on each uid side. This includes
shell-side ow areas in crossow and window zones as well as all leakage ow areas
and related information as detailed in Section 8.5. Also compute the tube-side ow
area, surface area, ratio of free ow to frontal area, and other pertinent dimensions.
2. Calculate the uid bulk temperature and uid thermophysical properties on each
uid side. Since the outlet temperatures are not known for the rating problem, they
are guessed initially. Unless it is known from past experience, assume the exchanger effectiveness as 50% for most single and multitube-pass shell-and-tube exchangers, or 60 to 75% for multishell-pass exchangers. For the assumed effectiveness,
calculate the uid outlet temperatures using Eqs. (9.9) and (9.10). Compute uid
mean temperatures on each uid side, depending on the heat capacity ratio C*,
as outlined in Section 9.1. Subsequently, obtain the uid properties (, cp , k, Pr,
and ) from thermophysical property books, handbooks, or Appendix A.
3. Calculate the Reynolds numbers (Re GDh =) and/or any other pertinent dimensionless groups (from the basic denitions) needed to determine the nondimensional heat transfer and ow friction characteristics (e.g., j or Nu and f, Eu, or Hg)
651
of heat transfer surfaces on each uid side of the exchanger. Subsequently, compute j or Nu and f, Eu or Hg factors. Correct Nu (or j) for variable uid property
effects in the second and subsequent iterations using Eqs. (9.11) and (9.12).
4. From Nu or j, compute the heat transfer coefcients for both uid streams from
the following equations:
h Nu
k
Dh
or
h jGcp Pr2=3
9:52
652
arrangement). Finally compute the surface area on the shell side from the following
equation.
As
q
Us FTlm
9:53
The required eective tube length of the exchanger is then calculated from L As = do Nt
and the number of baes required by using Eq. (8.126).
We now illustrate the rating methodology with an example.
Example 9.4 Determine heat transfer rate, outlet uid temperatures, and pressure
drops on each uid side for a TEMA E shell-and-tube heat exchanger with a xed
tubesheet and one shell and two tube passes. The tubes in the bundle are in 458 rotated
square arrangement. The uids are lubricating oil and seawater. Fouling factors for the
oil and water sides are 1:76 104 and 8:81 105 m2 K=W, respectively. The
geometric dimensions and operating properties are provided as follows. Assume mean
uid temperatures to be 638C and 358C for oil and water, respectively.
Shell-side inside diameter Ds 0:336 m
Tube-side outside diameter do 19:0 mm
Tube-side inside diameter di 16:6 mm
Tube pitch pt 25:0 mm
Tube bundle layout 458
Central bae spacing Lb;c 0:279 m
Inlet bae spacing Lb;i 0:318 m
Outlet bae spacing Lb;o 0:318 m
Bae cut c 86:7 mm or 25.8%
Tube material
admiralty (70% Cu, 30% Ni)
Operating conditions:
Oil ow rate m_ oil m_ s 36:3 kg=s
Oil inlet temperature Ts;i 65:68C
Oil side fouling factor
^ o; f 0:000176 m2 W=K
R
Fluid
Oil at 638C
Seawater at 358C
Density
s (kg/m3 )
Specic heat
cp (J/kg K)
Dynamic
Viscosity
(Pa s)
849
993
2094
4187
64:6 103
0:723 103
Thermal
Prandtl
Conductivity Number
k (W/m2 K
Pr
0.140
0.634
966
4.77
Use the DittusBoelter correlation [Eq. (7.80) in Table 7.6] for the tube-side heat
transfer coecient. Use the McAdams correlation [Eq. (7.72) in Table 7.6] for the tubeside friction factor. For the shell-side friction factor and Nusselt numbers, use the follow-
653
Fc 0:6506
Nr;cc 9
Ao;sb 0:001027 m2
Nb 14
Ao;tb 0:001995 m2
Ao;w 0:01308 m2
36:3 kg=s
m_ s
1108 kg=m2 s
Ao;cr 0:03275 m2
326
s
64:6 103
Now we compute Nus from the given correlation with Red Res . Note that we have not
calculated Tw , so we cannot calculate Prw . So in this iteration, we consider Prs Prw .
0:36
Nus 1:04Re0:4
d Prs
hid
Prs
Prw
0:25
1:04 3260:4 9660:36 125:0 W=m2 K
Nus k w 0:14 125:0 0:140 W=m2 K 0:14
921:0 W=m2 K
1
m
do
0:0190 m
654
rs
Ao;sb
0:001027 m2
0:3398
Ao;sb Ao;tb 0:001027 m2 0:001995 m2
rlm
0:0923
Ao;cr
0:03275 m2
rb
Ao;bp 0:00949 m2
0:2898
Ao;cr 0:03275 m2
Nss
Nss
1
0:1111
Nr;cc 9
Now we compute L
i and Lo for determining unequal bae spacing factor Js from
Table 9.2.
L
i
Js
Lb;i
Lb;o 0:318 m
L
1:14
o
Lb;c
Lb;c 0:279 m
1n
1n
Nb 1 L
L
14 1 1:140:4 1:140:4
o
i
0:9887
14 1 1:14 1:14
Nb 1 Li L
o
Finally, the adverse temperature gradient factor Jr 1 for Res 326 > 100.
Since all correction factors J are determined, the actual shell-side heat transfer
coecient is given by
hs ho hid Jc J Jb Js Jr 921:0 W=m2 K 1:018 0:8696 0:8669 0:9887 1
698:8 W=m2 K
This heat transfer coecient should be corrected for the uid property variations
as outlined in Section 7.6.1 once the wall temperature is calculated in the next
iteration.
655
Nt 102
51
2
2
2
d N 0:0166 m2 51 0:01104 m2
4 i t; p 4
7837 W=m2 K
0:0166 m
di
1
0:0190 m ln0:0190 m=0:0166 m
0:000176 m2 K=W
2 111 W=m K
698:8 W=m2 K
0:0190 m
1
0:0190 m
0:000088 m2 K=W
0:0166 m
7837 W=m2 K 0:0166 m
0:997 1
Cmax 76,012 W=K
656
0:1852
Cmin
Ct
75,785 W=K
63:08C
2
2
Tt;i Tt;o 32:2 37:48C
34:88C
2
2
Ts;m
Tt;m
Pressure Drop Calculations. To compute the idealized tube bundle pressure drop, we rst
calculate the ideal friction factor using the given formula:
d b
1:33 19:0 mm 1:72
fid 3:5 1:33 o Re0:476
3:5
3260:476 0:2269
s
pt
25:0 mm
where
6:59
6:59
1:72
0:52
1 0:14Res
1 0:14 3260:52
4fid G2s Nr;cc w 0:25 4 0:2269 1108 kg=m2 s2 9 0:25
1
5906 Pa
2gc s
m
2 1 849 kg=m3
b
pb;id
To calculate the pressure drop in the crossow section, we rst compute the correction
factors
b and
using the expressions from Table 9.3.
657
using D 3:7 for Res > 100 and rb 0:2898 as calculated for Jb earlier.
G2w
1754 kg=m2 s2
14 2 0:6 3
0:6527
2gc s
2 1 849 kg=m3
62,914 Pa
where
Gw
m_ s
Ao;cr Ao;w
1=2
36:3 kg=s
0:03275 m2 0:01308 m2 1=2
1754 kg=m2 s
Next let us determine pressure drop in inletoutlet sections using Eq. (9.51) after
computing
s from Table 9.3.
Lb;c
Lb;o
2n0
0
Lb;c 2n
0:279 m 20:2
0:279 m 20:2
1:5803
Lb;i
0:318 m
0:318 m
where n 0 0:2
Nr;cw
3
pio 2pb;id 1
b
s 2 5906 Pa 1
0:6524 1:5803
Nr;cc
9
16,237 Pa
Then
ps pcr pw pio 32,694 62,914 16,237 Pa 111,845 Pa 112 kPa
29:2% 56:3% 14:5%
Note that for this problem, the window section pressure drop is more than the crossow
section pressure drop, whereas the crossow section provides most of heat transfer. Thus
the window section results in excessive pressure drop with insignicant contribution to
heat transfer.
Tube-Side Pressure Drop. From Eq. (7.72) in Table 7.6,
0:046 37,6430:2 0:005593
f 0:046Re0:2
t
658
2pt
2 25:0 mm
4 fL
m_ 2t
2
2
pt
1
1
K
np
c
e
2gc t A2o;t di
18:1 kg=s2
4 0:005593 4:3 m
0:3
0:4
2
0:0166 m
2 1 993 kg=m3 0:01104 m2 2
Kc 0:3
Ke 0:4
for
The objectives of the approximate design (sizing) method for a given service of a shelland-tube heat exchanger are several-fold: (1) quick conguration and size estimation, (2)
cost estimation, (3) plant layout, or (4) checking the results of a sophisticated computer
program. The basis for this method is Eq. (3.184) rearranged as follows for the shell-side
or tube outside surface area:
As
q
q
Uo Tm Uo F Tlm
9:54
Here Uo Us is the overall heat transfer coecient based on tube outside or shell-side
surface area. By approximately but rapidly estimating q, Uo , F, and Tlm , one can arrive
at the approximate surface area requirement and subsequently the size of the exchanger,
as discussed next. Since this is a sizing procedure, either the heat duty and inlet temperatures are given or both inlet and outlet temperatures are specied. They are related by the
energy balance of Eq. (3.5) as follows by considering the shell uid as hot uid:
q m_ s cp;s Ts;i Ts;o m_ cp;t Tt;o Tt;i
9:55
The overall heat transfer coecient Uo of Eq. (9.54) is calculated from Eq. (3.31a) as
1
1
1
d lndo =di
d
d
o
o o
hi; f di hi di
Uo ho ho; f
2kw
9:56
Here hi , ho , hi; f ; and ho; f (the subscripts i and o denote tube inside and tube outside or
shell side) are selected from Table 9.4. It should be emphasized that the values given in
this table are based on the usual velocities or nominally allowable pressure drops; allow-
659
ance should be made on operating conditions that are quite unusual. Also, care should be
exercised as noted in the appropriate footnotes of this table.
The log-mean temperature dierence correction factor F should be estimated as
follows. The correction factor F 1 for a counterow exchanger or if one stream
changes its temperature only slightly in the exchanger. For a single TEMA E shell
with an arbitrary even number of tube-side passes, the correction factor should be
F > 0:8 if there is no temperature cross; a rough value would be F 0:9 unless it can
be determined from Fig. 3.13. Consider F 0:8 when the outlet temperatures of the two
streams are equal (thus avoiding the temperature cross). If Ts;o < Tt;o ; there exists a
temperature cross (we have assumed here that the shell uid is hotter than the tube
uid) in the multipass exchanger; and in this case, multiple shells in series should be
considered. They can be determined by the procedure outlined in Fig. 3.18.
For the known inlet temperatures and given or calculated outlet temperatures,
compute the log-mean temperature dierence Tlm from its denition of Eqs. (3. 172)
and (3. 173). Knowing all the parameters on the right-hand side of Eq. (9.54), the tube
outside total surface area As (including the n area, if any) can then be estimated from
this equation.
9.5.3.1 Exchanger Dimensions. To relate the As above to the shell inside diameter and
the eective tube length, we will use the information shown in Fig. 9.4 (Bell, 1998). It is
generated for one of the commonly used fully tubed shell-and-tube heat exchangers that
FIGURE 9.4 Tube outside (shell-side) surface area As as a function of shell inside diameter and
eective tube length for a tube bundle having 19.05-mm (34-in.)-outside-diameter plain tubes, 23.8mm (15
16-in.) equilateral triangular tube layout, single tube-side pass, and fully tubed exchanger with
xed tubesheets. (From Bell, 1998.)
660
TABLE 9.4 Typical Film Heat Transfer Coecients and Fouling Factors for Shell-and-Tube Heat
Exchangers
Fluid Conditions
Sensible heat transfer
Waterc
Ammonia
Light organicsd
Medium organicse
Heavy organics f
Gash
Light organicsd
Medium organicse
Heavy organics
Light multicomponent
mixtures, all condensabled
Medium multicomponent
mixtures, all condensablee
Heavy multicomponent
mixtures, all condensablef
Vaporizing heat transferp;q
Waterr
Ammonia
Light organicsd
Medium organicse
Heavy organicsf
h (W/m2 Ka;b
Fouling Resistance
(m2 K=Wa
5,0007,500
6,0008,000
1,5002,000
7501,500
12.5 104
01 104
02 104
14 104
250750
150400
210 104
210 104
100300
60150
80125
250400
500800
430 103
430 103
01 104
01 104
01 104
8,00012,000
4,0006,000
2,0003,000
10,00015,000
01 104
01 104
01 104
01 104
15,00025,000
01 104
1,5002,000
01 104
7501,000
2,0004,000
01 104
01 104
3,0004,000
1,5004,000
01 104
13 104
6002,000
25 104
1,0002,500
02 104
6001,500
14 104
300600
28 104
3,00010,000
4,00015,000
12 104
12 104
3,0005,000
1,0004,000
02 104
12 104
7503,000
02 104
1,0003,500
13 104
6002,500
13 104
7502,500
25 104
Liquid
Liquid
Liquid
Liquid
Liquid
Heating
Cooling
Liquid
Heating
Cooling
Pressure 100200 kN/m2 abs
Pressure 1 MN/m2 abs
Pressure 10 MN/m2 abs
661
Heavy organicsg
4001,500
28 104
3001,000
210 104
662
has the following geometry: 19.05 mm (34 in.) outside diameter plain tubes, 23.8 mm
(15
16 in:) equilateral triangular tube layout, single tube-side pass, and xed tubesheets. In
this gure, the eective tube length represents the actual tube length between tubesheets
for the straight tube exchanger and the length between the tubesheet and the tangent
line for the U-tube bundle. The solid black lines indicate the shell inside diameter. From
the estimated value of As above, one can calculate a number of combinations of the
eective tube length Leff and the shell inside diameter Ds . The desired range of Leff =Ds
(shown by dashed lines in Fig. 9.4) is between 3 and 15, with a preferable range between
6 and 10. Leff =Ds 3 results in poor shell-side ow distribution and high p for the
inlet and outlet nozzles. Leff =Ds 15 would be dicult to handle mechanically and
would require a longer footprint for the tube bundle repair/removal.
We now explain how to use Fig. 9.4 for dierent tube diameters and layouts,
tube-side multipass construction, and other tube bundle constructions. The eective
tube-side surface area for geometry dierent from that for Fig. 9.4 will be designated
as As0 . The ordinate of Fig. 9.4 is then renamed As0 : It is related to As calculated
from Eq. (9.54) as
As0 As F1 F2 F3
9:57
Once we calculate the correction factors F1 , F2 , and F3 as outlined next and As from Eq.
(9.54), A0s is computed from Eq. (9.57), and the combination of the eective tube length
and the shell inside diameter is then determined from Fig. 9.4 as before. Let us describe
how to calculate the correction factors.
F1 correction factor for the tube outside diameter and tube layout. F1 1 for
19.05 mm tubes having a 23.8 mm 308 tube layout. For other do and pt , obtain
the value from Table 9.5.
F2 correction factor for the number of tube passes. F2 1 for a one-tube-pass
design. The value of F2 for multiple tube passes (U-tube and oating head
bundles) can be obtained from Table 9.6.
F3 correction factor for various rear-end head designs (shell construction) given in
Table 9.7.
TABLE 9.5 Values of F1 of Eq. (9.56) for Various Tube Diameters and Layouts
Tube Outside
Diameter
[in. (mm)]
5
8
5
8
3
4
3
4
3
4
3
4
Tube Pitch
[in. (mm)]
Layout
F1
(20.6)
!!
0.90
(20.6)
! ^, &
1.04
(23.8)
!!
1.00
(23.8)
! ^, &
1.16
(19.05)
13
16
13
16
15
16
15
16
(19.05)
1 (25.4)
!!
1.14
(19.05)
1 (25.4)
! ^, &
1.31
(31.8)
!!
1.34
(31.8)
! ^, &
1.54
(15.88)
(15.88)
(19.05)
1 (25.4)
1 (25.4)
Source: Data from Bell (1998).
114
114
663
Up to 12 (305)
1.20
1.40
1.80
1.06
1.18
1.25
1.50
1.04
1.14
1.19
1.35
25 to 33 (635 to 838)
1.03
1.12
1.16
1.20
35 to 45 (889 to 1143)
1.02
1.08
1.12
1.16
48 to 60 (1219 to 1524)
Above 60 (above 1524)
1.02
1.01
1.05
1.03
1.08
1.04
1.12
1.06
Up to 12
(305)
1322
(330559)
2336
(584914)
3748
(9401219)
Above 48
(above 1219)
Fixed tubesheet
(TEMA L, M or N)
Split backing ring (TEMA S)
Outside packed oating head
(TEMA P)
U-tubea (TEMA U)
Pull-through oating head
(TEMA T)
1.00
1.00
1.00
1.00
1.00
1.30
1.30
1.15
1.15
1.09
1.09
1.06
1.06
1.04
1.04
1.12
1.08
1.40
1.03
1.25
1.01
1.18
1.01
1.15
9.5.4
The more rigorous thermal design method includes all elements discussed in preceding
sections. In a concise manner, the following is a step-by-step procedure for the design or
sizing problem.
1. For given heat transfer duty and uid streams inlet temperatures, compute the
outlet temperatures using overall energy balances and the uid mass ow rates
specied or selected. If outlet temperatures are given, compute the heat duty
requirement.
2. Select a preliminary ow arrangement (i.e., type of shell-and-tube heat exchanger)
based on common industry practice (see Section 10.2.1 for selection criteria),
mechanical integrity, and maintenance requirements.
664
9.6
In the preceding sections, rating problems for extended surface and shell-and-tube heat
exchangers are presented, as is a sizing problem for an extended surface exchanger. For
the sizing problem, no constraints were imposed on the design except for the pressure
drops specied. The objective of that problem was to optimize the core dimensions to
meet the heat transfer required for specied pressure drops.
Heat exchangers are designed for many dierent applications, and hence may involve
many dierent optimization criteria. These criteria for heat exchanger design may be
minimum initial cost, minimum initial and operating costs, minimum weight or material,
minimum volume or heat transfer surface area, minimum frontal area, minimum labor
(translated into a minimum number of parts), and so on. When a performance measure
has been dened quantitatively and is to be minimized or maximized, it is called an
objective function in a design optimization. A particular design may also be subjected
to certain requirements, such as required heat transfer, allowable pressure drop, limitations on height, width and/or length of the exchanger, and so on. These requirements are
called constraints in a design optimization. A number of dierent surfaces could be
incorporated in a specic design problem, and there are many geometrical variables
that could be varied for each surface geometry.y In addition, operating mass ow rates
and temperatures could be changed. Thus, a large number of design variables are
associated with a heat exchanger design. The question arises as to how one can eectively
adjust these design variables within imposed constraints and come up with a design
having an optimum objective function. This is what we mean by the optimum component
y
For a shell-and-tube exchanger, the geometrical variables are those associated with the tube, baes, shell, and
front- and rear-end heads. For an extended surface exchanger, the geometrical variable associated with a n are
the n pitch, n height, n thickness, type of n, and other variables associates with each n type.
Alternate
designs:
construction
types, flow
arrangements,
surface
selection, etc.
Designe'r
constraints
and design
variables
Problem specifications
including customer's
constraints and
design variables
Total constraints
and design
variables for the
optimization
problem
Problem formulation
Heat exchanger
design computer
programs
Thermophysical
properties
Changed geometry
and /or operating
conditions specified
by design variables
Geometrical
properties
Fixed
operating
conditions
Heat transfer
and pressure
drop evaluation
scaled
j and f
factors
Objective function
and constraints
evaluation
Optimization
package
665
Optimization
strategy for
redefining the
design variables
Optimum
solution
FIGURE 9.5 Methodology for heat exchanger optimization. (From Shah et al. 1978.)
design, sometimes also referred to as the most ecient design. If a heat exchanger is part
of a system, it could also be optimized based on the system objective function by varying
pertinent exchanger design variables as well as system variables in the optimization
routine.
A complete mathematical component or system-based optimization of heat exchanger design is neither practical nor possible. Many engineering judgments based on experi-
666
ence are involved in various stages of the design. However, once the general conguration and surfaces are selected, an optimized heat exchanger design may be arrived at if the
objective function and constraints can be expressed mathematically and if all the variables are changed automatically and systematically on some evaluation criteria basis.
A large number of optimization (search) techniques are available in the literature, and
quite a lot of commercial optimization software is available. A typical design and optimization procedure for a heat exchanger is summarized here with the owchart of Fig.
9.5 for completeness. The procedure is referred to as the case study method. In this
method, each possible surface geometry and construction type is considered to be an
alternative design, as indicated in Fig. 9.5. To make a legitimate comparison of these
alternatives, each design must be optimized for the application specied. Thus there may
be several independent optimized solutions satisfying the problem requirements.
Engineering judgment, a comparison of objective function values, and other evaluation
criteria are then applied to select a nal optimum solution for implementation.
Assume a liquid-to-gas heat exchanger to be required for a specic application having
minimum total cost. From the initial screening of surfaces (see Section 10.3), suppose
that two plate-n constructions (the louver-n and strip-n surfaces) and one at-tube
and wavy-n construction appear to be promising for the gas side. Then, for this problem, there are three alternative designs that need to be optimized.
As shown in Fig. 9.5, rst formulate the total number of constraints for the problem.
This includes the customers specied explicit constraints (such as xed frontal area,
the ranges of heat exchanger dimensions) and implicit constraints (such as required
minimum heat transfer, allowable maximum pressure drop). Once the basic surface
geometry for the design chosen is selected, the designer imposes some additional constraints, such as the minimum and maximum values for the n height, n thickness, n
pitch, n thermal conductivity, ow length, number of nned passages, gas ow rate, and
so on. The designer wants to vary all the design and operating variables within the ranges
specied such that the exchanger will meet the required heat transfer, maximum pressure
drop, and other constraints with minimum total cost.
To optimize the heat exchanger, the designer starts with one set of heat exchanger
surface geometrical dimensions which may not even satisfy all or some of the constraints
imposed. Subsequently, the various geometrical properties (such as heat transfer area,
free-ow area, hydraulic diameter) and thermal properties are evaluated based on the
input operating conditions. The heat transfer rate and pressure drop are then evaluated
by the procedure outlined for the rating problem (see, e.g., Sections 9.2.1, 9.4.3, and
9.5.2). Next, the output from heat exchanger calculations is fed to the optimization
computer program package, where the constraints and the objective function are
evaluated. Subsequently, new values for the design variables are generated and heat
exchanger calculations are repeated. The iterations are continued until the objective
function is optimized (minimized or maximized as desired) within the accuracy specied
and all constraints are satised. In some situations, it may not be possible to satisfy all
constraints. Engineering judgment is then applied to determine whether or not the
optimum design is satisfactory and which constraints to relax. One of the most important
but least known inputs for the heat transfer and pressure drop evaluation is the
magnitude for scaled-up or scaled-down (modied from the original) j and f factors. As
soon as one of the surface geometrical dimensions is changed (such as the n pitch,
height, or thickness) but others may stay unchanged, the surface is no longer geometrically similar to the original surface for which experimental j and f data are available.
In such cases, either theoretical or experimental correlations should be incorporated
REFERENCES
667
in the computer program to arrive at the scaled j and f factors for the new geometry.
Some of these correlations are presented in Section 7.5. The designer must use his or her
experience and judgment regarding the appropriate correlations to obtain the scaled j
and f factors. In addition, care must be exercised to avoid excessive extrapolations.
A review of Fig. 9.5 indicates that heat exchanger design (rating) program and
optimization software is needed for the optimization. In addition, a system simulation
program is added if the heat exchanger optimization has to be done based on the system
design approach.
Although the foregoing optimization procedure was outlined from a performance and
cost point of view, the heat exchanger can also be optimized as a component or as part of
a system based on a thermoeconomics point of view. This is discussed further in Section
11.6.
SUMMARY
The focus of this chapter is to provide step-by-step rating and sizing procedures for major
types of heat exchangers. All the information and design theory outlined in previous
chapters is applied and extended in formulating the design procedures discussed in this
chapter. The detailed thermal and hydraulic design of heat exchangers outlined in this
chapter is one of the major objectives of this book. After presenting how to determine the
mean temperature on each uid side in a heat exchanger, we have provided the rating and
sizing of extended-surface (plate-n and tube-n), plate, and shell-and-tube heat exchangers in depth with an example. Subsequently, we have also provided a general approach
to the optimization of a heat exchanger design.
REFERENCES
Bell, K. J., 1963, Final report of the cooperative research program on shell-and-tube heat
exchangers, Univ. Del. Eng. Stn. Bull. No. 5.
Bell, K. J., 1988a, Overall design methodology for shell-and-tube exchangers, in Heat Transfer
Equipment Design, R. K. Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere
Publishing, Washington, DC, pp. 131144.
Bell, K. J., 1988b, Delaware method for shell design, in Heat Transfer Equipment Design, R. K.
Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere Publishing, Washington, DC,
pp. 145166.
Bell, K. J., 1998, Approximate sizing of shell-and-tube heat exchangers, in Heat Exchanger Design
Handbook, G. F. Hewitt, exec. ed., Begell House, New York, Vol. 3, Sec. 3.1.4.
Chiou, J. P., 1980, The advancement of compact heat exchanger theory considering the eects of
longitudinal heat conduction and ow nonuniformity, in Compact Heat Exchangers: History,
Technological Advancement and Mechanical Design Problems, R. K. Shah, C. F. McDonald, and
C. P. Howard, eds., Book G00183, American Society of Mechanical Engineers, New York.
Kays, W. M., A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger
Publishing, Malabar, FL.
Marriot, J., 1977, Performance of an Alfaex plate heat exchanger, Chem. Eng. Prog., Vol. 73, No. 2,
pp. 7378.
Raznjevic, K., 1976, Handbook of Thermodynamic Tables and Charts, McGraw-Hill, New York.
Shah, R. K., 1981, Compact heat exchanger design procedures, in Heat Exchangers: ThermalHydraulic Fundamentals and Design, S. Kakac, A. E. Bergles and F. Mayinger, eds., Hemisphere
Publishing Corp., Washington, DC, pp. 495536.
668
Shah, R. K., 1988a, Plate-n and tube-n heat exchanger design procedures, in Heat Transfer
Equipment Design, R. K. Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere
Publishing, Washington, DC, pp. 255266.
Shah, R. K., 1988b, Counterow rotary regenerator thermal design procedures, in Heat Transfer
Equipment Design, R. K. Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere
Publishing, Washington, DC, pp. 267296.
Shah, R. K., and W. W. Focke, 1988, Plate heat exchangers and their design theory, in Heat Transfer
Equipment Design, R. K. Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere
Publishing, Washington, DC, pp. 227254.
Shah, R. K., and A. D. Giovannelli, 1988, Heat pipe heat exchanger design theory, in Heat Transfer
Equipment Design, R. K. Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere Publishing, Washington, DC, pp. 609653.
Shah, R. K., and T. Skiepko, 1999, Inuence of leakage distribution on rotary regenerator thermal
performance, Appl. Thermal Eng., Vol. 19, pp. 685705.
Shah, R. K., and A. S. Wanniarachchi, 1991, Plate heat exchanger design theory, in Industrial
Heat Exchangers, J.-M. Buchlin, ed., Lecture Series 1991-04, von Karman Institute for Fluid
Dynamics, Belgium.
Shah, R. K., K. A. Amiwala, and R. W. Mayne, 1978, Heat exchanger optimization, Heat Transfer
1978, Proc. 6th Int. Heat Transfer Conf., Vol. 4, pp. 185191.
Taborek, J., 1998, Shell-and-tube heat exchangers: single phase ow, in Handbook of Heat
Exchanger Design, G. F. Hewitt, ed., Begell House, New York, pp. 3.3.3-1 to 3.3.11-5.
REVIEW QUESTIONS
For each question, circle one or more correct answers. Explain your answers briey.
9.1
The uid mean temperature on each uid side in a gas-to-gas multipass heat
exchanger is generally computed as the:
(a) arithmetic mean temperature on one uid side and log-mean average on the
other uid side in each pass
(b) arithmetic mean temperature on both uid sides in each pass
(c) cant tell
(d) none of these
9.2
9.3
In the following industrial heat exchangers, we can always meet both the heat
transfer and pressure drop requirements on at least one uid side during the design
process:
(a) shell-and-tube exchanger
(c) plate-n exchanger
PROBLEMS
9.4
The following streams do not contribute signicantly to heat transfer on the shell
side of a shell-and-tube heat exchanger:
(a) A
(d) E
9.5
(b) B
(e) F
(c) C
9.6
669
PROBLEMS
9.1
670
Re
Re
10,000
8,000
6,000
5,000
4,000
3,000
2,500
2,000
1,500
1,200
1,000
800
600
500
0.00551
0.00593
0.00651
0.00690
0.00738
0.00805
0.00849
0.00900
0.00970
0.0104
0.0112
0.0124
0.0144
0.0160
0.0331
0.0340
0.0354
0.0363
0.0375
0.0394
0.0406
0.0426
0.0461
0.0496
0.0532
0.0587
0.0682
0.0755
8,000
7,000
6,000
5,000
4,000
3,000
2,000
1,500
1,200
1,000
800
600
500
400
300
0.00452
0.00471
0.00492
0.00522
0.00575
0.00682
0.00744
0.00830
0.00911
0.01045
0.01255
0.01415
0.0166
0.0205
0.0123
0.0126
0.0131
0.0137
0.0146
0.0162
0.0198
0.0231
0.0265
0.0306
0.0347
0.0429
0.0493
0.0592
0.0758
All materials are stainless steel with k 20:8 W=m K. The following geometrical properties have been evaluated for the air and gas sides.
Geometrical Properties
Minimum free-ow area Ao (m
Heat transfer area A (m2 )
L=rh
Aw m2
2
Air Side
Gas Side
0.8247
1469
0.455
1781
1.5638
2524
0.398
1614
550.7
(a) At what temperatures will you evaluate uid properties on each uid side?
Consider the following mean uid properties: Air side: specic heat
1:04 kJ=kg K, thermal conductivity 0.0431 W/m K and dynamic viscosity
0:283 104 Pa s; these properties on gas side are: 1.06 kJ=kg K, thermal
conductivity 0.0473 W/m K, and dynamic viscosity 0:305 104 Pa s:
(b) Evaluate the quality of design by determining: (i) The percentage of the
thermal resistance on the air and gas sides. Is this a thermally balanced design?
Use f for air and gas sides as 0.674 and 0.773, respectively, and w 0:1 mm.
(ii) The relative pressure drops p=p for each stream. Are the pressure drops
balanced? What fraction of the total pressure drop is due to entrance/exit
losses? Use the following densities: air-side inlet, outlet, and mean densities:
7.075, 4.962, and 5.833 kg/m3 ; gas-side inlet, outlet, and mean densities:
0.5094, 0.6875, and 0.5852 kg/m3 .
9.2
671
PROBLEMS
noow height. Assume that the surface on each uid side is a plain triangular n
11.1 of Kays and London (1998). The surface geometry and j and f data are given
below.
Re
10,000
8,000
6,000
5,000
4,000
3,000
2,500
j
0.00314
0.00333
0.00356
0.00372
0.00390
0.00412
0.00424
f
0.00878
0.00923
0.00971
0.00991
0.0103
0.0112
0.0119
Re
2,000
1,500
1,200
1,000
800
600
500
0.00436
0.00444
0.00471
0.00515
0.00599
0.00733
0.00840
0.0129
0.0149
0.0169
0.0190
0.0228
0.0294
0.0350
Fin density 437 m1 , plate spacing b 6:35 mm, ow passage hydraulic
diameter Dh 3:081 mm, n metal thickness 0:15 mm, material aluminum,
heat transfer surface area density 1204 m2 =m3 , and n area/total
area 0:756. Assume the plate thickness to be 0:5 mm. Try to accommodate an
integer number of ns on each uid side in an approximately 0:3 m 0:3 m 0:6 m
envelope. On one uid side of the exchanger, process air at 0.40 m3 /s ows with the
inlet temperature of 2388C. On the other uid side, makeup air ows at a 0.26 m3 /s
ow rate at 408C inlet temperature. Both uids are at atmospheric pressure at
inlet. Determine the outlet temperatures on each uid side as well as the pressure
drop on each uid side. Both the n and plates are made of aluminum having a
thermal conductivity of 190.4 W/m K.
9.3
A nned-tube exchanger is designed to cool water from 528C to 388C with ambient
air at 328C to be heated to 438C. The water and air mass ow rates are 2.52 and
13.13 kg/s, respectively. The allowable air-side pressure drop is 149 Pa. Estimate
the air-side frontal velocity to satisfy the desired performance. Consider
j=f 0:30, Rair =Rtot 0:7, and counterow performance. The nned tube
exchanger has Ao =Afr 0:56 and cp 1:005 and 4.187 kJ/kg K for air and
water, respectively. Use m 1:136 kg=m3 , Pr 0:705, and pi 101:4 kPa for air.
9.4
What will be the exit temperatures of the hot and cold streams and heat duty (heat
transfer rate) of a plate heat exchanger if 2.52 kg/s of hot water enters at 1048C
and 4.04 kg/s of cold water enters at 168C? These ow rates are expected at some
part-load condition, and the plant engineer is interested in determining the
temperatures to see how secondary cooling equipment downstream behaves.
Use the following data for your solution: the plate exchanger is a 1 pass1 pass
counterow exchanger (see Fig. 1.65a) with 14 ow channels for hot water and 14
ow channels for cold water (total 27 thermal plates). The eective width of plates
is 0.457 m and the gasket thickness or gap between plates is 3.0 mm. Note that the
hydraulic diameter will be equal to twice the gap dimension. The projected heat
transfer area for each plate on one side is 0.28 m2 . Use a fouling factor of
0.0002 m2 K/W for each uid stream, a plate thickness of 0.9 mm, and wall
thermal conductivity of 15.6 W/m K. The j and f factors for the plate surface
are given as j 0:2Re0:25 and f 0:6Re0:2 . Also, the following properties are
provided for water.
672
Property Hot
cp (kJ/kg K
(Pa s
k (W/m K
(kg/m3 )
Pr
9.5
Design a gas-to-gas two-uid heat exchanger with both uids assumed to be air.
The inlet temperature, pressure, and mass ow rate of one airstream are 487 K,
490 kPa, and 21 kg/s, respectively, and it must leave the heat exchanger at a
temperature of 619 K. The mass ow rate of the second airstream is the same as
that of the rst, and its inlet temperature and pressure are 690 K and 103 kPa,
respectively. The pressure drops are limited to 4.900 kPa and 2.575 kPa for the
cold and hot airstream. Additional data: Flow arrangement is a single-pass cross
ow (unmixedunmixed); heat transfer surface is a plain plate-n surface with
designation 19.86 (Kays and London, 1998). The heat transfer surface material is
aluminum with a plate thickness of 1 mm. List all your assumptions and provide
proper justication for each. If any additional assumption is needed, provide
explicit reasons for it.
9.6
Analyze a three-pass two-uid overall cross-counterow heat exchanger of a gasturbine plant with all three passes as equal unmixedunmixed crossow units (see
the right-hand side of Fig. 1.58a as a schematic). The desired overall heat exchanger eectiveness must be 0.766 in order to reach the required plant eectiveness.
The air mass ow rate is 21 kg/s and the inlet temperature 487 K. The mass ow
rate of the gas is the same, but the inlet temperature is 690 K. Determine the outlet
temperatures of both air and gas. Perform calculations involving determination of
thermophysical properties of uids at the true mean (integral) temperatures.
10
673
674
depending on the process. Thus, the overall objective of this chapter is to provide a good
understanding of heat exchanger selection in general as a component based on qualitative and quantitative criteria. However, system-based heat exchanger optimization is the
current industrial practice.
10.1.1
The exchanger in operation must withstand the stresses produced by the operating
pressure and the temperature dierences between two uids. These stresses depend
on the inlet pressures and temperatures of the two uids. The most versatile exchangers for a broad range of operating pressures and temperatures are shell-and-tube
exchangers for medium- to high-heat duties and double-pipe exchangers for lowerheat duties. They can handle from high vacuum to ultrahigh uid pressures [generally limited to 30 MPa (4350 psi) on the shell or annulus side and 140 MPa (20,000
psi) on the tube side]. Coupled with high pressures, shell-and-tube exchangers can
withstand high temperatures, limited only by the materials used; however, the inlet
temperature dierence is limited to 508C (1208F) from the thermal expansion point
of view when the exchanger design allows only limited thermal expansion, such as in
the E-shell design. These exchangers are used for gas, liquid, and phase-change
applications.
For liquidliquid or liquidphase change applications, if the operating pressures and
temperatures are modest (less than about 2.5 MPa and 2008C), gasketed or semiwelded
plate exchangers should be considered. For somewhat higher pressures and temperatures, fully welded or brazed plate exchangers may be the choice, depending on other
design criteria.
The plate-n extended surface exchanger is designed for low-pressure applications,
with the operating pressures on either side limited to about 1000 kPa (150 psig), except
for cryogenics applications, where the operating pressure is about 9000 kPa gauge
(1300 psig). The maximum operating temperature for plate-n exchangers is below
6508C (12008F) and usually below 1508C (3008F), to avoid the use of expensive materials.
There is no limit on the minimum operating temperature; plate-n exchangers are commonly used in cryogenic applications. Fins in a plate-n exchanger act as a ow-mixing
device for highly viscous liquids, and if properly designed, add surface area for heat
transfer with a reasonably high n eciency. In a plate-n exchanger, ns on the liquid
side are used primarily for pressure containment and rigidity. Fins on the gas side are
used for added surface area for heat transfer, with n eciencies usually greater than
80%.
675
The tube-n exchanger is used to contain the high-pressure uid on the tube side if
only one uid is at a high pressure. Fins on the liquid or phase-change side generally have
low heights, to provide reasonably high n eciencies. Turbulators may be used
within tubes for ow mixing. Tube-n exchangers with or without shells are designed
to cover the operating temperature range from low cryogenic temperatures to about
8708C (16008F).
For ultrahigh temperature [870 to 20008C (1600 to 36008F)] and near-atmospheric
pressures, as in high-temperature waste heat recovery, either rotary regenerators (870 to
11008C) or xed-matrix regenerators (up to 20008C) are used.
10.1.2
Cost
Cost is a very important factor in the selection of the heat exchanger construction
type. The cost per unit of heat transfer surface area is higher for a gasketed plate
exchanger than for a shell-and-tube exchanger. However, from the total cost (capital,
installation, operation, maintenance, etc.) point of view, PHEs are less expensive
than shell-and-tube exchangers when stainless steel, titanium, and other higher quality alloys are used. Since tubes are more expensive than extended surfaces or a
regenerator matrix, shell-and-tube (or broadly, tubular) exchangers are in general
more expensive per unit of heat transfer surface area. In addition, the heat transfer
surface area density of a tubular core is generally much lower than that of an
extended surface or regenerative exchanger. Rotary regenerators made of paper or
plastic are in general the least expensive per unit of heat transfer surface area.
10.1.3
Fouling and cleanability are among the most important design considerations for
liquid-to-liquid or phase-change exchangers and for some gas-to-uid exchangers.
Fouling should be evaluated for both design and o-design points. Periodic cleaning
and/or replacement of some exchanger components depend on the fouling propensity
of the uids employed. In applications involving moderate to severe fouling, either a
shell-and-tube or a gasketed plate heat exchanger is used, depending on the other
operating parameters. In a shell-and-tube exchanger, the tube uid is generally
selected as the heavily fouling uid since the tube side may be cleaned more easily.
A plate heat exchanger is highly desirable in those relatively low temperature applications [<3008C (5758F)] where severe fouling occurs on one or both sides, as plate
disassembly, cleaning, and reassembly is a relatively easy task. For highly corrosive
uid heating or cooling applications, shell-and-tube exchangers are used exclusively,
regardless of operating pressure and temperature conditions. Plate-n exchangers
usually have small hydraulic diameter passages and hence are more susceptible to
fouling. They are also relatively dicult to clean and are not employed in even
moderate fouling applications unless they can be cleaned chemically or thermally
by baking (see Section 13.4).
The fouling and cleanability problem is not as severe for gas-to-gas exchangers as for
liquid-to-liquid or phase-change exchangers, since in most applications gases are neither
very dirty nor have the fouling propensity of water. Regenerators have self-cleaning
characteristics because the hot and cold gases ow periodically in opposite directions
through the same passage. Hence, they can tolerate moderate fouling. If the application
has a potential for heavy fouling, a larger ow passage size is chosen, as in a xed-matrix
676
up to 200
up to 200
up to 200
up to 200
200300
200
up to 300
up to 300
up to 200
Partially welded
plate
Fully welded
plate
(AlfaRex)
Brazed plate
Bavex plate
Platular plate
Compabloc
plate
Packinox plate
Spiral
Compactness
m2 =m3
Plate-and-frame
(gaskets)
Type of
Heat
Exchanger
Liquidliquid,
two-phase
Gases, liquids
two-phase
Liquids
Gases, liquids,
two-phase
Gases, liquids,
two-phase
Liquidliquid,
two-phase
Liquidliquid,
gasliquid,
two-phase
Liquidliquid,
gasliquid,
two-phase
Liquidliquid,
gasliquid,
two-phase
Stream
Typesa
c/s, s/s,
Ti, Incoloy,
Hastelloy
s/s, Ti,
Hastelloy,
Inconel
s/s, Ti
Incoloy
s/s,
Hostelloy,
Ni alloys
s/s
s/s, Ti,
Ni alloys
s/s, Ti,
Incoloy,
Hastelloy
Materialb
25
40
30
60
35 to 200
50 to 350
195 to 220
200 to 900
25
300
200 to 700
up to 400
32
up to 300
40
25
35 to 200
up to 700
Maximum
Pressure (bar)c
Feature
Temperature
Range (8C)
Good
Good
Good
Good
Good
Mechanicald;m
Mechanicald
Mechanicald;o
Mechanicald
Goodj
Excellent
Goode
Goode
Corrosion
Resistance
Mechanicald;l
chemical
Chemicali
Chemical
Mechanical,d;g
chemicalh
Mechanicald
Cleaning
Methods
No
Yesf
Not usually
Yesn
In principle
No
No
No
Yesf
Multistream
Capability
No
Yes
Yes
Yes
Yes
Nok
Yes
Yes
Yes
Multipass
Capability
677
2005000
450
Printed-circuit
Polymer (e.g.
channel plate)
Plate-and-shell
Gases, liquids,
two-phase
S/s, Ni, Ni
alloys, Ti
PVDFq
PPr
Gasliquidp
Liquids
s/s, Ni, Ni
alloys, Ti
Ti, s/s
Al, s/s,
Ni alloy
Gases, liquids
two-phase
Gases, liquids,
two-phase
Gases, liquids,
two-phase
200 to 900
up to 350
> 400
70
> 400
200 to 900
up to 150s
Chemical
> 200
up to 500
Excellent
Good
Mechanicald;o ,
Chemicalu
Chemical
Excellent
Excellent
Excellent
Good
Water wash
Chemical
Chemical
90
Cryogenic
to 650
up to 10; 000
700800
Diusionbonded
plate-n
Marbond
8001500
Brazed
plate-n
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Not usually
Yes
Yes
Yes
678
regenerator, so that the impact of fouling is reduced, or cleaning by one of the methods
described in Section 13.4 may be employed.
10.1.4
Whereas in some applications, uid leakage from one uid side to the other uid
side is permissible within limits, in other applications uid leakage is absolutely not
allowed. Even in a good leak tight design, carryover and bypass leakages from the
hot uid to the cold uid (or vice versa) occur in regenerators. Where these leakages
and subsequent uid contamination is not permissible, regenerators are not used.
The choices left are either a tubular, extended surface, or some plate type heat
exchangers. Gasketed plate exchangers have more probability of ow leakage than
do shell-and-tube exchangers. Plate-n and tube-n exchangers have potential leakage problems at the joint between the corrugated n passage and the header or at
the tube-to-header joint. Where absolutely no uid contamination is allowed (as in
the processing of potable water), a double-wall tubular or shell-and-tube exchanger
or a double-plate PHE is used.
10.1.5
10.1.6
Fluid Type
679
Corrosive Service
Carbon steel
Ferritic carbonmolybdenum and
chromiummolybdenum alloys
Ferritic chromium steel
{
A thermally balanced design usually results in an optimum design from the cost viewpoint since the cost of the
extended surface per unit surface area is less than that of the prime surface, either tubes or plates.
680
exchanger type for phase-change exchangers parallels the guidelines provided for the gas
or liquid side of the exchanger.
Shell-and-Tube Exchangers
More than 65% of the market share (in the late 1990s) in process and petrochemical
industry heat exchangers is held by the shell-and-tube heat exchanger, for the following reasons: its versatility for handling a wide range of operating conditions with a
variety of materials, design experience of about 100 years, proven design methods,
and design practice with codes and standards. The selection of an appropriate shelland-tube heat exchanger is achieved by a judicious choice of exchanger conguration, geometrical parameters, materials, and the right design. Next we summarize
some guidelines on all these considerations qualitatively to provide the feel for the
right design for a given application. The major components of a shell-and-tube
exchanger are tubes, baes, shell, front-end head, read-end head, and tubesheets.
Depending on the applications, a specic combination of geometrical variables or
types associated with each component is selected. Some guidelines are provided
below. For further details on geometrical dimensions and additional guidelines,
refer to TEMA (1999).
10.2.1.1 Tubes. Since the desired heat transfer in the exchanger takes place across
the tube surface, the selection of tube geometrical variables is important from a
performance point of view. In most applications, plain tubes are used. However,
when additional surface area is required to compensate for low heat transfer coecients
on the shell side, low nned tubing with 250 to 1200 ns/m (6 to 30 ns/in.) and a n
height of up to 6.35 mm (14 in.) is used. While maintaining reasonably high n eciency,
low-height ns increase surface area by two to three times over plain tubes and decrease
fouling on the n side based on the data reported.
The most common plain tube sizes have 15.88, 19.05, and 25.40 mm (58, 34, and 1 in.)
tube outside diameters. From the heat transfer viewpoint, smaller-diameter tubes yield
higher heat transfer coecients and result in a more compact exchanger. However,
larger-diameter tubes are easier to clean and more rugged. The foregoing common
sizes represent a compromise. For mechanical cleaning, the smallest practical size is
19.05 mm (34 in.). For chemical cleaning, smaller sizes can be used provided that the
tubes never plug completely.
The number of tubes in an exchanger depends on the uid ow rates and available
pressure drop. The number of tubes is selected such that the tube-side velocity for water
and similar liquids ranges from 0.9 to 2.4 m/s (3 to 8 ft/sec) and the shell-side velocity
from 0.6 to 1.5 m/s (2 to 5 ft/sec). The lower velocity limit corresponds to limiting the
fouling, and the upper velocity limit corresponds to limiting the rate of erosion. When
sand and silt are present, the velocity is kept high enough to prevent settling.
681
The number of tube passes depends on the available pressure drop. Higher velocities
in the tube result in higher heat transfer coecients, at the expense of increased pressure
drop. Therefore, if a higher pressure drop is acceptable, it is desirable to have fewer but
longer tubes (reduced ow area and increased ow length). Long tubes are accommodated in a short shell exchanger by multiple tube passes. The number of tube passes in a
shell generally range from 1 to 10 (see Fig. 1.61). The standard design has one, two, or
four tube passes. An odd number of passes is uncommon and may result in mechanical
and thermal problems in fabrication and operation.
10.2.1.2 Tube Pitch and Layout. The selection of tube pitch is a compromise between
a close pitch (small values of pt =do ) for increased shell-side heat transfer and surface
compactness, and an open pitch (large values of pt =do ) for decreased shell-side plugging
and ease in shell-side cleaning. In most shell-and-tube exchangers, the ratio of the tube
pitch to tube outside diameter varies from 1.25 to 2.00. The minimum value is restricted
to 1.25 because the tubesheet ligamenty may become too weak for proper rolling of the
tubes and cause leaky joints. The recommended ligament width depends on the tube
diameter and pitch; the values are provided by TEMA (1999).
Two standard types of tube layouts are the square and the equilateral triangle, shown
in Fig. 10.1. The equilateral pitch can be oriented at 308 or 608 angle to the ow direction,
and the square pitch at 458 and 908.z Note that the 308, 458 and 608 arrangements are
staggered, and 908 is inline. For the identical tube pitch and ow rates, the tube layouts in
decreasing order of shell-side heat transfer coecient and pressure drop are: 308, 458, 608,
and 908. Thus the 908 layout will have the lowest heat transfer coecient and the lowest
pressure drop.
The square pitch (908 or 458) is used when jet or mechanical cleaning is necessary on
the shell side. In that case, a minimum cleaning lane of 14 in. (6.35 mm) is provided. The
square pitch is generally not used in the xed tubesheet design because cleaning is not
feasible. The triangular pitch provides a more compact arrangement, usually resulting in
a smaller shell, and the strongest header sheet for a specied shell-side ow area. Hence, it
is preferred when the operating pressure dierence between the two uids is large. If
{
The ligament is a portion of material between two neighboring tube holes. The ligament width is dened as the
tube pitch minus the tube hole diameter, such as the distance a shown in Fig. 10.1.
{
Note that the tube layout angle is dened in relation to the ow direction and is not related to the horizontal or
vertical reference line. Refer to Table 8.1 for the denitions of tube layouts and associated geometrical variables.
682
designed properly, it can be cleaned from six directions instead of four as in the square
pitch arrangement. When mechanical cleaning is required, the 458 layout is preferred for
laminar or turbulent ow of a single-phase uid and for condensing uid on the shell
side. If the pressure drop is limited on the shell side, the 908 layout is used for turbulent
ow. For boiling applications, the 908 layout is preferred, because it provides vapor
escape lanes. However, if mechanical cleaning is not required, the 308 layout is preferred
for single-phase laminar or turbulent ow and condensing applications involving a high
T rangey (a mixture of condensables). The 608 layout is preferred for condensing
applications involving a low T range (generally, pure vapor condensation) and for
boiling applications. Horizontal tube bundles are used for shell-side condensation or
vaporization.
10.2.1.3 Baes. As presented in Section 1.5.1.1, baes may be classied as either
longitudinal or transverse type. Longitudinal baes are used to control the overall
ow direction of the shell uid. Transverse baes may be classied as plate baes
or grid baes. Plate baes are used to support the tubes, to direct the uid in the tube
bundle at approximately right angles to the tubes, and to increase the turbulence and
hence the heat transfer coecient of the shell uid. However, the window section
created by the plate baes results in excessive pressure drop with insignicant contribution to heat transfer; ow normal to the tubes in crossow section may create owinduced vibration problems. The rod baes, a most common type of grid baes,
shown in Fig. 1.11, are used to support the tubes and to increase the turbulence.
Flow in a rod bae heat exchanger is parallel to the tubes, and hence ow-induced
vibration is virtually eliminated by the bae support of the tubes. The choice of bae
type, spacing, and cut are determined largely by the ow rate, required heat transfer,
allowable pressure drop, tube support, and ow-induced vibration. The specic
arrangements of baes in various TEMA shells are shown in Fig. 10.3.
Plate Baes. Two types of plate baes, shown in Fig. 1.10 are segmental, and disk and
doughnut. Single and double segmental baes are used most frequently. The single
segmental bae is generally referred to simply as a segmental bae. The practical range
of single segmental bae spacing is 15 to 1 shell diameter, although optimum could be 25
to 12. The minimum bae spacing for cleaning the bundle is 50.8 mm (2 in.) or 15 shell
diameter, whichever is larger. Spacings closer than 15 shell diameter provide added
leakagez that nullies the heat transfer advantage of closer spacings. If the foregoing
limits on the bae spacing do not satisfy other design constraints, such as pmax or
tube vibration, no-tubes-in-window or pure crossow design should be tried.
The segmental bae is a circular disk (with bae holes) with one disk segment
removed. The bae cut varies from 20 to 49% (the height c in Fig. 8.9 given as a
percentage of the shell inside diameter), with the most common being 20 to 25%. At
larger spacings, it is 45 to 50%, to avoid excessive pressure drop across the windows as
compared to the bundle. Large or small spacings coupled with large bae cuts are
undesirable because of the increased potential of fouling associated with stagnant ow
{
Here the T range represents the dierence in condensing temperature at the inlet minus condensing temperature at the outlet of an exchanger.
{
These are tube to bae hole, bae to shell, bundle to shell, and the tube pass partition leakages or bypasses
described in Section 4.4.1.1.
683
areas. If fouling is a primary concern, the bae cut should be kept below 25%. The bae
cut and spacing should be designed such that the ow velocity has approximately the
same magnitude for the cross ow and window ow sections. Alternate segmental baes
are arranged 1808 to each other, which cause shell-side ow to approach crossow in the
central bundle regiony and axial ow in the window zone. All segmental baes shown in
Fig. 1.10 have horizontal bae cuts. The direction of the bae cut is selected as follows
for shell-side uids: Either horizontal or vertical for a single-phase uid (liquid or gas),
horizontal for better mixing for very viscous liquids, and vertical for the following shellside applications: condensation (for better drainage), evaporation/boiling (for no stratication and for providing disengagement room), entrained particulates in liquid (to provide least interference for solids to fall out), and multishell pass exchanger, such as those in
Fig. 1.62 and the F shell.
Since one of the principal functions of the plate bae is to support the tubes, the terms
bae and support plate are sometimes used interchangeably. However, a support plate
does not direct the uid normal to the tube bank, it may be thicker than a bae, it has
less tube-to-bae hole clearance, and it provides greater stiness to the bundle. Support
plates with single-segmental baes are cut approximately at the centerline and spaced
0.76 m (30 in.) apart. This results in an unsupported tube span of 1.52 m (60 in.) because
each plate supports half the number of tubes. The double-segmental bae (Fig. 1.10),
also referred to as a strip bae, provides lower shell-side pressure drop (and allows larger
uid ows) than that for the single segmental bae for the same unsupported tube span.
The bae spacing for this case should not be too small; otherwise, it results in a more
parallel (longitudinal) ow (resulting in a lower heat transfer coecient) with signicant
zones of ow stagnation. Triple-segmental baes have ows with a strong parallel ow
component, provide lower pressure drop, and permit closer tube support to prevent tube
vibrations.
The lower allowable pressure drop results in a large bae spacing. Since the tubes
in the window zone are supported at a distance of two or more times the bae spacing,
they are most susceptible to vibration. To eliminate the possibility of tube vibrations
and to reduce the shell-side pressure drop, the tubes in the window zone are removed and
support plates are used to reduce the unsupported span of the remaining tubes. The
resulting design is referred to as the segmental bae with no-tubes-in-window, shown in
Fig. 1.10. The support plates in this case are circular and support all the tubes. The bae
cut and number of tubes removed varies from 15 to 25%. Notice that low-velocity
regions in the bae corners do not exist, resulting in good ow characteristics and less
fouling. Thus the loss of heat transfer surface in the window section is partially
compensated for. However, the shell size must be increased to compensate for the loss
in the surface area in the window zone, which in turn may increase the cost of the
exchanger. If the shell-side operating pressure is high, this no-tubes-in-window design
is very expensive compared to a similar exchanger having tubes in the window zone.
The disk-and-doughnut bae is made up of alternate disks and doughnut-shaped
baes, as shown in Fig. 1.10. Generally, the disk diameter is somewhat greater than the
half-shell diameter, and the diameter of the hole of the doughnut is somewhat smaller
than the half-shell diameter. This bae design provides a lower pressure drop compared
to that in a single-segmental bae for the same unsupported tube span and eliminates the
tube bundle-to-shell bypass stream C. The disadvantages of this design are that (1) all the
{
Various allowable clearances required for construction of a tube bundle with plate baes are provided by TEMA
(1999).
684
tie rods to hold baes are within the tube bundle, and (2) the central tubes are supported
by the disk baes, which in turn are supported only by tubes in the overlap of the largerdiameter disk over the doughnut hole.
Rod Baes. Rod baes are used to eliminate ow-induced vibration problems. For
certain shell-and-tube exchanger applications, it is desirable to eliminate the cross ow
and have pure axial (longitudinal) ow on the shell side. For the case of high shell-side
ow rates and low-viscosity uids, the rod bae exchanger has several advantages over
the segmental bae exchanger: (1) It eliminates ow-induced tube vibrations since the
tubes are rigidly supported at four points successively; (2) the pressure drop on the shell
side is about one-half that with a double segmental bae at the same ow rate and heat
transfer rate. The shell-side heat transfer coecient is also considerably lower than that
for the segmental bae exchanger. In general, the rod bae exchanger will result in a
smaller-shell-diameter longer-tube unit having more surface area for the same heat
transfer and shell-side pressure drop; (3) there are no stagnant ow areas with the
rod baes, resulting in reduced fouling and corrosion and improved heat transfer
over that for a plate bae exchanger; (4) since the exchanger with a rod (grid) bae
design has a counterow arrangement of the two uids, it can be designed for higher
exchanger eectiveness and lower mean (or inlet) temperature dierences than those of
an exchanger with a segmental bae design; and (5) a rod bae exchanger will generally be a lower-cost unit and has a higher exchanger heat transfer rate to pressure
drop ratio overall than that of a segmental bae exchanger. If the tube-side uid is
controlling and has a pressure drop limitation, a rod bae exchanger may not be
applicable. Refer to Gentry (1990) for further details on this exchanger.
Impingement Baes. Impingement baes or plates are generally used in the shell side
just below the inlet nozzle. Their purpose is to protect the tubes in the top row near the
inlet nozzle from erosion, cavitation, and/or vibration due to the impact of the highvelocity uid jet from the nozzle to the tubes. One of the most common forms of this
bae is a solid square plate located under the inlet nozzle just in front of the rst tube
row, as shown in Fig. 10.2. The location of this bae is critical within the shell to
minimize the associated pressure drop and high escape velocity of the shell uid after
the bae. For this purpose, adequate areas should be provided both between the nozzle
and plate and between the plate and tube bundle. This can be achieved either by
omitting some tubes from the circular bundle as shown in Fig. 10.2 or by modifying
the nozzle so that it has an expanded section (not shown in Fig. 10.2). Also, proper
positioning of this plate in the rst bae space is important for ecient heat transfer.
FIGURE 10.2 Impingement baes at the shell-side inlet nozzle. (From Bell, 1998.)
Shell Type
685
U-Tube Bundles
TEMA E
TEMA F
TEMA G
TEMA H
TEMA J
single nozzle
entry
TEMA J
double nozzle
entry
L
longitudinal
flow
TEMA X
cross flow
FIGURE 10.3 Shell-side ow arrangement for various shell types (Courtesy of Heat Transfer
Research, Inc., College Station, Texas).
686
Enough space should be provided between the tip of the plate and the tubesheet and
between the tip of the plate and the rst segmental bae. The most common cause of
tube failure is improper location and size of the impingement plate.
10.2.1.4 Shells. Seven types of shells, as classied by TEMA (1999), are shown in Fig.
1.6; they are also shown in Fig. 10.3 with baes. The E shell, the most common due to
its low cost and relative simplicity, is used for single-phase shell uid applications and
for small condensers with low vapor volumes. Multiple passes on the tube side increase
the heat transfer coecient h (if corresponding more increased p is within allowed
limits). However, a multipass tube arrangement can reduce the exchanger eectiveness
or F factor compared to that for a single-pass arrangement (due to some tube passes
being in parallelow) if the increased h and NTU do not compensate for the parallelow eect. Two E shells in series (in overall counterow conguration) may be used to
increase the exchanger eectiveness ".
As an alternative, a counterow arrangement is desirable (i.e., high ") for a two-tubepass exchanger. This is achieved by the use of an F shell having a longitudinal bae,
resulting in two shell passes. However, a TEMA F shell is rarely used in practice because
of heat leakage across the longitudinal bae and potential ow leakage that can occur if
the area between the longitudinal bae and the shell is not sealed properly. Also, the F
shell presents additional problems of fabrication and maintenance, and it is dicult to
remove or replace the tube bundle. If one needs to increase the exchanger eectiveness,
multiple shells in series are preferred over an F shell.
The TEMA G and H shells are related to the F shell but have dierent longitudinal
baes. Hence, when the shell-side p is a limiting factor, a G or H shell can be used;
however, " or F will be lower than that of a counterow exchanger. The split-ow G shell
has horizontal baes with the ends removed; the shell nozzles are 1808 apart at the
midpoint of the tubes. The double-split-ow H shell is similar to the G shell, but with
two inlet and two outlet nozzles and two longitudinal baes. The G and H shells are
seldom used for shell-side single-phase applications, since there is no advantage over E
or X shells. They are used as horizontal thermosiphon reboilers, condensers, and other
phase-change applications. The longitudinal bae serves to prevent ashing of the
lighter components of the shell uid, helps ush out noncondensables, provides increased
mixing, and helps distribute the ow. Generally, T and p across longitudinal baes
are small in these applications, and heat transfer across the bae and ow leakages at the
sides have insignicant inuence on the performance. The H shell approaches the crossow arrangement of the X shell, and it usually has low shell-side p compared to the E,
F, and G shells. For high-inlet-velocity applications, two nozzles are required at the inlet,
hence the H or J shell is used.
The divided-ow TEMA J shell has two inlets and one outlet or one inlet and two
outlet nozzles (a single nozzle at the midpoint of the tubes and two nozzles near the tube
ends). The J shell has approximately one-eighth the pressure drop of a comparable E
shell and is therefore used for low-pressure-drop applications such as in a condenser in
vacuum. For a condensing shell uid, the J shell is used, with two inlets for the gas phase
and one central outlet for the condensate and residue gases.
The TEMA K shell is used for partially vaporizing the shell uid. It is used as a kettle
reboiler in the process industry and as a ooded chiller (hot liquid in tubes) in the
refrigeration industry. Usually, it consists of an overall circular-cross-section horizontal
bundle of U tubes placed in an oversized shell with one or more vapor nozzles on the top
side of the shell (see one vapor nozzle in Fig. 1.6) to reduce liquid entrainment. The tube
687
FIGURE 10.4 An X shell exchanger with omission of top tube rows for better ow distribution
for a condensing application. (From Bell, 1998.)
bundle diameter ranges 50 to 70% of the shell diameter. The liquid (to be vaporized)
enters from below near the tubesheet through the left-hand nozzle and covers the tube
bundle. Pool and some convective boiling takes place on the shell side without forced
ow of the vaporizing uid outside the tubes on the shell side. The vapor occupies the
upper space in the shell without the tubes. The large empty space in the shell acts as a
vapor disengaging space; and if properly sized, almost dry vapor exits from the top
nozzle, thus eliminating the need for an external vaporliquid separator. Hence, it is
commonly used, although it is more expensive to fabricate, particularly, for highpressure applications. Generally, the kettle reboiler is considered as a pool boiling device;
however, convective (ow) boiling prevails in the tube bundle.
For a given ow rate and surface area, the crossow TEMA X shell has the lowest
shell-side pressure drop of all (except for K) shell congurations. Hence, it is used for gas
heating and cooling applications with or without nned tubes and for vacuum condensing applications. It is also used for applications having large shell ows. No transverse
baes are used in the X shell; however, support plates are used to suppress the owinduced vibrations. Flow distributions on the shell side could be a serious problem unless
proper provision has been made to feed the uid uniformly at the inlet. This could be
achieved by a bathtub nozzle, multiple nozzles, or by providing a clear lane along the
length of shell near the nozzle inlet as shown in Fig. 10.4.
The type of shell described in Fig. 1.6 has either one or two shell passes in one shell.
The cost of the shell is much more than the cost of tubes; hence, a designer tries to
accommodate the required heat transfer surface in one shell. Three or four shell passes
in a shell could be made by the use of longitudinal baes.y Multipassing on the shell side
with longitudinal baes will reduce the ow area per pass compared to a single pass on
the shell side in a single shell, resulting in a possibly higher shell-side pressure drop.
Multiple shells in series are also used for a given application for the following reasons:
. They increase the exchanger eectiveness ", or reduce the surface area for the same
". For the latter case, a subsequent reduction in tubing cost may oset the cost of an
additional shell and other components.
{
Positive or tight sealing between the longitudinal baes and the shell is essential to maintain the high exchanger
eectivenesses predicted.
688
689
Design Feature
Fixed
Tubesheet
Outsidepacked
Stung
box
Return
Bend
(U-Tube)
OutsidePacked
Latern
Ring
Inside
PullSplit
Through Backing
Bundle
Ring
No
No
Yes, with
bellows in
shell
Yes
Yes
Yes
Tes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yesa
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
With
special
tools
Yesb
Yesb
Yesb
Yesb
Yesb
No
No
No
No
Yes
Yes
Yes
Yes
Yes
No
No
No
Any
Anyc
Anye
2550
One or
twod
1535
Anye
1118
Any even
number
1118
95160
3550
passes can be employed. The TEMA L, M, and N rear-end heads are the counterparts of
TEMA A, B, and N front-end heads. The major disadvantages of the xed tubesheet
exchanger are (1) no relief for thermal stresses between the tubes and the shell, (2) the
impossibility of cleaning the shell side mechanically (only chemical cleaning is possible),
and (3) the impracticality of replacing the tube bundle. Fixed tubesheet exchangers are
thus used for applications involving relatively low temperatures [3158C (6008F) and
lower] coupled with low pressures [2100 kPa gauge (300 psig) and lower]. As a rule
690
of thumb, the xed tubesheet design is used for an inlet temperature dierence between
the two uids that is less than about 50 to 608C (1008F). If an expansion bellows is used,
this temperature dierence can be increased to about 80 to 908C (1508F). Expansion
bellows are generally uneconomical for high pressures [>4150 kPa gauge (600 psig)]. The
xed tubesheet exchanger is a low-cost unit ranked after the U-tube exchanger.
The dierential thermal expansion can be accommodated by a oating rear-end head
in which tubes expand freely within the shell, thus eliminating thermal stresses. Also, the
tube bundle is removable for mechanical cleaning of the shell side. Basically, there are
three types of oating rear-end heads: U-tube heads, internal oating heads (pullthrough/split-ring heads), and outside packed oating heads.
In the U-tube bundle, the thermal stresses are signicantly reduced, due to free
expansion of the U-tubes, and the rear-end head has an integral cover which is the
least expensive among rear-end heads. The exchanger construction is simple, having
only one tubesheet and no expansion joints, and hence it is the lowest-cost design,
particularly at high pressures. The tube bundle can be removed for shell-side cleaning;
however, it is dicult to remove a U tube from the bundle except in the outer row, and it
is also dicult to clean the tube-side bends mechanically. So a U-tube exchanger is used
with clean uids inside the tubes unless the tube side can be cleaned chemically. Flowinduced vibration can also be a problem for the tubes in the outermost row because of a
long unsupported span, particularly in large-diameter bundles.
The next-simplest oating head is the pull-through head T shown in Fig. 10.5. On the
oating-head side, the tubesheet is small, acts as a ange, and ts in the shell with its own
bonnet head. The tube bundle can easily be removed from the shell by rst removing the
FIGURE 10.5 Two-pass exchanger (BET) with a pull-through (T) rear-end head. (Courtesy of
Patternson-Kelley Co., Division of HARSCO Corporation, East Stroudsburg, Pennsylvania.)
FIGURE 10.6 (a) Sealing strips; (b) dummy tubes or ties rods; (c) sealing strips, dummy tubes, or
tie rods covering the entire length of the tube bundle.
691
front-end head. Individual tubes or the tube bundle can also be replaced if required.
Due to the oating-head bonnet ange and bolt circle, many tubes are omitted from the
tube bundle near the shell. This results in the largest bundle-to-shell circumferential
clearance or a signicant bundle-to-shell bypass stream C. So as not to reduce exchanger
performance, sealing strips (or dummy tubes or tie rods with spacers) in the bypass area
are essential, as shown in Fig. 10.6. They are placed in pairs every ve to seven tube
pitches between the bae cuts. They force the uid from the bypass stream back into the
bundle. However, localized high velocities near the sealing strips could cause owinduced tube vibration; hence, proper care must be exercised for the design. Since this
design has the least number of tubes in a bundle for a given shell diameter compared to
other oating-head designs, the shell diameter is somewhat larger, to accommodate a
required amount of surface area. One of the ideal applications of the TEMA T head
design is in the kettle reboiler, for which there is ample space on the shell side and the ow
bypass stream C is of no concern.
The large bundle-to-shell clearance can be minimized by bolting the oating-head
bonnet to a split backing ring (ange) as shown in Fig. 10.7. It is referred to as the TEMA
S rear-end head. The shell cover over the tube oating head has a diameter larger than
that of the shell. As a result, the bundle-to-shell clearances are reasonable and sealing
strips are generally not required. However, both ends of the exchanger must be disassembled for cleaning and maintenance. In both TEMA S and T heads, the shell uid is
held tightly to prevent leakage to the outside. However, internal leakage is possible due to
the failure of an internal hidden gasket and is not easily detectable. The TEMA T head
has more positive gasketing between the two streams than does the S head. Both TEMA
S and T head congurations are used for the tube-side multipass exchangers; the singlepass construction is not feasible if the advantages of the positive sealing of TEMA S and
T heads are to be retained. The cost of TEMA S and T head designs is relatively high
FIGURE 10.7 Two-pass exchanger (AES) with a split-ring (S) oating head. (Courtesy of
Patternson-Kelley CO., Division of HARSCO Corporation, East Stroudsburg, Pennsylvania.)
FIGURE 10.8 Two-pass exchanger (AEP) with an outside packed (P) oating head. (Courtesy of
Patternson-Kelley Co., Division of HARSCO Corporation, East Stroudsburg, Pennsylvania.)
692
FIGURE 10.9 Two-pass exchanger (BEW) having a packed oating head (W) with lantern rings.
(Courtesy of Patternson-Kelley Co., Division of HARSCO Corporation, East Stroudsburg,
Pennsylvania.)
compared to U-tube or xed-tubesheet units. The cost for the TEMA S head is higher
than for the TEMA T head. The split backing ring oating head is used extensively in the
petroleum industry for moderate operating pressures and temperatures. For very high
operating pressures and temperatures, the TEMA S head design has a special test ring
(TEMA, 1999).
In the outside-packed oating-head TEMA P design of Fig. 10.8, the stung box
provides a seal against the skirt of the oating head and prevents shell-side uid leakage
to the outside. This skirt (and the tube bundle) is free to move axially against the seal to
take thermal expansion into account. A split-ring ange near the end of the skirt seals the
back end of the chamber. Because of the specic design of this oating head, any leak
(from either the shell side or the tube side) at the gaskets is to the outside. Hence, the
TEMA P head is generally not used with very toxic uids. Also, the inlet and outlet
nozzles must be located at the stationary end; hence, this design could have only an even
number of tube passes. In this design, the bundle-to-shell clearance is large [about 38 mm
(1.5 in.)]; as a result, sealing strips are required. The TEMA P head exchanger is more
expensive than the TEMA W head exchanger.
The packed oating head with lantern ring or TEMA W head is shown in Fig. 10.9.
Here a lantern ring rests on the machined surface of the tubesheet and provides an
eective seal between the shell- and tube-side anges. Vents are usually provided in
the lantern ring to help locate any leaks in the seals before the shell-side and tube-side
uids mix. Although a single-pass design is possible on the tube side, generally an even
number of tube passes is used. The TEMA W head exchanger is the lowest-cost design of
all oating heads. Although its cost is higher than that of the U-tube bundle, this higher
cost is oset by the accessibility to the tube ends (by opening both rear- and front-end
heads) for cleaning and repair; consequently, this design is sometimes used in the
petrochemical and process industries.
A large number of combinations of front- and rear-end heads with dierent shell types
of Fig. 1.6 are possible, depending on the application and the manufacturer. Some
common types of combinations result in the following shell-and-tube heat exchangers:
AEL, AES, AEW, BEM, AEP, CFU, AKT, and AJW.
In light of the availability of dierent types of front- and rear-end heads, the tube
bundle of a shell-and-tube exchanger may simply be classied as a straight-tube or Utube bundle. Both have a xed tubesheet at the front end. The U-tube bundle has a shell
with a welded shell cover on the U-bend end. The straight tube bundle has either a xed
693
tubesheet or a oating head at the rear end. The former is referred to as a xed-tubesheet
bundle; the latter is referred to as a oating-head bundle.
With this background, two owcharts are presented in Fig. 10.10 to increase the
overall heat transfer rate or decrease the pressure drop on either the tube or shell side
during various stages of designing a shell-and-tube exchanger.
10.2.2
The chevron plate is the most common in PHEs. Hence, we will not discuss the reasoning behind why various other plate geometries have been used in PHEs. As described in
Section 1.5.2.1, PHEs have a number of advantages over shell-and-tube heat exchangers, such as compactness, low total cost, less fouling, accessibility, exibility in changing the number of plates in an exchanger, high q and ", and low uid residence time.
Because of these advantages, they are in second place to shell-and-tube heat exchangers
for market share in liquid-to-liquid and phase-change applications. The main reason for
their limited versatility involves the pressure and temperature restrictions imposed by
the gaskets. Replacing gaskets on one or both sides by laser welding of the plates (as in
694
a welded PHE) increases both the operating pressure and temperature limits of the
gasketed PHEs, and it allows the PHE to handle corrosive uids compatible with the
plate material. For low-heat duties (that translate into a total surface area up to 10 m2 ),
a more compact brazed PHE can replace the welded PHE, thus eliminating the frame,
guide bars, bolts, and so on, of welded or gasketed PHEs. A variety of other PHEs have
been developed for niche applications to cover specic operating conditions that cannot
be handled by the above-described PHEs. Some of these PHEs are described briey at
the end of Section 1.5.2.2.
Although the shell-and-tube heat exchanger is versatile and can handle all kinds of
operating conditions, it is not compact, not exible in design, requires a large footprint,
and is costly (total cost) compared to PHEs and other compact heat exchangers. Hence,
PHEs and many other heat exchanger designs have been invented to replace shell-andtube heat exchangers in individual narrow operating ranges. Refer to Sections 1.5.2 and
1.5.3 for details of these exchangers. Also, an excellent source of information on compact
heat exchangers for liquid-to-liquid and phase-change applications is a recent monograph by Reay (1999).
10.2.3
695
696
ns. Since there are no cuts in the surface, wavy ns are used in those applications
where an interrupted n might be subject to a potential fouling or clogging problem due
to particulates, freezing moisture, bridging louvers due to condensate, and so on.
Oset Strip Fins. This is one of the most widely used enhanced n geometries in platen heat exchangers (see Fig. 1.29d). The n has a rectangular cross section, and is cut
into small strips of length s . Every alternate strip is displaced (oset) by about 50% of
the n pitch in the transverse direction. In addition to the n spacing and n height, the
major variables are the n thickness and strip length in the ow direction. The heat
transfer coecients for the oset strip ns are 1.5 to 4 times higher than those of plain
n geometries. The corresponding friction factors are also high. The ratio of j=f for an
oset strip n to j=f for a plain n is about 80%. Designed properly, the oset strip n
exchanger would require a substantially lower heat transfer surface area than that of
plain ns at the same p. The heat transfer enhancement for an oset strip n is caused
mainly by the redeveloping laminar boundary layers for Re 10,000. However, at
higher Re, it acts as a rough surface (a constant value of f with decreasing j for
increasing Re).
Oset strip ns are used in the approximate Re range 500 to 10,000, where enhancement over the plain ns is substantially higher. For specied heat transfer and pressure
drop requirements, the oset strip n requires a somewhat higher frontal area than a
plain n, but a shorter ow length and overall lower volume. Oset strip ns are used
extensively by aerospace, cryogenic, and many other industries where higher heat transfer performance is required.
Louver Fins. Louvers are formed by cutting the metal and either turning, bending, or
pushing out the cut elements from the plane of the base metal (see Fig. 1.29e). Louvers
can be made in many dierent forms and shapes. The louver n gauge is generally
thinner than that of an oset strip n. Louver pitch (also referred to as louver width)
and louver angle (in addition, the n spacing and n height) are the most important
geometrical parameters for the surface heat transfer and ow friction characteristics.
On an absolute level, j factors are higher for louver ns than for the oset strip n at
the same Reynolds number, but the f factors are even higher than those for the oset
strip n geometry. Since the louver n is triangular (or corrugated), it is generally not as
strong as an oset strip n; the latter n has a relatively large at area for brazing, thus
providing strength. The louver ns may have a slightly higher potential for fouling than
oset strip ns. Louver ns are amenable to high-speed mass production manufacturing
technology, and as a result, are less expensive than oset strip ns and other interrupted
ns when produced in very large quantities. The n spacing desired can be achieved by
squeezing or stretching the n; hence it allows some exibility in n spacing without
changes in tools and dies. This exibility is not possible with the oset strip n.
A wide range in performance can be achieved by varying the louver angle, width, and
form. The operating Reynolds number range is 100 to 5000, depending on the type of
louver geometry employed. Modern multilouver ns have higher heat transfer coecients that those for oset strip ns, but with somewhat lower j=f ratios. However, the
performance of a well-designed multilouver n exchanger can approach that of an oset
strip exchanger, possibly with increased surface compactness and reduced manufacturing
cost. Multilouver ns (see Figs. 1.27, 1.28, and 1.29e) are used extensively by the automotive industry.
697
Perforated Fins. A perforated n has either round or rectangular perforations with the
size, shape, and longitudinal and transverse spacings as major perforation variables (see
Fig. 1.29f ). The perforated n has either triangular or rectangular ow passages. When
used as a plate-n surface, it is generally brazed. The holes interrupt the ow and may
increase h somewhat, but considerable surface area may also be lost, thus nullifying the
advantage. Perforated ns are now used only in a limited number of applications. They
are used as turbulators in oil coolers for mixing viscous oils, or as a high-p n to
improve ow distribution. Perforated ns were once used in vaporizing cryogenic uids
in air separation exchangers, but oset strip ns have now replaced them.
Pin Fins. These can be manufactured at very high speed continuously from a wire of
proper diameter. After the wire is formed into rectangular passages (e.g., rectangular
plain ns), the top and bottom horizontal wire portions are attened for brazing or
soldering with the plates. Pins can be circular or elliptical in shape. Pin n exchanger
performance is considerably lower, due to the parasitic losses associated with round
pins in particular and to the inline arrangement of the pins (which results from the highspeed manufacturing techniques). The surface compactness achieved by pin n geometry is much lower than that of oset strip or louver n surfaces. Due to vortex
shedding behind the round pins, noise- and ow-induced vibration may be a problem.
Finally, the cost of a round wire is generally more than the cost of a at sheet, so there
may not be a material cost advantage. The potential application for pin ns is at very
low Reynolds number (Re < 500, for which the pressure drop is of no major concern.
Pin ns are used in electronic cooling devices with generally free convective ows over
the pin ns.
10.2.3.2 Tube-Fin Surfaces. When an extended surface is needed on only one uid
side (such as in a gas-to-liquid exchanger) or when the operating pressure needs to be
contained on one uid side, a tube-n exchanger (see Section 8.2) may be selected, with
the tubes being round, at, or elliptical in shape. Also, when minimum cost is essential,
a tube-n exchanger is selected over a plate-n exchanger since the ns are not brazed
but are joined mechanically to the tubes by mechanical expansion. Flat or elliptical
tubes, instead of round tubes, are used for increased heat transfer in the tube and
reduced pressure drop outside the tubes; however, the operating pressure is limited
compared to that for round tubes. Tube-n exchangers usually have lower heat transfer
surface compactness than a plate-n unit, with a maximum heat transfer surface area
density of about 3300 m2 /m3 (1000 ft2 /ft3 ).
A tube-n exchanger may be designed for a wide range of tube uid operating pressures [up to about 3000 kPa gauge (450 psig) or higher] with the other uid being at low
pressure [up to about 100 kPa (15 psig)]. The highest operating temperature is again
limited by the type of bonding and the materials employed. Tube-n exchangers are
designed to cover the operating temperature range from low cryogenic temperatures to
about 8708C (16008F). Reasonable fouling can be tolerated on the tube side if the tubes
can be cleaned. Fouling is generally not a problem on the gas side (n side) in many
applications; plain uninterrupted ns are used when moderate fouling is expected.
Fluid contamination (mixing) of the two uids is generally not a problem since there
is essentially no uid leakage between them. Since tubes are generally more expensive
than extended surfaces, the tube-n exchanger is in general more expensive. In addition,
the heat transfer surface area density of a tube-n core is generally lower than that of a
plate-n exchanger, as mentioned earlier.
698
The tube-n construction is generally used in liquid-to-gas or phase-change uid-togas heat exchanger applications with liquid, condensing uid, or evaporating uid on the
tube side. Fins are generally used on the outside of tubes (on the gas side), although
depending on the application, ns or turbulators may also be used inside the tubes.
Round and at tubes (rectangular tubes with rounded or sharp corners) are most common; however, elliptical tubes are also used. Round tubes are used for higher-pressure
applications and also when considerable fouling is anticipated. Parasitic form drag is
associated with ow normal to round tubes. In contrast, the at tubes yield a lower
pressure drop for ow normal to the tubes, due to lower form drag, and thus avoid
the low-performance wake region behind the tubes. Also, the heat transfer coecient is
higher for ow inside at tubes than for circular tubes, particularly at low Re. The use of
at tubes is limited to low-pressure applications, such as automotive radiators, unless the
tubes are extruded with ribs inside (see the multiport tube in Fig. 1.27, also referred to as
microchannels) or with integral ns outside.
Flat Fins on a Tube Array. This type of tube-n geometry (shown in Fig. 1.31b) is most
commonly used in air-conditioning and refrigeration exchangers in which high pressure
needs to be contained on the refrigerant side. As mentioned earlier, this type of tube-n
geometry is not as compact (in terms of surface area density) as the plate-n geometries,
but its use is becoming widespread due to its lower cost. This is because the bond
between the n and tube is made by mechanically or hydraulically expanding the
tube against the n instead of soldering, brazing, or welding the n to the tube.
Because of the mechanical bond, the applications are restricted to those cases in
which the dierential thermal expansion between the tube and n material is small,
and preferably, the tube expansion is greater than the n expansion. Otherwise, the
loosened bond may have a signicant thermal resistance.
Many dierent types of at ns are available (see some examples in Fig. 1.33). The
most common are the plain, wavy, and interrupted. The plain at ns are used in those
applications in which the pressure drop is critical (quite low), although a larger amount
of surface area is required on the tube outside for the heat transfer specied than with
wavy or interrupted ns. Plain at ns have the lowest pressure drop than that of any
other tube-n surfaces at the same n density. Wavy ns are superior in performance to
plain ns and are more rugged. Wavy ns are used most commonly for air-conditioning
condensers and other commercial heat exchangers. A variety of louver geometries are
possible on interrupted at ns. A well-designed interrupted n would have even better
performance than a wavy n; however, it may be less rugged, more expensive to manufacture, and may have a propensity to clog.
Individually Finned Tubes. This tube-n geometry (shown in Fig. 1.31a) is generally
much more rugged than continuous n geometry but has lower compactness (surface
area density). Plain circular ns are the simplest and most common. They are manufactured by tension wrapping the n material around a tube, forming a continuous
helical n or by mounting circular disks on the tube. To enhance the heat transfer
coecient on the ns, a variety of enhancement techniques have been used (see Fig.
1.32). Segmented or spine ns are the counterpart of the strip ns used in plate-n
exchangers. A segmented n is generally rugged, has heavy-gauge metal, and is usually
less compact than a spine n. A studded n is similar to a segmented n, but individual
studs are welded to the tubes. A slotted n has slots in the radial direction; when
radially slitted material is wound on a tube, the slits open, forming slots whose width
699
increases in the radial direction. This n geometry oers an enhancement over tensionwound plain ns; however, segmented or spine ns would yield a better performance.
The wire loop n is formed by spirally wrapping a attened helix of wire around the
tube. The wire loops are held to the tube by a tensioned wire within the helix or by
soldering. The enhancement characteristic of small-diameter wires is important at low
ows, where the enhancement of other interrupted ns diminishes.
10.2.4
Regenerator Surfaces
700
FIGURE 10.11 Comparison of surface basic characteristics of two heat exchanger surfaces.
(From Shah, 1983.)
gas ows in compact heat exchangers and performance evaluation criteria (PECs) for
tubular surfaces. For extended surfaces, particularly the plate-n type, selection of
the surfaces on both uid sides are independent of each other, and generally, one
uid side is critical from the pressure drop requirement. Hence, we consider only one
uid side for the surface selection for plate-n surfaces.
10.3.1
Screening Methods
701
(not necessarily related to the surface j and f vs. Re characteristics) inuence the overall
performance of a heat exchanger.
In addition, if the exchanger is considered as part of an (open or closed system), the
exchanger surface (and/or other variables) may be selected based on the system as a
whole rather than based on the optimum exchanger as a component. Current methods of
surface selection for an optimum heat exchanger for a system include the use of sophisticated computer programs that take into account many possible eects. Such selection is
not possible in simplied approaches presented in the open literature. We focus on
considering simple but important quantitative screening methods for surface selection
on the gas side of compact heat exchangers since these exchangers employ a large variety
of high-performance surfaces.
The selection of a surface for a given application depends on exchanger design criteria. For a specied heat transfer rate and pressure drop on one uid side, two important
design criteria for compact exchangers (which may also be applicable to other exchangers) are the minimum heat transfer surface area requirement and the minimum frontal
area requirement. Let us rst discuss the signicance of these criteria.
To understand the minimum frontal area requirement, let us rst review how the uid
pressure drop and heat transfer are related to the ow area requirement, the exchanger
ow length, and the uid velocity. The uid pressure drop on one uid side of an
exchanger, neglecting the entrance/exit and ow acceleration/deceleration losses, is
given from Eq. (6.29) as
p
4fLG2
2gc Dh
10:1
C1 Re1
C2 Re0:5
10:2
where C1 and C2 are constants. Substituting Eq. (10.2) into Eq. (10.1) and noting
that Re GDh =, we get
p /
LG
LG1:5
10:3
Here G m_ =Ao : Therefore, for a specied constant ow rate m_ , the pressure drop is
proportional to the ow length L and inversely proportional to the ow area Ao or
A1:5
o .
The Nusselt number for laminar developed and developing temperature proles and
developed velocity proles is given by (see Sections 7.4.1.1 and 7.4.3.1)
Nu
C3
C4 Dh Pr Re=L1=3
10:4
702
where C3 and C4 are constants. Thus, the heat transfer coecient is independent of
the mass ow rate m_ or mass velocity G for the thermally developed laminar ow
and is proportional to G1=3 for the thermally developing laminar ow. We have not
considered simultaneously developing laminar ow (in which Nu / G1=2 ) since thermally developing ow provides a conservative estimate of Nu.
Considering p and h simultaneously, a decrease in G will reduce p linearly without
reducing h for fully developed laminar ow; for developing laminar ows, a reduction in
G will reduce p as in Eq. (10.3), with a slight decrease in h as given by Eq. (10.4).
Now as discussed earlier, there are a variety of enhanced surfaces available for selection. An undesirable consequence of the heat transfer enhancement is an increase in the
friction factor, which results in a higher pressure drop for an exchanger having a xed
frontal area and a constant ow rate. As noted in the preceding paragraph, reducing G
can reduce the pressure drop in compact exchangers without signicantly reducing the
heat transfer coecient h. For a specied constant ow rate, a reduction in G means an
increase in the ow area Ao for constant m_ and approximately constant (the ratio of
free-ow to frontal area). So as one employs the more enhanced surface, the required
ow area (and hence frontal area) increases accordingly to meet the heat transfer and
pressure drop requirements specied. Thus, one of the characteristics of highly compact
surfaces is that the resulting shape of the exchanger becomes more like a pancake, having
a large frontal area and a short ow length (e.g., think of the shape of an automotive
radiator in contrast to a shell-and-tube exchanger; see also Example 10.3 to show the
increase in free-ow area when employing a higher-performance surface). Hence, it is
important to determine which of the compact surfaces will meet a minimum frontal area
requirement.
The surface having the highest heat transfer coecient at a specied ow rate will
require the minimum heat transfer surface area. However, the allowed pressure drop is
not unlimited. Therefore, one chooses the surface having the highest heat transfer coecient for a specied uid pumping power. The exchanger with the minimum surface area
will have the minimum overall volume requirement.
From the foregoing discussion, two major selection criteria for compact surfaces with
gas ows are (1) a minimum frontal area requirement and (2) a minimum volume
requirement. For this purpose, the surfaces are evaluated based on the surface ow
area and volume goodness factors. We discuss these comparison methods after the
following example.
Example 10.1 Consider a gas turbine rotary regenerator (Fig. E10.1) having compact
triangular ow passages operating at Re 1000 and the pressure drop on the highpressure air side as 10 kPa. Determine the change in heat transfer and pressure drop
if this regenerator is operated at Re 500. The following data are provided for the
analysis:
j
3:0
Re
14:0
Re
Pr 0:7
hAh hAc
flow split 50 : 50
Ignore the eect of the ow area blockage by the hub and the radial seals to
determine the required changes. Assume that the mass ow rate of air (and hence
gas) does not change when Re is reduced. How would you achieve the reduction in
Re when the L and Dh values of the regenerator surface are being kept constant?
703
Hot gas
Cold air
FIGURE E10.1
SOLUTION
Problem Data and Schematic: The data are given in the problem statement for j, f ,
Pr, and thermal conductances and ow split. Also, L and Dh are to be kept constant.
Determine: Determine the change in heat transfer and pressure drop for this regenerator when the air-side (and hence gas-side) Re is reduced from 1000 to 500. How is
the reduction in Re achieved when the mass ow rate of the air is kept constant?
Assumptions: The ow is fully developed (thermally and hydrodynamically) laminar.
The wall thermal resistance and fouling resistance are negligible. The uid properties
do not change with the change in Re; and with a ow split of 50 : 50, we expect the
same eects on the air and gas sides with a change in Re.
Analysis: Since the ow is assumed to be fully developed laminar, Nu will be constant (see Section 7.4.1.1) and is given by its relationship to the j factor by Eq.
(7.33) and input data j Re 3:0 as
Nu j Re Pr1=3 3:00:71=3 2:66
Using the denition of Nu hDh =k and constant Dh for the change of Re from 1000
to 500, we get
Nu2 h1
1
Nu1 h2
Here subscripts 1 and 2 denote the cases for Re 1000 and 500, respectively. Thus,
the heat transfer coecient does not change with Re. This will also be the case for h
on the gas side. Thus, UA and hence the heat transfer rate q will not change.
704
When Re GDh = is changed from 1000 to 500 without changing the mass ow
rate and the regenerator geometry (L and Dh ), we get
Re2 G2
500
1
Re1 G1 1000 2
Now let us evaluate the pressure drop. From Eq. (6.29),
p /
L
G f Re
D2h
Since L, Dh , and f Re are constant for this regenerator, we can evaluate p2 with
Re 500 from
p2 G2 1
p1 G1 2
or
Ans:
r21 Af ;1 Ao;1 =
Therefore,
r2 p
2 1:41
r1
where is the ratio of free-ow area to frontal area, and r2 and r1 are disk radii for
Re 1000 and 500, respectively. Hence, the disk radius or diameter of this regenerator will need to be increased by 41%.
Ans.
Discussion and Comments: As discussed in Section 7.4.1.1, this problem clearly indicates that in fully developed laminar ow, the heat transfer coecient is not reduced
by reducing the uid velocity um G=, whereas the pressure drop is linearly
reduced with the reduction in the ow velocity. Hence, a reduction in ow velocity
without a reduction in the mass ow rate can be achieved by increasing the ow area
and hence the frontal area of the exchanger.
10.3.1.1 Surface Flow Area Goodness Factor Comparison. London (1964) dened the
ratio j=f as the surface ow area goodness factor. Using the denitions of j, Nu, f , and
Re, we get the ratio j=f as
j Nu Pr1=3
1
2
f Re
f
Ao o
Pr2=3 ntu m_ 2
2gc p
!
10:5
705
The term in parentheses on right-hand side of Eq. (10.5) is dependent only on the
operating parameters and is independent of the geometry and heat transfer surface
involved. Equation (10.5) can be rearranged as
A*o
Afr*
Ao
Pr
2=3
=2gc ntu
m_ 2 =p1=2
Afr
Pr2=3 =2gc ntu m_ 2 =p1=2
1
o j=f 1=2
1
o j=f 1=2
10:6a
10:6b
The left-hand sides of Eq. (10.6a) and (10.6b) are the dimensionless free-ow area A*o
and frontal area Afr*, respectively. Equations (10.5) and (10.6) show the signicance
of j=f as being inversely proportional to A2o Ao is the surface minimum free-ow
area) for specied operating conditions and o as constant. A surface having a higher
j=f factor is good because it will require a lower free-ow area and hence a lower
frontal area for the exchanger. The dimensionless j and f factors are independent of
the length scale of the geometry (i.e., the hydraulic diameter).y Thus, the ow area
Ao is independent of the hydraulic diameter, but dependent on the operating conditions (m_ and p, design condition (ntu), and uid type (Pr). Note that for many
compact surfaces, no signicant variation is found in the j=f ratio over the reported
test Reynolds number range, and hence Ao and Afr are not a strong function of the
surface type.
For fully developed laminar ow through simple geometries of Table 7.3, we
nd jH1 =f ranging from 0.263 for the equilateral triangular duct to 0.386 for the parallelplate duct. Thus, the parallel-plate duct relative to the triangular duct has 47%
0:386=0:263 1 higher j=f . Then from Eq. (10.6), we get Ao;1 =Ao;2
0:263=0:3861=2 0:825 (since o;1 o;2 1; where the subscripts 1 and 2 are for
the parallel-plate and triangular passages, respectively. Thus a parallel-plate exchanger
would have a 17.5% smaller free-ow area requirement. In the free-ow area goodness
factor comparison, no estimate of total heat transfer area or volume can be inferred.
Such estimates may be derived from the core volume goodness factors described next.
10.3.1.2 Core Volume Goodness Factor Comparisons. Two types of core volume goodness factor comparisons are suggested: hstd vs. Estd and o hstd vs. Estd . In the rst
method, a comparison is made for surfaces having the same hydraulic diameter. In the
second method, a comparison is made of the actual performance of surfaces having
equal or dierent hydraulic diameters. The heat transfer rate per unit temperature
dierence and per unit surface area q=ATw Tm z , and the uid pumping power
due to friction per unit of surface area are expressed as
h
cp
1
j Re
Pr2=3 Dh
10:7
As long as the surface geometrical similarity is maintained (i.e., the surface geometry is enlarged or reduced in size
by changing the hydraulic diameter), and j and f vs. Re characteristics remain the same. Hence, the j=f ratio is
independent of Dh at a given design Re.
{
The term heat transfer power is sometimes used to denote heat transfer rate q, watts.
706
P
1 3 1
f Re3
A 2gc 2 D3h
10:8
In either of the volume goodness factor comparisons, it is assumed that the surfaces
under comparison will provide the same performance. It means that the following
quantities are kept constant: (1) the same heat transfer rate, (2) the same pressure
drop, (3) the same temperature dierence between the wall and the uid, and (4)
the same uid ow rate. Remember that we want to arrive at a best surface on one
uid side of the exchanger from the minimum-volume-requirement point of view. The
heat transfer rate q and uid pumping power P due to friction on one uid side are
q o hATw Tm o hVTw Tm
P EA EV
m_ p
10:9
10:10
1
V
10:11
Thus the higher hy means a lower overall core volume requirement. Then the
excellence of a particular surface geometry in terms of the core volume is characterized
by a high position on a dimensional plot of h vs. E, as suggested by London and Ferguson
(1949). Considering the gas turbine application, they evaluated all uid properties for dry
air at standard conditions (std) of 2608C (5008F) and 1 atm pressure. However, for
other applications, these standard conditions for uid property evaluation could be
changed to any conditions for the desired uid for the hstd vs. Estd plot.
Since the regenerator is used for a gas-to-gas application, it is generally a thermally
balanced heat exchanger. In this case, the thermal resistances on both sides are of the
same order of magnitude and hence UA hA=2. Thus the comparison of hstd with Estd is
a realistic comparison for regenerators.
The hstd vs. Estd plot for fully developed ow with constant uid properties through
some constant cross-sectional ducts is presented in Fig. 10.12 for Dh 0:5 mm
(0.0016 ft). From this gure, it is found that hstd varies from 256.4 to 700.6 W/m2 K,
a factor of 2.7, with hstd 264:7 W=m2 K for the equilateral triangular duct.
When made for a xed Dh , a plot of Fig. 10.12, clearly shows the inuence of the
passage shape. The parallel plate heat exchanger may prove impractical, but it is clear
that there are several other congurations that possess signicant advantages over the
triangular and sine duct geometries. Based on this plot, the development and use of
{
Note that the heat transfer coecient h is the same as the heat transfer power per unit temperature dierence and
per unit surface area.
707
FIGURE 10.12 Theoretical laminar volume goodness factors for some simple duct geometries
(From Shah, 1983.)
rectangular passage geometry is being continued for applications that involve fully
developed laminar ows.
From Eqs. (10.7) and (10.8), it is evident that the dimensional hstd vs. Estd
performance is strongly dependent on the length scale of the surface geometry (i.e.,
Dh ). Thus, this comparison method reveals the benet of increased performance (reduced
surface area requirement) by going to a smaller Dh surface. This will also result in a much
more compact surface.
For the foregoing reasons, the plot of hstd vs. Estd is recommended for selection of a
heat exchanger surface for a new application for which there are no signicant system or
manufacturing constraints.
o hstd vs. Estd Comparisons. The preceding method of comparison was for surfaces
having the same Dh , , and o value (if there are any ns). When one wants to compare
the performance of extended surfaces, for which j and f data are available, one may be
interested in comparing the surfaces as they are. This is because we may not be able to
manufacture a surface whose geometry is scaled up or scaled down. Such a comparison
of actual surfaces could be made by a plot of o hstd vs. Estd . Here is the surface area
density or compactness, o hstd represents the heat transfer power per unit temperature
dierence and unit core volume, and Estd represents the friction power expenditure per
unit core volume. Note that this plot is modied from the hstd vs. Estd recommended
by Kays and London (1998) by including the eect of overall n eciency o of
the secondary surface. This eect is important for extended surface heat exchanger
applications.
The foregoing variables, for a given set of surfaces, are evaluated from the following
equations with uid properties determined at some standard conditions:
o hstd
Estd
cp
Pr2=3
o
4
j Re
D2h
3 4
f Re3
2gc 2 D4h
10:12
10:13
708
0:952
1=2
Ao;1 j=f
5:2=22:01=2
2
where the subscripts 1 and 2 are for triangular and rectangular ow passages. Thus,
a rectangular passage exchanger will require 4.8% (0.048) smaller ow area and
hencepfrontal
area for the same porosity. This translates approximately into 2.4%
1 0:952 reduction in the disk diameter.
From Eqs. (10.11) and (10.12), the core volume ratio is
V2 h1 Nu1 j Re Pr1=3 1 3:0
0:577
V1 h2 Nu2 j Re Pr1=3 2 5:2
Hence the reduction in the core volume is 42.3% 1 0:577:
Ans.
0:606
L1 Ao;2 V1 Ao;2 =Ao;1 0:952
{
As noted just before, Eq. (10.11) is valid for 1 2 . Hence, we have considered here as constant. This
means that the wall thickness for rectangular and triangular passages will be dierent. They can be calculated
using the formulas of Table 8.2 by equating hydraulic diameters, and for the same computing b for rectangular
passages for the known b for equilateral triangular passages.
709
1:05
Re1 m_ Dh =Ao 1 Ao;2 0:952
Thus, the operating Reynolds number will increase by 5% or will be 1050
1000 1:05:
Ans.
FIGURE E10.3 The j and f data comparison for an oset strip n and plain n at Dh 3:51 mm.
(From Webb, 1994.)
710
SOLUTION
Problem Data and Schematic: The j and f data are given in Fig. E10.3 for the oset
strip n and plain n of Fig. 1.29d and b respectively. For the plain n, Re 3000.
Determine: For identical m_ , hA, and P, determine As =Ap and Ao;s =Ao; p where the
subscripts s and p denote the values for strip n and plain n surfaces, respectively.
Assumptions: Assume constant uid properties and only one uid side of the exchanger.
Analysis: Since the j and f factors are higher for the oset strip n, the operating
Reynolds number for the oset strip n will be lower than that for the plain n for
the same m_ , hA, P, q, Th;i , and Tc;i . This Reynolds number needs to be evaluated
iteratively from the following relationship, which is derived from Eq. (10.5) with
constant ntu, m_ and p, and, the denition of Re m_ =Ao Dh = with m_ and Dh
constant]:
Res Ao; p j=f 1=2
s
From Fig. E10.3, the j and f factors at Re 3000 for the plain n are
jp 0:0038
fp 0:011
We need to assume a value of Res such that the corresponding js and fs satisfy
Eq. (1) for the values of Rep , jp , and fp above. Let us assume that Rep =Res 1:20. Hence,
Res
3000
2500
1:20
From Fig. E10.3, the j and f factors at Re 2500 for the strip n are
js 0:010
fs 0:042
0:83
Rep 0:0038=0:0111=2
or Rep =Res 1:20. Therefore, our guess of Rep =Res is correct (which was obtained
based on iterations) and we dont need to iterate further. Otherwise, continue guessing the values of Rep =Res until Eq. (1) is satised.
From Eq. (10.8), the uid pumping power P is given by
P
1 3 1
f Re3 A
2gc 2 D3h
Applying this equation, for equal pumping power (P p P s and Dh;p Dh;s , we get
3
As f Re p 0:011
1:203 0:453
Ap f Re3 s 0:042
Ans:
711
where the numerical values computed earlier are substituted. The minimum free-ow
area ratio, from Eq. (10.5), is given by
Ao;s j=f 1=2
p
1:20
Ao; p j=f 1=2
s
Ans:
The oset strip n will require 20% higher frontal area (for the same ) and 54.7%
1 0:453 less surface area than will the plain n for this problem. Using the denition
of the hydraulic diameter Dh 4Ao L=A, we get
Ls As Ao; p Dh;s
1
0:453
1 0:378
Lp Ap Ao;s Dh; p
1:20
since Dh;s Dh; p is given. Now if we assume the porosity is the same for the two
surfaces, the volume ratio is given by
Ao;s Ls
Vs
712
rate (heat duty), and pressure drop (or uid pumping power) for the derivation of these
relationships.
The ratios of free-ow area and frontal areas on one uid side for surface 1 to surface
2 are as follows using Eq. (10.6):
Ao;2
o;1 j=f 1 1=2
Ao;1
o;2 j=f 2
Afr;2 1 o;1 j=f 1 1=2
10:14
Note that this equation is valid for any (the same or dierent) hydraulic diameter of
the two surfaces and satises the heat transfer and pressure drop requirements
specied for a given uid ow rate (i.e., q, p, and m_ are constant for these
relationships).
For an equal uid pumping power requirement [i.e., P constant for Eq. (10.10)] and
equal heat duty q, from Eq. (10.9) we get the following equation for volume between
plates on one uid side with use of the denitions Nu hDh =k; Dh 4=, and
Nu j Re Pr1=3 :
V2 o h1 o;1 1 Dh;2 Nu1 o;1 1 j Re1 Dh;2 2
10:15
Using Dh from Re m_ =Ao Dh =, substituting it in Eq. (10.15), and using Eq.
(10.14), we get an alternative form of Eq. (10.15) as follows:
V2 f Re2 o;1 1
V1 f Re1 o;2 2
2
j1
j2
10:16
The ow length ratio on one uid side can be computed from L V=Afr as follows
using Eq. (10.15) and the denition of Re m_ =Afr Dh =:
L2 o;1 j1 Dh;2
L1 o;2 j2 Dh;1
10:17
The heat transfer surface area ratio A2 =A1 on one uid side is then calculated from the
denition A 4Ao L=Dh and Re / Dh =Ao , Eqs. (10.14) and (10.17), as follows:
A2 Ao;2 L2 Dh;1 L2 Re1
j=f 1 1=2 o;1 3=2 j1
10:18
Finally, the ratio of operating Reynolds numbers for the two surfaces can be determined from the denition Re m_ =Ao Dh = and the use of Eq. (10.14) as follows:
Re2
j=f 2 1=2 Dh;2
Re1
j=f 1
Dh;1
10:19
Since the relationships above are derived for constant q, p, m_ , and P, they are not
included in the list of operational parameters. Based on Eqs. (10.14)(10.18), we nd
that the ratios for Ao , Afr , and A are independent of Dh , and the ratios for V and L
are proportional to Dh for fully developed laminar ow. However, for other ows
713
(such as turbulent, transition, and developing laminar ows), one more constraint (in
addition to constant q, p, m_ ; P) needs to be specied by keeping constant the lefthand side of one of the equations (10.14)(10.19).
While the foregoing relationships are obtained for the case of keeping q, p and m_
constant, similar relationships can be obtained by keeping some variables constant and
others varying from the following set: Afr , V, L, A, Dh , Re, q, p, m_ ; P, and so on. Cowell
(1990) presents a number of such relationships.
Several important observations can be made from the foregoing relationships keeping
q, p and m_ constant. These are useful when we have dierent surfaces for selection and
we need to decide which to choose for a compact surface having fully developed laminar
ow.
. Since there is no hydraulic diameter involved in the ow area ratio relationship of
Eq. (10.14), the ow area on one uid side is independent of Dh for fully developed
laminar ow. Changing Dh without changing the wall thickness would result in a
slight change in the porosity and hence in Afr .
. From Eq. (10.15), V / Dh =j. Hence, the heat exchanger volume on one uid side
decreases with increasing porosity , increasing Colburn factor j and reducing
hydraulic diameter Dh .
. Based on Eq. (10.17), the ow length on one uid side decreases with increasing
Colburn factor j and decreasing hydraulic diameter Dh .
. Based on Eq. (10.18), the heat transfer surface area on one uid side decreases with
increasing j, Re, or Nu ( j Re Pr1=3 ), and decreasing Dh .
Note that j and f factors for any surface are independent of the hydraulic diameter
as long as the surface is geometrically scaled up or down.
To emphasize the foregoing points, let us consider a case of a rotary regenerator
having a 50 : 50% split for air and gas ows and having fully developed laminar ows,
or one uid side of a plate-n exchanger with fully developed laminar ow. If the
hydraulic diameter is reduced to one-half of the original value on one uid side, one
can show from Eqs. (10.14)(10.19) that Ao and A remain constant, V reduces to 12, L
reduces to 12, and Re reduces to 12, all changes for constant q, p, and m_ on one uid
side.
10.3.2
714
TABLE 10.4 Performance Evaluation Criteria for Enhanced Surfaces (with Constant di ) for SinglePhase Heat Exchangers
Fixed
PEC
Case
FG-1a
FG-1b
FG-2a
FG-2b
FG-3
FN-1
FN-2
FN-3
VG-1
VG-2a
VG-2b
VG-3
Geometry
m_
nt ; L
nt ; L
nt ; L
nt ; L
nt ; L
nt
nt
nt
nt L
nt L
nt L
q
Tmax
Objective
Increase q
Decrease Tmax
Increase q
Decrease Tmax
Decrease P
Decrease L
Decrease L
Decrease P
Decrease nt L
Increase q
Decrease Tmax
Decrease P
length L), ow rate m_ , uid pumping power P, heat transfer rate q, and uid inlet
temperature dierence Tmax . For a given PEC, the ratio of the design objective for
a surface of interest to that for a reference surface is then calculated as a function of
a similar ratio of a design variable.
The heat transfer rate in an exchanger is given by
q UA Tm PRTh;i Tc;i PR Tmax
10:20
Here P is the exchanger temperature eectiveness and R is the heat capacity rate
ratio. Reviewing this equation, the design objectives for the use of enhanced surface
may be as follows:
1. Increased UA for equal pumping power P and xed geometry [frontal area
(represented by nt ) and length L]. A higher UA means (a) higher q for a given
Tmax or Tm , and (b) lower Tmax or Tm for a given q.
2. For a xed ow area (i.e., xed nt in Table 10.4), (a) reduce the tube length L (and
hence A) for equal q and m_ or P, and (b) reduce P for equal q and m_ : In all cases,
Tmax or Tm is xed.
3. Reduce the surface area A and hence the volume and mass of the exchanger for
xed m_ and specied q (or Tmax and P.
Based on the foregoing objectives, three major categories of performance evaluation
criteria are developed: Fixed geometry (FG), xed ow area (FN), and variable geometry
(VG) criteria. For xed-geometry criteria, a plain surface is replaced by an enhanced
surface of equal length, a retrot application, resulting in higher q (or reduced Tmax )
and higher P. For xed-ow-area criteria, either P (and L) is reduced at constant q and m_
or m_ (and L) is reduced at constant P and q, employing an enhanced surface. For
715
variable-geometry criteria, the surface area A (/ nt L) is reduced and frontal area Afr
/ nt is increased for an enhanced surface for xed q, m_ , and P. Several PECs are
formulated based on these major criteria, as noted in various FG, FN, and VG cases
in Table 10.4.
The advantages of PEC comparison methods are that (1) the designer can select his or
her own criteria for comparison; (2) he or she can then compare the performance of a
surface to that of a reference surface directly, and (3) he or she does not need to evaluate
the uid properties since they drop out in computing the performance ratios. The
performance comparisons can include the eect of the thermal resistances of the wall
and of fouling and convection on the other uid side. The optimum surface selected by
this method may not be optimum in a two-uid heat exchanger when non-performancerelated overall heat exchanger constraints are imposed. These aspects are considered in
heat exchanger optimization, discussed in Section 9.6.
The algebraic relations for the PECs will now be summarized for comparing an
enhanced tubular surface to the corresponding plain tube for a shell-and-tube heat
exchanger having the tube length L per pass, tube diameters di and do , number of
tubes nt in each pass, and number of passes np . For this exchanger, the ow area and
surface area on the tube side are given by
Ao
2
d n
4 i t
A di Lnt np
10:21
The frontal area Afr on the tube side is then related to Ao for specied tube layout
and pitches. Heat transfer and ow friction characteristics are needed for comparing
the performance of an enhanced surface to the corresponding plain surface. We will
use a subscript p to designate the quantities for the plain surface (except for cp ), and
no subscript for the enhanced surface. From the denition of the j factor,
h jGcp Pr2=3
10:22
To evaluate the heat duty q of the enhanced surface, we need to compare hA of the
enhanced surface to hAp for the plain surface. From Eq. (10.22),
hA
j G A
hAp jp Gp Ap
10:23
Note that here and in this section Ap represents the surface area of the plain surface.
Using Eq. (6.30), the uid pumping power ratio is given by
P
f A G 3
P p fp Ap Gp
10:24
j=jp
f =fp 1=3
10:25
716
This equation is used for a comparison of two heat transfer surfaces under various
criteria of Table 10.4 when two of the three ratios on the left-hand side of Eq.
(10.25) are given and the third is to be determined.
However, when comparing an enhanced surface with a plain surface in a heat exchanger, the ratio needed for heat transfer performance is UA=UAp , which takes into
account wall thermal resistance, fouling resistance, and convection resistance on the
second uid side. Using Eq. (3.20) or (3.24) for UA and UAp , one can arrive at the
following expression modied from Webb (1981):
1 R*p
1 R*p
1 R*p
UA
"
#1=3
jp Gp Ap
UAp hAi; p
jp f P p A p 2
R*
R*
R*
j G A
hA
j fp P A
10:26
where R* and R*p are the total thermal resistances (excluding tube inside thermal
resistance) for enhanced and plain surfaces, respectively, normalized with respect to
the plain tube inside thermal resistance 1=hAi; p . They are given explicitly in Table
10.5. Note that Ai; p is simply designated as Ap (plain tube inside surface area) in the
last two equalities of Eq. (10.26) and throughout in Table 10.6.
In Eq. (10.26), R* includes Ap =A. Hence, R* can be presented as a function of Ap =A as
follows (a formula alternative to that in Table 10.5) when Ap =A is not unity, but is a
variable:
^ *o R
^ o;*f Ap hi; p hi; p w hi; p Ap
^ * Ao R
^ * Ap R
10:27
R* R
A
Aw
A
ho
kw
ho; f
A
Using Eq. (3.94), the ratio of the heat transfer rate in an enhanced to the plain surface
is given by
q
m_ P Tmax
qp m_ p Pp Tmax; p
10:28
Here we have assumed the uid cp to be the same for enhanced and plain surfaces
(and hence C=Cp m_ =m_ p , P is the temperature eectiveness, and
Tmax Th;i Tc;i is the inlet temperature dierence. Since the temperature eectiveness P is dependent on NTU, the NTU of the enhanced surface is related to that
for the plain surface by
NTU NTUp
UA m_ p
UAp m_
10:29
Enhanced
Heat Exchanger
Fouling resistance
Denition
Combined resistance
* p Rf*; p
R*p Ro;*p Rw;
717
Tmax
Tmax; p
q
qp
FG-1b
FG-2a
Tmax
1
Tmax; p
q
1
qp
m_
G
1
m_ p Gp
P
1;
Pp
L
A
1
Lp Ap
Tmax
1
Tmax; p
nt
A
fr 1
nt; p Afr; p
L
A
1
Lp Ap
nt
A
fr 1
nt; p Afr; p
m_
G
1
m_ p Gp
nt
A
fr 1
nt; p Afr; p
q
qp
FG-1a
L
A
1
Lp Ap
Determine:
Case
1 R*p
UA
UAp jp =j R*
since
P p fp
P
f
1 R*p
UA
UAp jp =j R*
since
P p fp
P
f
Solve G=Gp iteratively from the rst equation knowing fp vs. Rep and f vs. Re
characteristics. Compute jp and j from the known jp vs. Rep and j vs. Re
characteristics. Calculate UA=UAp from the foregoing equation. Then
follow the procedure of FG-1a to get q=qp from NTU.
continued over
and
f
Gp
UAp jp =j f =fp 1=3 R*
Pp m_ p
Tmax
Tmax; p
P m_
Compute NTU from Eq. (10.29) and subsequently, P, for the given ow
arrangement. Then, from Eq. (10.28),
q
P
qp Pp
Compute NTU from Eq. (10.29) and subsequently P for the given ow
arrangement. Then, from Eq. (10.28),
718
FN-1
L
Lp
m_
G
m_ p Gp
P
q
Tmax
1
P p qp Tmax; p
A
L
Ap Lp
m_
G
m_ p Gp
q
Tmax
1
qp Tmax; p
nt
A
fr 1
nt; p Afr; p
L
A
1
Lp Ap
nt
A
fr 1
nt; p Afr; p
q
1
qp
P
1
Pp
P
Pp
L
A
1
Lp Ap
nt
A
fr 1
nt; p Afr; p
Tmax
Tmax; p
FG-2b
FG-3
Determine:
Case
P
f
P p fp
G
Gp
3
2
P
f L
1
Pp
fp Lp
G
Gp
3
;
q
G P
1
qp
Gp Pp
1 R*p
UA
continued over
q
G P
1
qp
Gp Pp
1 R*p
UA
UAp jp =j f = fp P p =P1=3 R*
Pp m_ p
Tmax
Tmax; p
P m_
Compute UA=UAp as outlined for the FG-2a case. Compute NTU from
Eq. (10.29) and subsequently, P, for the given ow arrangement. Then from
Eq. (10.28),
719
L
Lp
P
Pp
FN-2
FN-3
q
Tmax
1
qp Tmax; p
m_
G
1
m_ p Gp
q
P
NTU
1)
1
Since
qp Pp
NTUp
A
L
Ap Lp
nt
A
fr 1
nt; p Afr; p
Then
Up
L
U
Lp
P
f L
P p fp L p
continued over
All formulas and the procedure for the FN-2 case apply here. After
computing the value of L=Lp , calculate P=P p from the equation
1 R*p
U
Up jp =j R*
Since G=Gp 1 the j and f factors are directly calculable. From Eq.
(10.24), P=P p f =fp L=Lp . From Eq. (10.28), UA=UAp 1 or
L=Lp Up =U. Substitution of these equations and given information
into Eq. (10.26) yields
1 R*p
UA
720
Tmax
Tmax; p
P
Pp
VG-2b
VG-3
L
A
1
Lp Ap
q
1
qp
m_
1
m_ p
P
1
Pp
nt
A
fr
nt; p Afr; p
G
Gp
3
All formulas and the procedure of the FG-2b case apply here. The comment of
case VG-2a is also valid here.
The procedure outlined for the FG-2a case applies here. For this case, the mass ow
rates are equal. This results in dierent nt or frontal area because Gs are dierent.
In case FG-2a, nt or frontal areas are the same and result in dierent mass ow
rates. Otherwise, all formulas and the procedure of the FG-2a case are identical
for this case.
q
Tmax
6 1
known G=Gp . Subsequently, compute P=P p from Eq. (10.24) for A=Ap 1:
Gp
m_
1
m_ p
Tmax
1
Tmax; p
P
1
Pp
A
1
Ap
L
A
1
Lp Ap
m_
1
m_ p
Substitute Ap =A from Eq. (6) into Eq. (5). The resultant equation has only one
unknown G=Gp since j=jp and f=fp are dependent on the value of G=Gp . Solve
for G=Gp from this implicit equation. Then A=Ap is known from Eq. (6).
Subsequently, compute nt =nt; p (or Afr =Afr; p and L=Lp from Eq. (10.30).
P
f A
1
Pp
fp Ap
1 R*p
UA
1
UAp
jp =j f =fp Ap =A2 1=3 R*Ap =A
When appropriate, it is assumed that the correlation for j vs. Re (and hence G) for the enhanced and plain surfaces have the same form [i.e., j=jp G=Gp n
where n is an exponent]. Similarly, it is assumed that f =fp G=Gp m , where m is some other exponent, and that R* and R*p are known.
q
qp
L
Lp
Hence, P=Pp 1
Afr
Afr; p
P
q
Tmax
1
P p qp Tmax; p
m_
1
m_ p
nt
nt; p
VG-2a
VG-1
Case
721
Thus, knowing the ratio UA=UAp from Eq. (10.26), the NTU of the enhanced
surface can be calculated using Eq. (10.29) for a given ow rate, and subsequently,
the temperature eectiveness for the given exchanger ow arrangement can be found
from the formulas of Table 3.6. The heat transfer rate is then computed from Eq.
(10.28).
Using Eq. (10.21) and m_ Ao G, the ratios of surface areas and ow rates for
enhanced and plain tubular surface for the same di are given by
A
n L
t
Ap nt; p Lp
n G
m_
t
m_ p nt; p Gp
10:30
Using the foregoing equations, specic algebraic formulas for the PECs of Table 10.4
are summarized in Table 10.6. It should be emphasized that the formulas of Table 10.6
can also be used to compare two dierent surfaces 1 and 2 for the specic PEC by adding
the subscript 1 to the enhanced surface values and replacing the subscript p by 2 as a base
surface. Also, the values R* and R*p in Eq. (10.26) are normalized with respect to the plain
tube inside thermal resistance. They can be normalized consistently with respect to any
uid-side thermal resistance.
Example 10.4 Would a selection of the following heat transfer surface provide better
performance than the existing design in which a given plain plate-n surface (i.e., surface
11.1, Table 10-3, Fig. 10-26, Kays and London, 1998) is used? The comparison is based
on the requirements of xed (1) ow rate, (2) uid pumping power, (3) heat transfer rate,
and (4) inlet temperature dierence between the hot- and cold-uid streams. The argument is supposed to be valid for any Reynolds number. The new heat transfer surface has
the following Colburn and Fanning friction factors:
j expa0 a1 r a2 r2 a3 r3 a4 r4 a5 r5 a6 r6
and
f expb0 b1 r b2 r2 b3 r3 b4 r4 b5 r5 b6 r6
where r ln Re and the numerical values of the coecients are as follows:
Coecient
Numerical Value
Coecient
Numerical Value
a0
a1
a2
a3
a4
a5
a6
0.1624564980E 04
0.1404062382E 04
0.4999486289E 03
0.9400171748E 02
0.9835078386E 01
0.5428407378E 00
0.1235104592E 01
b0
b1
b2
b3
b4
b5
b6
0.1242054696E 04
0.1077301312E 04
0.3855707180E 03
0.7291091700E 02
0.7674456065E 01
0.4263025064E 00
0.9766547339E 02
SOLUTION
Problem Data and Schematic: Two surfaces should be compared under the xed ow
rate, pumping power, heat transfer rate, and inlet temperature dierence between the
hot- and cold-uid streams. The entire range of applicable Re numbers should be
considered. The correlations for Colburn and Fanning friction factors for a new heat
722
transfer surface are provided in the problem formulation. The Colburn and Fanning
factors for the plain plate-n geometry 11.1 are given in Kays and London (1998) in
tabular form. A curve tting of these data (in the same form as the ones given in the
problem formulation) leads to the following set of coecients:
Coecient
Numerical Value
Coecient
Numerical Value
a0; p
a1; p
a2; p
a3; p
a4; p
a5; p
a6; p
0:1305722226E 05
0:1014346095E 05
0:3268456896E 04
0:5590860182E 03
0:5355816661E 02
0:2724931767E 01
0:5753732235E 01
b0; p
b1; p
b2; p
b3; p
b4; p
b5; p
b6; p
0:6188627536E 04
0:4958849985E 04
0:1647882636E 04
0:2906965749E 03
0:2872245480E 02
0:1507451536E 01
0:3283617030E 01
Determine: Determine which of the two heat transfer surfaces has better heat transfer/pressure drop performance.
Assumptions: The assumptions invoked in Section 3.2.1 are valid.
Analysis: We are given the information on only one uid side of the exchanger in
this problem. The problem formulation requires that the following relations be
satised.
P
hA
Tmax
m_
1
m_ p P p hAp Tmax; p
where the symbols without a subscript denote the variables for the new surface and
the symbols with the subscript p denote the plain plate-n surface (surface 11.1 from
Kays and London, 1998). This is the VG-1 PEC of Table 10.4 on one uid side of
the exchanger. According to Eq. (10.24) or Eq. (6) of Table 10.6, the pumping power
ratio (which is equal to unity, as indicated above) must be equal to
P
f A G 3
1
Pp
fp Ap Gp
2=3
Ap
j=jp
1=3
A
f =fp
From Eqs. (1) and (2), we can eliminate the surface area ratio, A=Ap . Hence, the
ratio of mass velocities for the two surfaces must satisfy the relation
G
Gp
fp j
f jp
1=2
723
Rep
fp Rep j Re
f Re jp Rep
1=2
3
This relation indicates that we can determine Re for which the new surface would
satisfy the aforementioned constraints for any given Rep for ow through the plain
plate-n surface.y The pair of Re numbers would then allow calculation of the
corresponding Colburn and Fanning factors, and subsequently, the evaluation of
heat transfer surface area reduction (or increase, if the new surface is not better
than the old one) that would be accomplished with the proposed change of geometry. Therefore, we should determine these pairs of Re numbers, say for a low Re
number range (say, around 500) and for a large Re number range (say, around
5000), and compare the heat transfer area changes. It should be noted that the
determination of these Re numbers may be a tedious iterative job, in particular if
one uses the graphical presentations of j and f factors as those given in Kays and
London (1998). In our case, we perform the calculations numerically. This calculation leads to the following results. For Rep 500, a very little change in Re would
occur (i.e., Re 505), but A=Ap 0:93. That means that the new surface would
require only 7% less heat transfer area for the same performance as the old one.
For Rep 5000, however, Re becomes 4870 and A=Ap 1:13! The proposed new
surface would require not less but 13% more heat transfer surface than the old one.
So it can be concluded that the proposed solution does not bring a signicant benet
for the low-Re range and may actually be worse for the high-Re range.
Discussion and Comments: The performance evaluation criteria may provide a useful
tool for assessing the performance of a selected heat transfer surface as well as for
the comparison of various solutions. In this case, the claim that a new design is
better is not substantiated. Our conclusion is based on the use of a variable-geometry criterion (VG-1; see Table 10.6). It should be pointed out, though, that highly
augmented heat transfer surfaces may reduce the required heat transfer area signicantly (50% and more) compared with plain plate-n heat transfer surfaces for the
VG-1 criterion.
10.3.3
All performance evaluation criteria discussed so far are based exclusively on the rst
law of thermodynamics. These criteria were devised utilizing mass and energy balances without involving the thermodynamic quality of the energy ows. However,
heat transfer and friction characteristics of heat transfer surfaces may easily be
related to the quality level of energy ows dened by the second law of thermodynamics.z That becomes very important in any system analysis, and feedback from
{
For a given Rep , jp , and fp are determined from the given surface data. Now one assumes a value of Re,
determines corresponding j and f factors and computes Re from Eq. (3). If this Re does not match with the
assumed Re, iterations are carried out with new values of Re, until Eq. (3) is satised, and corresponding j and
f factors are computed to be used subsequently in Eq. (2).
{
The body of knowledge usually called the second law of thermodynamics analysis always involves both the rst
and second laws of thermodynamics. However, it is customary to name the product of such an analysis by
indicating the second law of thermodynamics only (Bejan, 1988).
724
a system engineer may indicate a need for a change of design based on the second
law of thermodynamics. This approach is considered in Section 11.7, with an additional performance evaluation criterion.
10.3.4
The cost of an exchanger is usually an important selection criterion for a user. Let
us assume that all other pertinent evaluation criteria, if not already incorporated in
the cost evaluation/optimization routine, are satised. In that case, existing design
options will be selected based on the most cost-eective design. So a methodology
for cost estimation must be developed. Most heat exchanger manufacturers have
their own proprietary methods for cost estimation. Some approaches to this problem
have been reported in the literature in the past. We present here a simple procedure
of the ESDU (1994) used in some industries.
Heat exchanger cost may be related either to the heat transfer surface area of an
exchanger or to the heat exchanger duty required. In Fig. 10.13, this dilemma is presented
schematically as two options that may be perceived from the perspective of a heat
exchanger designer on one side and of a process engineer on the other. The simple
logic implied by Fig. 10.13 has been used to dene cost estimation evaluation.
The proposed methodology is based on empirical cost data compiled to evaluate all
the various feasible heat exchanger types. The decision variable is the cost of a heat
exchanger per unit of its thermal size, that is, per unit of the product UA q=Tm .
In Fig. 10.13, this quantity is denoted as CUA . An alternative solution would be the cost
of an exchanger per unit heat transfer surface area, CA of Fig. 10.13. The latter is less
attractive because it does not explicitly take into account the heat transfer duty, or what
is equivalent, the relevant thermal size of the exchanger, UA. It must be clear that the
overall heat transfer coecient, determined for a particular design, must inuence the
cost.
In Table 10.7, a selection of the cost data, represented by the values of CUA , is
compiled. This table is prepared for a particular exchanger purpose, namely, for an
application in which the heat exchange is accomplished between gas as a hot uid at
FIGURE 10.13 Cost of a heat exchanger vs. heat transfer area and/or heat duty.
725
TABLE 10.7 Cost Data CUA vs UA for Various Heat Exchanger Typesa
CUA [$/(W/K)]
q=Tm
or UA
(W/K)
Shell- andTube,
U 484
(W/m2 K
Double
Tube,
U 484
(W/m2 K
Printed
Circuit
U 1621
(W/m2 K
Plate-Fin,
U 491
(W/m2 K
Welded Plate
U
(W/m2 K
CUA
[$/(W K)]
103
5 103
3 104
105
106
3.98
1.00
0.29
0.17
0.106
2.5
0.75
0.31
0.31
0.31
12
2.4
0.6
0.42
0.28
3.1
0.513
0.210
0.115
349
1187
1068
1112
1173
4.9
1.22
0.42
0.28
0.22
medium pressure (say, 20 bar) and cold uid as treated water. Five dierent heat exchanger types may be used for this particular combination of working uids (see Chapters 1
and 2 for a description of each type and the assessment of the feasibility of the possible
selections): (1) shell-and-tube heat exchanger, (2) double-pipe heat exchanger, (3)
printed-circuit heat exchanger, (4) plate-n exchanger, and (5) welded plate exchanger.
Depending on the magnitude of q=Tm , dierent cost CUA values can be determined for
each of the heat exchanger types. An extensive set of CUA data for various heat exchangers is compiled by ESDU (1994), and partially summarized in Appendix D; they can be
used for this purpose. From Table 10.7, it is clear that the cost of a heat exchanger per
unit of its thermal size (i.e., per unit of UA), CUA , decreases with an increase in the heat
load q=Tm or heat exchanger size (UA).
The procedure for evaluation of a heat exchanger type based on the given cost
criterion is as follows:
1. Estimate the heat duty q from a heat balance using Eq. (2.1).
2. Determine q=Tm for the heat exchanger under consideration (a) by computing
Tm F Tlm with Tlm from Eq. (3.172) and the best estimate of F for (ESDU,
1994), or (b) from a known NTU and Cmin using q=Tm Cmin NTU [see Eqs.
(3.12) and (3.59)].
3. Repeat step 2 for each heat exchanger type.
4. From empirical data CUA vs. q=Tm , see Appendix D), estimate the CUA factor.
5. Calculate the cost of a particular heat exchanger type by multiplying CUA and
q=Tm :
6. Compare the costs for various heat exchanger types. If one of the types is much less
expensive than the other (by a factor of 1.5 to 2.0 or more), that design should be
selected. If the costs for all solutions are close to each other, a more detailed
analysis of each individual cost must be performed.
The procedure outlined above is utilized in a simplied way in Example 2.4. A more
elaborate analysis is the subject of Problem 10.8.
726
SUMMARY
Heat exchangers are designed for a variety of applications under varied operating
conditions. As a result, an optimum heat exchanger will be dierent depending on
the application. The most important selection criteria for a heat exchanger are
summarized next.
. The heat exchanger must function as designed for performance, durability, and
other criteria during its design life. As a result, the operating environment (i.e.,
pressure, temperature, fouling potential, uid leakage and contamination, material
compatibility, etc.), cost packaging, maintenance, and so on, are very important
variables. Based on these operating and design conditions, an engineer can select an
appropriate exchanger from Table 10.1 with additional considerations of the cost,
manufacturability, and other requirements.
. A large number of geometric variables are associated with shell-and-tube exchangers. Considerable discussion is provided in Section 10.2.1 for the choice of specic
geometrical variables. Similarly, for extended surface exchangers, a variety of n
geometries is available, and a qualitative discussion on the selection of particular
geometries is presented in Section 10.2.3. Surface selection for plate heat
exchangers and regenerators is discussed briey in Sections 10.2.2 and 10.2.4.
In many applications, the heat exchanger operates in a system or a thermodynamic
cycle. Therefore, quantitative criteria for component design and optimization have less
meaning since the heat exchanger should be designed for optimum system performance.
Hence, quantitative methods are presented in the text for screening various surfaces
to select the most appropriate ones as components. In this regard, two categories of
quantitative methods are summarized:
. The surface ow area and core volume goodness factor comparisons are presented
to screen and arrive at higher-performing extended surfaces. Geometrical scaling
laws are then summarized for a compact heat exchanger surface on one uid side
for changes in ow area, volume, surface area, and length of that surface for the
case of constant q, p, and m_ .
. Performance evaluation criteria are employed to compare the performance of an
enhanced tubular to a plain tubular exchanger surface.
REFERENCES
ALPEMA, 2000, The Standards of the Brazed Aluminum Plate-Fin heat Exchanger Manufacturer
Association (ALPEMA), 2nd Edition, AEA Technology plc, Didcot, Oxon, UK.
ASME, 1998, ASME Boiler and Pressure Vessel Code, 1998: Rules for Construction of Pressure
Vessels, Sec. VIII, Div. 1, American Society of Mechanical Engineers, New York.
Bejan, A., 1988, Thermodynamic Design in Advanced Engineering Thermodynamics, Wiley, New
York, pp. 594669.
Bell, K. J., 1981, Construction features of shell-and-tube heat exchangers, in Heat Exchangers:
Thermal-Hydraulic Fundamentals and Design, edited by S. Kakac, A. E. Bergles and F.
Mayinger, Hemisphere Publishing Corp., Washington, DC, pp. 721763.
Bell, K. J., 1998, Approximate sizing of shell-and-tube heat exchangers, in Heat Exchanger Design
Handbook, G. F. Hewitt, exec. ed., Begell House, New York, Vol. 3, Sec. 3.1.4.
Cowell, T. A., 1990, A general method for the comparison of compact heat transfer surfaces, ASME
J. Heat Transfer, Vol. 112, pp. 288294.
REVIEW QUESTIONS
727
ESDU, 1994, Selection and costing of heat exchangers, Engineering Science Data, Item 92013,
ESDU, Int., London, UK.
Gentry, C. G., 1990, RODbae heat exchanger technology, Chem. Eng. Prog. July, pp. 4857.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger
Publishing, Malabar, FL.
Lancaster, J. F., 1998, Materials of construction, in Handbook of Heat Exchanger Design, G. F.
Hewitt, ed., Begell House, New York, Sec. 4.5.
London, A. L., 1964, Compact heat exchangers, Part 2, Surface geometry, Mech. Eng., Vol. 86,
June, pp. 3134.
London, A. L., and C. K. Ferguson, 1949, Test results of high-performance heat exchanger surfaces
used in aircraft intercoolers and their signicance for gas-turbine regenerator design, Trans.
ASME, Vol. 71, pp. 1726.
Reay, D. A., 1999, Learning from Experiences with Compact Heat Exchangers, CADDET Analyses
Series 25, Centre for the Analysis and Dissemination of Demonstrated Energy Technologies,
Sittard, The Netherlands.
Sekulic, D. P., A. J. Salazar, F. Gao, J. S. Rosen, and H. S. Hutchins, 2003. Local transient behavior
of compact heat exchanger core during brazing, Int. J. Heat Exchangers, Vol. 4, No. 1.
Shah, R. K., 1978, Compact heat exchanger surface selection methods, Heat Transfer 1978, Proc. 6th
Int. Heat Transfer Conf., Vol. 4, pp. 193199.
Shah, R. K., 1983, Compact heat exchanger surface selection optimization and computer-aided
thermal design, in Low Reynolds Number Flow Heat Exchangers, edited by S. Kakac, R. K.
Shah and A. E. Bergles, pp. 845874, Hemisphere Publishing Corp., Washington, DC.
Shah, R. K., 1995, Heat exchangers, in Encyclopedia of Energy Technology and the Environment,
A. Bisio and S. G. Boots, eds. Wiley, New York, Vol. 3, pp. 16511670.
TEMA, 1999, Standard of the Tubular Exchanger Manufacturers Association, 8th ed. Tubular
Exchange Manufacturers Association, New York.
Webb, R. L., 1981, Performance evaluation criteria for use of enhanced heat transfer surfaces in heat
exchanger design, Int. J. Heat Mass Transfer, Vol. 24, pp. 715726.
Webb, R. L., 1994, Principles of Enhanced Heat Transfer, Wiley, New York.
Webb, R. L., and A. E. Bergles, 1983, Performance evaluation criteria for selection of heat transfer
surface geometries used in low Reynolds number heat exchangers, in Low Reynolds Number
Flow Heat Exchangers, S. Kakac, R. K. Shah, and A. E. Bergles, eds., Hemisphere Publishing,
Washington, DC, pp. 735752.
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
10.1
10.2
10.3
10.4
728
10.5
To cool concentrated hydrochloric acid with a heavy oil (low h), the following
exchanger(s) would be feasible:
(a) plate exchanger with titanium plates
(b) plate-n exchanger with ns on oil side
(c) TEMA AEW with oil in the shell
(d) TEMA CEN with oil in the shell.
10.6
10.7
10.8
10.9
In general, all shell-and-tube exchangers have transverse plate baes except for
the following TEMA shell:
(a) G
(b) J
(c) K
(d) X
10.10
The horizontal bae cut is used for the following shell-side uids:
(a) single-phase uids
(b) condensing uids
(c) evaporating uids
(d) very viscous liquids
(e) slurries
10.11
10.12
REVIEW QUESTIONS
729
(b) reduced number of baes in the exchanger compared to an equivalent segmental bae exchanger
(c) increased rigidity of the tube bundle
(d) reduced shell-side fouling
(e) reduced shell-side pressure drop
10.13
10.14
The following shell types are commonly used for single-phase uids on the shell
side:
(a) E
(b) F
(c) G
(d) H
(e) J
(f ) K
(g) X
10.15
The following shell types are commonly used for two-phase or multiphase uids
on the shell side:
(a) E
(b) F
(c) G
(d) H
(e) J
(f ) K
(g) X
10.16
In a single tubesheet design, the least likelihood of leakage between the shell and
tube uids is with the following front-end heads:
(a) A
(b) B
(c) C
(d) N
10.17
The most important criteria for the selection of rear-end heads are:
(a) to control the uid velocities
(b) operating pressures
(c) thermal stresses between tubes and shell
(d) shell- or tube-side cleaning requirement
(e) high shell-side heat transfer coecient
(f) cant tell
10.18
The bundle-to-shell bypass stream C could be signicant in the following rearend head constructions, and as a result, sealing strips are usually required:
(a) L
(b) M
(c) N
(d) P
(e) S
(f ) T
(g) U
(h) W
10.19
10.20
Arrange the following shell types from high to low p on the shell side for a
liquid at a specied ow rate (assume turbulent ow) and inlet temperature, for
the same total surface area, and for single-segmental baes/support plates at the
same spacing:
(a) E
(b) F
(c) G
(d) H
(e) J
(f ) X
730
10.21
Arrange the following rear-end heads in order of the most easy to the most
dicult for cleaning and inspection of the shell side:
(a) M
(b) P
(c) S
(d) T
(e) U
(f ) W
10.22
Arrange the following rear-end heads from the least-cost to highest-cost designs:
(a) P
(b) S
(c) T
(d) U
(e) W
10.23
10.24
Circle the following statements as true or false for preferred tube layout when
shell-side mechanical cleaning is required.
(a) T F
A 608 tube layout is used for shell-side laminar ow.
(b) T F
A 458 tube layout is used for shell-side turbulent ow.
(c) T F
A 458 tube layout is used for condensing uid on the shell side.
(d) T F
A 608 tube layout is used for boiling uid on the shell side.
10.25
10.26
REVIEW QUESTIONS
731
10.27
For a specied heat transfer and pressure drop requirement, strip ns require the
following compared to plain ns:
(a) larger frontal area
(b) larger frontal area and core volume
(c) shorter ow length but larger core volume
10.28
In a plain n compact heat exchanger, the pressure drop on the side of concern is
500 Pa. If the ow is developing laminar, what is the approximate p if the ow
length is doubled?
(a) 1000 Pa
(b) 2000 Pa
(c) 700 Pa
(d) cant tell
10.29
10.30
Rotary regenerators:
(a) employ interrupted surfaces
(b) have surface area densities greater than 400 m2 /m3
(c) are used for moderately fouling gases
(d) are more expensive per unit surface area compared to plate-n and tube-n
surfaces
10.31
10.32
If the j and f characteristics for surface P are both 20% higher than those of
surface Q, which surface would you select for your heat exchanger? Why?
(a) surface P
(b) surface Q
(c) cant tell
10.33
Usually, the maximum operating conditions for design of a metal plate-n unit is
1000 kPa or 8008C due to:
(a) joining techniques between the ns and plates
(b) manufacturing technology limitations
(c) cost factors
10.34
Which compact heat exchanger has the highest heat transfer surface area density
and the lowest cost per unit surface area?
(a) shell-and-tube
(b) plate
(c) plate-n
(d) rotary regenerator
10.35
10.36
In a compact heat exchanger having fully developed laminar ow, the following
relationships exist when we compare two surfaces. Consider q, m_ , and p as
given and constant for these comparisons in parts (a), (c), (d), and (f). Circle the
following statements as true or false.
(a) T F
The ow length is inversely proportional to the hydraulic diameter
and directly proportional to the j factor.
(b) T F
The ow areas for two surfaces are the same if the L=Dh ratio and
f factors are the same for the given uid, ow rate, and pressure
drop.
732
(c) T F
(d) T F
(e) T F
(f ) T F
PROBLEMS
10.1
10.2
You have designed two heat exchangers to do the same thermal job [i.e., they
have the same q and total tube-side mass ow rate m_ t (kg/s)]. A comparison of
these heat exchangers shows the following:
Property
Heat transfer area A m
Number of tubes Nt
2
Smooth Tube
Rough Tube
13.62
100
9.80
120
PROBLEMS
2.5
18
0.008
733
1.25
18
0.016
Assume constant and identical uid properties for both heat exchangers.
(a) Calculate P=P s , where the subscript s is for the smooth tube.
(b) Calculate p=ps .
(c) Suppose that P=P s is 0.8; how would you alter the rough tube exchanger
geometry to make P=P s 1, keeping the same rough tube length? Give
quantitative estimates considering no change in the friction factor.
(d) How would you expect UA to change for part (c)? Give explicit reasons
whether it will increase, decrease, or will have no change.
10.3
An aircraft oil cooler has oset strip ns on the air side with 790 ns/m, a 0.15 mm
n thickness, a 0.25 mm plate thickness, and a 9.5 mm plate spacing. Air enters
the heat exchanger with 5 m/s frontal velocity 1:579 105 m2 =s,
1:20 kg=m3 . Assume that you have an option of using 3- and 12-mm strip
lengths for the n. Compare the performance of these two n geometries.
(a) Calculate Dh , *, and . Hint: Draw a unit cell and use associated simple
geometrical relationship.
(b) Compute j and f for each surface. Hint: Use Manglik and Bergles correlations, Eqs. (7.124) and (7.125).
(c) Determine h2 =h1 , E2 =E1 , and j=f 2 = j=f 1 . Here the subscript 1 refers to the
12 mm strip length.
(d) For both strip ns, compare the values of E and P for the same Afr and the
same hA. What would be the ratio of surface areas, A2 =A1 ?
(e) What would be the approximate value of u1 for the 3 mm strip n exchanger
for the same hA and same P? Assume that fold fnew for the 3 mm strip n.
Hint: Use the functional relationship of P from Eq. (10.8). Knowing the new
value of u1 , compare it with the old value of u1 and estimate the new value of
h and subsequently the new value of A for the same hA.
(f) Discuss the results of parts (d) and (e).
10.4
As a heat exchanger designer, you have designed a plain plate-n heat exchanger
to transfer the heat specied within the pressure drop allowed. The design point
for the plain plate-n corresponds to Re 1000. You have an option to employ
an oset strip n of the same hydraulic diameter as an alternative. These surfaces
are shown in Fig. 10.3.
(a) Determine the design Reynolds number for the oset strip n for equal uid
pumping power per unit surface area (equal E).
(b) Determine h2 =h1 for equal E, where subscripts 1 and 2 designate plain and
oset strip ns, respectively. Hint: Use the denition of j and appropriate data.
(c) Based on the answer for part (b), how much reduction is achieved in the core
volume on the one uid side under consideration using the oset strip n?
(d) Determine the area goodness factor ratio for these surfaces. Can you tell which
surface will require a higher frontal area? Why?
734
10.5
10.6
10.7
Multilouver Fin
800
1.25
6
4.8
0.075
30
24
9.5
3.5
24
Compute the change in the heat transfer surface and the air pumping power
requirements by replacing the louver n with the oset strip n under the xed
ow area (FN-2) performance evaluation criterion. Assume the mean air velocity
through both n geometries to be 10 m/s. Use the appropriate correlations for the j
and f factors from Chapter 7. Assume that the thermal resistances of the wall and
the coolant are zero. Ignore fouling on both uid sides. Use the following properties of air: 1:058 kg=m3 ; cp 1:008 kJ=kg K, k 0:0288 W=m K, and
20:4 106 Pa s.
10.8
A high viscosity liquid with cp 1:9 kJ=kg K and mass ow rate 0.6 kg/s enters a
heat exchanger at 758C having an inlet pressure of 3.1 MPa. This liquid is cooled
to 358C by water having an inlet temperature of 188C and a mass ow rate of
2.1 kg/s. No signicant fouling should be expected. Select the most feasible heat
exchanger type using a cost estimate.
11
The main objectives of this chapter are twofold: (1) to present and discuss important
factors that aect heat exchanger performance, and (2) to introduce a basic analysis
for the thermodynamic design and optimization of heat exchangers. A quest for
answers regarding the rst objective will help us to identify the important factors
that aect heat exchanger eectiveness, to quantify the eects of these factors, and
to provide guidelines for a qualitative assessment of the eectivenesses of the exchangers with dierent ow arrangements but with a given, identical design task. The
second objective is to dene a gure of merit for assessing the thermodynamic
eciency of a heat exchanger and to present an approach to thermoeconomic considerations.
In Section 11.1, the dierences between a heat exchanger as a component and as
part of a system are identied. In Section 11.2, a detailed modeling of a heat
exchanger using energy balances only (i.e., the rst law of thermodynamics) is provided for the determination of heat exchanger eectiveness and temperature distributions. In Section 11.3, a combined approach based on both the rst and second laws
of thermodynamics is introduced to quantify inherent irreversibilities in a heat
exchanger. The most important source of irreversibility is heat transfer across the
nite temperature dierences, which is discussed rst. Fluid mixing and uid friction,
as additional sources of irreversibility, are studied next. A temperature cross phenomenon is then discussed in detail in Section 11.4 by evaluating entropy generation
in a 12 TEMA J shell-and-tube heat exchanger. Using all the analysis tools presented in the rst four sections, a heuristic approach to an assessment of heat
exchanger eectiveness is developed in Section 11.5. In Section 11.6, energy, exergy,
and cost balances important for analysis and optimization of heat exchangers are
presented. Finally, a thermodynamic criterion for evaluation/selection of heat transfer surfaces is summarized in Section 11.7.
11.1 INTRODUCTION
Traditionally, modeling of a heat exchanger is based on energy balances (i.e., on the
consequences of both the rst law of thermodynamics and the mass conservation
principle), so only the concepts of heat transfer rate and enthalpy rate change would
Fundamentals of Heat Exchanger Design. Ramesh K. Shah and Duan P. Sekulic
Copyright 2003 John Wiley & Sons, Inc.
735
736
suce for such an analysis.{ For an adiabatic heat exchanger (see the assumptions in
Section 3.2.1), the enthalpy rate change of one uid stream must be equal to the enthalpy
rate change of the other, being at the same time equal to the exchanger heat transfer rate.
This simple energy balance statement will be used in subsequent sections in the dierential form to model spatial distributions of temperatures of both uid streams. In
Chapter 3, we also used the energy balances (both dierential and overall), but only to
determine the heat exchanger eectiveness without determining the temperature distributions.
Let us start with a review of the analysis of heat exchanger design methods discussed
in Chapter 3 that relies on a relationship that can be presented in generalized form as
follows:
heat transfer
rate q
1 0
1
effectiveness=
heat capacity
temperature
@
A
@
A
correction
rate or thermal
difference
factor
conductance
8
"Cmin Tmax in "-NTU method
>
>
>
>
>
< P1 C1 Tmax in P-NTU method
11:1
>
FUA Tlm
in LMTD method
>
>
>
>
:
UA T max in -NTU method
In our analysis, the enthalpy rate change is the rate change of enthalpy of a uid stream caused by heat transfer
interaction between the two uid streams. From the thermodynamic point of view, note that the heat transfer rate
is an energy interaction (not a change in the uid property), while the enthalpy rate change represents a change of
uid property caused by existing heat interaction. This distinction is important for understanding of the interpretation of heat exchanger performance and will be emphasized in more detail later through the introduction of
several advanced concepts of thermodynamics.
INTRODUCTION
737
eectiveness relationships are already known or can be determined using the matrix
formalism mentioned in Section 3.11.4), an engineer needs only a relationship between
the eectiveness/correction factor and design parameters, without detailed insight into
local temperature distributions. Thus, a misleading conclusion may be reached: that the
only information a designer should possess concerning a ow arrangement is the relationship between an eectiveness factor and design parameters (e.g., P1 -NTU1, "-NTU,
or F-P relationship).
Analysis presented so far does suce for a design procedure for a heat exchanger with
an already dened eectiveness relationship. However, a very important and still unanswered question should be addressed as well. Why does an eectiveness factor (say,
heat exchanger eectiveness) have a high (or low) value for a given ow arrangement
(especially for a complex one) compared to the corresponding value for another ow
arrangement (for the same set of design parameters)? For example, we do know that a
crossow heat exchanger has less exchanger eectiveness than for a counterow exchanger (for the same set of design parameters NTU and C*). The only rational explanation
that we can oer at this point (in addition to the intuitive ones) is that the eectiveness
relationship for a crossow exchanger simply provides a smaller numerical value for " or
P for the given heat capacity rate ratio and NTU than does " or P for a counterow
exchanger. In addition, "/NTU for a xed heat capacity rate ratio C* is dierent for
counterow than for crossow, for NTUmin < NTU < 1. For NTU 4, this gradient is
almost identical. For NTUmin 0:4, all ow arrangements provide almost identical
eectiveness values for a given set of design parameters [see Eq. (3.89)]. Why is that
so? The reasoning will become clear when we present an astonishingly simple heuristic
approach based on the second law analysis of exchanger ow arrangements. Also, we
present a thermodynamic performance gure of merit, the eciency of a heat exchanger
from a system viewpoint. Consequently, these analysis tools will help us in assessing
relative magnitudes of exchanger eectiveness for complex ow arrangements for the
selection of an appropriate ow arrangement for a specied task. This understanding will
also become valuable in nding an optimum heat exchanger design from a system viewpoint.
11.1.1
Heat exchangers in numerous engineering applications are only one of many components
of a system. Thus, the design of a heat exchanger is inevitably inuenced by system
requirements and should be based on system optimization rather than component optimization. An objective function for such system-based optimization is inuenced by the
main features of heat exchanger operation. For a given set of input data (e.g., ow rates
and inlet temperatures), exchanger geometry, and other pertinent information, the output data (e.g., the outlet temperatures) will depend on heat transfer and uid ow
phenomena that take place within the boundaries of the heat exchanger. So even though
one seeks a system optimum, in the process of determining that optimum, one must fully
understand the features of the exchanger as a component.
Since heat exchangers are used in many systems, we do not attempt any specic
system analysis or process integration. We discuss only the basic thermodynamic aspects.
Despite exact mathematical/numerical results obtained through system-based optimization, the designer should know that the heat exchanger design (sizing) problem studied is
a complex problem that has no single exact solution at all. In all but trivial cases, a
738
designer must deal with uncertainty margins of the input data, in addition to numerous
assumptions. Usually, a range of data (say, for a cost analysis in an optimization
routine), and not a single set of parameters, must be considered. As shown in Fig. 2.1,
for every case considered in the top left box of the problem specication, one arrives at an
optimum solution at the end of the process of Fig. 2.1. Hence, there can be many (and not
only one) optimum solution for a given exchanger sizing problem as provided by
dierent heat exchanger manufacturers. With that in mind, the reader should understand
the limitations of results obtained by modern computer software for design, optimization, and system integration.
11.1.2
739
determination of the exchanger eciency factor ", P, and F [see Eq. (11.1)], through
various heat exchanger basic design methods; we did not pay any particular attention to
uid temperature distributions and their relationship to exchanger performance. Let us
now consider the distribution of local temperatures and temperature dierences in a heat
exchanger having simple counterow and parallelow arrangements. Both these arrangements correspond to two limiting cases of the same geometrical situation: The two uid
streams are owing in geometrically parallel orientation but in opposite or same directions to each other, thus providing the largest and smallest heat exchanger eectiveness
values (see Figs. 3.7 and 3.8 and Table 3.3). It is assumed, by denition, that both uids
change their respective temperatures only in ow directions (i.e., the local temperature
distribution is uniform for a uid across a ow cross section). Subsequently, we consider
a more complex situation with a cross ow of working uids and the possibility of local
mixing along the ow direction.
11.2.1
11:2
Adiabatic wall
.
mc
Fluid 1
dq
1 p,1
x + dx
m 2 c p, 2
Fluid 2
Adiabatic wall
dq
H1 + dH1
H1
H1 + dH1
H1
H 2 + dH 2
H2
dq
(a)
H2
H 2 + dH2
(b)
(c)
FIGURE 11.1 Energy balance control volumes for a counterow arrangement. The control
volumes are represented by rectangular areas. (a) Control volume for uid 1 dierential energy
balance; (b) control volume for uid 2 dierential energy balance; (c) control volume for both uids
1 and 2 dierential energy balance. Enthalpy rates H_ j m_ cp j Tj , j 1 or 2.
740
where i1 1 or 1, for the same or opposite (positive or negative) direction of uid 1
with respect to the positive direction of the x axis, respectively.
For uid 2 only (the elementary control volume in Fig. 11.1b):
dT
i2 m_ cp 2 T2 i2 m_ cp 2 T2 2 dx UT1 T2 dA 0
dx
11:3
where i2 1 or 1, for the same or opposite (positive or negative) direction of uid 2
with respect to the positive direction of the x axis, respectively.
For both uids 1 and 2 (the elementary control volume in Fig. 11.1c):
dT
dT
i2 m_ cp T m_ cp T
0
i1 m_ cp T m_ cp T
dx
dx
dx
dx
1
2
11:4
Note that an assumption of uniform distribution of the total heat transfer surface area
A along the ow length L means that dA=dx A=L; that is, U dA UA=L dx. Let us,
without restricting the model generality, x the direction of uid 1 to be in a positive axial
direction (i1 1) while i2 1 [i.e., i2 1 for counterow (see Fig. 11.1), or i2 1
for parallelow]). Rearranging Eqs. (11.2)(11.4), we obtain
dT1 UA
T2 T1
dx
L
dT2 UA
i2 m_ cp 2
T1 T2
dx
L
dT1
dT
m_ cp 1
i2 m_ cp 2 2 0
dx
dx
m_ cp 1
11:5
11:6
11:7
Note that only two of the three balance equations are sucient to dene the two
temperature distributions. For example, either Eqs. (11.5) and (11.6) or Eqs. (11.5) and
(11.7) can be utilized. Subsequently, distribution of the temperature dierence along the
heat exchanger can be determined.
To close the problem formulation, a set of boundary conditions is required at the heat
exchanger terminal points. For a parallelow exchanger, the inlet temperatures for both
uids are known at x 0. For the counterow exchanger, the known inlet temperatures
are on the opposite sides of the exchanger, at x 0 and x L, respectively. Explicitly,
these conditions are
T1 T1;i
at x 0
T2 T2;i at
x 0 for parallelflow
x L for counterflow
11:8
T T1;i
T2;i T1;i
x
L
11:9
741
and design parameters NTU1 and R1 , as dened by Eqs. (3.101) and (3.105). Hence,
d1
NTU1 1 2 0
d
11:10
d2
i2 NTU1 R1 1 2 0
d
11:11
d1
d2
i2
0
d
d
11:12
R1
The boundary conditions are as follows:
1 0
2 1
at 0
(
0 for parallelflow
at
1 for counterflow
11:13
11:14
The set of relationships given by Eqs. (11.5)(11.8) or (11.10)(11.14) dene a mathematical model of the heat transfer process under consideration in terms of temperature
distributions for both uids. For example, one can solve Eqs. (11.5), (11.7), and (11.8) to
obtain temperature distributions (as presented in Figs. 1.50 and 1.52). This can be done
for virtually any combination of design parameters (NTU1 and R1 ) for both parallelow
and counterow arrangements without a need for separate mathematical models.
Some mathematical aspects of the solution procedure and thermodynamic interpretation of the results will be addressed in the examples that follow. A rigorous and unied
solution of the parallelow heat exchanger problem dened above is provided in
Example 11.1 (Sekulic, 2000). The relation between the heat exchanger and/or temperature eectiveness as a dimensionless outlet temperature of one of the uids and a thermodynamic interpretation of these gures of merit is emphasized in Example 11.2. An
approach to modeling more complex situations, such as a 12 TEMA J shell-and-tube
heat exchanger or various crossow arrangements, is left for an individual exercise (see
Problems 11.1 through 11.10 at the end of the chapter).
Example 11.1 Determine temperature distributions of two parallel uid streams in
thermal contact. The uid streams have constant mass ow rates and constant but
dierent inlet temperatures. Show that a unied solution procedure can be formulated
for both parallelow and counterow arrangements.
SOLUTION
Problem Data and Schematic: The two uid streams ow in a parallel geometric orientation as presented in Fig. E11.1A. Both parallelow and counterow arrangements are
considered (i.e., uid 2 can be in either one of two opposite directions). Inlet temperatures, mass ow rates, and uid properties are known. All the geometric characteristics of
the ow passages are dened as well.
Determine: The local temperatures of both uids as functions of the axial distance along
the uid ow direction.
742
Fluid 1
Fluid 2
(parallelflow)
i 2 = +1
x
= x/L
Fluid 2
x = L (counterflow)
i2 = 1
=1
Assumptions: It is assumed that thermal interaction between the uids takes place under
the assumptions described in Section 3.2.1
Analysis: Any two of three dierential balances given by Eqs. (11.10)(11.12) together
with the boundary conditions given by Eqs. (11.13) and (11.14) describe the theoretical
model for analysis. Let us dene the model of this heat transfer process using Eqs. (11.10)
and (11.12)(11.14).
A general solution will be obtained utilizing the Laplace transforms method (Sekulic,
2000), although several other methods can be used as well (see Section 3.11). The rationale for using this particular method is that it can be applied eciently to a number of
more complex situations, such as for a crossow arrangement with both uids unmixed
(see Problem 11.2).
Applying Laplace transforms to Eq. (11.10) yields
d1
l
0
NTU1 1 2
d
!s
NTU1
s
s NTU1 2
743
and
1 s 1 0 i2 s
2 s 2 0 0
R1 s
s NTU1
2 0
s2 sNTU1 1 i2 R1
1 s explicitly in terms of s.
Substitute Eq. (6) into Eq. (3) to get
Now, applying inverse Laplace transforms on Eqs. (3) and (6), we get (for R1 6 1 if
i2 1)
NTU1 1i2 R1
1 sg 2 0 1 e
1 l1 f
1 i2 R 1
and
2 l1 f2 sg 2 0
1 i2 R1 eNTU1 1i2 R1
1 i2 R 1
Parameter 2 0 in Eqs. (7) and (8) depends on both design parameters (NTU1, R1 )
and the value of i2 . The value of 2 0 can be determined for the parallelow arrangement (i2 1) directly from the boundary condition at the uid 2 inlet [i.e.,
2 0 2;i 1]. For the counterow arrangement (for i2 1), the value of 2 0
can be obtained by collocating Eq. (8) at the uid 2 inlet (i.e., at 1), and solving for
2 0. Consequently,
2 0
1 R1
1 R1 eNTU1 1R1
for i2 1
1 R1
>
for i2 1 counterflow
:
1 R1 e1R1 NTU1
2 0
10
In Eq. (10), one should rst dene the uid stream direction parameter (i.e., i2 1),
and then select the numerical values for design parameters.
Finally, combining Eqs. (7), (8), and (10), the general solution for temperature distributions for both parallelow and counterow exchangers can be written as follows:
744
1
1 R1
1 eNTU1 1i2 R1
1 i2 R1 1 R1 e1=21i2 NTU1 1R1
11
1 R1
1 i2 R1 eNTU1 1i2 R1
2
1 i2 R1 1 R1 e1=21i2 NTU1 1R1
Inserting i2 1 for parallelow and i2 1 for counterow into Eq. (11), we get the
following temperature distributions:
Flow Arrangement
Flow Indicator i2
1
2
Parallelow
1 eNTU1 1R1
1 R1
1 R1 eNTU1 1R1
1 R1
Counterow
1
1 eNTU1 1R1
1 R1 eNTU1 1R1
1 R1 eNTU1 1R1
1 R1 eNTU1 1R1
The temperature distributions for both parallelow and counterow arrangements and
for several sets of parameters are presented in Fig. E11.1B.
All the results presented so far for i2 1 require that 0 R1 < 1 (i.e., R1 6 1).
In the case of a balanced counterow heat exchanger, R1 1 and i2 1, the
original mathematical model given by the set of equations (11.10) and (11.12) transforms
into
d1 d2
NTU1 2 1
d
d
1.00
R1= 0
1.00
0.60
1
2
R1 = 0
2
R1
0
0.80
0.80
0.6
12
R1 = 0
0.6
0.6
0.6
1.0
0.60
2
0.40
0.40
0.20
1.0
0.20
NTU1 = 2.0
Counterflow
0.00
0.00
0.20
0.40
(a)
0.60
0.80
1.00
NTU1 = 2.0
Parallelflow
0.00
0. 00
0.20
0.40
0.60
0.80
1.00
(b)
FIGURE E11.1B Temperature distributions in (a) counterow and (b) parallelow heat
exchangers.
745
with the boundary conditions as given by Eqs. (11.13) and (11.14) for counterow.
The solution of this problem leads to linear dimensionless temperature distributions.
The temperature distributions for uids 1 and 2 are
1
NTU1
1 NTU1
2
NTU1 1
1 NTU1
13
Discussion and Comments: In this example, it has been shown how to nd temperature
distributions in a heat exchanger with parallel streams (in geometrical sense). This example demonstrates that both parallelow and counterow arrangements represent the two
subproblems of a single heat transfer problem that dier only in the stream direction.
Still, the character of temperature dierence distributions in these two situations is
radically dierent, as shown in Fig. 11.2. Note that this modeling is based on energy
balances performed on two control volumes selected arbitrarily from a total of three
balances (either for one or the other uid, or for both uids). All results of this modeling
are the consequence of conservation principles.
11.2.2
In Section 3.3.1, heat exchanger eectiveness and the maximum possible heat transfer
rate qmax are introduced by denition [see Eqs. (3.37) and (3.42)]. This has provided the
basis for the formulation of heat exchanger eectiveness in terms of terminal temperatures of the uids and their heat capacity rates as in Eq. (3.44). However, that approach
requires a priori denition of a hypothetical innite surface area of the heat exchanger.
On the other hand, by knowing the temperature distributions of a given heat exchanger,
we can devise the concept of heat exchanger eectiveness without invoking the concept of
a hypothetical counterow heat exchanger of innite surface. We can show that the
denition of heat exchanger eectiveness is obtained using the rst law of thermodynamics only (Sekulic, 2000), without invoking explicitly the second law of thermodynamics. The true meaning of heat exchanger eectiveness as a dimensionless outlet
temperature of the uid stream having the smaller heat capacity rate is a direct consequence of this interpretation. Moreover, the maximum possible heat transfer rate
Traditional "
Based on a comparison of Dened utilizing the rst
qmax Cmin Th;i Tc;i
meaning
the actual heat transfer
law of thermodynamics
rate exchanged in the
explicitly, and the
Cc Tc;o Tc;i
"
T1;o T1;i
T2;i T1;i
C1 Cmin
Dimensionless outlet
temperature of a uid
with smaller heat
capacity rate, C1 < C2
746
exchanged in a heat exchanger can subsequently be derived, not postulated. The two
interpretations are summarized in Table 11.1.
In Example 11.2, these important thermodynamic consequences of the analysis of
temperature distributions in a heat exchanger will be illustrated using both parallelow
and counterow heat exchangers as an example. It should be reiterated that the interpretation given is universally valid, regardless of the complexity of the ow arrangement
involved.
Example 11.2 Show that heat exchanger eectiveness and/or temperature eectiveness
represent the nondimensional outlet temperature of one of the two uid streams of a
heat exchanger. Use the parallelow and counterow arrangements as examples.
Demonstrate that heat exchanger eectiveness can be interpreted as a ratio of actual
heat transfer rate to the heat transfer rate of a hypothetical exchanger with an innitely
large thermal size (NTU ! 1), as emphasized in Chapter 3, but without invoking
explicitly the second law of thermodynamics.
SOLUTION
Problem Data and Schematic: As presented in Example 11.1.
Determine: Demonstrate the equivalence between the heat exchanger eectiveness denition and the dimensionless outlet temperature of one of the uids in a heat exchanger.
Assumptions: As invoked in Example 11.1.
Analysis: The outlet temperature of uid 1 for parallelow and counterow can be
obtained from Eq. (11) of Example 11.1 at 1 (the outlet of uid 1).
1 1
1 R1
1 eNTU1 1i2 R1
1 i2 R1 1 R1 e1=21i2 NTU1 1R1
1 eNTU1 1R1
1 R1
1 eNTU1 1R1
1 R1 eNTU1 1R1
Equations (2) and (3) are identical to Eqs. (I.2.1) and (I.1.1) of Table 3.6. So the dimensionless outlet temperatures are equal to parallelow ("pf ) and counterow ("cf ) heat
exchanger eectivenesses, respectively. We have obtained expressions for heat exchanger
eectiveness without invoking the concept of an ideal heat exchanger. Hence, the true
meaning of the eectiveness is simply the dimensionless outlet temperature of the uid
with the smaller heat capacity rate (note that R1 C* in this case). This conclusion is
general and valid for any ow arrangement.
747
lim
NTU1 !1
1 1
lim
NTU1 !1
1 eNTU1 1R1
10
1
NTU
1R
1
1
10
1 R1 e
Invoking the denition of the dimensionless temperature from Eq. (11.9), the following
result can be obtained from Eq. (4):
lim T1;o T2;i
NTU1 !1
NTU!1
lim m_ cp 1 T1;o T1;i m_ cp 1 T2;i T1;i qmax
NTU!1
The actual heat transfer rate in a two-uid single-phase heat exchanger of any ow
arrangement is as follows:
q m_ cp 1 T1;o T1;i
Finally, dividing the right-hand side of Eq. (7) with qmax from Eq. (6), and comparing
the result with the denition of the outlet temperature of uid 1, one can obtain
T1;o T1;i
q
"
1 1 P1
T2;i T1;i
qmax
This constitutes the proof required. Note that for a counterow arrangement with
R1 C * 1, the heat exchanger eectiveness becomes " 1 1 NTU=1 NTU;
see Example 11.1 for the corresponding temperature distributions.
Discussion and Comments: The heat exchanger/temperature eectiveness has its true
meaning as a dimensionless outlet temperature of the uid with the smaller heat capacity
rate (heat exchanger eectiveness, ") or the dimensionless outlet temperature of a given
uid (temperature eectiveness, say P1 for uid 1). The numerical value of the heat
exchanger eectiveness is between 0 and 1 and indicates how close the outlet temperature
of one uid can approach the inlet temperature of the other uid. The traditional meaning of the heat exchanger eectiveness [although a thermodynamic interpretation based
on Eq. (8) is perfectly valid and insightful] involves the concept of a hypothetical
innitely large counterow heat exchanger. Refer to Sekulic (2000) for a detailed
discussion concerning an analysis of this approach; discussion of a concept of the
thermodynamic eciency for a heat exchanger is given in Section 11.6.5.
748
1 R1 expNTU1 1 i2 R1
1 R1 exp1=2NTU1 1 R1 1 i2
11:15
Note that Eq. (11.15) is valid for both parallelow (i2 1) and counterow
(i2 1) arrangements for 0 R1 < 1. In Fig. 11.2, a graphical representation of Eq.
(11.15) is given for both counterow and parallelow arrangements and for several
1.0
Parallelflow
NTU1
0.4
Counterflow
0.8
NTU1
0.4
0.8
1.2
0.6
0.8
1.2
0.4
0. 2
5.0
5.0
0.0
0.0
0.2
0.6
0.4
0.8
1.0
The terms parallelow and counterow are used throughout the book following the practice by Kays and
London (1998), which has been extensively used throughout the world for almost last ve decades. As demonstrated in this section using a straightforward rst law analysis, a more favorable terminology from a semantic
point of view would be unidirectional and bidirectional ows (i.e., in both arrangements, uid streams are owing
parallel in a geometric sense but oriented in the same or opposite directions).
749
values of NTU1 and R1 0:6. Refer to Sekulic (2000) for more detailed data. A study of
these temperature dierence distributions reveals the following conclusions.
. As we have learned in Chapter 3, counterow is the best and parallelow the worst
ow arrangement from the eectiveness viewpoint for given NTU and C* (or
NTU1 and R1 ). For the same heat capacity rate ratio (say, R1 0:6), and in
particular at large NTU1 values, the magnitude of temperature dierence change
along the ow direction () is substantially larger for parallelow than it is for
counterow (compare corresponding curves in Fig. 11.2). The two temperature
distributions are identical for R1 0 (not shown in Fig. 11.2) and dier more at
large heat capacity rate ratios.
. For a given NTU for counterow, features a more pronounced variation along
as R1 (or C*) decreases from 1 to 0. This means that there will be larger local
temperature dierences with decreasing R1 (or C*) if all other parameters remain
the same. If the temperature dierence distributions are close to each other or
identical, they will have similar or identical eectiveness.
. If the temperature dierence distributions dier substantially, the heat exchanger
eectiveness will dier considerably as well.
In the limiting case of R1 1, the temperature dierence distribution for counterow
arrangement is uniform throughout the exchanger, 2 1 1=1 NTU1 , as
derived in Example 11.1. For the same heat capacity rate ratio in the parallelow
arrangement, the change in local temperature dierence is the largest possible.
Comparing the corresponding heat exchanger eectiveness (see Figs. 3.7 and 3.8) for
these conditions, one can easily conclude that the temperature eectiveness of these two
ow arrangements diers the most.
The distribution of local temperature dierences has a profound inuence on exchanger eectiveness (either " or P1 ). A nite temperature dierence between the two uid
streams is the driving potential for heat transfer, but large temperature dierences lower
the exchanger or temperature eectiveness, and ultimately may contribute to a lower
system eciency, as we demonstrate in subsequent sections.
11.2.4
750
Four distinct situations are possible with respect to mixing of the uids, as emphasized in Section 1.6.1.3. These are also shown graphically in sketches for Eqs. (II.1)(II.4)
of Table 3.6. In Table 11.2, various temperature distributions are shown to be either oneor two-dimensional j f or or j f ; , j 1, 2, depending on uid mixing.
Our goal now is to show how one can formulate the models and subsequently solve them
to nd these temperature elds and/or corresponding outlet temperatures. As a byproduct of this analysis, one can easily devise temperature and/or heat exchanger eectiveness. In Tables 3.3 and 3.6, the formulas for the heat exchanger eectiveness are
listed. Here we discuss the analytical models for determining both temperature elds
and eectiveness. This will provide an insight into the inuence of uid mixing on heat
exchanger eectiveness, discussed in Section 11.3.
Referring to Fig. 11.3, one can write energy balances for control volumes as follows:
d m_ 1 cp;1 T1
|{z}
fluid enthalpy rate
into the control volume
@T1
_
dx
d m1 cp;1 T1
dq
0
|{z}
@x
|{z} heat transfer rate
11:16
and
d m_ 2 cp;2 T2
|{z}
fluid enthalpy rate
into the control volume
@T
d m_ 2 cp;2 T2 2 dy 0
@y
|{z}
heat transfer rate in
dq
|{z}
11:17
Note that d m_ j cp; j dCj , j 1, 2, in Eqs. (11.16) and (11.17) implies that constant
thermophysical properties assumption is invoked. It is assumed that no mixing takes
place on either side of the heat transfer surface. Consequently, both uids will have twodimensional temperature elds. In Eqs. (11.16) and (11.17), dq represents heat transfer by
convection from the hot uid to the wall; and in the steady-state formulation, that heat
will be transferred by conduction through the wall and by convection to the cold uid.
L1
L1
Control volumes
dx
L2
y+dy
y
dy
Fluid 1
dm2cp,2T2
x x+dx
y
h1
h2
x
L2
Fluid 2
751
1 1
2 2 ;
d1
1
1
d
C* NTU
@2
2 1
@
0 NTU
0 C* NTU
1 0 1
2 ; 0 0
1 1 ;
2 2 ;
@1
1 2
@
@2
2 1
@
0 NTU
0 C* NTU
1 0; 1
2 ; 0 0
Temperature eld
Independent
variables
Boundary
conditions
C NTU
2 d
1 0; 1
2 0 0
0 NTU
0 C* NTU
@1
1 2
@
NTU
d2
1
1 d
2
d
NTU 0
1 1 ;
2 2
Model
1 0 1
2 0 0
0 NTU
0 C * NTU
C NTU
d1
1
1
2 d
d
C* NTU 0
NTU
d2
1
1 d
2
d
NTU 0
1 1
2 2
MixedMixed
The dimensionless variables are dened as follows: j Tj T2;i =T1;i T2;i , x=L1 NTU, and y=L2 C * NTU. Horizontal and/or vertical lines in schematic gures
symbolize ow of the respective uid through the ow areas characterized with one-dimensionality (along the ow direction). The absence of lines in the direction of ow symbolizes
uid mixing in the direction transverse to the ow direction. Fluid 1 is assumed to have a smaller heat capacity rate Cmin .
Schematic
Dierential equations
UnmixedUnmixed
Information
752
Using the rate equations for convection and conduction, dq from the hot uid to the cold
uid can be expressed as follows:
Tw;1 Tw;2
dq o;1 h1 T1 Tw;1 dx dy kw
dx dy o;2 h2 Tw;2 T2 dx dy
w
|{z}
|{z}
|{z}
convection from hot fluid to wall
convection from wall to cod fluid
conduction within the wall
11:18
Equation (11.18) indicates that there is neither energy generation nor axial conduction
through the separating wall, as idealized in Section 3.2.1. The products hj o; j Tj ,
j 1, 2 of Eq. (11.18) represent the heat transfer rates exchanged per unit heat transfer
area between uids 1 or 2 and the wall separating the uids. In the case where the thermal
resistance of the separating wall is neglected, only convective terms exist in Eq. (11.18),
with Tw;1 Tw;2 Tw (i.e., the heat transfer surface has a uniform wall temperature
orthogonal to the ow directions). Note that the balances presented by Eqs. (11.16)
(11.18) do not involve the overall heat transfer coecient. They are formed by applying
the thermodynamic convention for each of the thermal energy ow rates (positive if
entering the system, otherwise negative). Note also that the assumptions of uniform
distribution of the heat transfer area and uniform wall thermal resistance are invoked.
Equation (11.18) is rewritten as follows:
T1 Tw;1
dq
o;1 h1 dx dy
Tw;1 Tw;2
dq
kw =w dx dy
Tw;2 T2
dq
o;2 h2 dx dy
11:19
Adding up the temperature dierences from three equations of Eq. (11.19), dening
dA dx dy, and using Eq. (3.18) for denition of the overall heat transfer coecient
U, but neglecting the fouling thermal resistances [since we did not include them in the
formulation of Eq. (11.18), which we could have included readily if desired], we get
dq U dAT1 T2
11:20
Substituting Eq. (11.20) into Eqs. (11.16) and (11.17) and simplifying, we can get the
following partial dierential equations:
@1
1 2
@
@2
2 1
@
11:21
753
with smaller heat capacity rate or as its complementary value. Note also that both
dimensionless temperatures j are assumed to be locally dependent on both independent
coordinates and :
1 1 ;
2 2 ;
11:22
0 C* NTU
11:23
Two boundary conditions (for uniform inlet temperatures) accompany the set of Eqs.
(11.21):
1 0; 1
2 ; 0 0
11:24
The set of equations Eqs. (11.21) and (11.24) represents the mathematical model of a
crossow heat exchanger. Four particular cases of crossow (see Section 1.6.1.3, and
Tables 3.6 and 11.2) dier from each other with respect to the presence or absence of uid
mixing on each uid side within the heat exchanger core (see Problems 11.5 and 11.6). In
Table 11.2, a summary containing all four models, deduced from Eqs. (11.21)(11.24), is
presented. Each of these models can be solved and closed-form analytical solutions can
be obtained using various solution methods (see Section 3.11). The solution of the
general unmixedunmixed case is asked in Problem 11.2. A particular case of an
unmixedmixed crossow arrangement is considered in detail in the following example.
The mixedmixed case is considered in Problem 11.7.
Example 11.3 Determine temperature dierence elds in a heat exchanger with a
mixedunmixed crossow arrangement. Assume that the uid with the smaller heat
capacity rate is mixed.
SOLUTION
Problem Data and Schematic: The ow arrangement under consideration corresponds to
the model and schematic given in the third column of Table 11.2 (Cmax uid unmixed).
Determine: Temperature dierence as a function of both axial and transverse coordinates
( and , as dened in Table 11.2).
Assumptions: The assumptions are as presented in Section 3.2.1.
Analysis: We rst determine the temperature elds for both uids followed by a temperature dierence distribution relationship within the heat exchanger core. The analytical model consists of two equations, one partial and one ordinary dierential equation,
as presented in the third column of Table 11.2 (the details of the model development are
the subject of Problem 11.6). Uniform temperatures are considered at inlets as corresponding boundary conditions. The solution to this analytical model will provide the
desired temperature elds. Let us rst solve the partial dierential equation for uid 2.
Subsequently, using the temperature eld solution for uid 2 and replacing its explicit
form in the ordinary dierential equation for uid 1, we will nd the temperature
754
distribution for uid 1. Finally, the dierence between the two uid temperatures will
provide the solution of the problem.
The solution of the partial dierential equation for uid 2 from Table 11.2 can be
obtained as follows using the Laplace transforms technique:
l
@2
2
lf1 g!s
@
!s
s2 ; s 2 ; 0 2 ; s
1
s
1
2
where s represents a complex variable that replaces . Rearranging Eq. (2) with
2 ; 0 0 from Table 2, we get
2 ; s 1
ss 1
An inverse Laplace transform of Eq. (3) provides the temperature eld for uid 2:
l1 2 ; sgs! 2 ; 1 1 e
Note that the explicit form of 1 still has to be determined. The ordinary dierential
equation for uid 1 of Table 11.2 can now be written as follows:
d1
1
1
d
C * NTU
C NTU
0
1 1 e d
After determining the integral on the right-hand side of Eq. (5) and rearranging, the
dierential equation for the uid 1 temperature distribution becomes
d1
k1 0
d
where k 1 expC * NTU=C * NTU. The boundary condition for Eq. (6) is
1 0 1
Finally, the relationship for the temperature dierence distribution can easily be determined from Eqs. (8) and (9) as
; 1 2 ; ek
10
755
Discussion and Comments: As expected, the temperature distribution for unmixed uid 2
is two-dimensional and that for mixed uid 1 is one-dimensional, both dependent on k,
which in turn depends on NTU and C*. Knowing the temperature distribution of the
uid with the smaller heat capacity rate (uid 1 in this case), one can easily determine
heat exchanger eectiveness (solve Problem 11.3 for understanding the details).
Similarly, the analysis of a crossow arrangement for the mixed uid having a larger
heat capacity rate can be performed, and heat exchanger eectiveness can subsequently
be determined (see Problem 11.4). Even more complex situations with nonuniform inlet
temperatures are the subject of Problems 11.8 and 11.9.
756
of states during a reverse change of state from the nal to the initial state without
additional energy interaction(s). An additional energy interaction is due to an absence
of reversibility in a real world when thermal phenomena are involved. In practical terms,
this means that the presence of irreversibilities is accompanied by thermodynamic losses,
ultimately leading to poorer thermal performance than predicted by an idealized reversible process. In a narrower sense, the same term is used to describe the losses in energy
terms I_irr caused by the presence of irreversibilities. The value of irreversibility cannot be
negative and it is not a system property as discussed next.
Irreversibility can be expressed in energy terms as a product of entropy generation S_irr
and a temperature weighting factor To (i.e., I_irr To S_ irr ). It can be shown that in many
engineering applications the weighting factor can be interpreted as the temperature of the
surroundings, which is identied as a thermodynamic reference state for measuring the
thermal energy potential of the system at hand. It should be noted that entropy generation is not a property of a system, while the entropy S is. Entropy is dened as a system
property by a statement
that its change in an ideal, reversible process must be equal to the
transfer of an entity dq=T that accompanies any heat transfer dq across the system
boundary where the local temperature is T. Hence, this abstract system property indicates that heat transfer must be accompanied by an entropy change. As a consequence, a
reversible adiabatic process can be identied by zero entropy change. If a process is not
reversible (as with any heat transfer across a nite temperature dierence), the situation
is radically dierent. Entropy change S is either
equal (reversible process) or larger
(irreversible process) than the entropy transfer ( dq=T, nonproperty){ that accompanies
heat transfer dq, the dierence being attributed to entropy generation S_ irr [see Eq.
(11.36)]. The amount of entropy generation is the quantitative measure of the quality
level of energy transfer. Entropy generation of zero corresponds to the highest quality of
energy transfer and/or energy conversion (a reversible process), and entropy generation
greater than zero represents poorer quality. All real processes are characterized by
entropy generation greater than zero.
The concept of exergy or available energy e is introduced to describe the maximum
available energy that can be obtained from a system in a given state. Each uid stream
that enters or leaves a heat exchanger carries exergy rate. Due to irreversible processes in
a system (e.g., a heat exchanger), the available energy of a uid decreases and the
dierence between the input and output exergy rates is equal to the lost exergy (lost
available energy), which in turn is identical to the irreversibility in energy terms (a Guy
Stodola theorem, i.e., exergy destroyed lost available energy temperature weighting
factor entropy generation; that is, e_ W_ lost To S_ irr :
11.3.1
Temperature and temperature dierence distributions within the heat exchanger inuence thermodynamic irreversibility. Thermodynamics teaches us that a measure of the
eciency of any thermal process can be assessed by gaining an insight into the irreversibility level incurred in an associated heat transfer process (Bos njakovic, 1965). This
irreversibility can be identied by determining the corresponding entropy generation. So
there is a deeper rationale for turning our attention toward temperature dierences
within a heat exchanger to determine heat transfer performance behavior. The driving
potential for heat transfer in a heat exchanger is the nite local temperature dierence
{
Note that the entropy is a system property and entropy transfer or entropy generation is not a system property.
757
between the uids exchanging heat, and we should expect that it greatly inuences the
exchanger eectiveness as well. Consequently, it is plausible to conclude that these temperature dierences are related to both the exchanger eectiveness and thermodynamic
eciency of an exchanger (one such thermodynamic gure of merit is dened in Section
11.6.5). How these temperature dierences inuence the irreversibility level is explained
next in terms of entropy generation.
The thermodynamic irreversibility manifested within a heat exchanger as an adiabatic
open system can be identied in terms of entropy generation by total entropy change
(entropy measure of irreversibility, i.e., entropy generation rate S_ irr ) of both uid
streams:
S_irr S_ m_ 1 s1 m_ 2 s2
11:25
Now we will evaluate S_ irr only due to nite temperature dierences, considering the uids
as pure simple single-phase compressible substances. Since ds dh=T (where h is the
specic enthalpy), the entropy rate change for uid 1 in a heat exchanger operating under
steady-state conditions for an ideal gas or an incompressible liquid is given by
m_ 1 s1
o
i
o
m_ cp dT
T1;o
m_ dh
m_ cp 1 ln
T1;i
T 1
T
i
1
11:26
Similarly, the entropy rate change for uid 2 in the exchanger will be
m_ 2 s2 m_ cp 2 ln
T2;o
T2;i
11:27
Note that we do not need to distinguish at this point whether the uid is hot or cold.
Moreover, this distinction is not necessarily relevant for calculating the entropy measure
of irreversibility. What matters, though, is that the two uids have dierent temperatures.
Hence, the concepts of hot and/or cold will not necessarily be used here. Consequently,
Eq. (11.25) can be rewritten as follows:
S_irr
2
X
j1
m_ j sj m_ 1 cp;1 ln
T1;o
T2;o
m_ 2 cp;2 ln
T1;i
T2;i
11:28
The two terms in Eq. (11.28) have the opposite signs since the uids have dierent
temperatures (T1;i 6 T2;i , i.e., either T1;o T1;i and T2;o T2;i or T1;o T1;i and
T2;o T2;i ). Two important thermodynamic points have to be reiterated here. First,
the fact that the two uids have dierent temperatures is of far more importance than
that one may conveniently be described as hot and the other as cold. This is because the
hot/cold dichotomy is introduced by convention. In a heat exchanger, as will be demonstrated later (see Section 11.4.3), the same uid may change the role of a hot/cold uid
side over some ow length! So, in this chapter, we will, as a rule, refer to a uid as uid 1
or uid 2 whenever a general case has to be considered in which any of the two uids may
either be hot or cold. If a uid is identied as having higher/lower temperature (such as in
a particular given example), we denote as the hot uid that has a temperature at its inlet
port higher than the temperature of other uid at its inlet port. Second, a more advanced
thermodynamic analysis advocated in this chapter involves the concept of entropy; hence
758
it involves not only temperature dierences but also temperature ratios and the products
of absolute temperature and entropy dierences [see, e.g., Eq. (11.28) or (11.53)]. As a
consequence, proper care must be taken regarding the use of absolute temperatures (K or
8R) for all temperatures associated with entropy and also exergy later. To emphasize this
fact, we use in this chapter, as a rule, temperatures on the absolute Kelvin (or Rankine)
scale and not on the commonly used degree Celsius (or Fahrenheit) scale.
The heat transfer rate between the two uid streams in thermal contact under adiabatic conditions is equal to the respective enthalpy rate changes (see Chapters 2 and 3):
q m_ 1 h1 m_ 2 h2
11:29
For better clarity, we consider uid 1 as the hot uid and uid 2 as the cold uid. Hence,
the enthalpy rate changes for the hot and cold uids are hh cp;h Th;i Th;o ,
hh cp;h Tc;o Tc;i for Eq. (11.29). Changing the subscripts 1 and 2 of Eqs. (11.28)
and (11.29) to h and c, combining them, and rearranging, we get,
Th;lm Tc;lm
S_ irr
1
1
q
Th;lm Tc;lm
Th;lm Tc;lm
11:30
where
Th;lm
Th;i Th;o
lnTh;i =Th;o
Tc;lm
Tc;o Tc;i
lnTc;o =Tc;i
11:31
Here Th;lm represents the log-mean temperature of the hot uid as dened using inlet and
outlet temperatures Th;i and Th;o . The Tc;lm is dened similarly. In contrast, the arithmetic
mean temperatures of the hot and cold uids are Th;m Th;i Th;o =2 and
Tc;m Tc;i Tc;o =2, and the log-mean temperature dierence between hot and cold
uids in a heat exchanger is given by Eq. (3.172).
The entropy generation is related to the dierence in the uid temperatures. Equation
(11.30) is written for the exchanger as a whole. On the local level, entropy generation is
related to local temperature dierences [such as Eq. (11.15)]. Hence, the dierence
between mean temperatures of two uids [the numerator of Eq. (11.30)] directly inuences the entropy measure of the irreversibility manifested within the heat exchanger. As
a consequence, the heat exchanger irreversibility for a given heat transfer rate can be
reduced by reducing temperature dierences between the uids, which in turn will
increase the exchanger eectiveness ".{ A heat exchanger characterized by smaller temperature dierences between the uids generates a smaller irreversibility in a given system
compared to a heat exchanger (for the same heat transfer rate) that has larger temperature dierences between the uids. Since the entropy measure of irreversibility is related
directly to thermodynamic system eciency (see Section 11.6.5), this statement leads to
an anticipated conclusion about the possible detrimental inuence of this source of
irreversibility on the overall system eciency.
Thermodynamic irreversibility represented by entropy generation as in Eq. (11.28)
can be formulated in terms of heat exchanger thermal design parameters. Using the
{
Note that the log-mean temperature dierence Tlm of Eq. (3.172) is proportional to (Th;lm Tc;lm ). Hence, a
smaller value of Tlm means a larger value of " for an exchanger.
759
denitions of heat exchanger eectiveness and heat capacity rate ratio Eqs. (3.44) and
(3.56)] and considering C1 Cmin , one can show that
T1;o
1 "#1 1 1 P1 #1 1
T1;i
T2;o
1 C*"# 1 1 R1 P1 # 1
T2;i
11:32
where # T1;i =T2;i represents the inlet temperature ratio. Substituting these expressions
in Eq. (11.28) results in
S_irr
S* C* ln1 "#1 1 ln1 C *"# 1
Cmax
S_ irr
S* R1 ln1 P1 #1 1 ln1 R1 P1 # 1
C2
11:33a
11:33b
where S_ irr 6 0 for # 6 1 and S_ irr 0 for # 1. Note that normalizing S_ irr by Cmax or C2 ,
as indicated in the leftmost term of Eq. (11.33), is a matter of arbitrary choice. The
entropy generation is equal to zero for the inlet temperature ratio equal to 1. Entropy
generation [Eq. (11.33)] for dierent ow arrangements is dierent for the same inlet
temperature ratio, heat capacity rate ratio C * or R1 , and NTU (Sekulic, 1990b). This is
because of dierent heat exchanger eectivenesses, " " (NTU, C*), for dierent ow
arrangements [and xed values of NTU and C* (or NTU1 and R1 )]. It should be emphasized that the control volume for the entropy generation of Eq. (11.33) [see Eq. (11.25)] is
drawn outside the exchanger core or matrix through inlet/outlet ports. Hence, the S*
expression of Eq. (11.33) is valid for an exchanger with any ow arrangement by employing its appropriate "-NTU or P-NTU formula. Some additional features of Eq. (11.33)
are discussed in Section 11.4.1.
11.3.2
760
A virtual stream
mo = mi
mi =
j=1 m j i
,
(a)
Fluid 1: Uniform or nonuniform
temperature inlet, Manifold
partitioned
Stream 1
Uniform temperature
at outlet. Manifold is
not partitioned
Stream n
(b)
(c)
FIGURE 11.4 Flow passages with uid mixing: (a) passage or duct with uid mixing; (b) outlet
manifold; (c) two adjacent ow passages with/without uid mixing.
the mixed uid and the available thermal potential is destroyed. This is certainly an
irreversible phenomenon that leads to a corresponding entropy increase. A practical
consequence of this thermodynamically detrimental process is the equalization of uid
local temperatures across the ow passage cross section, ultimately leading to a reduced
heating/cooling manifested at the uid exits. As a consequence, the respective temperature eectiveness (or the heat exchanger eectiveness) is also reduced.
Let us consider a uid stream being mixed while owing right to left through a duct, as
shown in Fig. 11.4a. The objective of the analysis is to determine entropy generation
associated with uid mixing using a very simplied approach. This situation is often
present in heat exchangers. For example, in the outlet header, mixing is accomplished
between dierent streams of the same uid ideally with no heat transfer with the surroundings as shown in Fig. 11.4b. In the heat exchanger core region, a mixed uid
exchanges heat with the other uid, either mixed or unmixed. A signicant simultaneous
heat transfer and mixing (see Fig. 11.4c for the unmixedmixed case) prevent us from
determining the sole contribution of mixing to the total irreversibility for a control
volume under consideration. That is the reason why we formulate a more general case
but consider a mixing-only case as follows.
A mixed uid (uid 1 in Fig. 11.4a) ows through the passage while simultaneously
being mixed in a direction transverse to the overall ow direction. For the sake of
clarity (but with an inevitable loss of rigor), let us assume that resultant heat exchange
between the uid and the environment can be modeled as heat transfer between each of
n virtual streams of the mixed uid at the inlet that merge into one mixed stream at
the outlet. Mass rate, energy/enthalpy rate, and entropy rate balances (continuity,
energy, and entropy equations) for the control volume of the bulk ow of uid 1 are
as follows:
m_ o
Continuity equation:
n
X
m_ j;i
j1
Energy equation:
m_ o ho
n
X
m_ j;i hj;i
j1
Entropy equation:
m_ so
n
X
n
X
qj
j1
m_ sj;i
j1
n
X
q
j1
761
dmcv
0
d
11:34
dEcv
0
d
11:35
dScv
S_ irr > 0
d
11:36
where qj , j 1, n, represent the equivalent heat transfer rates between virtual streams
(having average individual bulk temperatures Tj along the respective ow paths) and
the surroundings (the other uid side). Note that S_ irr > 0 in Eq. (11.36) is a consequence
of the second law of thermodynamics for a real system.
For a steady-state ow, Eqs. (11.34)(11.36) reduce to{
m_ o
n
X
m_ j;i
11:37
j1
m_ o ho
n
X
m_ j;i hj;i
j1
m_ o so
n
X
n
X
11:38
qj
j1
m_ sj;i
j1
n
X
q
j1
S_ irr
11:39
Taking into account the change of local temperatures along the ow paths would,
indeed, require writing the balances in a dierential form and integrating them along the
ow path. The form of these relationships would depend on actual heat transfer conditions. For isolating the mixing eect only, we consider an adiabatic mixing case.
The simplest physical situation of practical interest corresponds to the conditions
encountered in the headers/manifolds or in the parts of a heat exchanger where the
heat transfer q can be neglected (as is truly for example, in the exit zone of a TEMA J
heat exchanger; Fig. 11.6). In such an adiabatic situation, Eqs. (11.38) and (11.39) can be
simplied as follows:
m_ o ho
n
X
m_ j;i hj;i
11:40
j1
S_irr m_ o so
n
X
m_ sj;i
j1
n
X
m_ sj
11:41
j1
For a simple compressible substance, using the expression of Eq. (11.26) for m_ j sj for
each stream, S_ irr of Eq. (11.41) reduces to
S_ irr
n
X
j1
m_ cp j ln
To
Tj;i
11:42
762
Hence, S_ irr 0 for uniform inlet temperature Tj;i Ti To , j 1, n) and S_irr 6 0 when
thermal dissimilarity is present (i.e., nonuniform temperature Tj;i 6 To ). For example,
mixing the two streams of uid 1 at the exit of a 12 TEMA J shell-and-tube heat
0
00
exchanger (see Fig. 11.6 where To T1;o , and Tj;i are denoted as T1;o
and T1;o
) is a source
of irreversibility due to the fact that these outlet temperatures of the shell uid (uid 1)
leaving the two zones A and B are not the same.
From Eq. (11.42), it is obvious that entropy generation caused by mixing would never
be equal to zero if thermal dissimilarity is present while mixing the streams, even for a
single uid. The mixing process actually eliminates the presence of local temperature
dierences; hence it is an inherently irreversible process. This statement holds true regardless of the presence or absence of heat transfer to the environment (or other uid) during
mixing. Consequently, it is expected that a heat exchanger featuring mixing of either uid
within the heat exchanger core or within the header zones should have less eectiveness
compared to an exchanger with the same design parameters but without mixing.
11.3.3
The importance of uid pressure drop and uid pumping power P in the heat exchanger
is discussed in Chapter 6. One of the important components of the uid pressure drop in
the heat exchanger is the uid friction associated with ow over the heat transfer surface.
We derive the irreversibility associated with this uid friction in this section. Since the
control volume is drawn at the inlet and outlet pipes/tanks, this analysis does take into
account the contribution of both skin friction and form drag that is important in many
exchangers.
To identify the irreversibility caused only by uid friction, let us assume a uid
owing through a ow passage of an arbitrary cross section. The ow is caused solely
by the pressure dierence between the two points along the uid path. The entropy
generated with such a ow is equal to the entropy change between the two points
along the ow path, say between inlet and outlet. If the enthalpy change contribution
to entropy change can be neglected (steady and adiabatic ow), the entropy change is as
follows using the T ds relationship: dh 0 T ds v dp (where s and v are specic
enthalpy and specic volume, respectively):
o
i
d S_
o
i
m_ ds
o
i
m_
v
dp
T
11:43
S_ irr
m_
o
v dp
i
Tlm
p lnTo =Ti
m_
To Ti
11:45
763
In both Eqs. (11.44) and (11.45), we have S_ irr 0 for p 0 and S_ irr 6 0 for p 6 0.
Hence, the entropy generation caused by uid friction is never equal to zero for p > 0,
as in a heat exchanger. In a heat exchanger with two uids, the irreversibility contribution of each of the two uids has to be included [i.e., two terms of the form of Eq. (11.44)
or (11.45) have to be calculated].
11.4.1
Let us rewrite Eq. (11.33) in a symbolic form as a function of relevant design parameters
as follows:
S_ irr
S * f C*; "; # f C *; NTU; #; flow arrangement
Cmax
S_irr
S * f R1 ; P1 ; # f R1 ; NTU1 ; #; flow arrangement
C2
11:46a
11:46b
The second equality in Eq. (11.46) is written by taking into account Eq. (3.50). So S* is a
function of the heat capacity rate ratio, NTU, inlet temperature ratio, and ow arrangement. In Fig. 11.5, this relationship is presented for counterow and parallelow
arrangements for uids having equal heat capacity rates (C* 1), and an inlet temperature ratio equal to 0.5 (i.e., # 0:5). It can be shown that corresponding curves for
numerous other ow arrangements will be located between the two limiting cases presented in Fig. 11.5 (Sekulic, 1990b). It should be emphasized that these curves (except for
parallelow) have at least one distinct maximum, as explained next. For an exchanger at
small NTU values, when NTU ! 0, the magnitude of entropy generation will tend to
zero (i.e., S* ! 0. This is certainly an expected result because at NTU 0 there is no
heat transfer since UA 0 despite the temperature potential for it (represented by the
given inlet temperature dierence). On the other side, if NTU ! 1, the temperature
dierences along the heat exchanger tend to their minimum possible values (e.g., T 0
for C* 1 for a counterow exchanger). Consequently, S* decreases to a limiting
asymptotic value (equal to zero for C* 1 for a counterow exchanger). Hence, a
curve having minimum values at both ends will have at least one maximum value in
between 0 < NTU < 1.
This analysis thus provides the following conclusions for many ow arrangements
having only one maximum for S* [including counterow but excluding parallelow
(Sekulic, 1990a); see Section 11.4.3 for an exception; refer to Shah and Skiepko (2002)
for other exceptions]:
764
Parallelflow arrangement
0.12
0.10
0.08
S*
0.06
Counterflow arrangement
0.04
C*= 1.0
Inlet temperature ratio = 0.5
0.02
0.00
0
10
NTU
FIGURE 11.5 Entropy generation in parallelow and counterow exchangers with C* 1.
8
>0
>
>
>
>
>
< 0
@S*
@NTU >
0
>
>
>
>
:
<0
at small NTU
at NTU NTU*
at large NTU
lim S * 0
NTU!0
11:47
NTU!1
The most interesting feature of the heat transfer irreversibility behavior implied by
Eq. (11.47) is an existence of at least one maximum of S* for a nite-size heat exchanger
at NTU* (or NTU*1 ).
Let us determine the value of NTU* and the corresponding eectiveness at that
operating point (the same can be done for the number of transfer units dened as
NTU1 and the corresponding temperature eectiveness P1 ). The entropy generation
maximum is characterized by{
@S*
0 at
@NTU
NTU
Smax
NTU*
11:48
Performing the calculation as indicated in Eq. (11.48) on Eq. (11.33), one can show (see
Problem 11.13) that the maximum of entropy generation corresponds to{
"
Smax
1
1 C*
or
P1
Smax
1
1 R1
765
11:49
Note that Eq. (11.49) is identical to the corresponding relationships in Eq. (3.114). Thus,
the number of heat transfer units at the maximum irreversibility in a heat exchanger is
exactly the same as the limiting value of NTU ( NTU*) at the onset of an external
temperature cross. At that operating point, the outlet temperatures of both uids are
equal. Hence, beyond this NTU > NTU*, there will be a temperature cross, and the hotuid outlet temperature will become lower than the cold-uid outlet temperature. As
dened in Sections 3.2.3 and 3.6.1.2, the temperature cross derives its name from ctitious
or actual crossing of the temperature distributions of the hot and cold uids in an
exchanger. If there is no actual crossing of hot- and cold-uid temperature distributions
within an exchanger and Tc;o > Th;o , we refer to it as an external temperature cross as
found in the temperature distributions of a counterow exchanger in Example 3.2 or the
12 TEMA E shell-and-tube heat exchanger of Fig. 3.17a for the high-NTU case (the
solid lines). We will call it an internal temperature cross if there is an actual crossing of
hot- and cold-uid temperature distributions within an exchanger. There are two possibilities for the internal temperature cross: (1) Tc;o > Th;o (as in the 12 TEMA E exchanger of Fig. 3.17b at high NTU, or (2) Tc;o;local > Th;o;local , where the subscript local
means one of the multiple outlets on one or both uid sides of an exchanger (see the
temperature distributions in Fig. 11.7 for the 12 TEMA J shell-and-tube heat exchanger
of Fig. 11.6). Note that there can be an external temperature cross without an internal
temperature cross (such as in a counterow exchanger, as shown in the temperature
distribution of Example 3.2). Let us discuss the implications of the external and internal
temperature crosses further in the following two subsections.
11.4.2
Results obtained in Section 11.4.1 [i.e., an interpretation of the physical meaning of Eq.
(11.49), and the relation between the equalization of outlet temperatures and maximum
entropy generation] lead to yet another interesting analogy. This analogy can be
explained by studying Eq. (11.49) in a dimensional form.
First, let us reconrm (see also Section 3.6.1.2) the statement of equality of the outlet
temperatures at the operating point indicated by Eq. (11.49). By invoking denitions of
exchanger eectiveness [Eq. (3.44)] and heat capacity rate ratio [Eq. (3.56)], Eq. (11.49)
can be rewritten as follows:
T1;o T1;i
1
11:50a
11:50b
766
11:51a
C1 C2 To C1 T1;i C2 T2;i
11:51b
Therefore,
Equation (11.51) clearly indicates an identical result as before, but this time it can be
interpreted as having been obtained for a quite dierent physical situation of adiabatic
mixing of the two uid streams (with heat capacity rates C1 and C2 and inlet temperatures T1;i and T2;i ). Such imaginary mixing will lead to the outlet temperature To for
the mixture of the two given streams (C1 C2 ). From thermodynamics, we know that
adiabatic mixing leads to total destruction of the available thermal energy potential
(implied by the dierence of the temperatures of the uids) that exists at the onset of
mixing. Thus, this process must be characterized by maximum entropy generation.
In conclusion, there is an analogy between a heat exchange process at the operating
point that corresponds to the condition of an equalization of outlet temperatures and the
adiabatic mixing process of the same two uids. This analogy demonstrates that entropy
generation in a heat exchanger at that operating point must be the maximum possible.
This constitutes an additional physical explanation of the thermodynamic signicance of
an external temperature-cross operating point in a heat exchanger.
A clear thermodynamic meaning of the result given by Eq. (11.49) can easily be
conrmed by comparing Eqs. (11.28) and (11.42), that is, comparing the entropy
generations obtained for two completely dierent processes: (1) a heat transfer process
in a heat exchanger characterized with equal outlet temperatures of the uids involved
[i.e., T1;o T2;o To ; see Eq. (11.28)], and (2) a mixing process of the two uids
[Tj Tj;i ; j 1, 2; see Eq. (11.42)]. The entropy generation rates for two physically
quite dierent processes are found to be identical.
Analysis provided so far shows clearly how the facts related to the detrimental inuence of uid mixing on heat exchanger performance t nicely into the consistent thermodynamic picture. Hence, the results for heat exchanger eectiveness presented in Chapter
3 (Tables 3.3 and 3.6) have deeper physical explanations. Let us now show how this kind
of thermodynamic analysis can be used to understand the behavior of a relatively complex heat exchanger ow arrangement. Moreover, we demonstrate why such analysis has
importance for practical design considerations.
11.4.3
It has been emphasized (see Section 3.6.1.2) that contrary to a general design requirement
to transfer heat only from one uid to the other, and not vice versa, in some heat
exchangers, reverse heat transfer takes place. For example, consider the 12 TEMA J
shell-and-tube heat exchanger of Fig. 11.6. The existence of a temperature cross leads to a
situation in which an addition of more surface area in the second tube pass does not
contribute a signicant increase in heat transfer because of reverse heat transfer taking
place in the second pass. Note that we reached this conclusion in Section 3.6.1.2 based on
assumed (i.e., at that point not analytically derived) temperature distributions. In this
767
T1, o
T2, i
T1,o
Tube segment b
Tube segment d
x = - L/2
T2, o
x=0
Tube segment a
Tube segment c
Shell zone A
Shell zone B
T1, i
L
chapter (see Section 11.4.2), we provided a thermodynamic interpretation of this undesirable phenomenon through temperature distributions.
We elaborated in detail how one can determine temperature distributions and subsequently evaluate the inuence of local temperature dierences and a mixing process on
heat exchanger performance. Let us now show how the internal temperature cross can
cause peculiar behavior in the P1 -NTU1 or "-NTU results. For a given heat capacity rate
ratio, with increasing NTU1, the temperature eectiveness P1 reaches a maximum
beyond which an increase in NTU1 causes P1 to decrease rather than increase monotonically as may be expected.{ This behavior is illustrated in Fig. 3.16. Figures 11.7 and
11.8 summarize the results of a thermodynamic analysis of this exchanger. This analysis
includes both the rst and second laws of thermodynamics. The approach is straightforward. It starts with a determination of temperature and heat transfer rate distributions
obtained through an application of the rst law of thermodynamics (Figs. 11.7 and 11.8a;
model formulations in Problem 11.1). Subsequently, entropy generation is determined
utilizing both the rst and second laws of thermodynamics [see Section 11.3.1 and Eq.
(11.33)].
Figure 11.7 shows three temperature distributions for NTU1 0:87, 1.83, and 5.0 for
R1 2. Figure 11.8 provides data regarding the corresponding distribution of dimensionless heat transfer rates [the total, and those determined at dierent tube sections (a, b,
c, and d) and shell zones (A and B) of the Fig. 11.6 exchanger] and entropy generations.
To demonstrate this analysis, let us consider a numerical example from the results of
Kmecko (1998).
Example 11.4 For a 12 TEMA J shell-and-tube heat exchanger, establish the number
of temperature crosses and explain the meaning of the existence of a maximum eective{
There are also dierent and unexpected P1 vs. NTU1 curve behaviors for n-pass n-pass (n > 2) plate heat
exchangers and complex ow arrangements (Shah and Skiepko, 2002).
768
IV2 ,d
0.8
0.8
2I V2 ,d
2,b
II
0.6
0.6
0.4
ITC
2 ,o
0.2
III
22 ,c
0.4
0.2
ITC
1 1, A
0
-0.2
ETC
11, B
-0.4
2 ,b
2 2 , a
0. 2
2 ,a
2, c
1,A
11 ,B
0.4
-0.4
-0.2
0.2
0.4
(b)
(a)
1
0.8
d
2 ,d2
0.6
0.4
1,B
ETC
22II,b
0.2
2, c
22,a 2 ,a
1,A
I
ITC
0
-0.4
-0.2
0. 2
0.4
(c)
ness at nite NTUs. The range of operating points under consideration is dened with
R1 2 and a range of NTU values between NTU1 0:87 and 5.0. The inlet temperature
ratio is equal to 2.
SOLUTION
Problem Data and Schematic: The schematic of a 12 TEMA J shell-and-tube heat
exchanger is given in Fig. 11.6. The heat capacity rate ratio R1 2. The range of
NTU values is 0.87 to 5.0. The inlet temperature ratio # 2:0.
Determine: The number of temperature crosses for this exchanger and explain the existence of maximum eectiveness at nite NTUs.
769
q*
d
q*
0.4
R1 = 2
0.2
q*
q*
R1 = 1
*
S */S max
q *B
q*A
0.99
q*a
q*c
-0.2
0
8 10
NTU1
(a)
12
14
0.98
6
NTU1
(b)
10
FIGURE 11.8 (a) Heat transfer rate (R1 2), and (b) dimensionless entropy generation rate
(R1 1 and 2, # 2) in a 12 TEMA J shell-and-tube heat exchanger. Note that the NTU1 axis
is extended slightly beyond 0 to show clearly the ends of the curves separate from the y axis. (From
Kmecko, 1998.)
770
the ITC point 1;B 2;c . Note that the temperature eectiveness of uid 1 is P1 1;o ,
and P2 can then be calculated from P2 P1 R1 .
If NTU1 is increased from 0.87 to 1.83, a new operating point will be reached. Under
that condition, temperature distributions will become as given in Fig. 11.7b. This gure
reveals a peculiar increase of the uid 2 dimensionless temperature (2;c ) in the second
part of the tube segment c: instead of decreasing, the original hot tube uid 2 dimensionless (or dimensional) temperature increases. This is due to reverse heat transfer taking
place to the left of the ITC point in Fig. 11.7b. As expected, the resulting exit temperature
of uid 1 is higher than that for the operating point presented in Fig. 11.7a. This means
that the temperature eectiveness of uid 1 has increased from P1;a 1;o;a to
P1;b 1;o;b (the subscripts a and b denote the cases from Fig. 11.7a and b, with
NTU 0:87 and 1.83, respectively). If we increase the number of transfer units even
further, say from 1.83 to 5.0 (see Fig. 11.7c), the increase in the uid 2 dimensionless
temperature 2;c in the tube segment c becomes very pronounced (see, e.g., 2;c at
0:5 in Fig. 11.7c). The shell uid temperature still continues to increase in both
shell zones, but much more so in zone B, 1B in Fig. 11.7c. However,
1;o 1;A;o 1;B;o =2 of Fig. 11.7c is less than 1;o of Fig. 11.7b. As a result, the
temperature eectiveness of uid 1 will become smaller than the corresponding value of
Fig. 11.7b. This means that with an increase of NTU1 starting from 1.83, the temperature
eectiveness decreases from its maximum value at NTU1 1:83 and R1 2:0, as shown
in Fig. 11.8a, in the q* P1 vs. NTU1 curve. To understand this peculiar behavior, let
us now consider the distribution of heat transfer rates within the heat exchanger in
various shell zones and tube segments, as presented in Fig. 11.8a.
It is interesting to notice from Fig. 11.8a that with an increase in NTU1, an initial
increase in total dimensionless heat transfer rate, q* q=m_ cp 1 T2;i T1;i P1 , is
followed by a maximum at NTU1 1:83 and decrease at higher NTU1. Adding more
heat transfer area beyond that for the maximum eectiveness decreases the heat
transfer rate exchanged between the two uids. How could that be possible? The
components of this heat transfer rate are presented in the same diagram (Fig.
11.8a) for both the shell uid (zones A and B) and the tube uid (tube segments a
to d) having appropriate subscripts with q*. From this gure, it is clear that only the
tube segment d (the inlet segment of the rst tube pass) contributes to the heat
exchanger performance by its positive dimensionless heat transfer rate slope in the
direction of increasing NTU1. The heat transfer contribution in tube segments a and b
(and correspondingly, in the shell zone A) is diminishing rapidly, while the heat
transfer in tube segment c becomes reversed and is increasing in the negative direction.
Consequently, a large portion of the heat exchanger actually fails to contribute to the
original design goal of increasing q with P1 at large NTU1 values, beyond
NTU1 1:83 for this case!
Discussion and Comments: The behavior of the analyzed heat exchanger can be interpreted in terms of entropy generation. For this exchanger, S_ irr can be calculated using Eq.
(11.33) for # T1;i =T2;i 2:0 and the P-NTU formula of Eq. (III.11) of Table 3.6 for
given NTU1 and R1 1 and 2 as shown in Fig. 11.8b. For this exchanger, the entropy
generation features two maximums and one local minimum. The two equalizations of
outlet temperatures (ETCs, one at smaller and the other at larger NTU1) correspond to
two maximums in entropy generation for the case of R1 1 and 2. In this situation, the
rst is at NTU1 0:87 and the second at NTU1 5:0, as shown in Fig. 11.8b. At the
operating point characterized by the maximum temperature eectiveness, the heat
771
exchanger operates at a local minimum entropy generation. Note that the existence of the
second maximum at large NTU1 is not of great practical importance per se (a 12 TEMA
J shell-and-tube heat exchanger is never designed as a single-pass unit at such large
NTU1). Rather, it indicates that the entropy generation between NTU1 0:87 and 5.0
may both decrease and increase. In other words, it makes no thermodynamic sense (or
practical sense either) to design a 12 TEMA J heat exchanger for a large NTU (in the
analyzed example, a large NTU means that NTU > 1:83).
The foregoing discussion demonstrates why study of a heat exchanger design should
be accompanied by a study of internal heat transfer intricacies and a thermodynamic
analysis. A very practical conclusion has been reached. The tube segments a and b are
contributing virtually nothing to the heat exchange process at high NTUs, and the tube
segment c has reverse heat transfer. This conclusion would never have been reached if
study of this design were conducted looking at the heat exchanger as a black box. Of
course, shell-and-tube heat exchangers are hardly ever designed for NTU > 1:5 with a
single shell pass.
We nd the following interesting observations from a detailed review of some "-NTU
curves of exchangers presented in Chapter 3 and many "-NTU curves for other ow
arrangements from Shah and Pignotti (1989).
. When there is an external temperature cross only (as in a counterow exchanger) or
only an external temperature cross can be found by modifying the original ow
arrangement (such as modifying the ow arrangement of Fig. 3.17b to 3.17a), the
exchanger eectiveness will continue to increase monotonically with increasing
NTU.
. When an internal temperature cross cannot be eliminated [due to the exchanger
geometry (such as the 12 TEMA J exchanger of Fig. 11.6) or even if by modifying
the ow arrangement], the exchanger eectiveness will decrease with increasing
NTU beyond the point of S*min at NTU > 0. For such exchangers, the S_ irr versus
NTU curve has at least two maximums, and the minimum value of S* occurs
between the two maximums (see, e.g., Fig. 11.8b).
The following exchangers have an internal temperature cross and the exchanger
eectiveness decreases with increasing NTU beyond the point of S*min at NTU > 0
based on the extensive P-NTU results of Shah and Pignotti (1989): a crossow
exchanger with both uids mixed, 22, 23, and 24 overall parallelow PHEs,
and the following TEMA shell-and-tube heat exchangers: 13 E (two passes in
parallelow) , 14 E, 12 G (overall parallelow), 12 H (overall parallelow), 12
J, and 14 J.
772
and ow friction phenomena to assess the heat exchanger performance, even without a
detailed quantitative analysis. We use this argument to show how a simple heuristic
approach can be used to assess heat exchanger performance (Sekulic, 2003).
Our goal is to compare the pairs of ow arrangements having the same #, NTU, and
R. The objective is to predict through a heuristic analysis which of the two arrangements
in a pair has better eectiveness without computing the eectiveness.
Let us compare two single-pass crossow arrangements: (1) the arrangement with
uid 1 unmixed and connected in identical order between the rows A and B and uid
2 split into two equal individually mixed streams as presented in Fig. 11.9a, and (2) the
conguration with an inverted order coupling of uid 1 shown in Fig. 11.9b. Thus the
only dierence between the two arrangements is the coupling of uid 1 streams between
the two rows of uid 2. Let us also assume that the P1 -NTU1 relationships for the two
arrangements are not known but that the two exchangers have identical NTU1 and R1 .
The question is as follows: Which of the two arrangements has better performance?
To attempt to answer the question, let us recall that the presence of larger temperature
dierences and/or uid mixing in an exchanger inevitably increases thermodynamic
irreversibility (i.e., entropy generation), and correspondingly decreases the exchanger
eectiveness, if all other conditions remain the same. As far as mixing in individual
rows is concerned, the two arrangements are identical. However, the local temperature
dierences in the two ow arrangements are dierent due to overall ow congurations.
Consequently, the two exchangers must have dierent eectivenesses.
Hence, let us compare qualitatively the magnitude of the local temperature dierences
in these two arrangements by concentrating on the two lateral streams S1 and S2 of uid
1 in Fig. 11.9a and b. In the case of an identical order coupling (Fig. 11.9a), the stream S1
leaves the rst row A of the uid 2 from a zone where uid 2 enters that row, and
subsequently enters the second row B of the same uid again in the zone where uid 2
enters that row (that is why we call this coupling identical). Stream S2 leaves the rst row
A of uid 2 with a temperature that has been changed less when compared to the change
experienced by stream S1 (due to the fact that uid 2 has already experienced some heat
transfer through row A before it meets stream S2 ). Moreover, stream S2 enters the second
row B of uid 2 in the zone where the exit of uid 2 is located in that row. The heat
transfer will be accomplished at established temperature dierences, dened by the given
operating and design conditions.
In the case of the inverted order, though (Fig. 11.9b), stream S1 from the exit of
the rst row A of uid 2, from a zone where uid 2 enters the row, is led into the
S2
S2
Ao Bi
S2
Ao Bi
S1
Fluid 1
B
A
S1
Fluid 2
(a)
S2
Fluid 1
S1
B
Fluid 2
S2
(b)
FIGURE 11.9 Single-pass crossow: uid 2 split into two equal streams individually mixed and
owing through rows A and B: (a) uid 1 unmixed and connected in identical order; (b) uid 1
unmixed and connected in inverted order. The outlet and inlet zones of the two uid 2 stream rows
are denoted by Ao and Bi .
773
second row B of uid 2 to the zone from which uid 2 leaves that row. Similarly,
stream S2 is led from the rst row A of uid 2, from a zone where uid 2 leaves the
row, to the second row B of uid 2, into the zone where uid 2 enters the row. So
stream S1 , the stream that has experienced a larger temperature change in the rst
row A of uid 2 than stream S2 , will meet uid 2 in two dierent zones of the
second row B of uid 2 for the two analyzed arrangements. As a consequence, in the
case of the identical order coupling, the heat transfer between uids 1 and 2 will be
accomplished in the second row B of uid 2 at smaller temperature dierences at the
entrance than that in the case of the inverted order. Based on this observation of the
smaller temperature dierence at the inlet of the second tube row or a pass, we
could infer that it will have lower temperature dierence irreversibility, which means
higher exchanger eectiveness.
This conclusion has been reached without any knowledge about the corresponding
eectiveness relationships; it is based on the qualitative analysis of the heat transfer
process involved. Calculating the eectivenesses for the two arrangements at the same
inlet operating conditions, using appropriate P-NTU relationships from Shah and
Pignotti (1989), can easily provide a simple proof. For example, if NTU1 3 and
R1 0:8, the temperature eectiveness of the identical order and inverted order arrangements are 0.7152 and 0.6668, respectively. The aforementioned conclusion has been
veried for pairs of heat exchangers that dier only with respect to a particular source
of irreversibility (a uid coupling between the passes, mixing, nite temperature dierence magnitude at a location where they may be at maximum).
A systematic analysis of the applicability of this heuristic approach can be conducted
for other heat exchangers. For crossow arrangements, it was shown that even in the
cases for which the closed-form expressions of the eectiveness relationship does not
exist, the simple heuristic approach usually works.
In Fig. 11.10, two two-pass arrangement pairs are compared.{ For both the two-pass
cross-parallelow exchangers and the two-pass cross-counterow arrangements, the
identical order coupled passes (Fig. 11.10a and c) have larger eectiveness values compared to the corresponding inverted-order coupled passes (Fig. 11.10b and d).
Finally, let us consider an exchanger with the most complex two-pass cross-counterow arrangement, with both uids unmixed and connected either in identical (Fig.
11.11a) or inverted order (Fig. 11.11b). For these ow arrangements, there are no
closed-form analytical solutions for eectivenessNTU relationships. The solutions for
NTU1 3 and R1 0:8 for the two cases are (1) both uids coupled in identical order
(Fig. 11.11a), P1 0:757, and (2) both uids coupled in inverted order (Fig. 11.11b),
P1 0:736. Consequently, the identical order arrangement has obviously higher eectiveness. We can reach the same conclusion without utilizing the complex semianalytical
relationship if we simply use the heuristic approach described above. This is presented in
Example 11.5.
Example 11.5 Provide a heuristic argument that a two-pass cross-counterow arrangement with both uids unmixed and connected in identical order has a larger uid 1
{
Note that the identical and inverted orders between rows/passes appear to be dened dierently sometimes (see,
e.g., Fig. 11.9b vs. Fig. 11.10b, both having inverted order). However, they are dened consistently as identical
order if both uid streams enter the second row/pass in the same order as they enter the rst row/pass; if they enter
the second row/pass in a dierent order (one of the streams enters from the other end of the row/pass), we refer to
it as an inverted order. Review Sections 1.6.1.3 and 1.6.2.1 for further clarication.
774
Fluid 1
Fluid 1
Fluid 2
Fluid 2
(a)
(b)
Fluid 1
Fluid 1
Fluid 2
Fluid 2
(c)
(d)
SOLUTION
Problem Data and Schematic: Schematics of the two ow arrangements are given in Fig.
11.11.
Determine: Which of the two ow arrangements has the higher temperature eectiveness
for the same operating point.
Assumptions: The assumptions are as discussed in Section 3.2.1.
S2
S2
B
S1
S1
Fluid 1
Fluid 1
Fluid 2
(a)
Fluid 2
(b)
FIGURE 11.11 Two-pass cross-counterow exchangers with both uids unmixed and connected
in (a) identical order and (b) inverted order.
775
Analysis: Assume that uid 1 is the hot uid. In that case, the uid 1 stream S1 leaves pass
A with its temperature higher than the temperature of stream S2 . In the case of an
identical order coupling (Fig. 11.11a), uid 1 stream S1 (the warmer of the two) is
channeled to meet uid 2 at the exit of pass B, at the point where the corresponding
uid 2 stream has already been heated by owing through that pass. The uid 1 stream S2
(the colder of the two), however, is guided to meet uid 2 at its inlet into pass B, where it
has the lowest temperature (uid 2 is assumed to be the cold uid). So these temperature
dierences would be smaller for the identical order coupling (Fig. 11.11a) than for the
case of an inverted order coupling (Fig. 11.11b). This is because in the inverted order
coupling of Fig. 11.11b, streams S1 and S2 of uid 1 are channeled to meet the corresponding streams of uid 2 in pass B at locations characterized with opposite inlet/outlet
sections when compared to pass A. The same reasoning can be conducted following the
two bordering streams of uid 2. So we should expect the identical order coupling to be
more favorable than the inverted, the result already indicated above.
Discussion and Comments: A note regarding the presence of mixing is warranted at this
point. It seems to be more obvious to assess two ow arrangements with respect to uid
mixing than with respect to local temperature dierences. A good example is the
sequence of decreasing eectiveness values for, say, single-pass crossow arrangements
given in Table 11.2, for the given NTU and R. For example, for the set of values
NTU1 3 and R1 0:8, the temperature eectiveness P1 takes the values 0.7355,
0.6791, 0.6655, and 0.6254, for unmixedunmixed, unmixed (uid 2)mixed (uid 1),
unmixed (uid 1)mixed (uid 2), and mixedmixed ow arrangements, respectively. The
increasing inuence of mixing is obvious. Note that for unmixedmixed and mixed
unmixed arrangements, uid 1/uid 2 is either mixed or unmixed, respectively, and
because R1 0:8 < 1, it is clear that mixing of the uid with the larger heat capacity
rate contributes more to performance deterioration, as expected. It should be mentioned
that the heuristic approach might become more dicult to conduct in complex multipass
heat exchangers. However, if performed correctly, it should provide at least guidance
regarding the actual performance.
776
Let us assume that the particular design requirements have been formulated for a process
industry application. Since we are dealing here with a specic application, we identify the
uids as hot and cold instead of using more general terminology. The temperature of a
hot-uid stream in a process has to be reduced from Th;i to Th;o while simultaneously a
cold-uid stream temperature in the same process has to be increased from Tc;i to Tc;o , as
777
shown in Fig. 11.12.{ We specify Th;o < Tc;i as shown in Fig. 11.12. Note that it is
possible that in some applications Tc;o may be desired to be even greater than Th;i . For
the sake of simplicity, both uids are considered to be simple substances and have heat
capacity rates Ch and Cc . In addition, cold and hot thermal sources (usually called
utilitiesthe heat sources or sinks, or yet other process streams) are available if the
required enthalpy changes cannot be accomplished utilizing the cold stream alone.
Note that the general validity of this analysis is not restricted by adopting an assumption
that the uids are simple compressible substances within the given temperature ranges
and away from the onset of phase-change phenomena.
Utilizing Eq. (2.1), and writing the outlet temperatures explicitly for the two streams
leaving the process, one obtains
Tj;o Tj;i
q
H_
Tj;i
Cj
Cj
11:52
where j h or c. The enthalpy rate change, H_ , equals the heat transfer rate q in the
exchanger. Relationships dened by Eq. (11.52) can be interpreted as line segments in a
(T; H_ ) graph, each with a slope equal to the reciprocal value of the corresponding heat
capacity rate. These graphs are presented in Fig. 11.12 (for a selection of given inlet and
outlet temperatures and heat capacity rates) for the three coupled exchangers shown
underneath. They are cold utility, main, and hot utility exchangers.{ Note that equalization of heat capacity rates would lead to parallel temperatureenthalpy rate change lines.
In Fig. 11.12, the temperatureenthalpy rate change lines are not parallel, therefore, the
corresponding heat capacity rates are not equal. The greater the imbalance between the
heat capacity rates, the more pronounced is the nonuniform temperature dierence
(Th;hex Tc;hex ) distribution along the enthalpy rate change axis.
The situation presented in Fig. 11.12 can be interpreted as a general case situated
between the two limiting designsto accomplish the design goal (to reduce the hot-uid
enthalpy and to increase the cold-uid enthalpy)a designer may distribute the total
heat load between (1) the main heat exchanger, (2) the cold utility heat exchanger, and
(3) the hot utility heat exchanger, shown in Fig. 11.12. This may be done in such a way as
to exploit the heat recovery in the overlapping region of the temperatureenthalpy rate
change diagram (note: Th;o < Tc;i and Tc;o < Th;i ). An important question for the exchanger designer is the following: What would be the optimum thermal size, UA, of the main
heat exchanger?
In a limiting case, the size of the main heat exchanger may be very large, say close to
an innitely large thermal size (i.e., NTU ! 1). The outlet temperature of the hot uid
from that exchanger (Th;o;hex ) will become very close to the inlet temperature of the cold
stream (Tc;i ). That is, the minimum temperature dierence between the uids
Tmin Th;o;hex Tc;i;hex Th;o;hex Tc;i (the so-called pinch) becomes very close to
zero. Note that an increase in size of the main heat exchanger in the temperature
enthalpy diagram (Fig. 11.12) may easily be presented by shifting horizontally the
{
It should be emphasized that the required heating and cooling of cold- and hot-uid streams of Fig. 11.12 may
need multiple (and not one) heat exchangers.
In Fig. 11.12, the main exchanger has the hot process stream on one uid side and the cold process stream on the
other uid side, so that thermal energy is recovered from the hot process stream, heating the cold process stream
(as desired) without any additional expense of utility streams. The cold utility exchanger cools the hot stream
coming out of the main exchanger and the hot utility exchanger heats the cold stream coming out of the main
exchanger. The utility streams are water, steam, or air in general.
{
778
Hot utility
C * or R1
or P1
Th,i
NTU or NTU1
Tc,o
Hot fluid
T
Th,o,hex
Tc,o,hex
T min
Cold fluid
Tc,i = Tc,i,hex
Th,o
Cold utility
qm,hex
qcu,ex
qhu,ex
H
Th,i
Cold utility
exchanger
Main exchanger
System
control
volume
Tc, o
Th,o
Tc,i
Component control
volumes
Hot utility
exchanger
FIGURE 11.12 Heat exchanger temperatureenthalpy rate dierence diagram. The energy
balance control volumes (CVs) denote system component boundaries. The subscript hex
denotes the main heat exchanger under consideration.
temperature curves toward each other until Tmin 0 (note that qm;hex on the abscissa
becomes the largest possible). This hypothetical design would reduce the requirement for
utilities to a minimum (qhu;ex and qcu;ex become the smallest possible). Note that this
idealized design cannot eliminate the need for either a hot or cold utility (the hot and cold
uid lines will not necessarily overlap entirely) for the given Cc and Ch , but it
would reduce both utilities to the minimum. Still, this design is not a realistic one (i.e.,
it would not be a workable design) because it requires an innitely large main heat
exchanger.
Another limiting case would be to invest virtually nothing into the design of the main
heat exchanger and to accomplish the required task by (1) cooling the hot uid exclusively with a cold utility, and (2) heating the cold uid with a hot utility. The temperatureenthalpy diagram in that case would feature the absence of the overlapping of the
TH_ lines, such as by shifting the TH_ line for the cold uid to the right until Tc;i is
vertically in the same location as Th;i . This solution would require a maximum possible
energy cost (no heat recovery at all) but at the same time would reduce the investment in
the main heat exchanger to zero (qm;hex 0). For this example, it is assumed that hot and
cold utilities would be available without an additional need for capital investment. This
design is clearly workable (see Example 11.6 on the next page).
779
It is obvious that an optimum design must be somewhere between these two limiting
cases. The objective functions may be, say, the total cost (involving the energy cost and
number and size of the units required), the physical size of the exchanger, the exergy
losses (see the following sections), and the like. Our intention in this section, however, is
not to study thermoeconomic optimization and/or to extend the analysis considering a
whole system to perform an integration of a heat exchanger network using either pinch
analysis or exergy (or entropy generation) analysis. Our goal is less ambitiousto
formulate the design optimization objective problem related to a single exchanger or a
couple of them, and to show how an optimization problem arises.
In a more general case, most notably for the case of a heat exchanger network, the
temperatureenthalpy rate change diagram becomes a valuable tool in constructing
composite curves. Namely, instead of presenting temperatureenthalpy rate change
lines for only two streams of a single heat exchanger, one composite TH_ curve for all
hot uid streams and another such composite curve for all cold uid streams within the
corresponding temperature ranges can be depicted as well. Each such line has a slope
related to the sum of all heat capacity rates of the involved streams within the given
temperature range. This analysis, though, is beyond the scope of our interest. More
details about the procedure of constructing the composite curves, very useful for the
pinch technology method, can be found in the literature devoted to process synthesis and
integration (e.g., Linnho et al., 1982; Gunderson and Naess, 1988; Sama, 1995a).
11.6.2
780
(589 K). The task can be accomplished either by using only one two-uid heat exchanger
(and utilizing only the stream with the higher temperature) or using two two-uid heat
exchangers connected in a two-stage array with the cold uid owing through both
exchangers. The available two hot streams (the utilities) should be used in the two
exchangers. Determine the optimal distribution of heat transfer surface areas of the
two exchangers to get the minimum total heat transfer area sucient to heat the cold
stream to the desired temperature. Compare the two heat exchanger solution with the
design having one heat exchanger. Assume constant and uniform overall heat transfer
coecients in both units to be equal to 454 W=m2 K.
SOLUTION
Problem Data and Schematic: Two heat exchangers arranged in a two-stage array with
only one cold stream owing through both units is shown in Fig. E11.6. For a minimum
surface area requirement, both exchangers are considered to be counterow. The input
data are given in Fig. E11.6.
Tc,A,o = Tc,B,i
Tc,B,o = 533 K
Tc,A,i = 311 K
A
Hot fluid, stage A
Th,A,o
Th,B,i = 589 K
Determine: The heat transfer surface area of each of the two exchangers for a minimum
total heat transfer surface area. Compare that solution to a single-exchanger solution.
Assumptions: All appropriate assumptions for modeling each of the exchangers of
Section 3.2.1 are invoked. Additionally, it is assumed that both units have equal overall
heat transfer coecients.
Analysis: We need to formulate a functional relationship between total surface area Atot
and cold-uid interstage temperature T Tc;A;o Tc;B;i (shown in Fig. E11.6) for a
minimum total surface area: Find
minfAtot f Tg
where
Atot AA AB
and
T Tc;A;o Tc;B;i
781
Therefore, the problem is characterized with the following variables: (1) the interstage
temperature T of the cold uid is a decision variable, (2) the outlet temperatures (Th;A;o
and Th;B;o ) of the two hot uids are dependent variables, and (3) the independent variables (i.e., parameters of optimization) are: the inlet and outlet temperatures (Tc;A;i and
Tc;B;o ) of the cold uid, the inlet temperatures (Th;A;i and Th;B;i ) of the hot uids, the heat
capacity rates of both uids, and the overall heat transfer coecients of both units.
To dene the objective function, Atot f T, rst let us compute the surface areas AA
and AB . To accomplish that task we have to utilize the rst law of thermodynamics by
formulating energy rate balances (i.e., equating the heat transfer rates and uid enthalpy
rates imposed by the rst law of thermodynamics for each uid side):
UAj Tlm; j Cc; j T Tc; j;i
where j A or B. Hence, the individual surface areas AA and AB from Eq. (3) are
AA
Cc;A T Tc;A;i
UA Tlm;A
AB
Cc;B Tc;B;o T
UB Tlm;B
As emphasized above, Eq. (3) is based on energy and mass conservation principles.
Hence, all the subsequent results are the consequences of the rst law of thermodynamics
(energy equation) and the continuity equation only.
The total area is dened as follows:
Atot AA AB
where
Tlm;A
UA Tlm;A
UB Tlm;B
Tlm;B
In Eqs. (5) and (6), the outlet temperatures Th;A;o and Th;B;o are dependent on the
interstage temperature T and can be found again from energy conservation principles.
This time, however, we should formulate enthalpy rate balances for uids in both units;
that is, for units A and B,
Cc;A T Tc;A;i Ch;A Th;A;i Th;A;o
Therefore,
Th;A;o Th;A;i
Cc;A
T Tc;A;i
Ch;A
Th;B;o Th;B;i
Cc;B
T
T
Ch;B c;B;o
Substitution of Th;A;o and Th;B;o of Eq. (8) in Eqs. (5) and (6) yields an objective
function Atot f T, where T is a decision variable. Formally, to solve the problem,
we should nd an extremum of this function with respect to the decision variable (i.e.,
@Atot =@T 0). Performing this partial dierentiation, we get (Alt-Ali and Wilde, 1980)
Cc;B
@Atot Cc;A
A
1
0
@T
UA A A A RA A UB B RB B
782
where
A T Tc;A;i
B Tc;B;o T
A Th;A;i Tc;A;i
B Th;B;i T
RA
Cc;A
Ch;A
RB
Cc;B
Ch;B
10
Cc;B =UB
A
B RB B Th;A;i Tc;A;i Th;B;o Topt
11
where Topt represents the value of the interstage temperature of the cold uid for
@Atot =@T 0 (i.e., the optimum interstage temperature).
Finally, the optimum interstage temperature of the cold uid Topt can be obtained
from Eq. (11) after introducing the denitions of , , and Rs as follows:
Topt
Cc;A =UA
Th;A;i Th;A;i Tc;A;i Th;B;i Tc;B;o
Cc;B =UB
1=2
12
For the given problem, inserting the numerical values of the variables in Eq. (12), we
get
"
Topt
#1=2
13
Now, replacing T Topt 344 K into Eq. (4) and introducing other known variables, the heat transfer surface areas become AA 48:1 m2 and AB 395:5 m2 . Total
heat transfer area is Atot AA AB 444:6 m2 . In a trivial case, if only one heat
exchanger is used (with the hot uid having an inlet temperature of 589 K), the heat
transfer surface area can easily be calculated from Eq. (3) for j 1. In such a case, this
area would be 464.5 m2, which is 4.5% more than Atot 444:6 m2 for the optimum area
for two exchangers.
Discussion and Comments: This example demonstrates how a proper distribution of heat
transfer surface area can lead to an optimal design. This conclusion has been reached
utilizing only energy and mass ow rate balances. The fact that a single heat exchanger
would have a larger heat transfer surface area (compared to a two-heat-exchanger solution) does not mean that the solution with two exchangers would be the desired design
solution. Cost (in addition to other considerations) may be a key decision factor for the
best achievable design. Note that temperature dierences between the uids at the terminal points of both heat exchangers have an important role in the determination of the
optimum solution. However, the temperature dierences dened only on the basis of
783
total area minimization (i.e., utilizing only the rst law of thermodynamics and mass
balances) do not necessarily minimize the entropy generation in the assembly or lead to
the most economic solution. For this to be accomplished, a combined thermodynamic
and economic analysis has to be performed as discussed in the following section.
11.6.3
Let us now illustrate how cost rate balances may inuence the design solution. Of course,
the simultaneous use of energy balances is mandatory as well. This time, the objective
function will be the total annual cost as a function of the cold-uid exit temperature,
keeping the heat duty xed in a heat exchanger. Obviously, variation of the exit temperature will change exchanger " or P1 as well as cause a change in the coolant mass ow
rate for xed q. Thus, C* or R1 will change. This will yield a dierent NTU and hence A.
Consequently, a search for the most favorable design based on an economic criterion
would require resizing the exchanger.
Both economics and thermodynamics will inuence the solution. It is important to
notice that the corresponding thermodynamic part of the analysis would imply only the
rst law of thermodynamics. How that has to be accomplished can be illustrated best by
using an example. The example represents a slightly modied problem presented by
Peters and Timmerhaus (1980).
Example 11.7 Design a condenser in a distillation unit to operate at optimum total
annual cost. The particular unit under consideration uses cooling water to condense
vapor. It operates at a minimum total annual cost if water leaves the condenser at
528C (325 K), and at the given inlet uid conditions and with installed optimal heat
transfer area. The outlet temperature of the cooling water, however, may increase to
578C (330 K) due to changing environmental considerations with a corresponding
decrease in the cooling-water mass ow rate and appropriate resizing of the exchanger.
Assuming that all inlet variables (except for the mass ow rate of water) must stay
unchanged, determine how large the change of total annual cost would be in excess of
the established optimum value for the given design, if the mentioned changes take place.
The additional information is as follows. The heat exchanger condenses m_ 2000 kg/h
of vapor, which has an enthalpy of phase change of 4 105 J/kg. Condensation occurs at
778C (350 K). The inlet temperature of the cooling water is 178C (290 K), and the specic
heat of water at constant pressure is 4:2 103 J=kg K. The overall heat transfer coecient is 280 W/m2 K. The distillation unit must operate for o 6500 hr/y. The unit cost
of cooling water is 2 105 c/kg, where c is the monetary unit. The unit cost for the heat
exchanger per installed unit of heat transfer area is 300 c/m2. The annual cost of heat
exchanger operation is 20% of the cost of installed heat exchanger area.
SOLUTION
Problem Data and Schematic: In the condenser, the cooling water increases its temperature along the ow path, and the condensing stream has a constant temperature, as
shown in Fig. E11.7A. All pertinent data are provided in this gure.
Determine: The increase in total cost in excess of the optimum value when the exit
temperature of the cooling water increases by 5 K.
784
Condensing
vapor
Cooling
water
Tc,i = 290 K
105
C /kg, CA = 300 C
/m2,
cA = 0.2/yr
Assumptions: It is assumed that Eq. (3.12) is valid. The mean temperature dierence in
that equation is assumed to be equal to the log-mean temperature dierence determined
using terminal temperature dierences. Assume the total annual cost of heat exchanger
operation to be Ctot Cw CA , where Cw is the cost of the total amount of cooling
water, and CA is the xed cost of the heat exchanger installed.
Analysis: Let us start with evaluating the heat transfer rate q:
q m_ h hg 2000 kg=h4 105 J=kg 8 108 J=h
The total annual cost of the condenser is equal to
Ctot Cw CA
Therefore,
q
q
c A CA
cp Tc;o Tc;i
U Tm
"
#
8 108 J=h
c=kg
4:2 103 J=kg KTc;o 290 K
Ctot o Cw m_ 2 cA CA A o Cw
6500 h=yr2 105
"
#
8
1
2 8 10 J=h1 h=3600 s
0:2 300 c=m
yr
280 W=m2 K Tm
24,762
c 0:4762 105 c
Tm
Tc;o 290 K yr
yr
where
Tm Tlm
Th Tc;i Th Tc;o
Tc;o Tc;i
T 290 K
c;o
Th Tc;i
Th Tc;i
350 290 K
ln
ln
ln
Th Tc;o
Th Tc;o
350 K Tc;o
785
The relationship of Eq. (2) is presented in Fig. E11.7B. The minimum value of total cost,
Ctot; min 1898 c=yr, is at Tc;o 325 K, as suggested in the problem formulation. This
can be conrmed easily by nding the rst derivative of the objective function, Eq. (2),
with respect to the coolant exit temperature, Tc;o (i.e., @Ctot =@Tc;o 0). This minimum
point (see Fig. E11.7B), corresponds to the designed heat exchanger heat transfer area of
20 m2 [calculated from the rate equation, Eq. (2.2)]. If the outlet temperature of the
coolant increases to 330 K, the total cost will increase to Ctot 1927 c=yr, a 1.5% higher
value than the value at the optimal point. To keep the duty unchanged, the heat transfer
area changes to approximately 22 m2, and the coolant mass ow rate will change from
1.512 kg/s to 1.323 kg/s.
Discussion and Comments: In this example, the objective function is the total annual cost
of heat exchanger operation. The optimum (minimum) total cost corresponds to the
outlet coolant temperature equal to 325 K. Any change in this temperature incurred
through a change of heat exchanger size (area) causes some increase in the total cost.
At the same time, both the coolant mass ow rate and the exchanger heat transfer area
must change to keep the heat load xed. In this analysis, we utilized only the rst law of
thermodynamics to formulate energy balances (used implicitly for setting up the cost rate
balance).
This example illustrates how an optimum design may lead to an economic penalty if a
heat exchanger operates at o-design. In this case, only the exchanger surface area and
the cooling water utility are considered for cost. The optimization problem required a
minimum annual variable cost as a function of the exit temperature. Equations (2.1) and
(2.2) dened the heat exchanger model. The temperature at optimum (i.e., at the minimum cost) is conrmed to be 325 K from Fig. E11.7B. The o-design condition requires
a heat exchanger size change, as a result of a change in the exit temperature (for other
variables constant except for the mass ow rate). Note that this economic optimum was
not related in any way to a thermodynamic irreversibility minimization.
1 960
1 950
Ctot,modif.
C tot
1 940
1 930
1 920
Ctot,min
1 910
1 900
1 890
320
322
324
326
328
330
332
Tc,o (K)
FIGURE E11.7B Total annual variable operating cost of a heat exchanger vs. coolant exit
temperature.
786
11.6.4
In previous sections, we demonstrated the utilization of both energy and cost rate
balances. These analyses, though, have not considered the quality of energy rate ows.
Under the term energy rate ow, we consider either the enthalpy rate or the heat transfer
rate. Consequently, we should examine more closely the role of the irreversibility analysis
for heat exchanger thermal design and optimization. In this section, we formulate the
exergy rate balance. In the following two sections, we use this concept to dene a
thermodynamic gure of merit (thermodynamic eciency of a heat exchanger) and to
introduce the cost of irreversibility.
Let us assume that a heat exchanger represents a component of a process or power
plant. The plant can be optimized for the total annual cost of operation. However, to nd
the cost of the related irreversibilities (see Section 11.6.6), we should dene the irreversibility on an energy basis in a quantitative manner. This cost can be calculated as a sum
of the cost of compensation for the irreversibility and the capital investment. Irreversibility
may be measured by its energy measureeither the entropy generation multiplied by the
appropriate weighting temperature factor (Ahrendts, 1980), or exergy (Kotas, 1995). So
far in our analysis we measured irreversibilities in terms of entropy generation. As is
known from thermodynamics (Moran and Shapiro, 1995), irreversibility can conveniently be dened in terms of energy rate units using the exergy rate as follows (excluding
chemical exergy):
e_j;k m_ j hj;k href Tref sj;k sref
11:53
The exergy rate represents the rate of the available energy of a given uid stream with
respect to the conveniently selected reference state. Each uid stream (entering or leaving
the heat exchanger; see Fig. 11.13) carries certain exergy, dened by Eq. (11.53). In this
equation, the subscript j denotes either uid 1 or uid 2, or alternatively, the hot or cold
uid, k denotes inlet or outlet, and ref denotes a state of the respective uid at the
thermodynamic condition dened by a selected reference state (often, but not always, the
state of the environment). Exergy does not obey a conservation principle because it
includes not only the properties of a thermodynamic system at an exchanger terminal
port, but also the reference thermodynamic state. In our case, the term thermodynamic
system refers to whichever uid stream exposed to heat transfer in a heat exchanger. In
many applications, that state can simply be a state of the system in thermodynamic
equilibrium with the environment. The usefulness of the concept of exergy is in providing
a reference level regarding the maximum possible useful energy potential that would be
available from a particular energy source with respect to the surroundings. The quality of
energy ow rate can be interpreted by a simple example as follows. The quality of the
same amount of available energy from a uid at 5008C is higher than for the same uid at
508C if both have to be used (of course at dierent ow rates) to heat another uid
entering at 08C in a heat exchanger. The rst uid stream would have a larger exergy rate
and would therefore be able to transfer more heat over a wider temperature span. In
other words, its potential to do work is high and its use at the low temperature level
would be wasteful if there is a second stream at a lower temperature (< 5008C) that is
available to do the same job.
Applied to a situation in a heat exchanger, the exergy balance can be closed only by
introducing into the balance the exergy destruction d. Consequently, the exergy rate
balance for a heat exchanger (Fig. 11.13) may be written as follows:
d e_1;i e_1;o e_2;i e_2;o
11:54
Hot fluid
. with inlet temperature Th,i , pressure ph,i , enthalpy
rate Ih,i = f(Th,i , ph,i ), and exergy rate Eh,i = f(Th,i , ph,i, Tref , pref )
787
E h,,i
Tref > Th,i
Heat
exchange, q
Th,o
Th,i
Tc,o
Tc,i
Hot fluid
. with inlet temperature Tc,i, pressure pc,i , enthalpy
rate Ic,i = f(Tc,i, pc,i) and exergy rate Ec,i = f(Tc,i, pc,i, Tref , pref)
FIGURE 11.13 Exergy rate ows through a heat exchanger.
where the two expressions in parentheses on the right-hand side of the equality represent
individual exergy destruction rates for each of the two uids. The total exergy destruction
rate given by Eq. (11.54), d in watts (Btu/hr), represents the quantitative energy measure
of irreversibility. As demonstrated by Bosnjakovic (1972), this irreversibility can also
be calculated from entropy generation using the GouyStodola theorem as follows:
d Tref S_ irr
11:55
Equation (11.55) obviously implies a need for insight into the heat transfer thermodynamic intricacies (expressed in terms of entropy generation).
We started the analysis in this chapter with the argument that insight into temperature
distributions must be gained and that the heat exchanger designer should understand
how the exchanger performance depends on the most important sources of irreversibility,
S_ irr . Now we can clearly see that this understanding may contribute to a reduction of
exergy destruction in a system. A system engineer knows how to relate that destruction
directly to the monetary value of the capital investment and operating costs and to
optimize the system by providing appropriate changes in heat exchanger design. By
writing exergy rate balances (or calculating entropy generation), energy rate measures
of irreversibilities can be determined and subsequently their monetary value calculated
(see Section 11.6.6).
11.6.5
Let us now introduce a thermodynamic gure of merit{ in the form of exergy eciency
for the performance of a heat exchanger in a system. The performance level may be
{
The second law eciency is dened in many dierent ways and they are in general referred to as the thermodynamic gure of merit.
788
dened using the concept of exergy. The exergy balance for a heat exchanger, based on
Eq. (11.54), states simply that exergy entering a heat exchanger (carried in by both uid
streams) must be equal to the sum of exergy leaving the exchanger and the destruction of
exergy caused by heat exchanger operation (as a consequence of irreversibilities), that is
e_i e_o d
where
11:56
11:57
11:58
In a particular case, when uids 1 and 2 are identied as cold and hot uids, respectively,
Eq. (11.58) indicates that the exergy increase in the cold uid plus the exergy destruction
must be equal to the exergy decrease in the hot uid utilized in this process. Equation
(11.58) in such a case can be divided by the dierence in the exergy rates on the righthand side of the equality, to obtain
e_c;o e_c;i
d
1
e_h;i e_h;o e_h;i e_h;o
11:59
e_c;o e_c;i
d
1
_
_
_
eh;i e_h;o
eh;i eh;o
11:60
8
e_c;o e_c;i
>
>
>
>
< e_h;i e_h;o
>
>
e_h;o e_h;i
>
>
: _
ec;i e_c;o
11:61
The eciency dened by Eq. (11.61), 1, takes care of dierent objectives of a heat
exchanger used in a system. Namely, if the purpose of a heat exchanger is to increase the
exergy rate (i.e., energy availability or energy quality) of a cold-uid stream (at
the expense of a decrease of exergy rate of the hot-uid stream), is calculated using
the rst expression of Eq. (11.61). An objective of heating a uid by cooling another uid
is an example for the rst expression of of Eq. (11.61) (the heat source may be waste
thermal energy, a heat exchanger in a heat pump, etc.). On the contrary, if the purpose of
the exchanger is to cool the hot stream at or below the reference temperature (refrigeration for temperatures below the environment temperature), the second expression of Eq.
(11.61) should be used.
The change of design parameters may lead to dierent behaviors of the heat exchanger eectiveness and thermodynamic eciency for the same heat exchanger. Let us
demonstrate this fact by an example.
789
Example 11.8 Assess the following two designs for a counterow heat exchanger. The
heat exchanger is to heat 1 kg/s of air from 2278C (500 K) and 3 bar to 3078C (580 K).
The hot-uid stream is also air at the inlet temperature of 3278C (600 K) and 2 bar. The
two options available are as follows: The heat exchanger is designed for a hot-uid mass
ow rate of either 1.24 kg/s or 4.94 kg/s. Decide which design will yield better exergy
eciency in the system. The surroundings is at 300 K and 1 bar.
SOLUTION
Problem Data: Operating conditions: (a) cold uid is air with Tc;i 500 K, Tc;o 580 K,
m_ c 1 kg=s, pc;i 3 bar; (b) hot uid is air with Th;i 600 K, ph;i 2 bar, m_ h 1:24 or
4.94 kg/s. The heat exchanger is a counterow unit. The surroundings is at Tref 300 K,
and pref 1 bar.
Determine: Which of the two uid ow rates will lead to a design with higher exergy
eciency?
Assumptions: The assumptions are as listed in Section 3.2.1. Air is assumed to behave as
an ideal gas. Pressure drops are negligible, i.e., pc;i pc;o and ph;i ph;o . Idealize the same
overall heat transfer coecient in both designs. Assume surroundings to be relevant for
exergy denition (i.e., at reference conditions as follows: Tref 300 K and pref 1 bar).
Analysis: A study of input data reveals the following. The heat capacity rates of the two
uids are not known, but we do know that they are not the same because the mass ow
rates are dierent. The specic heats of both uids are not known. Note that the specic
heat of an ideal gas is a function of the temperature only. So, to determine the heat
capacity ratio, we have to know the exit temperature of the hot uid. However, the exit
temperature of the hot uid cannot be determined a priori because the specic heat is not
known. However, both uids are of the same type (i.e., air) and we can idealize that the
specic heats at constant pressure of both uids will not dier signicantly from each
other. The mass ow rate of the hot uid is specied as larger than that for the cold uid
(i.e., m_ h 1:24 kg/s or 4:94 kg=s > m_ c 1 kg/s). Therefore, in both cases, the heat capacity rate of the hot uid will be larger than the corresponding value for the cold uid.
Hence, cold air is the Cmin uid, and " "c . The heat exchanger eectiveness represents
in this case a dimensionless outlet temperature of the cold uid (" "c , i.e., the cold uid
has a smaller heat capacity outlet rate, Cc Cmin ; see Section 11.2.2). Since both terminal
temperatures of the cold uid are known and xed, the heat exchanger eectiveness "c
and heat transfer rate q for both designs are the same (a subject of Review Question 3.26).
However, because the heat capacity rate ratios are not the same, the two designs must
have dierent NTUs. Dierent NTUs for the same U and Cmin must correspond to
dierent heat transfer areas.
One can arrive at the same conclusion by noting that the heat transfer rate, the inlet
conditions for both uids, and the smaller heat capacity rate all are predetermined. From
the denition of heat exchanger eectiveness based on its thermodynamic interpretation
(see Example 11.2), it becomes clear that the heat exchanger eectiveness must be identical
in both cases. Therefore, the change in NTU is due to dierent heat transfer surface areas.
Thus, the two designs considered would have the same " and q but dierent As. So the
two exchangers would extract the same heat transfer rate from the hot uid but at
dierent temperature levels and with dierent local temperature distributions. Hence,
790
the irreversibility level (responsible for thermodynamic performance) would not be the
same. The question is: Which of the two designs (the one with the larger or the one with
the smaller mass ow rate of the hot uid) will provide higher exergy eciency?
To be able to use one of the two denitions of the exergy eciency of Eq. (11.61), one
should recognize the purpose of the device. According to the problem formulation, the
goal is to heat the cold uid at the expense of the hot uid. Therefore, the exergy
eciency is dened by the rst of the two equations given by Eq. (11.61). In other
words, the increase in the exergy rate of the cold uid will be accomplished at the expense
of a decrease in the exergy rate of the hot uid:
where
e_c;o e_c;i
e_h;i e_h;o
with
hj;k j;k air; T Tj;k
0
sj;k j;k
air; T Tj;k ; p pj;k
In Eqs. (2) and (3), j h or c and k i or o. The condition of the reference state (and
values of stream properties at the condition of the surroundings) is denoted by ref.
The design problem formulated in this example is a sizing problem in which both
NTUs and one outlet temperature are unknown. The outlet temperature of the hot uid
in terms of Tc;i can be determined from the denition of " and C* as follows:
Th;o Tc;i 1 C *"Tc;i Th;i
The calculation of exergy eciency must be performed numerically using Eqs. (1)(4)
because thermophysical properties depend on the temperature or on both temperature
and pressure [see Eqs. (3) and (4)]. The results of the calculation of exergies and are
listed below. These results are obtained using EES software (2000). Air is treated as an
ideal gas. C* and " are calculated from input data and subsequently, NTU is calculated
for these values of " and C* for a counterow exchanger.
m_ h
(kg/s)
Th;o
(K)
C*
"
NTU
e_c;i
(kW)
e_c;o
(kW)
e_h;i
(kW)
e_h;o
(kW)
1.24
4.94
536
584
0.8
0.2
0.8
0.8
2.94
1.79
142.1
142.1
178.9
178.9
191.7
762.5
152.6
721.6
0.941
0.899
A review of the results above indicates that the exergy eciency is larger for the design
that requires larger NTU (smaller mass ow rate of the hot uid). Both designs are
characterized by the identical heat exchanger eectiveness.
Discussion and Comments: The results obtained conrm a statement made at the beginning of this chapter that heat exchanger (or temperature) eectiveness does not necessarily provide sucient information about exchanger performance. The dierences
between the terminal temperatures on the hot uid inlet side are the same for both
exchangers but the temperature dierence between the uids at the cold uid inlet side
increases from 36 K (for a smaller mass ow rate or a larger NTU) to 84 K (for a larger
mass ow rate or a smaller NTU). The results clearly demonstrate that smaller local
791
temperature dierences along the respective ow paths lead to larger exergy eciency (as
a consequence of the smaller entropy generation for the smaller m_ h case). Moreover, in
the case of the smaller exergy eciency, the exit temperature of the hot uid (584 K) is
quite close to the cold-uid outlet temperature (580 K). To understand the implication,
review the S* vs. NTU curve for C * 1 for a counterow exchanger in Fig. 11.5. The
trend for the S* vs. NTU curve for other values of C* is identical except for C * 0 and
S*max is of the same order of magnitude for all C* values (Sekulic, 1990) for a given #.
Hence, for this problem, one can see that the operating point approaches from the left to
S*max at a high airow rate, and the operating point for a low airow rate will be considerably to the right of S*max . Thus, S* and S_ irr S *Cmax will be higher for the high-airowrate case. So this high-airow-rate case is an unfavorable design solution from the exergy
point of view. An engineers ultimate decision, though, will also depend on a number of
additional considerations (pressure drop considerations, cost analysis, etc.). These considerations may include a trade-o between the physical size of the heat exchanger (cost of
exchanger) and the mass ow rate of the hot uid (operating cost), or alternately, the
minimum cost of thermodynamic irreversibility, as discussed in the next subsection.
11.6.6
With the thermodynamic background developed so far, we can now point out how to
relate an exergy ow rate to the cost rate. The heat exchanger transforms the total exergy
rate input, carried in by both streams, into an exergy rate output that is always smaller
than the one at the input due to the inevitable presence of irreversibilities (caused by heat
transfer rate at nite temperature dierences, uid mixing, friction phenomena, etc).
Each exergy rate ( j streams into and out of the heat exchanger) has an associated cost
rate C_ j as
C_ j Cj e_j
11:62
where Cj denotes the cost per unit of exergy (c=J). The cost per unit of exergy of
uid ows that enter a heat exchanger has to be determined before an exergy costing
can be performed. For example, if a particular uid stream has to be utilized in a
system, a designer must know the cost of the utilization of that stream. This cost has
to be determined from an analysis located upstream from the location of the
stream utilization (Bejan et al., 1995). Now we can write a cost rate balance for a
heat exchanger:
X
X
11:63
C_ j;i
C_ j;o Z_ cap Z_ op
jh;c
jh;c
In Eq. (11.63), Zcap and Zop denote the capital investment cost and operating expenses{
of a heat exchanger that must be balanced by a dierence between the sum of the cost
rates of all inlet uid ows and the cost rates of all outlet uid ows. It should be noted
that cost balancing as presented by Eq. (11.63) has to be performed for a heat exchanger
as a component in a system. Note also that cost rates do not obey the conservation
principle.
{
For simplicity, we ignore here the installation cost, which can be comparable to capital and operating costs for
some shell-and-tube heat exchangers and PHEs, and can be added as a third term on the right-hand side or can be
included in the capital cost..
792
Before one attempts to perform an optimization, a careful utilization of a commonsense second law of thermodynamics approach to the design of the entire system has to be
conducted (Sama et al., 1989; Sama, 1995b). This approach is based on a selection of
thermodynamic rules such as the ones given in the summary of this chapter based on the
results of Section 11.3 and this section for a heat exchanger design. If input cost rates of
the form of Eq. (11.62) are known, and if both the capital investment and operating
expenses are known, Eq. (11.63) can be used directly to determine the output cost rate of
a stream. Once determined, a cost rate balance can be combined with the cost rate
balances of other components and used to formulate an objective function (Bejan
et al., 1995) for a system. Such an objective function combines thermodynamics and
economics, exploiting the concepts of exergy and/or entropy generation for a system;
that is the reason this branch of engineering is referred to as thermoeconomics (Bejan et
al., 1995). Equations (11.62) and (11.63) clearly show how the cost rate balance can be
related to the exergy rate balances (assigning a dollar sign to exergy ows) for a heat
exchanger as a component of a system. Determination of cost per unit exergy ow is
dicult and a single cost model could not be applied. This complex thermoeconomic
topic is beyond the scope of our presentation.
A practical methodology to account for the cost of irreversibilities in a heat exchanger
as a stand-alone unit has been developed by London (1982). This methodology has at its
core an identication of the individual costs of irreversibilities incurred during heat
exchanger operation, which is in fact related to both the capital (initial design) and operating costs. This is because some small irreversibilities may be very expensive (such as
airow friction irreversibility for an automotive radiator), and some large irreversibilities
may be less expensive (such as liquid-side friction irreversibility in a gasliquid heat
exchanger). Londons approach does not use the exergy rates explicitly but does use the
exergy destruction rate, referred to as the energy measure of the irreversibilities. Individual
energy measures of irreversibility are determined by multiplying individual entropy
generation rates with a temperature-weighting factor, Tref . Total irreversibility destruction
is related to total entropy generation through the relationship given by Eq. (11.55). Details
of Londons procedure are also presented by London and Shah (1983).
With the thermodynamic background developed so far, we can outline a procedure to
evaluate in monetary terms the various irreversibilities in a heat exchanger as shown in
Fig. 11.14. For example, ve irreversibilities of signicance in a heat exchanger may be
the nite temperature dierence between hot and cold uids, uid mixing at the
exchanger ports (if applicable), pressure drops on the hot and cold sides, and
irreversibility associated with heat leakage between the exchanger and the environment.
The total energy measure of irreversibility normalized by the heat exchanger duty is then
given by
d I_irr
I_
I_
I_
I_
irr
irr irr
11:64
irr
q T
q mixing q p;h q p;c q leak
q
where I_irr;i Tref S_ irr;i , with the subscript i representing individual irreversibilities of Eq.
(11.64). These are summarized in Table 11.3 for an ideal gas or an incompressible liquid,
all in measurable system operation quantities (temperatures, pressures, mass ow rates,
etc.). These irreversibilities have an energy monetary value that is dependent on the
system in which a particular heat exchanger is used. Assigning a monetary value to
irreversibility is analogous to assigning a cost rate to exergy rates as in Eq. (11.62).
Once the monetary values or costs of various irreversibilities are determined, the analyst
793
is in the position of deciding which particular irreversibilities are most costly and should
rst be reduced for a cost-eective heat exchanger. Hence, the industrial approach to the
design of a heat exchanger is to reduce the most costly irreversibilities rather than reducing all irreversibilities in a heat exchanger [i.e., not to minimize d of Eq. (11.64), but to
minimize only those irreversibilities on each uid side that are the most expensive]. Note
that reducing one irreversibility may increase or require the addition of another, or
involve an increase in capital investment. These considerations lead to the development
of trade-o factors, useful criteria for arriving at an optimum heat exchanger design as a
component. Thus, coming back to Fig. 11.14, the design and optimization of a heat
exchanger involves the process of evaluating various irreversibilities on an entropy
basis and then converting them into an energy basis (either in the form of exergy ows
or energy measures of irreversibility), assigning monetary values, minimizing the most
costly irreversibilities and as a result developing trade-o factors, and continuing this
process until the optimum heat exchanger is developed. To understand this process,
London (1982) and London and Shah (1983) have provided a detailed example of a
condenser in a thermal power plant with a clear demonstration of the accounting
procedure above, including how to develop the trade-o factors. Problems 11.17 and
11.18 are based on this example. Even more intricate problems of second-law-based
thermoeconomic optimization of exchangers (including both single- and two-phase
heat exchangers) are proposed by Zubair et al. (1987), estimating the economic value
of entropy generation in the heat exchanger caused by nite temperature dierences and
pressure drop. That method permits an engineer to trade the cost of entropy generation
on each uid side of the heat exchanger against its capital expenditure.
Now let us relate the cost rate analysis for a system [Eqs. (11.62) and (11.63)] with a
similar analysis based on minimizing the most costly irreversibilities using the approach
of Fig. 11.14 and the associated Eq. (11.64). The exergy cost analysis of a system, based
on Eq. (11.63), includes all irreversibilities present individually within all components
[say, a compressor, a condenser, an expansion device, an evaporator, and an accumulator/dehydrator in an automotive air-conditioning system, for which cost rates are
calculated using Eq. (11.62)] if the uid variables are determined at the boundaries of
each component and each component is considered to be a black box. Hence, optimization of the system is performed by taking into account the irreversibility losses of all
components without distinguishing various components of the total irreversibility in
794
Heat leak
j c or h
Hot fluid:
Tref 1 m_ p
Tref 1 m_ p
I_irr
I_irr
;
cold
fluid:
q p;h Th;lm q
q p;c Tc;lm q
h
c
qleak
Tref
qleak Tref
I_irr
I_irr
; from environment:
To environment:
1
1
q leak
q
Th
q leak
q
Tc
Tj;i Tj;o
1
1
I_irr
where Tj;lm
Tref
q T
lnTj;i =Tj;o
Tc;lm Th;lm
Tref m_ cp
Tc;o
Tc;o
I_irr
1 ln
q mixing
Tref
Tref
q
795
each component. Equation (11.64) is used to minimize the most costly irreversibilities to
produce a cost-eective heat exchanger, and it does not necessarily lead to an optimum
design of the system as a whole. So a system-based optimization would use an objective
function based on Eq. (11.63), while a heat exchanger component optimization would be
based on an analysis of individual irreversibilities. It should be noted that the cost rates in
Eq. (11.63) can be calculated using Eq. (11.62) in such a way as to determine exergy rates
[using Eq. (11.53)] by excluding from the entropy change all less costly irreversibility
contributions. In such a way, the system-based optimization will include only the most
costly irreversibilities in the same manner as would be done by using an individual
irreversibility calculation for each component. However, the opposite is not possible;
an optimization based on Eq. (11.64) only cannot provide an optimum system as a whole.
Example 11.9 Determine the annual cost of exergy destruction in a heat exchanger
caused by nite temperature dierences. The heat exchanger heats a cold uid stream
having a heat capacity rate of 5:8 106 W=K from 198C (292 K) to 278C (300 K). The
hot uid reduces its temperature from 308C (303 K) to 258C (198 K). The reference
temperature (surroundings) is 178C (290 K). The annual capital cost of the equipment
involved is 103 c=kW of energy at 10% annual interest rate. The average yearly energy
cost is 150 c=kW yr.
SOLUTION
Problem Data and Schematic: The terminal temperatures of the uids entering and
leaving the heat exchanger and the cold-uid heat capacity rate are as follows:
Tc;i 292 K, Tc;o 300 K, Th;i 303 K, Th;o 298 K, and Cc 5:8 106 W=K. The
reference surrounding temperature is To 290 K (178C). The annual capital cost of
the equipment is Ceqp 1000 c=kW and the annual interest rate r 0:1/yr. The average
annual cost of energy is Cq 150 c=kW yr. See Fig. 11.13 as a representative sketch.
Determine: The annual cost of exergy destruction in the exchanger caused by nite
temperature dierences.
Assumptions: All appropriate assumptions of Section 3.2.1 are invoked here.
Analysis: The exergy destruction is given by Eq. (11.55):
d Tref S_ irr
Also, according to Eq. (11.64), taking into account only the nite temperature dierence
contribution
d I_irr
2
q
q
T
In other words, the energy measure of irreversibility in the heat exchanger caused by
nite temperature dierences is equal to corresponding exergy destruction since no other
irreversibility contributions are included. The entropy generation caused by nite temperature dierences is given by Eq. (11.30):
Tlm;h Tlm;c
S_ irr q
Tlm;h Tlm;c
796
where
Tlm; j
Tj;o Tj;i
lnTj;o =Tj;i
Equation (4) provides the magnitudes of the logmean temperatures of the two uids
(equal to the logarithmic mean between the inlet and outlet temperatures of each uid)
with j h and c.
Inserting problem data into Eq. (4), we get
Tlm;h
Th;o Th;i
298 303 K
300:5 K
ln Th;o =Th;i
ln298=303
Tlm;c
Tc;o Tc;i
300 292 K
296:0 K
ln300=292
ln Tc;o =Tc;i
1
1
296:0 K 300:5 K
2347 W=K
The exergy destruction caused by nite temperature dierences, from Eq. (1), is
d Tref S_ irr 290 K 2347 W=K 0:681 106 W
Now, keeping Eq. (2) in mind and taking the cost data into account, we can calculate the
cost of this exergy destruction:
C dCeqp r Cq 0:681 106 W 1 0:1 0:15c=W yr 0:17 106 c=yr
Discussion and Comments: This example illustrates how one can determine in a simplied
way a cost value of the lost work (exergy destruction) caused by irreversibilities in a heat
exchanger. In this example, only the nite temperature dierence irreversibility contribution is considered. Problems 11.17 and 11.18 include all other relevant irreversibility
contributions for a representative heat exchanger application.
11.7 PERFORMANCE EVALUATION CRITERIA BASED ON THE SECOND
LAW OF THERMODYNAMICS{
All performance evaluation criteria for heat transfer surfaces presented in Section 10.3.2
are based on the rst law of thermodynamics (i.e., energy and mass balances). In a
dierent approach to a performance evaluation criterion, we introduce the thermodynamic quality of heat transfer and uid ow processes to evaluate the heat transfer surface performance. Such an evaluation requires an assessment of the irreversibility level of
heat transfer and uid ow phenomena. Hence, a combined rst and second law of
{
The body of knowledge usually called the second law of thermodynamics analysis always involves both the rst
and second laws of thermodynamics. However, it is customary to name the product of such an analysis by
indicating the second law of thermodynamics only.
797
thermodynamics analysis becomes necessary. Here, we emphasize only the possible formulation of a PEC that reects these considerations. For details of this approach and a
review of the literature, refer to Bejan (1988).
The second law of thermodynamics performance evaluation criteria are based on objective functions that include both heat transfer and pressure drop (uid friction) irreversibilities and hence gauge the combined eect of these irreversibilities. Separation of the
two irreversibilities can subsequently be performed if necessary. Note that heat transfer
and pressure drop irreversibilities should be translated into costs separately because the
unit costs of these irreversibilities are in general not equal.
The entropy rate balance for a duct/channel control volume (i.e., a ow passage of a
heat exchanger) denes total entropy generation between the uid inlet and outlet as
follows:
S_ irr S_ irr;T S_ irr;p S_ irr;T 1
11:65
where the S_ irr;T and S_ irr;p terms denote the contributions to overall entropy generation
incurred by either nite temperature dierence between the uid and the wall (T) or
pressure drop (p) (see Problem 11.19). The irreversibility distribution ratio,
S_ irr;p =S_ irr;T expresses, by denition, the trade-o between the two contributions.
An explicit form of Eq. (11.65) depends on heat transfer and uid ow conditions and
idealizations (e.g., boundary conditions, free-ow area geometry, ow regime, selection
of dimensionless parameters). One such expression in terms of dimensionless parameters
for a constant-cross-section duct and constant-property uid can be written as
S*
Nq2 Dh =L
4j Pr2=3
2f Ec
Dh =L
11:66
where the dimensionless heat transfer rate Nq and Eckert number Ec are dened as
Nq q=m_ cp Tm and Ec u2 =cp Tm Jgc , respectively. Equation (11.66) represents a
dimensionless form of the corresponding entropy generation expression derived by
Bejan (1988). These dimensionless groups are conveniently dened using uid bulk
mean temperature Tm [physical modeling involving the derivation of Eq. (11.66) is the
subject of Problem 11.20]. Note that the dimensionless group Ec is usually dened based
on a temperature dierence, not the bulk temperature. Such a denition can easily be
introduced in Eq. (11.66), leading to the introduction of an additional temperature ratio
parameter. For simplicity, the Ec number is dened as given above. Equation (11.66)
assumes constant and known heat transfer and mass ow rates.
For a specied constant mass ow rate m_ and given duct length L, a change in the
hydraulic diameter Dh causes a change in S*. From the algebraic structure of Eq. (11.66)
and Reynolds analogy (see Section 7.4.5), it becomes clear that the two terms on the
right-hand side of Eq. (11.66) have opposite trends with respect to a change in Re (which
aects j and f factors) or the hydraulic diameter. Thus, this objective function [i.e., Eq.
(11.66)] may have an extremum (a minimum.) It has been recognized that what is good
for the reduction of friction irreversibility (by decreasing surface area) is apparently bad
for the reduction of nite temperature irreversibility (i.e., for an increase in exchanger
eectiveness), and vice versa. An optimum trade-o between these two inuences may
exist. Consequently, such an objective function may be used as a basis for dening a new
thermodynamic performance evaluation criterion. An optimum geometry (or ow
regime) can ultimately be dened for a given selection of characteristic parameters. We
can illustrate this approach by the following example.
798
Example 11.10 Thermoeconomic optimization of a large energy system requires minimization of irreversibility costs of a plant. Within the scope of that analysis, a heat
exchanger designer must decide which geometry of a heat exchanger passage cross
section would contribute the least to the overall irreversibility. The length and freeow area of the duct representing the passage are known and xed, as well as the uid
(air) inlet thermal state and mass ow rate. Temperature of the heating uid is constant
and equal to 1008C (373 K). The wall thermal resistance can be neglected. Duct geometry
options include (1) square, (2) rectangular (aspect ratio
* 18), and (3) circular cross
sections. Determine which cross section would be best from the point of view of irreversibility minimization. Compare the ndings with an analysis of the magnitude of heat
transfer and pressure drop for each of the duct shapes. The following data are available.
Free-ow area of the cross section is 5 103 m2 , the duct length 5 m, the mass ow rate
of air 5 102 kg=s, and the uid inlet temperature 300 K. The thermophysical properties of the air are as follows: density, 1.046 kg/m3; specic heat at constant pressure,
1.008 kJ/kg K; dynamic viscosity, 2:025 105 Pa s; Prandtl number, 0.702; and thermal conductivity, 2:91 102 W=m K.
SOLUTION
Problem Data and Schematic: The duct geometries for this problem are shown below
along with the input data for the problem.
799
TABLE E11.10 Thermal and Hydraulic Characteristics and Entropy Generation of Various
Ducts
Variable or Parameter
Equation
Square
4Ao =P
m_ =Ao
GDh =
Nusq Nurect 0:024Re0:8 Pr0:4
Nucirc 0:023Re0:8 Pr0:4
j St Pr2=3 Nu Pr1=3 =Re
kNu=Dh
PL
hA=m_ cp
1 eNTU
q "m_ cp Tw Ta;i
Ta;i q=m_ cp
Ta;i Ta;o =2
frect 0:0791Re0:25 1:0875 0:1125
*
fcirc 0:00128 0:1143Re1=3:2154
p 4fLG2 =2gc Dh
q=m_ cp Tm
u2 =cp Tm Jgc
Nq2 Dh =L=4j Pr2=3
2f Ec=Dh =L
* Sp
*
ST
Dh (m)
G (kg/m2 s)
Re
Nu
j
h (W/m2 K)
A (m2)
NTU
"
q (W)
Ta;o 8C
Tm 8C
f
pPa
Nq
Ec
*
ST
*
Sp
S*
Rectangular
2
Circular
7:071 10
10
34,919
89.78
4:444 10
10
21,946
61.92
2
7:979 102
10
39,402
94.77
2:893 103
36.94
1.4142
1.038
0.6458
2,376
74.1
50.6
5:642 103
3:174 103
40.55
2.25
1.83
0.8396
3,089
88.3
57.7
6:976 103
2:706 103
34.56
1.25
0.87
0.5810
2,138
69.4
48.2
5:534 103
76.3
0.1456
2:799 104
2:043 102
2:240 104
2:065 102
150.0
0.1852
2:738 104
1:899 102
4:292 104
1:942 102
66.3
0.1320
2:820 104
2:021 102
1:963 104
2:041 102
2
L
4Tm m_ cp St Ao cp Tm Jgc Dh
From the last line of Table E11.10 it is obvious that the rectangular duct generates the
minimum entropy. Therefore, the rectangular duct would be considered as most favorable from an entropy-generation point of view.
Discussion and Comments: To understand the inuence of various sources of irreversibility on duct performance, the results of the analysis are summarized below. For the
three duct geometries, the following quantities are compared: heat transfer rate, pressure
drop, dimensionless entropy generation caused by temperature dierence, dimensionless
entropy generation caused by pressure drop, and total dimensionless entropy generation.
The results are presented qualitatively in terms of the highest heat transfer rate, the
lowest pressure drop, and the lowest entropy generations, all marked with the symbol
::
^ . The lowest heat transfer rate, the highest pressure drop, and the highest irreversibility
::
::
_ . Finally, the medium values are marked
.
are each marked with the symbol
Variable or Parameter
q
p
S*T
*
Sp
S*
Square
Rectangular
Circular
::
::
::
_
::
:
:
_
::
^
::
_
::
^
::
_
::
^
::
_
::
^
::
:
:
^
:
:
800
It is obvious that the rectangular duct has the best performance in terms of both heat
transfer rate and irreversibility level. The penalty for high heat transfer performance is
paid by a very high pressure drop. It is interesting to note that the worst geometry from
the heat-transfer-rate point of view is the circular duct, but from the entropy-generation
point of view, it is a square duct. The reason for that is related to the dierent orders of
magnitude of the entropy generation caused by temperature dierence and pressure drop.
If the ultimate goal is to reach the minimum entropy generation for a given geometry,
say a circular duct, an optimization based on the objective function given by Eq. (11.66)
has to be performed. Solving @S*=@Re will lead to the determination of an optimum
Re, or the optimum duct hydraulic diameter for a given mass ow rate. For the circular
duct of this problem, this optimum diameter of the duct (the hydraulic diameter
considered as the only degree of freedom, the other variables xed) is found as very
large: 135 mm.
The optimum trade-o between heat transfer and friction (pressure drop) irreversibilities may or may not exist. Also, the existence of a minimum of entropy generation is
not always present for the range of parameters selected. It must also be emphasized that
optimization based on a thermodynamic criterion cannot be a goal per se in a design
eort related to an isolated heat exchanger. Usually, the ultimate goal is the minimization
of cost within the framework of a system analysis.
SUMMARY
In this chapter, several interdisciplinary issues are discussed. These include (1) the use
of energy and mass balances (rst law of thermodynamics only) and mathematical
modeling to obtain temperature distributions and temperature dierence distributions
in various ow arrangements, (2) the application of the rst and second laws of thermodynamics (combined) to identify irreversibility sources that have a detrimental eect on
heat exchanger performance, (3) a heuristic approach to the assessment of heat exchanger eectiveness, (4) thermodynamic analysis, including exergy and thermoeconomic
accounting for heat exchanger optimization, and (5) performance evaluation criteria
based on the minimization of entropy generation. Our goal has been to understand
why certain heat exchanger designs would lead to a higher or lower eectiveness and/
or thermodynamic (exergy) eciency than would a similar one, and to develop the skills
needed for an approach to optimization of a heat exchanger as part of a system. Finite
temperature dierences, uid mixing, and uid friction are important irreversible phenomena associated with exchanger performance. Details on these irreversibilities are
presented in the text.
The most important guidelines for the design of a heat exchanger as a component in a
system are as follows. An approach to optimum design must be based on sound engineering judgment, along with utilization of a commercial or proprietary software (if any)
with understanding. Such an approach should be performed utilizing not only the energy
balances (implied by the rst law of thermodynamics), but also entropy-generation
balances (implied by the combined rst and second laws of thermodynamics). An exergy
balance or an entropy-generation calculation has to be accompanied by economic evaluations. The components of exergy balances or total entropy-generation rates should
have assigned monetary values. Exergy cost balances may be used to dene an objective
function in a search for the optimum design of the system in which the analyzed heat
exchanger is a component. The methodology outlined in Fig. 11.14 should be adopted
for optimization of a heat exchanger in a system if the minimization of the most costly
801
REFERENCES
Ahrendts, J., 1980, Reference states, Energy, Vol. 5, pp. 667677.
Alt-Ali, M. A., and D. J. Wilde, 1980, Optimal area allocation in multistage heat exchanger systems,
ASME J. Heat Transfer, Vol. 107, pp. 199201.
Bejan, A., 1988, Advanced Engineering Thermodynamics, Wiley, New York.
Bejan, A., G. Tsatsaronis, and M. Moran, 1995, Thermal Design and Optimization, Wiley, New
York.
Bosnjakovic, F., 1965, Technical Thermodynamics, Holt, Rinehart and Winston, New York (6th
German ed., Technische Thermodynamik, Steinkopf, Dresden, 1972).
EES, 2000, F_Chart Software, Engineering Equation Solver, Middleton, WI.
Gregorig, R., 1965, Exergieverluste der Warmeaustauscher (Exergy Losses in a Heat Exchanger),
Teil 1, Reibung (Friction), Chem. Ing. Techn. Vol. 37, pp. 108116; Teil 2, Endlichen Temperaturunterschiedes (Finite Temperature Dierences), pp. 524527.
802
Gunderson, T., and L. Naess, 1988, The synthesis of cost optimal heat exchanger networks: an
industrial review of the state of the art, Comput. Chem. Eng., Vol. 12, pp. 503530.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger Publishing, Malabar, FL.
Kmecko, I., 1998, Paradoxical Irreversibility of Enthalpy Exchange in Some Heat Exchangers, M.S.
thesis, University of Novi Sad, Novi Sad, Yugoslavia.
Kotas, T. J., 1995, The Energy Method of Thermal Plant Analysis, Krieger Publishing, Melbourne,
FL.
Linnho, B., D. W. Townsend, D. Balard, G. F. Hewitt, B. E. A. Thomas, A. R. Guy, and R. H.
Marshand, 1982, User Guide on Process Integration for Ecient Use of Energy. Institution of
Chemical Engineers and Pergamon Press, Oxford.
London, A. L., 1982, Economics and the second law: an engineering view and methodology, Int. J.
Heat Mass Transfer, Vol. 25, pp. 743751.
London, A. L., and R. K. Shah, 1983, Costs of irreversibilities in heat exchanger design, Heat
Transfer Eng., Vol. 4, No. 2, pp. 5973; discussion by W. Roetzel, in Vol. 5, No. 34, 1984,
pp. 15, 17, and Vol. 6, No. 2, 1985, p. 73.
Moran, M. J., and H. N. Shapiro, 1995, Fundamentals of Engineering Thermodynamics, Wiley, New
York.
Peters, M. S., and K. D. Timmerhaus, 1980, Plant Design and Economics for Chemical Engineers,
McGraw-Hill, New York.
Sama, D. A., 1995a, Dierences between second law analysis and pinch technology, J. Energy
Resour. Technol., Vol. 117, pp. 186191.
Sama, D. A., 1995b, The use of the second law of thermodynamics in process design, J. Energy
Resour. Technol., Vol. 117, pp. 179185.
Sama, D. A., S. Qian, and R. Gaggioli, 1989, A common-sense second law approach for improving
process eciencies, Proc. Int. Symp. Thermodynamic Analysis and Improvement of Energy
Systems, Beijing, International Academic Publishing, Pergamon Press, New York, pp. 520532.
Shah, R. K., and Skiepko, T., 2002, Entropy generation extrema and their relationship with heat
exchanger eectiveness number of transfer units behavior for complex ow arrangements, Heat
and Mass Transfer 2002, Proc. 5th ISHMT-ASME Heat Mass Transfer Conf., Tata McGrawHill., New Delhi, India, pp. 910919.
Sekulic, D. P., 1990a, A reconsideration of the denition of a heat exchanger, Int. J. Heat Mass
Transfer, Vol. 33, pp. 27482750.
Sekulic, D. P., 1990b, The second law quality of energy transformation in a heat exchanger, ASME
J. Heat Transfer, Vol. 112, pp. 295300.
Sekulic, D. P., 2000, A unied approach to the analysis of unidirectional and bi-directional parallel
ow heat exchangers, Int. J. Mech. Eng. Educ., Vol. 28, pp. 307320.
Sekulic, D. P., 2003, A heuristic approach to an assessment of heat exchanger eectiveness, to be
published in Int. J. Mech. Eng. Education, Vol. 31.
Shah, R. K., and A. Pignotti, 1989, Basic Thermal Design of Heat Exchangers, Report Int-8513531,
National Science Foundation, Washington, DC.
Sontag, R. E., and G. J. Van Wylen, 1982, Introduction to Thermodynamics, Wiley, New York.
Zubair, S. M., P. V. Kadaba, and R. B. Evans, 1987, Second-law-based thermoeconomic optimization of two-phase heat exchangers, ASME J. Heat Transfer, Vol. 109, pp. 287294.
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
REVIEW QUESTIONS
803
11.1
Circle the following statements as true or false and provide detailed reasons.
(a) T F
A workable solution of a heat exchanger design problem requires
neither irreversibility analysis nor economic analysis.
(b) T F
The true meaning of the concept of temperature eectiveness cannot
be derived without invoking explicitly the second law of thermodynamics.
11.2
11.3
11.4
11.5
11.6
A heat exchanger with equal exit temperatures of the two uids is characterized
with:
(a) minimum entropy generation
(b) maximum entropy generation
(c) minimum heat exchanger eectiveness
(d) maximum heat exchanger eectiveness
11.7
11.8
804
11.9
(d) T
(e) T
(f) T
(g) T
(h) T
(i) T
(j) T
The total entropy on hot plus cold uid sides in an industrial exchanger increases
due to:
(a) heat transfer in the exchanger
(b) pressure drops in the exchanger
(c) leakage of hot uid to cold uid
(d) fouling
(e) all of these
(f) none of these
PROBLEMS
11.1
11.2
@2
2 1
@
2 ; 0 0
PROBLEMS
805
2 ; 1 e ek
K 1 expC * NTU
11.4
Consider a crossow heat exchanger with the smaller heat capacity rate
uid unmixed and the other uid mixed throughout. Show that the
heat exchanger eectiveness of this ow arrangement is given by
" 1 expMC*=C*, where M 1 expNTU. Determine the
numerical values for NTU that correspond to the maximum entropy generation
of this exchanger for C * 1 and 0.1.
11.5
11.6
Formulate corresponding reduced mathematical models for determining temperature distributions of both uids of a crossow heat exchanger using the general
model obtained in Problem 11.5. Consider the following particular cases:
(a) The uid with the larger heat capacity rate is mixed; the other uid is unmixed,
(b) The uid with the smaller heat capacity rate is mixed; the other uid is
unmixed,
(c) Both uids are mixed.
11.7
11.8
806
PROBLEMS
807
Flow Arrangement
Flow Arrangement
808
178C (290 K). The temperature of the water at the heat exchanger outlet is 278C
(300 K). The pressure drop on the coolant side of the heat exchanger (including
manifolds and connecting pipes) is 50 kPa. It is assumed that the water pump
operates with negligible losses. In addition, the exchanger (including the connecting piping) heat losses to the environment constitute approximately 2% of the
condenser heat transfer rate. Determine the magnitude and relative importance of
all irreversibilities associated with the operation of the heat exchanger.
11.18 Reconsider the analysis of Problem 11.17 (a steam electric power plant condenser), but now include the economic aspect of the entropy generation estimation.
Assign a monetary value to the entropy-generation contributions in terms of
busbar energy delivery costs. The relevant additional data are as follows:
Variable
Value
1600
80
700
85
80
700
3
4000
11.19 Show that entropy generation caused by heat transfer and uid friction associated
with ow through a duct can be described by Eq. (11.65). Elaborate explicitly all
the assumptions needed for derivation of this relationship.
11.20 A heat exchanger passage carries a constant property uid. The cross-sectional
area is Ao and the wetted wall perimeter is P. The mass ow rate of the uid is
xed. The heat transfer rate between the uid and the wall is across a mean
temperature dierence T, and it is considered to be constant along the ow
direction of a short passage under consideration. Show that dimensionless entropy
generation can be written in the form of Eq. (11.66) if the uid represents a simple
compressible substance with constant thermophysical properties. No phase
change is present.
11.21 It can be shown that dimensionless entropy generation caused by heat transfer at
nite temperature dierences and uid friction takes the following form for ow
through an isothermal duct:
2=3
2=3
# 1e4jPr
S_ irr
# 11 e4jPr L=Dh ln
#
m_ cp
L=Dh
2=3
1 f
# 1e4jPr L=Dh 1
Pr2=3 Ec ln
2=3
2 j
#e4jPr L=Dh
Using mass, energy, and entropy rate balances, show that this result is correct if
the ow is assumed to be fully developed. Demonstrate that entropy generation
may have a global minimum for a selected set of operating parameters. Perform
the analysis for both air and water as working uids.
12
One of the common assumptions in basic heat exchanger design theory is that uid be
distributed uniformly at the inlet of the exchanger on each uid side and throughout the
core. However, in practice, ow maldistribution{ is more common and can signicantly
reduce the desired heat exchanger performance. Still, as we discuss in this chapter, this
inuence may be negligible in many cases, and the goal of uniform ow through the
exchanger is met reasonably well for performance analysis and design purposes.
Flow maldistribution can be induced by (1) heat exchanger geometry (mechanical
design features such as the basic geometry, manufacturing imperfections, and tolerances), and (2) heat exchanger operating conditions (e.g., viscosity- or density-induced
maldistribution, multiphase ow, and fouling phenomena). Geometry-induced ow
maldistribution can be classied into (1) gross ow maldistribution, (2) passage-topassage ow maldistribution, and (3) manifold-induced ow maldistribution. The
most important ow maldistribution caused by operating conditions is viscosityinduced maldistribution and associated ow instability.
In this chapter, we consider geometry-induced ow maldistribution in Section 12.1
and operating conditioninduced ow maldistribution in Section 12.2. Next, mitigation
of ow maldistribution is discussed in Section 12.3. Finally, header design for compact
heat exchangers is summarized in Section 12.4.
Flow maldistribution is dened as nonuniform distribution of the mass ow rate on one or both uid sides in any
of the heat exchanger ports and/or in the heat exchanger core. The term ideal uid ow passage/header/heat
exchanger would, as a rule, denote conditions of uniform mass ow distribution through an exchanger core.
809
810
(4) shallow bundle eects,{ and (5) general equipment and exchanger system eects
(Kitto and Robertson, 1989).
The most important causes of ow nonuniformities can be divided roughly into three
main groups of maldistribution eects: (1) gross ow maldistribution (at the inlet face of
the exchanger), (2) passage-to-passage ow maldistribution (nonuniform ow in neighboring ow passages), and (3) manifold-induced ow maldistribution (due to inlet/outlet
manifold/header design). First, we discuss gross ow maldistribution. Subsequently, the
passage-to-passage ow maldistribution is addressed, followed by a few comments
related to manifold-induced ow maldistribution.
12.1.1
811
dual uniform uid streams (i.e., having an N-step function velocity distribution). The
other uid side is considered as having uniform ow distribution for such an analysis. If
ow nonuniformity occurs on both uid sides of a counterow or parallelow exchanger,
the exchanger is divided into a sucient number of subexchangers such that the ow
distributions at the inlet on both uid sides are uniform for each subexchanger. For all
other exchangers, the solution can only be determined numerically, and the solutions of
Sections 12.1.1.1 and 12.1.1.2 are not valid in that case.
12.1.1.1 Counterow and Parallelow Exchangers. In this section, we derive an
expression for the exchanger eectiveness and hence heat transfer performance for
counterow and parallelow exchangers having an N-step velocity distribution function
on the uid 1 side and perfectly uniform ow distribution on the uid 2 side, as shown
in Fig. 12.1a. Here uid 1 can be either the hot or cold uid, and in that case, uid 2 will
be the cold or hot uid. Subsequently, we apply this analysis to a heat exchanger having
a two-step velocity distribution function at the inlet.
Heat Transfer Analysis. Let us consider a counterow exchanger with an N-step inlet
distribution function of uid 1, shown in Fig. 12.1a. The same analysis would be valid
for a parallelow heat exchanger. Fluid 2 is considered uniform. We may model this
exchanger as an array of N subexchangers, each obeying the standard assumptions of
Section 3.2.1. Hence,
q
N
X
12:1
qj
jA
where q represents the total heat transfer rate, and qj , j A, B, . . . , N, the fractions of
heat transfer rate in N hypothetical subexchangers, each having uniform mass ow rates
on both sides, as shown in Fig. 12.1b. The assumptions invoked, including the auxiliary
ones introduced above, lead to the following results:
q C1 T1;i T1;o
P1
T1;i T1;o
T1;i T2;i
and qj C1; j T1;i T1;o j
and
P1; j
T1;i T1;o j
T1;i T2;i
j A; B; . . . ; N
12:2
j A; B; . . . ; N
12:3
FIGURE 12.1 Idealized two-step function ow nonuniformity on uid 1 side and uniform ow on
uid 2 side of a counterow exchanger.
812
Substituting Eqs. (12.2) and (12.3) into Eq. (12.1) and rearranging, the expression for
uid 1 temperature eectiveness becomes
P1
N
1 X
C P
C1 jA 1; j 1; j
12:4
where
C1
N
X
12:5
C1; j
jA
Note that Eqs. (12.4) and (12.5) are valid for a maldistributed uid regardless of
whether it is hot or cold, Cmin or Cmax , or denoted as 1 or 2. The subscript 1 in these
equations may be replaced by a designator of the maldistributed stream, say as in Pms and
Cms . The temperature eectiveness of uid 1 for each of the subexchangers of Eq. (12.4) is
computed knowing individual NTU and heat capacity rate ratio:
C1; j
P1; j P1; j NTU1; j ;
C2; j
j A; B; . . . ; N
12:6
where P1; j on the right-hand side of Eq. (12.6) is computed for each exchanger using
the expression provided in Table 3.6 as follows:
P1; j
8
1 expNTU1; j 1 R1; j
>
>
>
< 1 R expNTU 1 R
counterflow
>
1 expNTU1 1 R1; j
>
>
:
1 R1; j
parallelflow
1; j
1; j
1; j
12:7
Application of Eq. (12.6) requires the values of NTU1; j , C1; j , and C2; j for each subexchanger. To determine these variables, we should invoke the standard assumptions of
Section 3.2.1 to get the free-ow area and heat capacity rate ratio as
Ao
N
X
12:8
Ao; j
jA
C1; j m_ 1; j u1; j Ao; j
C1
u1
Ao 1
m_ 1
j A; B; . . . ; N
12:9
C2
u1
Ao 2
j A; B; . . . ; N
12:10
Note that the sets of relations given by Eq. (12.9) and (12.10) may be reduced by one
equation each by utilizing Eq. (12.5) for uid 1 (and similarly for uid 2).
The number of heat transfer units and the capacity rate ratios for Eq. (12.7) can be
determined from their denitions as follows:
NTU1; j
R1; j
C1; j
C1 A1 C1; j
C1; j
C2; j
j A; B; . . . ; N
813
12:11
12:12
Note also that Dh 4Ao L=A with Dh and L identical for all subexchangers
j A; B; . . . ; N for the counterow/parallelow heat exchanger. A1; j =A1 of Eq. (12.11)
is obtained then from the denitions of Dh as
Ao; j
Aj
j A; B; . . . ; N
12:13
Ao 1
A 1
In addition,
Ao; j
Ao
Ao; j
Ao 2
j A; B; . . . ; N
12:14
P1;ideal P1
P1;ideal
or "*
"ideal "
"ideal
12:15
where P1;ideal represents the temperature eectiveness for the case of having no ow
maldistribution.
The inuence of gross ow maldistribution is shown in Fig. 12.2 for a balanced
(C * 1) counterow heat exchanger in terms of "* for a two-step inlet velocity
distribution function (i.e., for two subexchangers). For a particular value of umax =um
FIGURE 12.2 Performance deterioration factor "* for a balanced heat exchanger, C* 1,
N 2. (From Shah, 1981.)
814
and given NTU, we nd the greatest reduction in the heat exchanger eectiveness occurring when two-step function ow maldistribution occurs in equal ow areas (50 : 50%).
The eect of ow maldistribution increases with NTU for a counterow exchanger. Note
that the reduction in the temperature eectiveness P1 of Eq. (12.4), obtained using Eq.
(12.15), is valid regardless of whether the maldistributed uid is the hot, cold, Cmax , or
Cmin uid.
We can idealize an N-step function velocity distribution into an equivalent two-step
function velocity distribution. Based on the analysis of passage-to-passage ow maldistribution presented in Section 12.1.2, it is conjectured that the deterioration in the
exchanger eectiveness is worse for the two-step function velocity distribution. Hence,
conservatively, any ow maldistribution can be reduced to a two-step function, and its
eect can readily be evaluated on the exchanger eectiveness, which will represent the
highest deterioration.
As indicated above, the discussion of the eect of the gross ow maldistribution in
this section refers to a heat exchanger with counterow arrangement and balanced ow
(C* 1). Hence, the increased eect of ow maldistribution with increasing NTU is
valid only for this special situation. If either C * 6 1 or if the ow arrangement is parallelow, the inuence of ow maldistribution may decrease with increasing NTU. This can
be determined from Eq. (12.15), assuming the validity of the appropriate eectiveness
NTU relationships for each subexchanger and for the heat exchanger as a whole. Also,
the actual ow rate conditions in most practical cases would not correspond to a
balanced heat exchanger case.
Pressure Drop Analysis. There is no rigorous theory available for predicting a change in
the pressure drop due to ow maldistribution in the exchanger. This is because for
nonuniform ow distribution, the static pressures at the core inlet and outlet faces will
not be uniform, and hence, constant pressure drop across the core is not a valid
assumption. The following is a suggested approximate procedure. This approach is
not based on a rigorous modeling of the actual ow conditions and must be used
very cautiously. Consider a two-step function velocity distribution at the core inlet
on uid 1 side as shown in Fig. 12.1a for N 2. Subexchangers in Fig. 12.1b for
N 2 are in parallel. Using Eq. (6.28), evaluate the pressure drop pj for a specic
subexchanger which has the highest uid velocity in the ow passages. Also compute
puniform for uid 1 considering the ow as uniform at the core inlet in Fig. 12.1.
Therefore, as a conservative approach, this largest pj (i.e., pmax ) will be the pressure
drop on the uid 1 side having imposed ow nonuniformity. The increase in pressure
drop due to ow nonuniformity is then
pincrease pmax puniform
12:16
It should be emphasized that the entrance and exit losses in addition to the core friction
contribution will be higher (in the evaluation of pmax ) than those for uniform ow.
If ow nonuniformity occurs on both sides of an exchanger, the procedure outlined
above is applied to both sides, since the pressure drops on both sides of a two-uid
exchanger are relatively independent of each other, except for the changes in uid density
due to heat transfer in the core. Hence, the analysis above is applicable to any ow
arrangement.
Example 12.1 A counterow heat exchanger has a severe ow maldistribution due to
poor header design. On the uid 1 side, 25% of the total free-ow area has the ow
815
velocity 50% larger than the mean ow velocity through the core as a whole. The number
of heat transfer units of the heat exchanger is NTU1 3. The total heat capacity rates
through the exchanger are nearly the same (i.e., the heat exchanger is balanced).
Determine the reduction in the temperature eectiveness of uid 1 and an approximate
increase in the pressure drop due to ow maldistribution. Assume fully developed
laminar ow on both uid sides (i.e., U remains constant).
SOLUTION
Problem Data and Schematic: A schematic of the heat exchanger under consideration is
similar to that in Fig. 12.1 with only A and B subexchangers. The following data are
known.
NTU1 3
u1;A 1:5u1
C1 C2
Ao;A 0:25Ao;1
0:25
Ao 1 A1
Ao;A
Ao;A
0:25
Ao 1
Ao 2
Also,
Ao;B
A1;B
0:75
Ao 1 A1
Ao;B
Ao;B
0:75
Ao 1
Ao 2
u1;A
1:5
u1
0:25
1
1 0:25 0:75
C2
Ao 2
Ao 1
C2
C2
816
3 0:25
2:00
C1;A
C1 A1 C1;A
0:375
NTU1;B
3 0:75
3:60
C1;B
C1 A1 C1;B
0:625
The heat capacity rate ratios, required for eectiveness calculations, are computed
using Eq. (12.12) as follows:
R1;A
C1;A C1;A C2 C1
1
0:375
1 1:50
C2;A
C1 C2;A C2
0:25
R1;B
C1;B C1;B C2 C1
1
0:625
1 0:8333
0:75
C2;B
C1 C2;B C2
1 eNTU1;A 1R1;A
1 exp2:001 1:50
0:5584
NTU
1R
1;A
1;A
1
1:5 exp2:001 1:50
1 R1;A e
P1;B
1 eNTU1;B 1R1;B
1 exp3:601 0:8333
0:8315
1 R1;B eNTU1;B 1R1;B 1 0:8333 exp3:601 0:8333
C1;A
C1;B
P1;A
P 0:375 0:5584 0:625 0:8315 0:7291
C1
C1 1;B
NTU1
3
0:750
1 NTU1 1 3
Finally, the quantitative measure of the reduction in the eectiveness due to maldistribution is [see Eq. (12.15)]
P*1
0:0279
P1;ideal
0:750
Ans:
Discussion and Comments: From the results, it becomes clear that a relatively large ow
maldistribution on the uid 1 side in this particular case causes a deterioration of the
temperature eectiveness of approximately 2.8%. With all other parameters xed, a heat
exchanger with high NTU will suer more pronounced eectiveness deterioration (see
Fig. 12.2).
817
12.1.1.2 Crossow Exchangers. A direct extension of the approach used for a counterow/parallelow exchanger to that for a crossow exchanger with dierent combinations of uid mixing/unmixing on each uid side is not necessarily straightforward.
Only when ow nonuniformity is present on the unmixed uid side with the other uid
side as mixed can a simple closed-form solution be obtained, as outlined next.
MixedUnmixed Crossow Exchanger with Nonuniform Flow on the Unmixed Side. Let
us consider a single-pass crossow exchanger having the unmixed uid (uid 1)
maldistributed. The inlet velocity distribution is represented with an N-step function
(Fig. 12.3). Fluids 1 and 2 can be arbitrarily hot and cold, or vice versa.
The total heat transfer rate in the exchanger is given by
q
N
X
qj
12:17
jA
where the qj represent individual heat transfer rates/enthalpy rate changes as follows{:
qA P1;A C1;A T1;i T2;i C2 T2M;A T2;i
qB P1;B C1;B T1;i T2M;A C2 T2M;B T2M;A
..
.
12:18
12:19
FIGURE 12.3 Idealized two-step function ow nonuniformity on the unmixed uid 1 side and
uniform ow on the mixed uid 2 side of a crossow exchanger.
{
For the sake of simplied notation, the absolute value designators are omitted.
818
Our objective is to determine the relationship between the uid 1 temperature eectiveness P1 and the temperature eectiveness and heat capacity rates of the subexchangers of Fig. 12.3b. From Eq. (12.19), we get
P1
q
C1 T1;i T2;i
12:20
Replacing q in Eq. (12.20) with q from Eq. (12.17), and utilizing the relationships
provided by Eq. (12.18), we get
P1
T1;i T2M;A
T1;i T2M;N1
1
12:21
P1;A C1;A P1;B C1;B
P1;N C1;N
T1;i T2;i
T1;i T2;i
C1
Temperature dierence ratios in Eq. (12.21) can be eliminated by manipulating relationships from Eq. (12.18) as follows:
T1;i T2M;A
P1;A C1;A
1
T1;i T2;i
C2
T1;i T2M;B
P1;A C1;A
P1;B C1;B
1
1
T1;i T2;i
C2
C2
...
Y
T1;i T2M;N1 N1
P1;k C1;k
1
T1;i T2;i
C2
kA
12:22
j1
N
Y
X
P1;k C1;k
1
P1;A C1;A
1
P1; j C1; j
C2
C1
jB
k1
12:23
To reemphasize, the uid 1 side is unmixed and the uid 2 side is mixed for the expression
of a crossow exchanger above. The temperature eectiveness of the ideal heat exchanger
of Fig. 12.3 as a whole, and those of the subexchangers, can be expressed in terms of
corresponding heat capacity rate ratios and numbers of transfer units as follows (see
Table 3.6):
P1;ideal
P1; j
C2
C
1 exp 1 1 eNTU1
C1
C2
T1;i T1;o j
C1; j
C
2 1 exp
1 eNTU1; j
T1;i T2;i
C1; j
C2
12:24
j A; B; . . . ; N
(12.25)
where
NTU1
UA
C1
NTU1; j
UAj
C1; j
j A; B; . . . ; N
12:26
819
P1;A C1;A
1
P1;A C1;A P1;B C1;B 1
C2
C1
12:27
820
solutions for the eects of two-dimensional ow nonuniformities on thermal performance and pressure drop in crossow plate-n compact heat exchangers.
Example 12.2 Analyze a crossow heat exchanger with uid 1 unmixed and uid 2
mixed having pronounced maldistribution on uid 1 side and NTU1 3. The total
heat capacity rates of the two uids are nearly the same. Determine the temperature
eectiveness of the maldistributed uid 1 if 25% of the total free-ow area has the ow
velocity 50% larger than the mean velocity through the core on the uid 1 side corresponding to the uniform ow case.
SOLUTION
Problem Data and Schematic: A schematic of the heat exchanger under consideration is
similar to that in Fig. 12.3 with only A and B subexchangers. The following data are
known:
NTU1 3
u1;A 1:5u1
C1 C2
Ao;A 0:25Ao;1
C1;B
0:625
C1
NTU1;A 2:00
NTU1;B 3:60
Now P1;A and P1;B are computed using Eq. (12.25) as follows after incorporating
C2 C1 :
C1;A
C
P1;A 1 1 exp
1 eNTU1;A
C1;A
C1
1
0:7385
1 exp 0:375 1 e2:00
0:375
C1;B
C
1 1 exp
1 eNTU1;B
C1;B
C1
P1;B
1
1 exp 0:625 1 e3:60
0:7288
0:625
821
Now, compute the temperature eectiveness of uid 1 given by Eq. (12.27) as (with
C1 C2
P1
P1;A C1;A
C1;A
C1;B
P1;A C1;A
1
P1;A
P1;A C1;A P1;B C1;B 1
P1;B
1
C1
C1
C1
C1
C1
Ans:
This actual eectiveness is, indeed, smaller than the one calculated for an idealized
situation, 0.6063 vs. 0.6133. Finally, the fractional deterioration in the temperature
eectiveness is given by Eq. (12.15) as
P*
0:0114
P1;ideal
0:6133
822
precisely when small dimensions are involved [e.g., a rotary regenerator with
Dh 0:5 mm (0.020 in.)]. Since dierently sized and shaped passages exhibit dierent
ow resistances and the ow seeks a path of least resistance, a nonuniform ow through
the matrix results. This phenomenon usually causes a slight reduction in pressure drop,
while the reduction in heat transfer rate may be signicant compared to that for nominal
(average) size passages. The inuence is of particular importance for continuous-ow
passages at low Re (i.e., laminar ow) as found in compact rotary regenerators. For a
theoretical analysis for passage-to-passage ow maldistribution, the actual nonuniform
surface is idealized as containing large, small, and/or in-between size passages (in parallel) relative to the nominal passage dimensions. The models include (1) a two-passage
model (London, 1970), (2) a three-passage model, and (3) an N-passage model (Shah and
London, 1980). Although triangular and rectangular passage cross sections have been
studied, similar analysis can be applied to any cross-sectional shapes of ow passages.
The analysis to follow can also be utilized for analyzing ow maldistribution in
viscous oil cooler with constant-wall-temperature boundary conditions (i.e., condensation or vaporization taking place on the other uid side). See Section 12.2.1 for further
details. Let us rst dene the two-passage-model ow nonuniformity. From the methodological point of view, this approach is the most transparent and oers a clear idea of
how the modeling of ow nonuniformity can be conducted. Also, the two-passage model
predicts a more detrimental eect on heat transfer and pressure drop than that of an Npassage (N > 2) model.
12.1.2.1
Two-Passage Model. Let us consider that a heat exchanger core characterizes ow nonuniformity due to two dierent ow cross sections diering in either (1) cross-section
size of the same passage type, (2) dierent cross-sectional shapes of ow passages, or
(3) a combination of both. The two most common types of idealized passage-to-passage
nonuniformities are plate spacing and n spacing, shown in Fig. 12.4a and b, respectively. For the analysis, the actual heat exchanger core will be assumed to be a collection of two (or more) distinct sets of uniform ow passages, passages 1 and passages 2
FIGURE 12.4 Two-passage nonuniformity model: (a) plate-spacing nonuniformity; (b) nspacing nonuniformity. Note that passages dier in size. The nominal size of the passage may be
large, small, or in between, depending on how it is dened. (From London, 1968.)
823
(or N passages). Our objective here is to determine the reduction in heat transfer and
pressure drop due to this passage-to-passage ow nonuniformity.
The following assumptions are invoked for setting up the model.
. Flow is hydrodynamically and thermally fully developed (Nu constant,
f Re constant).
. Thermophysical properties of the uids are constant and uniform.
. Entrance and exit pressure losses are negligible (the core friction component is
dominant).
. Static pressures are constant and uniform across the cross section at the entrance
and exit of this multipassage exchanger.
. The total ow rate through all nonuniform ow passages is identical to that going
through all nominal ow passages.
. The lengths of all ow passages are the same.
The pressure drop for all ow passages (regardless of the size, shape, and distribution
of ow passages) will be the same in the core based on the fourth assumption above:
pj pi po j
12:28
where
pj fj
4L
Dh
!
j
m u2m
2gc
!
j 1; 2; n
12:29
where j denotes the ow passage type. Invoking the denitions of the Reynolds number
and mass ow rate, Eq. (12.29) is regrouped as
pj
2L f Re
m_
gc Ao D2h j j
12:30
For a two-passage model, applying Eq. (12.30) for j 1 and 2 and taking the ratio
and rearranging, we get
m_ 1 f Re2 Dh;1 2 Ao;1
12:31
since p1 p2 from Eq. (12.28). Equation (12.31) provides the ow fraction distribution in the two types of ow passages. Normalizing ow rates with m_ n , hydraulic diameters with Dh;n , and free-ow areas with Ao;n , Eq. (12.31) becomes
m_ 1 =m_ n f Re2 Dh;1 =Dh;n 2 Ao;1 =Ao;n
12:32
where m_ n m_ 1 m_ 2 and all variables with a subscript n denote nominal values (selected
by the choice of an analyst), either the passage geometry 1, the passage geometry 2, or
some nominal passage geometry in between (for normalization of Dh and Ao used in the
equation) for a two-passage nonuniformity.
824
To compute the ow area ratios in Eq. (12.32), we maintain approximately the same
frontal area of the heat exchanger core with actual and nominal ow passages. There are
two choices for selecting the nominal passage geometry, and accordingly, the values of
Ao; j =Ao;n j 1; 2 will be dierent. They are as follows.
1. The number of ow passages for the nominal geometry is the sum of the number of
ow passages for passage types 1 and 2 and the frontal area is the same. In this
case,
Ao;1
Ao;2
A^o;1
A^o;2
1
2
12:33
^
Ao;n
Ao;n
Ao;n
A^o;n
where A^o;1 , A^o;2 , and A^o;n are the ow area for one passage of passage types 1, 2,
and n, respectively, and 1 and 2 are the corresponding fractions of the number of
passages of types 1 and 2.{ This case applies when comparing sharp and rounded
triangular (or any two similar passages), where the frontal area remains constant
for the same total number of ow passages, regardless of which is the nominal ow
passage. However, the free-ow area will be dierent for the nominal ow passages
since the ow areas of sharp and rounded corner passages are dierent (see
Example 12.3).
2. In the alternative case, the total number of ow passages for the nominal passages
could be different from the actual number of ow passages for the same frontal
area. This case applies when we compare the two-passage model (e.g., large and
small rectangular or triangular passages with 50% : 50% or any other percent
distribution) with the nominal passage geometry having approximately the
same frontal area.{ In this case, the number of ow passages for the nominal
passage geometry will be different from the sum of the number of ow passages
of passage types 1 and 2. The ow area ratios Ao;1 =Ao;n and Ao;2 =Ao;n are given by
Ao;1
1 A^o;1
Ao;2
2 A^o;2
12:34
Ao;n 1 A^o;1 2 A^o;2
Ao;n 1 A^o;1 2 A^o;2
where the denitions of A^o;1 , A^o;2 , 1 , and 2 are the same as dened above after
Eq. (12.33). Note that we may use A^o; j =A^o;n in Eq. (12.34) instead of A^o; j , j 1 or
2, since the fraction is primarily known.
The pressure drop ratio (the ratio of the pressure drop for either of the two passage
types, either 1 or 2, to the nominal passage pressure drop) can be calculated, using
Eq. (12.30), as
p1 f Re1 m_ 1 Dh;n 2 Ao;n
12:35
{
Note that 1 1 2 . The parameter , by denition, represents a ratio of the number of the ith shaped passage
P
to the total number of passages. If more than two passages are involved, the following relation holds:
i 1.
Also note that Ao;1 Ao;2 =Ao;n 6 1 in general. This is because we have presumed the same frontal area, and as a
result, the wall thickness is dierent for dierently shaped passages (see Example 12.3).
{
The number of passages must be an integer, so the frontal area for a particular selection of passages may not
necessarily be the same when compared to a two-passage model with a nominal passage model. However, in a
compact heat exchanger with a very large number of ow passages, the dierence will be negligible.
825
Note that the ow area of two nominal passages is the same as the two (large and small)
passages of the nonuniform core (see Fig. 12.4 for two examples).
As we know, since the uid seeks the path of least ow resistance, if we replace some
nominal passages with dierent ow passages having larger and smaller ow areas, a
larger fraction of the ow will go through the larger ow area passages. Then for a
constant ow rate, the pressure drop (and hence heat transfer) will reduce for this
exchanger with mixed passages.{ This means that p1 =pn p2 =pn will be less
than unity. This gain (reduction) in the pressure drop due to passage-to-passage nonuniformity is
pgain 1
p1
pn
12:36
Let us now determine a change in heat exchanger eectiveness due to passage-topassage nonuniformity. Heat transfer through dierently shaped passages would be
dierent, which in turn would produce dierent temperature dierences between uids
1 and 2. Hence, one cannot consider dierent passages, let us say two passages A and B,
in parallel to arrive at an eective h as the average of conductances hA and hB . To arrive
properly at an eective value of h for a two-passage geometry heat exchanger, the
passage geometrical properties, uid physical properties, exchanger ow arrangement,
and "-NTU relationship must be considered. A procedure is outlined in the following
subsections for the two most important cases of a two-passage geometry for a counterow exchanger with C* 1 (a rotary regenerator case) and an exchanger with C* 0
(an oil cooler case with constant wall temperature). Refer to Shah and London (1980) for
an analysis of other ow arrangements.
For both these cases, the heat transfer results are presented in terms of the number of
transfer units ntuj for each type of passage on the maldistributed uid side as follows:
hA
Nu 4L
4kL NuAo
ntuj
j 1; 2
12:37
m_ cp j
RePr Dh j
cp m_ D2h j
For the nominal passage, dene ntun by Eq. (12.37) with j n. The normalized
ntuj =ntun , based on Eq. (12.37), is
ntuj
Nuj m_ n Dh;n 2 Ao; j
12:38
ntun Nun m_ j Dh;j
Ao;n
Nuj and Nun in this equation should be obtained from the results of Table 7.3 for the
appropriate thermal boundary conditions for fully developed laminar ow. Note that for
H1 or *
H2 , while the
a counterow exchanger with C * 1, the boundary conditions are *
T for the C * 0 case.
boundary condition is *
counterflow heat exchanger with C* 1. In this case, "j and ntuj are related as
follows using Eq. (3.85):
"j
ntuj
1 ntuj
j 1; 2; n
12:39
This is what we have shown through Eqs. (12.28)(12.36) that the performance (q and p) of a continuous-owpassage regenerator matrix will be lower when there are small and large passages in parallel compared to the
performance of nominal (average)-size passages of the same shape.
826
where j depends on whether the heat exchanger unit considered has all uniform (nominal)
passages, "n j n, or it refers to a maldistributed heat exchanger that consists of two
subexchangers (with the eectivenesses "1 or "2 for the passage geometries j 1 or j 2,
respectively). Note that since ntuj is dened using a heat transfer coecient (not the
overall heat transfer coecient U as in NTU), the heat exchanger eectiveness of Eq.
(12.39) must be dened based on the passage wall temperature:
To; j Ti
ntuj
"j
1 ntuj
Tw Ti
j 1; 2; n
12:40
where Tw represents the mean wall temperature of the heat transfer surface, Ti is the inlet
temperature of uids in both subexchangers and nominal exchanger, and the distribution
of Tw; j vs. x are parallel to the distribution of Tj vs. x as shown in Fig. 1 of Shah and
London (1980). The temperature Tw is assumed to be the same for both passage geometries at the inlet (thus leading to the same inlet temperature dierence for both passage
types). The ntuj for Eq. (12.40) are computed from Eq. (12.38) for a specied value of
ntun and known ow fraction distribution from Eq. (12.32).
The average eectiveness of the maldistributed heat exchanger can be calculated from
the eectiveness of two subexchangers using a simple energy balance and assuming
constant specic heat of the uids as follows:
m_ "ave m_ 1 "1 m_ 2 "2
12:41
It must be emphasized that the analysis presented here is for one uid side of the
exchanger (either the hot- or cold-uid side of a rotary regenerator). To nd the resultant
eect on the exchanger performance, the eect of the other uid side needs to be taken
into account, as will be shown in Example 12.3. The eective ntu on one uid side is then
given by
ntueff
"ave
1 "ave
12:42
ntueff
ntun
12:43
We need to compute ntueff for both uid sides and subsequently calculate NTUeff for the
exchanger to determine a reduction in the exchanger eectiveness due to passage-topassage nonuniformity, as shown for a specic exchanger in Example 12.3.
London (1970) determined ntucost and pgain for plate-spacing and n-spacing type
nonuniformities and concluded that the deviation in passage size causes a more severe
reduction in the number of transfer units than does the pressure drop gain.
Specic results from the two-passage model for the passage-to-passage nonuniformity
are presented in Fig. 12.5 for rectangular passages. This two-passage model consists of
50% of the ow passages large (c2 > cn ) and 50% being small (c1 < cn ) compared to the
nominal passages, and the nominal aspect ratios n* 1, 0.5, 0.25, and 0.125. In Fig.
H1 and *
T boundary conditions and for a
12.5a, a reduction in ntu is presented for the *
827
(a)
(b)
FIGURE 12.5 Deterioration factors for two-passage nonuniformities in rectangular passages: (a)
percentage loss in ntu as a function of c , n*, and thermal boundary conditions; (b) percentage
reduction in p as a function of c and n*. (From Shah, 1981.)
828
nominal (design) ntun of 5.0. Here, ntucost , a percentage loss in ntu, and the channel
deviation parameter c are dened as
ntucost 1
ntueff
ntun
ntu*cost
ntueff
100
1
ntun
c 1
c1
cn
12:44
where ntueff is the eective ntu for the two-passage model passage-to-passage nonuniformity, and ntun is the ntu for nominal (or reference) passages. It can be seen
from Fig. 12.5a that a 10% channel deviation (c 0:10, which is common for a highly
compact surface) results in 10 and 21% reduction in ntuH1 and ntuT , respectively, for
n* 0:125 and ntun 5:0: In contrast, a gain in the pressure drop due to the passage-topassage nonuniformity is only 2.5% for c 0:10 and n* 0:125, as found from Fig.
12.5b. Here p*gain is dened as
p*gain
pactual
100
1
pnominal
12:45
The following observation may be made from Fig. 12.5a and additional results preT
sented by Shah and London (1980): (1) the loss in ntu is more signicant for the *
H1 boundary condition; (2) the loss in ntu increases with
boundary condition than for the *
higher values of nominal ntu; and (3) the loss in ntu is much more signicant than the
gain in p at a given c .
N-Passage Model. The previous analysis was extended for an N-passage model by Shah
and London (1980). In the N-passage model, there are N dierent-size passages of the
same basic shape, either rectangular or triangular. The results of Fig. 12.5a and b for
rectangular passages are also applicable to an N-passage model in which there are N
dierent-size passages in a normal distribution about the nominal passage size with a
proper denition of the channel deviation parameter c as follows:
c
"
N
X
i
i1
c
1 i
cn
!2 #1=2
12:46
Here i is the fractional distribution of the ith shaped passage. For N 2 and i 0:5,
Eq. (12.46) reduces to Eq. (12.44) for c .
Similar results are summarized in Fig. 12.6 for the N-passage nonuniformity model
associated with equilateral triangular passages. In this case, the denition of the channel
deviation parameter c is modied to
c
N
X
i1
"
i
rh;i
1
rh;n
!2 #1=2
12:47
where rh;n is the hydraulic radius of the nominal passages, rh;i is the hydraulic radius of
the ith passage, and they are related for a two-passage model as follows:
2r2h;n r2h;1 r2h;2 , but this particular case corresponds to an equilateral triangular
passage. Qualitative trends of the results in Fig. 12.6 are similar to those in Fig. 12.5
for rectangular ow passages.
829
FIGURE 12.6 Percentage loss in ntu and percentage reduction in p as functions of c for Npassage nonuniformities in equilateral triangular passages. (From Shah, 1985.)
Note that the percentage reduction in ntu and p vs. c curves for N 2 and N > 2
are identical (as shown in Figs. 12.5 and 12.6), except that the value of c is higher for a
two-passage model compared to the N-passage model for the same value of cmax =cn .
Hence, the two-passage model provides the highest deterioration in performance.
heat exchanger with C* 0. In this case, "j and ntuj are related as follows using
Eq. (3.84):
"j 1 entuj
j 1; 2; n
12:48
830
where j depends on whether the heat exchanger unit considered has all nominal passages,
"n j n, or it refers to a maldistributed heat exchanger that consists of two subexchangers (with the eectivenesses "1 or "2 for the passage geometries j 1 or j 2,
respectively). The average eectiveness of the passage-to-passage maldistributed heat
exchanger can be calculated using Eq. (12.41). Similar to the previous case, the cost of
the inuence of passage-to-passage nonuniformity on ntu is dened as follows:
ntueff
ntun
12:49
1
1 "ave
12:50
ntucost 1
where
ntueff ln
Characteristic
Sharp Corner
Passage
Rounded Corner
Passage
f Re
13.333
15.993
NuH1
A^o =A^o;
3.111
4.205
0.868
Dh =Dh;
1.125
For this regenerator, the ow split gas : air 50% : 50%. NTUn 2:5. Idealize negligible
wall resistance, C * 1 and Cr* ! 1.
SOLUTION
Problem Data and Schematic: The passage-to-passage ow nonuniformity is caused by
diering ow resistance of sharp and rounded corner ow passages shown in Fig. E12.3.
All data for thermal and hydraulic characteristics of the ow passages are provided in the
table above, and 1 2 0:5, NTUn 2:5, and C * 1.
831
FIGURE E12.3
A^o;sharp
0:5 1 0:5
A^o;sharp
Ao;2
A^o;round
2
0:5 0:868 0:434
Ao;n
A^o;sharp
The ratio of mass ow rates through these passages, using Eq. (12.31), is given by
m_ 1 f Re2
m_ 2 f Re1
Dh;1
Dh;2
2
2
Ao;1 15:993
1
0:5
1:092
Ao;2 13:333 1:125 0:434
0:522 and
)
m_ n 1 m_ 1 =m_ 2 1 1:092
m_ n
m_ 2
0:478
m_ n
Ans:
The ratio of the pressure drop for the sharp corner triangular passages to the nominal
pressure drop can be determined as follows using Eq. (12.35):
psharp p1 f Re1 m_ 1 Dh;n 2 Ao;n 13:333
1
0:522 12
1:044
pn
pn f Ren m_ n Dh;1 Ao;1 13:333
0:5
Ans:
Similarly, ratio of the pressure drop within the rounded triangular passages to the pressure drop through the nominal passages should, ideally, be 1.044. We could have calculated this ratio using an analogous relationship to the one given by Eq. (12.35) by
replacing the subscript 1 with 2.
832
pround p2 f Re2 m_ 2 Dh;n 2 Ao;n
pn
pn f Ren m_ n Dh;2 Ao;2
2
15:993
1
1
0:478
1:044
13:333
1:125
0:434
The heat exchanger eectiveness due to the nonuniformity depends on the number of
transfer units for the respective passages. The number of transfer units for sharp and
rounded corner passages, normalized with respect to nominal passages, are determined
using Eq. (12.38) as
ntuj
Nuj m_ n Dh;n 2 Ao; j
12 0:5 0:9579
ntun
ntun Nun m_ 1 Dh;1 Ao;n 3:111 0:522
2
nturound ntu2 Nu2 m_ n Dh;n 2 Ao;2 4:205
1
1
0:434 0:9697
ntun
ntun Nun m_ 2 Dh;2 Ao;n 3:111 0:478
1:125
For the given NTUn 2:5, we obtain ntun ntuh ntuc 5:0 for C* 1 [see, for
example, Eq. (9.23)]. With ntun 5, we get
ntu1 5 0:9579 4:7893
Consequently, the heat exchanger eectivenesses for the two types of passages for C * 1
would be
"j
ntuj
1 ntuj
j 1; 2
Therefore, "1 for sharp-corner triangular passages and "2 for rounded-corner triangular
passages are
"1
4:7893
0:8277
1 4:7893
"2
4:8483
0:8290
1 4:8483
The average heat exchanger eectiveness can be calculated from a simple energy balance,
using Eq. (12.41), as
"ave
m_ 1
m_
" 2 " 0:522 0:8277 0:478 0:8290 0:8283
m_ n 1 m_ n 2
"ave
0:8283
833
The eective NTU for this regenerator, using Eq. (5.54), is given by
NTUeff
1
1
2:412
1=ntueff;h 1=ntueff;c 1=4:8241 1=4:8241
NTUeff
2:412
0:7069
1 NTUeff 1 2:412
In contrast, the eectiveness of the heat exchanger with nominal uniform passages is
"n
NTUn
2:5
0:7143
1 NTUn 1 2:5
"n "ave
0:7143 0:7069
100
100 1:0%
"n
0:7069
Ans:
Discussion and Comments: It should be noted that in pressure drop analysis, sharp-corner
passages are considered as nominal passages. Thus, the pressure drop of the matrix with
nonuniform ow passages is 4.4% larger than that for the ideal matrix with uniform
sharp corners. This is because the sharp-corner triangular passage (used for the comparison) has a lower f Re than that for the rounded-corner triangular passage. In contrast,
there is a reduction (1.0%) in regenerator eectiveness due to nonuniform ow because
of the poor performance of the rounded-corner ow passages. The comparison was
performed by comparing the performance of nonuniform passages with nominal
sharp-corner passages which have a lower heat transfer coecient and a lower friction
factor. If we would have considered the rounded-corner triangular passages as nominal
passages, there would have been a reduction in the pressure drop and a slight gain in heat
exchanger eectiveness.
12.1.2.2 Passage-to-Passage Flow Nonuniformity Due to Other Eects. Finally,
passage-to-passage ow nonuniformity for very compact surfaces may be induced by
brazing and/or fouling in addition to manufacturing imperfection. Both controlled
atmosphere brazing and vacuum brazing have a negligible eect on j and f data if
the plates/tubes/primary surface is clad and ns are unclad, and the ratio of the joint
area to free-ow area is less than 10%. For ultracompact surfaces/ow passages, this
ratio may not be small (i.e., ow area blockage and brazing-induced surface roughness
may not be negligible, and accurate experimental j and f data are essential in this case).
Gross blockage due to brazing may increase the pressure drop substantially. The inuence of surface roughness induced by salt dip brazing (currently an outdated technology
due to environmental concerns) is generally nonnegligible (i.e., can increase p considerably with only a slight increase in h or j ) in highly compact surfaces (Shah and
London, 1971). Controlled atmosphere brazing, a state-of-the-art manufacturing process for compact heat exchangers (Sekulic, 1999), provides a very uniform ow passage
834
Whereas manifolds are integral in plate heat exchangers due to construction features,
manifolds are common and attached separately in many other applications. In the PHEs,
the uids enter and exit the manifolds laterally and ow within the core axially; here the
axial direction is dened as the main direction of uid ow within the PHE passages (see
Fig. 12.7a and b). In other applications, the uids enter and exit the core also axially, or a
combination of axial and lateral entry and exit. In the PHEs, the manifolds are of two
basic types: dividing ow and combining ow. In dividing-ow manifolds, uid enters
laterally and exits the manifold axially. The velocity within the manifold, parallel to the
manifold axis, varies from the inlet velocity to zero value. Conversely, in combining-ow
FIGURE 12.7 Manifold congurations: (a) U-ow or parallelow conguration; (b) Z-ow, Sow, or reverse-ow conguration. Pressure prole in (c) U-ow conguration, (d) Z-ow
conguration. In these congurations, (e) typical ow distribution, and ( f ) typical temperature
distribution. [Parts (e) and (f ) from Thonon, 2002.]
835
manifolds, uid enters axially from the PHE core and exits at the end of the manifold
laterally, with the velocity within the exit manifold varying from zero to the outlet
velocity. When interconnected by lateral branches, these manifolds result in paralleland reverse-ow systems, or U- and Z-ow (or S-ow) arrangements, as shown in Fig.
12.7a and b. Because the inlet and outlet manifolds have the same eective-diameter
pipes connected by the lateral branches, this construction has built-in inherent ow
nonuniformity, as evidenced by the resultant typical pressure proles as shown in Fig.
12.7c and d, and the mass ow rate distribution as shown in Fig. 12.7e.
Modeling of a manifold requires determination of both axial and lateral velocity and
static pressure distributions. Available solutions of the manifold ow models may be
either analytical (in simplied cases, Bajura and Jones, 1976; Edwards et al., 1984; Shen,
1992) or numerical (Majumdar, 1980; Thonon et al., 1991; Heggs and Scheidt, 1992) by
considering the inlet and outlet manifolds connected by the ow channels in a PHE. The
key problem in analytical modeling is the diculty of identifying a relevant streamline on
which to calculate energy and pressure losses and apply the Bernoulli equation. The
state-of-the-art design procedures utilize commercial and/or proprietary CFD codes.
The accurate modeling can be done only numerically and the reader is referred to the
references noted in this section.
Such manifold-induced ow maldistribution has traditionally been analyzed by a simplied approach without explicitly including the ow resistance in the lateral branches.
Bajura and Jones (1976) dened a set of generalized equations for manifold systems. The
model consists of (1) pressure-ow equations, and (2) ow distribution equations, including the related boundary conditions. These equations are obtained by writing continuity
and momentum equations for dividing and combining ow branch control volumes near
branching points, and the discharge equation, which provides the relationship of the
pressure dierences between the manifold and the lateral branch. So this model is based
on an application of rst principles, using continuity and momentum equations for header
ows and discharge equation for lateral ows. The experimental results for simple dividing
and combining ows were in very good agreement with this theory.
Shen (1992) extended the work of Bajura and Jones (1976) numerically to include the
eect of ow friction in the manifolds and the momentum losses associated with turning
the ow in the lateral branches. Datta and Majumdar (1980) conducted a numerical
analysis of both U- and Z-ow arrangements and found that the ow distribution
within the heat exchanger core is dependent on the following three dimensionless groups:
(1) the branch-to-manifold ow area ratio Ao*, (2) the manifold friction parameter F, and
(3) the lateral branch pressure loss coecient Kb , all dened as follows:
Ao*
Nc A2o;b
A2o;m
fm Lm Dp
Nc A2o;b
Kb
p
u2m =2gc
12:51
Here Nc is the number of channels on one uid side in a PHE, Ao;b is the free-ow area for
a branch or an exchanger and Ao;m is the ow area of the manifold, Dp is the port or
manifold diameter, and the subscript m is for manifold. Note that the pressure loss
coecient Kb of Eq. (12.51) is equal to the bracketed term of Eq. (6.28). The main
conclusions of the inuence of these parameters are as follows (Datta and Majumdar,
1980):
. The relative variation in the lateral ow distribution increases with increased Ao*.
. A reverse-ow manifold provides relatively more uniform ow distribution than a
parallel-ow manifold for otherwise identical conditions.
836
. In parallel- and reverse-ow manifolds, maximum ow occurs through the last port
and rst port, respectively.
. The eect of friction parameter is in general less signicant than that of the area
ratio Ao*.
A few general conclusions from these studies for more uniform ow through manifold
systems and some design guidelines for manifolds are as follows:
. Flow maldistribution is insignicant in PHEs with less than 20 ow channels on a
given uid pass.
. Flow maldistribution in the PHEs due to the manifold system (U- or Z-ow)
increases with increasing ow rate, increasing the number of plates in a given
pass and decreasing the liquid viscosity.
. In a U-ow manifold system, the maximum ow occurs through the rst port, and
in the Z-ow manifold system through the last port. Neither arrangement provides
uniform ow through the PHE or lateral branches. However, atter (relatively
more uniform) ow distribution is obtained with the U-ow manifold system
than with the Z-ow system (see Fig. 12.7e).
. To minimize ow maldistribution in a PHE, the ow area of the inlet manifold
(area of the actual or simulated pipe before lateral branches) should be larger than
the ow area of the lateral branches (heat exchanger core). The larger the port
diameter, the more uniform ow through the heat exchanger core. Alternatively,
ow maldistribution in a PHE plate pack (core) increases as the fraction of the total
pressure drop in the manifold becomes signicant.
. The ow area of a combining-ow manifold in Fig. 12.12b (the outlet manifold/
pipe in Fig. 12.7a and b) should be larger than that for the dividing-ow manifold in
Fig. 12.12a (the inlet manifold/pipe in Fig. 12.7a and b) for a more uniform ow
distribution through the core in the absence of heat transfer within the core. If there
is heat transfer in lateral branches (core), the ow areas should be adjusted rst for
the density change and then the ow area of the combining manifold should be
made larger than that calculated previously.
. Flow reversal is more likely to occur in a Z-ow system, which is subjected to poor
ow distribution.
. Based on the limited tests, a 2-pass 2-pass Z-ow arrangement can be treated as if
each pass were in a separate exchanger.
Thonon et al. (1991) and Heggs and Scheidt (1992), among others, have analyzed heat
transfer in a PHE with U- and Z-ow arrangements having 60 channels (30 channels on
each uid side). They found that when both uids enter the same end of the PHE with
either a U- or a Z-ow arrangement, the reduction in the exchanger performance is small
(ca. 2%) compared to the ideal uniform-ow case and may be neglected for practical
purposes. Typical temperature distributions are shown in Fig. 12.7f. However, two uids
can enter at dierent ends in Z- or U-ow arrangements. Based on Fig. 12.7e, the Z-ow
has a smaller ow rate/velocity at the entrance-end ow channels and a large ow rate/
velocity at the exit-end ow channels. When both uids in the Z-ow arrangement enter
from the dierent ends, there are two possibilities: (1) the ow rate in the end channels
can be the largest, or (2) the ow rate in the end channels can be the smallest. A similar
837
situation exists for the U-ow arrangement with uids entering from the dierent ends.
Heggs and Scheidt (1992) show that the eect of ow maldistribution on the PHE
performance is severe (up to 15% for the case that they analyzed) when the two uids
enter the exchangers from dierent ends.
From the foregoing results, we nd that if both uids enter from the same end in a
PHE, the manifold system has a signicant negative impact on ow maldistribution and
pressure drop, and a less degrading eect on overall heat transfer. However, if the uids
enter from dierent ends, both signicant ow and temperature maldistribution can
occur in a PHE. While in Section 1.5.2.1, we mentioned one of the advantages of
PHEs as having the same residence time for all uid particles on any uid side, the
foregoing results indicate that severe ow and temperature maldistributions can occur
due to inherent construction features of a PHE, resulting in dierent residence times.
This can have an impact on the use of a PHE in the chemical and food industries if the
residence time or temperatures are to be controlled over the entire surface area.
838
Liquid
cooling
pmin
Tm
C
Tw
xw
Liquid
heating
mmin
FIGURE 12.8 (a) Temperature distribution of a viscous liquid in the entrance region of a tube,
and (b) pressure drop vs. mass ow rate for a single tube (ow passage) in laminar liquid ow
cooling. (From Mueller, 1974.)
839
rate lower than the increase in the ow rate. As a result, the pressure drop will increase
monotonically with the liquid ow rate, as shown by a dashed-line curve for the liquid
heating case in Fig. 12.8b. In this case, there will not be any ow instability as found for
liquid cooling.
For gases, the viscosity increases with the temperature. Hence, ow maldistribution
can occur with the constant-tube-wall-temperature case when heating the gas and not
when cooling the gas, a phenomenon just opposite that for the liquids. With increasing
temperature drop for liquids or the increasing temperature rise for gases, the ow
maldistribution becomes more pronounced for the constant-wall-temperature boundary
condition.
Example 12.4 A viscous liquid ows under steady, fully developed laminar ow conditions through a tubular heat exchanger having two tube rows connected to inlet and
outlet pipes through lateral headers. The pressure drops within the headers are negligible.
The dynamic viscosity of the liquid decreases exponentially with an increase in temperature while the other properties may be considered as being nearly constant. The temperature of the wall of the channels is either lower or higher than the temperature of
liquid due to evaporating/condensing uid stream on the tube outside. Thus, the wall
temperature is uniform and constant along the tube length. The ow rate in each tube
corresponds to the ow rate between that for points A and B in Fig. 12.8b. Determine
which of the following conditions may exist in this heat exchanger: (1) dierent mass ow
rates of the viscous liquid may be established in the two tube rows for the cases of both
heating and cooling of the liquid, (2) a condition such as item (1) is possible only in the
case of heating, or (3) such a condition is possible only in the case of cooling.
SOLUTION
Problem Data and Schematic: A liquid has the following property: T; if T1 < T2 ,
then 1 T1 > 2 T2 , and if T1 > T2 , then 1 T1 < 2 T2 . Here
subscripts 1 and 2 denote any two temperatures T1 6 T2 of the liquid within the range
of temperatures considered. The uid ows through the two tube rows (A and B in Fig.
E12.4A). The liquid temperature T1 is either lower or higher than the uniform wall
temperature Tw .
Determine: Whether or not two dierent mass ow rates may be established in the two
tube rows connected to the same inlet and outlet headers under the conditions of either
cooling and/or heating of the viscous liquid.
Assumptions: The ow is steady laminar. The pressure drops along the headers are
negligible. Consequently, the pressure drops along the tube rows between the headers
will be the same. The thermophysical properties of the liquid, except for the viscosity,
are assumed to be constant. We assume that the liquid viscosity varies with the temperature as C1 expC2 =T where C1 and C2 are constants. Entrance and exit pressure
drops are negligible. Thermal resistance on the tube side is controlling (i.e., the temperature of the wall is constant, that is, the heat transfer coecient outside the channels is
very large).
840
FIGURE E12.4A
Analysis: The pressure drop under fully developed laminar ow through tubes can be
obtained from Eq. (6.67a) or directly from Eq. (13.2) by treating P independent of Dh as
p
1 1 16L
_
m
f
Re
D3h 2gc P
Hence, we may conclude from Eq. (1) that the pressure drops depend on the products of
viscosity and mass ow rate for each of the two tube rows:
pj / m_ j
where j j A or B represents an average viscosity between inlet and outlet of either
tube row A or B.{ For liquids, the dynamic viscosity may be approximated by an
exponential decreasing function in terms of the local temperature (Mueller, 1987):
j C1 eC2 =Tj
Tj Tj;i
1 eNTUj / 1 eC3 x=m_ j
Tw Tj;i
j A or B
Note that letters A and B in Figs. 12.8b and E12.4A are not related and denote dierent entities.
841
From Eq. (2) it is clear that pressure would dier along the ow direction at each
location within tube rows A and B everywhere but at the terminal ports. Note that the
pressure gradient along the ow p=L is proportional to viscosity [see Eq. (1) above],
which changes along the uid ow direction as indicated by Eqs. (3) and (4). So even
though the pressure drop between the inlet and outlet for each tube row is going to be the
same, the local pressure distributions along the ow direction are not necessarily the
same for the two tube rows [see Eq. (2)] in the ow instability region AB of Fig. 12.8b
(specied input). From Eq. (2) it must be clear that the same pressure drop for both tube
rows, j A and B, may be reached by having diering viscosities and diering mass ow
rates in these two channels, at least in principle. We may say that if a larger mass ow rate
and corresponding lower viscosity are established in one tube row, the same product of
these two entities may be obtained for a lower mass ow rate and larger viscosity that
may be established in the second tube row. The question is whether such conditions are
possible for both heating and/or cooling. An answer would lead to a solution of the
problem.
So, based on the relationships of Eqs. (2)(4), the temperature and pressure distributions for a case of liquid cooling (i.e., TA;i TB;i > Tw ) are presented in Fig. E12.4B.
From these plots we can conclude that if a viscous liquid is cooled, diering tube rows
may have diering mass ow rates despite the same pressure drop in the ow instability
region of p vs. m_ relationship. This is because the liquid viscosity increases with a
decrease in its temperature. This means that ow misdistribution in dierent tube rows
will be present in such a situation.
In the case of heating of the liquid, its viscosity decreases with an increase in temperature, and the situation is going to be dierent. Namely, the liquid with an assumed larger
mass ow rate would have lower local temperatures than those of the liquid that would
have the smaller mass ow rate based on Eq. (4). If that is the case, the viscosity along the
ow direction for the uid with a higher mass ow rate (lower temperatures) would have
higher viscosities along the channel due to the relationship given by Eq. (3). The situation
for the uid with smaller mass ow rate (higher temperatures) would lead to the presence
of lower viscosities along the ow direction. So, in both cases, both the mass ow rate
and the viscosity are either increased or decreased for a considered tube row. However, to
have an invariant pressure drop for both channels, the product of the mass ow rate and
viscosity must stay invariant as shown by Eq. (2). That would be quite opposite from
what would actually happen based on the analysis above. Therefore, such conditions
could not be satised: It is not possible to have heating of a viscous liquid that would lead
to maldistribution if the pressure drop must stay the same between the headers. That
T
TA,i = TB,i
mB
pA = pB
mA > mB
mB
Tw
0
x=0
mA > m B
TA,o
TB,o
x=L
FIGURE E12.4B
0
x=0
x=L
842
means there is no instability region for a p vs. m_ relationship for viscous liquid heating
(see the dashed curve in Fig. 12.8b).
Discussion and Comments: The analysis presented indicates that operating conditions
may cause the ow maldistribution in a certain operating range of liquid ow rate. For
example, a liquid with viscosity decreasing with an increase of temperature may feature
diering mass ow rates in dierent tube rows within the heat exchanger, if the uid is
cooled and the pressure drops are kept the same for these channels at their respective
terminal ports under certain operating conditions. If the uid is heated, the mass ow
rates in diering tubes must be approximately the same to keep the same pressure drops.
It should be noted that this analysis was conducted for a viscous liquid that is characterized with a decreasing viscosity with increasing temperature. Gases may feature the
opposite behavior (although the changes of the properties would be even smaller).
Such a situation is considered in Problem 12.8.
Mueller (1974) proposed the following procedure to determine the minimum pressure
drop, pmin , above which the ow maldistributioninduced instability would not be
possible. The case considered is for a viscous liquid of a known inlet temperature
being cooled as it ows through the length of a tube of known constant temperature Tw .
1. From viscosity data, determine the slope m of the curve ln vs. 1/T, where T is
temperature on the absolute temperature scale.
2. With the known slope m and liquid viscosities at the inlet (i ) and wall (w )
temperatures, determine the average liquid viscosity (ave ) using Fig. 12.9. This
gure is based on the assumption that the uid temperature reaches the wall
temperature within the tube. If the uid exit temperature is still larger than the
wall temperature, the average viscosity should be modied. The details are provided by Mueller (1974).
3. With these viscosities, determine xw of Fig. 12.8 from
xw
L
21 ave =w
12:52
m_ cp
0:4
kL
12:53
128
m_ xw ave L xw w
gc D4h
12:54
843
FIGURE 12.9 Viscosity ratio chart for various slopes of m of C1 eC2 =T : (From Mueller, 1974.)
If the pressure drop for the given design is found to be less than that calculated by
Eq. (12.54), the uid ow length should be increased (either increasing the duct
length or considering a multipass design) to eliminate the ow instability. If the
ratio of the average and wall temperature viscosities is larger than 0.5, the maximum pressure drop can occur at a ow rate larger than the breakthrough mass
ow rate. If the exchanger pressure drop exceeds pmin , ow instability will not be
a problem. However, ow maldistribution is still possible, as considered in the next
subsection.
When an exchanger has multiple (N) tubes, and the total ow rate on the tube side
is less than N m_ min (where m_ min is dened in Fig. 12.8 per tube), dierent ow rates will
establish in dierent tubes depending on their operating average temperature.
Conceptually, the ow rate in individual tubes could be calculated based on the preceding method. Hence, one can visualize the ow maldistribution due to dierent uid
viscosity in dierent tubes. While Mueller (1974) provides a method for computing
pmin or m_ min per tube and recommends operation of the multitubular exchanger at
ow rates higher than m_ min , Putnam and Rohsenhow (1985) provide a method for
operation when m_ < m_ min for a specic case.
12.2.1.2 Flow Maldistribution When No Flow Instability Present. When the ow rate
m_ > N m_ min (in reality, each tube or ow passage should have the ow rate greater than
m_ min , there is no ow instability. However, there will be passage-to-passage ow
844
m_ 2 1
12:55
For turbulent ow, use f / Re0:2 or a similar relationship in Eq. (12.29) to get the ratio
m_ 1 =m_ 2 in terms of the 2 =1 ratio (see, e.g., an equation for this ratio in Problem 12.6). In
a similar manner, we can analyze developing laminar ow with the correct values for the f
factors in Eq. (12.29).
Once the ow fraction distribution is found from Eq. (12.55), the pressure drop ratio
for fully developed laminar ow case can be determined by Eq. (12.35), where the subscript n represents the case corresponding to the mean temperature for both tube rows
(i.e., for the exchanger on that uid side). For other ow types, calculate the pressure
drop ratio from the following equation, derived from Eq. (12.29):
p1 f1 m;1 um;1 2
12:56
The heat transfer case can be analyzed for the constant-wall-temperature boundary
condition using Eq. (12.38) to determine ntuj for the given ntun and the computed ow
fraction distribution from Eq. (12.55). Subsequently, follow Eqs. (12.48)(12.50) to
determine the impact on the heat transfer performance due to ow maldistribution in
a viscous cooler. Be sure to modify these equations as appropriate when the ow is
turbulent or developing laminar ow on the ow maldistributed side.
845
846
FIGURE 12.10 Typical compact heat exchangers with (a) normal headers and (b) turning
headers.
(return) pipe. The header is variously referred to as a tank, box, or distributor. Manifolds
have a bit more complex function, especially for compact heat exchangers. An incoming
stream must be distributed uniformly into a heat exchanger core for the lowest core p
and the highest achievable q. Basically, a manifold is a ow channel/duct with one (side
or central) inlet and multiple sidewall outlets to the heat exchanger core, or vice versa.
Compact heat exchangers having normal and turning headers are illustrated in Fig. 12.10
and oblique ow headers in Fig. 12.11. Examples of manifolds with ow distributions are
shown in Fig. 12.12.
FIGURE 12.11 Oblique-ow headers: (a) parallelow; (b) counterow; (c) free discharge header.
(From London et al., 1966.)
Dividing flow
Combining flow
(a)
(b)
Reverse flow
(c)
847
Parallel flow
(d)
FIGURE 12.12 Major types of manifolds: (a) dividing-ow manifolds, (b) combining-ow
manifolds. The corresponding inlet/outlet congurations (see also Fig. 12.7); (c) reverse-ow
conguration, (d) parallelow conguration.
Two important requirements may be identied for header and/or manifold design.
They should be designed so that they result in (1) uniform distribution of the uid stream
within a heat exchanger core, and (2) minimal pressure drop within the header/manifold,
since in general we do not get any heat transfer for that pressure drop expenditure.
The design of the inlet header is more critical. An area increase from the inlet pipe to
the core face may be 5 to 50 times. It is impossible to maintain streamline ow in headers
for such a large area enlargement. Hence, the ow is normally separated in an inlet
header with either a completely detached or a singly attached jet on one wall. Flow
separation results in increased pressure drop in the inlet header and nonuniform ow
distribution at the core face. In addition to ow separation, the shape of inlet header
could produce high-velocity regions, and this could lead to localized erosion at the core
face (tube entrance from the tank), particularly for liquid ows.
The design of the outlet header should match that of the inlet header (or vice versa) so
that the pressure drop across the core is uniform, resulting in the uniform ow distribution. To minimize pressure losses due to ow separation, area contraction in the outlet
header should be smooth. Also, sharp turns in the outlet header should be avoided.
Turns and bends create centrifugal forces, resulting in nonuniform pressure at the core
face, which may lessen ow uniformity in the core.
From the foregoing viewpoints, the header design objective is to provide for acceptably uniform ow through the core with an acceptable header geometry and acceptably
low pressure drops since this spent p in the headers is not associated with heat transfer
between two uids (the heat exchanger core is designed for this heat transfer).
In baed shell-and-tube exchangers and in some multipass crossow heat exchangers,
good uid mixing takes place on the shell side or within the core. Flow nonuniformity at
the core inlet generally does not degrade the performance of the exchanger if there is
good mixing within the core. The design of headers is important for those exchangers in
which there is very little uid mixing in the core, and for the gas side of a gas-to-uid heat
exchanger.
In the highly compact gas-to-uid heat exchanger, the header design on the gas side is
more dicult because the exchanger core shape is characterized by a large ow frontal
area and a short ow length for the gas ow path. In such cases, either no header is used,
as in an automobile radiator (on the air side), or an oblique ow header of the type
848
Oblique-Flow Headers
The design theory for oblique-ow headers for heat exchangers has been derived by
London et al. (1968), based on the work of Perlmutter (1961). This theory is based on
the study of ow conditions and corresponding header shape, assuming steady, constantdensity, inviscid ow. Three single-pass header congurations are of particular interest,
as shown in Fig. 12.11.
In an oblique-ow header, the uid inlet ow direction with respect to the core face is
at an angle dierent from 908 (i.e., normal ow, as in a normal-ow header). A special
class of oblique-ow headers has an inlet ow direction parallel to the core face area. The
main feature of this type of header is the minimization of header volume and ow
separation. The three main types of oblique headers are:
1. Parallelow headers (Fig. 12.11a). The uid inlet and outlet are on opposite sides of
the core and the uid ows in the same directions through both headers.
2. Counterow headers (Fig. 12.11b). The uid inlet and outlet are on the same side of
the core and the uid ows in counterow through the headers.
3. Free discharge headers (Fig. 12.11c). Only the inlet header exists and uid discharges freely at the outlet without ducting.
In Table 12.1, the model predictions for pressure distributions, theoretical shapes, and
pressure drops are compiled for the three types of oblique headers based on the results of
London et al. (1968). The assumptions adopted are given in Example 12.5. A simple box
conguration is considered for the outlet header (if the outlet header exists) from the
construction and cost points of view; then inlet headers for parallelow and counterow
oblique headers require a special shape to achieve the uniform ow distribution through
the core; and the shapes are derived theoretically as summarized in Table 12.1. All
geometrical characteristics and the notation are presented in Fig. 12.11.
If a designer has freedom to select the header type, a counterow header would be the
best option for the lowest p in headers, followed by the free discharge header and
parallelow header. Reviewing the results in Table 12.1, note that p for the inlet header
is higher than that for the outlet header in all three cases, and p for the outlet headers is
largely associated with the nonuniform velocity distribution at the exit shown in Fig.
12.11a and b.
In addition to the three congurations of oblique headers presented in Table 12.1 and
Fig. 12.11, various other congurations are possible (such as headers at dierent incoming angles, or with turning vanes or guide vanes to turn the ow and minimize ow
nonuniformity). No systematic studies are reported in the literature for such headers,
due to many geometrical variations and lack of any available theory. One of the best
alternatives is to conduct three-dimensional CFD analysis when accurate ow distribution and pressure drop predictions are required.
Example 12.5 Determine the geometry of the inlet header, z=yo as a function of
X* x=L, of the parallelow oblique header of Fig. 12.11a (note that the outlet header
849
pi pX * 2
zi 2
X *2 i
4
o yo
Hi
Hi
u2m
2gc
Ho
z
1 X*
zi
u2m
2gc
X*
x
L
z
1 X*
z
1 X*
1 1 X *2
Hi
4 o yo
Theoretical Modela
Inlet:
pi
1
Hi
pt
1
Total:
Hi
pt
H
1 1:467 o
Hi
Hi
pi 1
Inlet:
Hi
3
H
d po
Outlet :
0:645 o
Hi
Hi
pt
Total:
0:595
Hi
Total:
po
H
0:645 o
Hi
Hi
pi
2 Ho
1
Hi
12 Hi
Outlet:
Inlet:
Pressure Drop
Counterow headerc
Parallelow header
Header Type
TABLE 12.1 Theoretical Shape, Pressure Proles, and Pressure Drops of Oblique-Flow Headers
850
is a box type) for the case when inlet and outlet velocity heads are equal Hi Ho and
i =o 1:4. Use the inlet header prole given in Table 12.1 in which z=yo is dependent on
yo =zi , the ratio of outlet to inlet header height.
SOLUTION
Problem Data and Schematic: The inlet header is a parallelow oblique header as shown
in Fig. 12.11a with appropriate notations. The following data are given:
Ho
u2m
2gc
!
Hi
o
u2m
2gc
!
i
i
1:4
o
Determine: The inlet header geometry [i.e., z=yo x=L], shown in Fig. 12.11a.
Assumptions: The following assumptions are made (London et al., 1968): (1) the inlet and
outlet header uid mass densities are individually constant, (2) the inlet header velocity
distribution is uniform, (3) the ow velocity and pressure in the inlet header are functions
of x only, (4) the inlet and outlet header uid ows are inviscid, (5) the outlet header
pressure is a function of x only, (6) the velocity is uniform through the core, and (7) the
depth of the header (the third dimension) is unity.
Analysis: From Table 12.1, the geometry of an inlet parallelow oblique header is given
by
Z
z
1 X*
2
yo =4i =o X *2 yo =zi 2 1=2
Let us rst demonstrate briey how this equation may be derived (for details, see London
et al., 1968). By applying the Bernoulli equation for a streamline through the inlet header,
core, and outlet header (Fig. 12.11a), we can relate inlet pressure to a pressure at any x
position as follows:
pi px
i 2
u u2i
2gc
Invoking assumptions concerning constant uid density, uniform inlet velocity, and
uniform velocity v through the core (in the negative z direction in Fig. 12.11a), the
continuity equation will have the following form{:
ui zi uzx vx vL x vx vL uo yo
o
i
Note that x-direction mass ow rate through any cross section x plus the mass ow rate through the core in the y
direction from x 0 to x is equal to mass ow rate entering the inlet header at x 0. Alternatively, x-direction
mass ow rate through any cross section x is the same as the mass ow rate going through the core in the y
direction from x x to L.
851
2
i 2
u u2i Ho X *2
2gc
4
Finally, combining Eqs. (3) and (4) to eliminate u from the second equality in Eq. (4) and
keeping in mind the denition of velocity heads from Table 12.1 and as indicated above
by Eq. (5), we obtain Eq. (1). A detailed derivation of Eq. (1) is the subject of Problem
12.9.
We will now derive the ratio yo =zi as a function of the velocity head ratio Hi =Ho from
the continuity equation considering mass ow rates to be the same at the entrance of the
inlet and outlet headers:
Ao um i Ao um o
Therefore,
um;i o Ao;o o yo
um:o i Ao;i
i z i
or
um;i
um;o
2
2 2
o
yo
i
zi
From the denition of the velocity heads (see Table 12.1), we get
Hi
u2m =2gc i
Ho u2m =2gc o
or
Hi
i um;i 2
H o o um;o
Ho i zi
Substituting yo =zi 2 from Eq. (8) into Eq. (1), we get the geometry of the inlet header as
z
1 X*
Since the inlet and outlet velocity heads must be equal (imposed by the problem formulation; i.e., Hi Ho ), Eq. (9) can be rewritten as follows using i =o 1:4 as given
z
1 X*
1 X*
10
852
We can now calculate the coordinates of the inlet header shape z=yo :
X*
0.2
0.4
0.6
0.8
1.0
z=yo
0.714
0.545
0.363
0.208
0.089
Discussion and Comments: The inlet header prole calculated demonstrates a need to
design this header with a variable header cross section according to the functional relationship given by Eq. (10) for the case when the outlet header is box type and uniform
ow is through the core. In this case, the inlet and outlet velocity heads would be equal,
and the header loss would be, from Table 12.1, pt =Hi 1 1:467Ho =Hi
1 1:467 1 2:47 velocity heads.
12.4.2
Normal-Flow Headers
Manifolds
Two major types of manifolds, as noted in Section 12.1.3, are shown in Fig. 12.12a and b:
dividing-ow manifolds and combining-ow manifolds. In dividing-ow manifolds, uid
enters either axially or laterally and exits the manifold laterally. The axial velocity within
the manifold varies from the inlet velocity to the zero value. Conversely, in combining
ow manifolds, uid enters laterally and exits at the end of the manifold either axially or
laterally.
Inlet and outlet manifolds in a heat exchanger can be arranged as (1) reverse-ow
conguration and (2) parallelow conguration, as shown in Fig. 12.12c and d, interconnected by lateral branches represented by ow passages in the heat exchanger core
(see also Fig. 12.7).
Modeling a manifold requires determination of both axial and lateral velocity and
static pressure distributions. Available solutions of the manifold ow models may be
either analytical (in simplied cases, Bajura and Jones, 1976) or numerical (Majumdar,
SUMMARY
853
1980). The key problem in analytical modeling is the diculty in identifying a relevant
streamline on which to calculate energy and pressure drop losses and apply the Bernoulli
equation. The state-of-the-art design procedures utilize commercial and/or proprietary
CFD codes. Still, simple analytical modeling has merit for assessment purposes (Shen,
1992).
SUMMARY
This chapter deals with several important ow maldistributions (geometry and operating
conditions induced) in heat exchangers. Design theory is provided to determine quantitatively the inuence of two major geometry-induced ow maldistributions: (1) gross
ow maldistribution and (2) passage-to-passage ow maldistribution.
. Simple modeling for gross ow maldistribution is possible only for counterow,
parallelow, and one crossow unmixedmixed arrangements. For other ow
arrangements, the inuence of ow maldistribution can be evaluated numerically.
In many situations, gross ow maldistributions do not reduce heat transfer signicantly but may cause a signicant increase in pressure drop. Some specic gross
ow maldistributions in an unmixedunmixed crossow exchanger can increase the
exchanger eectiveness. This is mentioned in the paragraph just before Example
12.2.
. Compact heat exchangers with continuous-ow passages are highly susceptible to
passage-to-passage ow maldistribution important in laminar ows. This maldistribution reduces the pressure drop slightly but may reduce heat transfer signicantly. Usually, the pressure drop reduction is neglected in the design. The simplest
model of passage-to-passage ow maldistribution is the two-passage model, where
two dierent-size passages are in parallel. The two-passage model reduces exchanger eectiveness and hence heat transfer performance more than that for an Npassage (N > 2) model of passage-to-passage nonuniformity. Hence, for a conservative design, a two-passage model is most appropriate to determine the eect of
passage-to-passage ow maldistribution.
Among operating conditioninduced ow maldistributions, the most important is viscosity-induced ow maldistribution. It can induce ow instability in a multitube or multicontinuous passage exchanger if the ow rate is below some critical value (m_ min in Fig.
12.8). If m_ > m_ min , the viscosity-induced ow maldistribution problem reverts to the
problem of passage-to-passage nonuniformity.
It should be mentioned that no generalized recommendations can be made for preventing negative inuences of ow maldistribution. Each case should be considered
separately.
Header and manifold design is very important in controlling the level of ow
maldistribution within the core and reducing undesired pressure drop in headers and
manifolds, particularly for compact heat exchangers. Specic information for designing
the header cross-section prole is presented in Section 12.4 for oblique-ow headers.
REFERENCES
Bajura, R. A., and E. H. Jones, Jr., 1976, Flow distribution manifolds, ASME J. Fluid Eng., Vol. 98,
pp. 654666.
854
Chiou, J. P., 1980, The advancement of compact heat exchanger theory considering the eects
of longitudinal heat conduction and ow nonuniformity, in Compact Heat Exchangers: History,
Technological Advancement and Mechanical Design Problems, R. K. Shah, C. F. McDonald, and
C. P. Howard, eds., Book G00183, HTD-Vol. 10, American Society of Mechanical Engineers,
New York, pp. 101121.
Chiou, J. P., 1982, The eect of nonuniformities of inlet temperatures of both uids on the thermal
performance of crossow heat exchanger, Heat Transfer 1982, Proc. 7th Int. Heat Transfer Conf.,
Vol. 6, pp. 179184.
Chowdhury, K., and S. Sarangi, 1985, The eect of ow maldistribution on multipassage heat
exchanger performance, Heat Transfer Eng., Vol. 6, No. 4, pp. 4554.
Cichelli, M. T., and D. F. Boucher, 1956, Design of heat exchanger heads for low holdup, AIChE
Chem. Eng. Prog., Vol. 52, No. 5, pp. 213218.
Datta, A. B., and A. K. Majumdar, 1980, Flow distribution in parallel and reverse ow manifolds,
Int. J. Heat Fluid Flow, Vol. 2, pp. 253262.
Fleming, R. B., 1966, The eect of ow distribution in parallel channels of counterow heat
exchangers, in Advances in Cryogenic Engineering, Vol. 12, K. D. Timmerhaus, ed., Plenum
Press, New York, pp. 352363.
Heggs, P. J., and H.-J. Scheidt, 1992, Thermal performance of plate heat exchangers with ow
maldistribution, in Compact Heat Exchangers for Power and Process Industries, American
Society of Mechanical Engineers, HTD-Vol. 201, pp. 8793.
Hewitt, G. H., G. L. Shires, and T. R. Bott, 1994, Process Heat Transfer, CRC Press, Boca Raton,
FL.
Kitto, J. B., and J. M. Robertson, 1989, Eects of maldistribution of ow on heat transfer equipment performance, Heat Transfer Eng., Vol. 10, No. 1, pp. 1825.
Kutchey, J. A., and H. L. Julien, 1974, The measured inuence of ow distribution on regenerator
performance, SAE Trans., Vol. 83, SAE Paper 74013.
London, A. L., 1968, Laminar ow gas turbine regenerators the inuence of manufacturing
tolerances, T.R. No. 69, Department of Mechanical Engineering, Stanford University, Stanford,
California, 1968.
London, A. L., 1970, Laminar ow gas turbine regenerators: the inuence of manufacturing
tolerances, ASME J. Eng. Power, Vol. 92, Ser. A, pp. 4556.
London, A. L., G. Klopfer, and S. Wolf, 1968, Oblique ow headers for heat exchangers the ideal
geometries and the evaluation of losses, T.R. No. 63, Department of Mechanical Engineering,
Stanford University, Stanford, California.
London, A. L., G. Klopfer, and S. Wolf, 1968, Oblique ow headers for heat exchangers, ASME J.
Eng. Power, Vol. 90, Ser. A, pp. 271286.
Majumdar, A. K., 1980, Mathematical modeling of ows in dividing and combining manifolds,
Appl. Math. Model., Vol. 4, pp. 424434.
Mueller, A. C., 1974, Criteria for maldistribution in viscous ow coolers, Heat Transfer 1974, Proc.
5th Int. Heat Transfer Conf., Vol. 5, pp. 170174.
Mueller, A. C., 1977, An inquiry of selected topics on heat exchanger design, AIChE Symp. Ser. 164,
Vol. 73, pp. 273287.
Mueller, A. C., 1987, Eects of some types of maldistribution on the performance of heat exchangers, Heat Transfer Eng., Vol. 8, No. 2, pp. 7586.
Mueller, A. C., and J. P. Chiou, 1988, Review of various types of ow maldistribution in heat
exchangers, Heat Transfer Eng., Vol. 9, No. 2, pp. 3650.
Perlmutter, M., 1961, Inlet and exit header shapes for uniform ow through a resistance parallel to
the main stream, ASME J. Basic Eng., Vol. 83, pp. 361370.
Putnam, G. R., and W. M. Rohsenow, 1985, Viscosity induced nonuniform ow in laminar ow
heat exchangers, Int. J. Heat Mass Transfer, Vol. 28, pp. 10311038.
REVIEW QUESTIONS
855
Ranganayakulu, C. H., K. N. Seetharamu, and K. V. Sreevatsan, 1996, The eects of inlet uid ow
nonuniformities on thermal performance and pressure drops in crossow plate-n compact heat
exchanger, Int. J. Heat Mass Transfer, Vol. 40, pp. 2738.
Sekulic, D. P., 1999, Behavior of aluminum alloy micro layer during brazing, in Recent Res. Dev.
Heat Mass Momentum Transfer, Vol. 2, pp. 121140.
Sekulic, D. P., C. Pan, F. Gao, and A. T. Male, 2001, Modeling of molten cladding ow and
diusion of Si across a clad-core interface of an aluminum brazing sheet, DVS Berichte, Vol.
212, pp. 204219.
Shah, R. K., 1981, Compact heat exchangers, in Heat Exchangers: Thermal-Hydraulic Fundamentals
and Design, S. Kakac, A. E. Bergles, and F. Mayinger, eds., Hemisphere Publishing, Washington, DC, pp. 111151.
Shah, R. K., 1985, Compact Heat Exchangers, in Handbook of Heat Transfer Applications, 2nd Ed.,
Eds. W. M. Rohsenow, J. P. Hartnett and E. N. Ganic, Chapter 4, Part III, pp. 4174 to 4311,
McGraw-Hill, New York.
Shah, R. K., and A. L. London, 1971, Inuence of Brazing on Very Compact Heat Exchanger
Surfaces, Paper 71-HT-29, American Society of Mechanical Engineers, New York.
Shah, R. K., and A. L. London, 1980, Eects of nonuniform passages on compact heat exchanger
performance, ASME J. Eng. Power, Vol. 102, Ser. A, pp. 653659.
Shen, P. I., 1992, The eect of friction on ow distribution in dividing and combining ow manifolds, ASME J. Fluids Eng., Vol. 114, pp. 121123.
Thonon, B., 2002, Private communication, CAE-GRETh, Grenoble, France.
Thonon, B., P. Mercier, and F. Feidt, 1991, Flow distribution in plate heat exchanger and consequences on thermal and hydraulic performances, Proc. 18th Eurotherm Conference, Springer
Verlag, Hamburg, Germany.
Wilson, D. G., 1966, A Method of Design for Heat-Exchanger Inlet Headers, Paper 66-WA/HT-41,
Americal Society of Mechanical Engineers, New York.
REVIEW QUESTIONS
Where multiple choices are given, circle one or more correct answers. Explain your
answers briey.
12.1
12.2
12.3
12.4
856
(b) manufacturing tolerances that are a signicant fraction of the surface hydraulic diameter
(c) fouling
(d) deposition of condensable substances
12.5
12.6
12.7
12.8
12.9
Using the knowledge of irreversibilities gained in Chapter 11, explain why the
temperature maldistribution increases exchanger eectiveness for an unmixed
unmixed crossow exchanger when the hotter portion of the hot uid maldistributed inlet temperature is near the exit end of the cold uid, whose temperature is
uniform.
REVIEW QUESTIONS
857
858
12.19 The pressure drop is increased signicantly compared to that for the uniform ow
case if the following ow maldistributions exist on one uid side of an exchanger:
(a) gross ow maldistribution
(b) viscosity-induced maldistribution
(c) manifold-induced maldistribution (d) passage-to-passage maldistribution
12.20 The primary function of the headers/manifolds is:
(a) to avoid ow separation within the header/manifold
(b) to provide uniform ow distribution over the core face
(c) to result in the lowest possible pressure drop within the headers/manifolds
(d) to yield uniform temperature distribution at the core inlet
12.21 The design of the inlet header/manifold is more/less critical than the design of the
outlet header/manifold:
(a) more
(b) less
(c) cant tell
12.22 Flow separation in the headers, caused by an area change in the free-ow area,
leads to:
(a) increased pressure drop
(b) decreased pressure drop
(c) high-velocity regions
(d) localized erosion
12.23 The header/manifold design problem is more important in a:
(a) shell-and-tube heat exchanger
(b) compact heat exchanger
(c) spiral plate heat exchanger
12.24 Oblique-ow headers are characterized by:
(a) The inow is orthogonal to the heat transfer core face.
(b) The inow is parallel to the heat transfer core face.
(c) The inow is at an angle dierent from 908 to the core face.
(d) The inlet and outlet headers are on the same side of the exchanger with side
inlets.
(e) Only the inlet header is always present.
12.25 For specied heat transfer performance in an exchanger, important design considerations for normal and oblique ow headers are to:
(a) match inlet and outlet header designs
(b) minimize the core pressure drop
(c) allow nonuniform ow through the core to reduce the core pressure drop
(d) Any header design is acceptable as long as the heat exchanger core has perfectly manufactured ow passages.
12.26 Circle the following statements as true or false.
(a) T F
More uniform ow distribution through the core is achieved by a
reverse-ow manifold system than by a parallelow manifold system
(b) T F
More uniform ow distribution through the core is achieved if the
ow area of the combining-ow header is smaller than that of the
dividing-ow header.
(c) T F
Maintaining a larger pressure drop in an exchanger core than in the
headers is important to provide uniform ow distribution through
the core.
PROBLEMS
859
On the uid 1 side of a counterow exchanger, 80% of the total free-ow area is
fouled such that the velocity through that portion of the ow area constitutes only
60% of the mean velocity through the uid 1 side core as a whole. The ratio of the
heat capacity rates of uids 1 and 2 for the heat exchanger as a whole is equal to 1.
The number of transfer units of the heat exchanger is 3.5. Determine the temperature eectiveness of uid 1 if ow on the uid 2 side is uniform.
12.2
The crossow unmixed (uid 1) mixed heat exchanger of Example 12.2 has to be
analyzed for the inuence of ow maldistribution in the entire range of possible
nonuniformities of the mass ow rate on the uid 1 side. Determine the change in
the temperature eectiveness of the maldistributed uid (uid 1) if the X fraction
of the total free-ow area on the uid 1 side is characterized by a ow velocity
860
larger by Y% than the mean velocity through the core. Consider the cases
for which X 13, 0.5, 23, and Y 25, 50, and 75, respectively, for each of the X
values.
12.3
Consider a gas turbine rotary regenerator made up of a deepfold surface (rectangular passages of * 0:125). Due to the manufacturing process, some of the
passages are close to a trapezoidal 858 rather than a rectangular cross
section. Assume that the matrix is made up of 50% rectangular and 50% trapezoidal passages, as shown in Fig. P12.3. Determine the mass ow fraction distribution and the reduction in heat exchanger eectiveness and p due to passage-
FIGURE P12.3
12.4
Characteristic
Rectangular Passage
Trapezoidal Passage
Re
NuH1
Ao; =Ao;
Dh; =Dh;
20.585
6.490
1
1
15.659
3.256
1
0.9961
FIGURE P12.4
PROBLEMS
Characteristic
* 14
* 16
f Re
NuH1
18.233
5.331
19.702
6.049
861
Note:
Dh
2b*
1 *
Ao; 1=4
1:5 for one passage each only
Ao; 1=6
12.5
12.6
m_ II
1=2n n=2n
De;I 1n=2n Ao;I
aII
II
aI
I
De;II
Ao;II
where aI 0:8, aII 3:44 and n 0:25 and the subscripts I and II denote plate
groups I and II.
Consider a water-to-water counterow single-pass PHE with the aforementioned two groups of plates. Water ow rates on the hot and cold sides are 18
and 10 kg/s, respectively, and the inlet temperatures are 40 and 208C, respectively.
The total fouling resistance and wall resistance are given as 0.00004 and 0.000003
m2 K=W, respectively. The following is additional information.
Plate
Fluid Properties
0:00081 Pa s
995:4 kg=m3
k 0:619 W=m K
cp 4177 J=kg K
Pr 5:47
Consider the same uid properties for water for each plate group and identical
equivalent diameter for both plate groups.
862
(a) Determine the ow distribution of hot and cold uids in plate groups I and II.
(b) Outline a step-by-step procedure to calculate the heat duty of this exchanger.
(c) Compute the heat duty of this PHE. Use the following equation to calculate
the heat transfer coecients:
k 0:646
h 0:724
Re0:583 Pr1=3
Dh 30
12.7
Consider an air-cooled tubular exchanger with engine oil owing in the tubes.
Because of the dierent amount of heat transfer taking place in the tubes, assume
that oil ows at 300 K in 50% of the tubes and at 380 K in the remaining 50% of
the tubes. The objective of this problem is to determine the inuence of viscosityinduced ow maldistribution considering laminar ow in the tubes. Assume that
C* 1 and ntum 1. Since no results are presented in Table 3.6 or 3.3 on the "NTU relationship for a crossow heat exchanger with a nite number of tubes and
dierent NTUs associated with each tube, assume that the exchanger is counterow.
(a) Determine the ow fraction distribution in the two types of passages.
(b) Calculate the change in p due to viscosity-induced ow maldistribution.
Does it represent a loss or a gain in comparison to the nominal or base
case with oil ow in both tubes at 340 K?
(c) Determine the change in exchanger eectiveness due to the ow nonuniformity. Does it represent a loss or a gain?
Use the following viscosity data for the oil and assume its density to be constant.
T K
300
340
380
Pa s
0.486
0.053
0.014
12.8
A gas ows through the circular tubes (lateral branches) under laminar condition
through the U-ow manifold conguration of Fig. 12.7. The two lateral headers
are large enough to feature a negligible pressure drop along the uid ow direction
within each of the headers. For this gas, the dynamic viscosity increases linearly
with temperature. The temperature of the tube wall is either lower or higher than
the temperature of the gas owing inside it, but it is uniform and constant along
the ow length, due to high heat transfer on its outside surface. Determine which
conditions will exist in this heat exchanger: (1) dierent mass ow rates of gas may
be established in dierent tubes for both heating and cooling of the gas owing
through the tubes; (2) such a condition is possible only in the case of heating; or
(3) such a condition is possible only in the case of cooling.
12.9
Derive an analytical expression that describes the geometry of the inlet parallelow oblique header given in Table 12.1. That expression should be written in the
form of a relationship between the local dimensionless header wall coordinate
Z z=yo and the dimensionless coordinate in the ow direction X * x=L, as
presented by Eq. (1) in Example 12.5. The main steps in the derivation are discussed in Example 12.5, but the complete derivation is to be performed for the
solution of this problem.
13
863
864
than that for the metals used for heat transfer surfaces. Fouling is an extremely complex
phenomenon characterized by combined heat, mass, and momentum transfer under
transient conditions. Liquid-side fouling occurs on the exchanger side where liquid is
being heated, and gas-side fouling occurs on the gas cooling side; however, reverse
examples can be found.
Fouling is very costly since it (1) increases capital costs due to the need to oversurface
the heat exchanger and for cleaning; (2) increases maintenance costs resulting from
cleaning, chemical additives, or troubleshooting; (3) results in loss of production due
to shutdown or reduced capacity; and (4) increases energy losses due to reduced heat
transfer, increased pressure drop, and dumping of dirty streams present. Gas-side fouling
can also be a potential re hazard in a fossil-red exhaust environment, resulting in
catastrophic lost production and repair costs. In some applications, increased pressure
drop due to fouling may reduce gas ows aecting adversely heat transfer and increasing
solvent concentration (such as during waste heat recovery from paint oven exhausts)
which is not acceptable environmentally.
Fouling signicantly reduces heat transfer with a relatively small increase in uid
pumping power in systems with liquid ows and high heat transfer coecients. For
systems having low heat transfer coecients, such as with gases, fouling increases the
uid pumping power signicantly with some reduction in heat transfer. Note that plugging will also increase pressure drop substantially but doesnt coat the surface and still
may be considered as fouling in an application.
Let us rst discuss only qualitatively the inuence of a deposit on a heat transfer
surface. We consider either fully developed laminar or turbulent ows. Using the results/
correlations for laminar ow (Nu constant; see Table 7.3) and turbulent ow [the
DittusBoelter correlation, Eq. (7.79) in Table 7.6], we express the heat transfer coecient as follows:
8
Nu k
>
>
>
< D
with Nu constant
0:8
>
k
4m_
>
>
:
Pr0:4
0:023
Dh
P
13:1
for turbulent flow
f
Re
>
< D3h 2gc P
4L G2
p f
Dh 2gc >
>
1 0:046 4L 0:2 4m_ 1:8
>
>
: 3
2gc
P
Dh
865
applied for turbulent ow: f 0:046Re0:2 . The most important physical outcome of
fouling is the ow cross section getting plugged and resulting in a reduced hydraulic
diameter of ow passages. Therefore, for a given mass ow rate m_ , uid ow length L,
heat transfer area A PL, and known uid properties, one gets from Eqs. (13.1) and
(13.2),
h/
1
Dh
p /
1
D3h
13:3
The functional relationships given by Eq. (13.3) are obtained assuming total wetted
perimeter as constant regardless of the change in the hydraulic diameter. In practice,
when Dh for an exchanger changes, A may change as well as for a circular tube. In that
case, since P di Nt Dh Nt for a tubular exchanger (Nt total number of tubes),
p / 1=D4h from Eq. (13.2) for laminar ow and p / 1=D4:8
h for turbulent ow instead
of 1=D3h of Eq. (13.3) for a tubular exchanger. Alternatively, the p expression for
turbulent ow in a circular tube can be expressed as follows using Eq. (6.29) with the
denitions G m_ =Ao , Ao =4di2 and Dh di :
p f
4L G2
4L
1
32Lm_ 2 f
m_ 2
f
di 2gc
di =42 di4 2gc
2 gc di5
13:4
Substitution of f 0:046Re0:2 will change the exponent of m_ in Eq. (13.4) to 1.8 and the
exponent of di to 4.8. Also, the surface roughness change due to fouling on the f factors
should be included as an additional eect (generally, we neglect the eect of surface
roughness on the heat transfer coecient for conservatism). Actual inuences of fouling
on the heat transfer coecient and pressure drop are substantially more complex than
those presented by Eqs. (13.3) and (13.4), due to inherently transient nature of fouling
processes.
The pressure drop ratio pf =pc of a fouled and a clean exchanger for a constant
mass ow rate, from Eq. (13.4), is given by
pf
ff
pc
fc
Dh;c
Dh; f
5
13:5
If we consider that fouling does not aect the friction factor (i.e., fc ff ) and also
consider that reduction in the tube inside diameter due to fouling is only 10 to 20%,
the resulting pressure drop increase will be approximately 69% and 205%, respectively,
according to Eq. (13.5), regardless of whether the uid is liquid or gas [note that in
contrast, h / 1=Dh , as shown in Eq. (13.1) or (13.3)]. This increased p can be translated
into increased uid pumping power using Eq. (6.1); and for liquids, the density being
signicantly higher than that for gases, a substantially higher p due to fouling can be
allowed for liquids for a reasonable increase in liquid pumping power. Also, the equipment cost of uid pumping power is lower for liquids than for gases for a given amount of
pumping power.{
Now let us review the impact of fouling on exchanger heat transfer. As fouling will
reduce the free-ow area and hence the passage Dh , it will increase the convection heat
y
For example, for a midsized automobile, the cost of a 300-W fan for the radiator airow was $35 to 40, compared
to $20 to 25 for an equivalent power radiator coolant water pump in 2001.
866
transfer coecient h [of Eq. (13.1)] between the uid and heat transfer surface (which
may be covered with a fouling layer), for two reasons: increased ow velocity with a
reduction in the ow area, and increased surface roughness due to the fouling layer. Both
these eects would increase the pressure drop substantially. Fouling layers (deposits) on
one or both uid sides increase thermal resistance to heat ow from the hot uid to cold
uid by conduction through the fouling layers (see Fig. 3.4), which also have lower
thermal conductivity. The added thermal resistances in general reduce the exchanger
overall UA substantially compared to the increase in h due to fouling, as mentioned
above.{ Due to a large uncertainty, transient nature, variations in the fouling resistance
^ f 1=hf ), and no accurate means of its measurement, the increase in h due to fouling is
(R
ignored or lumped into the reported values of fouling resistances. Hence, the heat transfer coecients hh and hc for hot and cold uids are determined for unfouled clean
surfaces for the UA computation of fouled surfaces. From the overall thermal resistance
Eq. (3.20) or (3.24), we nd that fouling deposits will reduce UA and hence q more
signicantly in liquids than in gases. This is because liquids have h an order of magnitude
higher than that for gases in general. To understand this, consider a process plant
heat exchanger with clean U 1500 W/m2 K or the overall unit thermal resistance
^ f ;h R
^ f ;c 3 104 [a reasonable
^ o 6 104 m2 K=W. If the fouling resistances R
R
value from TEMA (1999)] are considered, 50% extra heat transfer
^ o;new 6 3 104 m2 K=W and
area is chargeable to fouling since R
^ o;new . In contrast, for a gas-to-gas clean compact heat exchanger, consider
q A Tm =R
^ o 3 103 m2 K=W. For the same fouling resistances,
Uc 300 W=m2 K or R
4 2
^
^
Rf ;h Rf ;c 3 10 m K=W, the heat transfer surface area chargeable to fouling is
only 10%.
Based on the foregoing discussion, fouling in liquids has a signicant detrimental
eect on heat transfer with some increase in uid pumping power. In contrast, fouling
in gases reduces heat transfer somewhat (5 to 10% in general) but increases pressure drop
and uid pumping power signicantly (up to several hundred percent) from the cost
point of view.
It should be emphasized that the same magnitude of a fouling factor (or fouling unit
thermal resistance){ can have a dierent impact on performance for the same or dierent
applications. For example, the same fouling factor may represent heavy fouling in a clean
service (such as a closed-loop refrigerant system) or low fouling in a dirty service (such as
a renery crude preheat train). As another example, the same fouling factor in two
dierent plants may have radically dierent fouling rates because of dierent feedstocks,
preprocessing, or equipment design.
For example, see the added thermal resistance terms 1=o hf A for the hot and cold uid sides in Eq. (3.20) or
(3.24), which may reduce 1=UA more than the increase in hh and/or hc due to fouling, depending on their relative
magnitudes in the equation.
z
The concept of fouling resistance introduced in Section 3.2.4 is explained further in Section 13.2.6.
867
by adding thermal resistances of fouling layers in the thermal circuit using empirical data,
as explained through Fig. 3.4 and discussed further in Section 13.3. The problem, though,
is that this simplied modeling approach does not (and cannot) reect a real transient
nature of the fouling process. The current practice is to use fouling factors or fouling unit
thermal resistances from TEMA Standards (1999) (see Section 13.3 and Table 9.4 for
tubular and shell-and-tube heat exchangers). However, probably a better approach
would be to perform cost analysis for cleaning frequency by taking into account any
initial overdesign (by including fouling resistances). This overdesign may provide added
heat transfer performance initially due to larger surface area and ow area than
required for a clean exchanger but will reduce the ow velocity and hence may accelerate
initial fouling in some applications. Let us now consider in detail dierent types of
fouling mechanisms, sequential events in fouling, and modeling of a fouling process as
an example.
13.2.1
Fouling Mechanisms
There are six types of liquid-side fouling mechanisms: (1) precipitation or crystallization
fouling, (2) particulate fouling, (3) chemical reaction fouling, (4) corrosion fouling, (5)
biological fouling, and (6) freezing (solidication) fouling. Only biological fouling does
not occur in gas-side fouling since there are in principle no nutrients in the gas ows. In
reality, more than one fouling mechanisms is present in many applications and their
synergistic eect makes the fouling even worse than predicted/expected with a single
fouling mechanism present. Note that there are additional examples of fouling that
may not fall in the foregoing categories, such as accumulation of noncondensables in
a condenser. In addition, plugging will also increase pressure drop substantially, but
doesnt coat the surface and still may be considered as fouling in applications. Refer
to Melo et al. (1988) and Bott (1990) for a detailed study of fouling.
In precipitation or crystallization fouling, the dominant mechanism is the precipitation
of dissolved salts in the uid on the heat transfer surface when the surface concentration
exceeds the solubility limit. Thus, a necessary prerequisite for an onset of precipitation is
the presence of supersaturation. Precipitation of salts can occur within the process uid,
in the thermal boundary layer, or at the uidsurface (foulinglm) interface. It generally
occurs with aqueous solutions and other liquids of soluble salts which are either being
heated or cooled. When the solution contains normal solubility salts (the salt solubility
and concentration decrease with decreasing temperature such as wax deposits, gas
hydrates and freezing of water/water vapor), the precipitation fouling occurs on the
cold surface (i.e., by cooling the solution). For inverse solubility salts (such as calcium
and magnesium salts), the precipitation of salt occurs with heating the solution.
Precipitation/crystallization fouling is common when untreated water, seawater,
geothermal water, brine, aqueous solutions of caustic soda, and other salts are used in
heat exchangers. This fouling is characterized by deposition of divalent salts in cooling
water systems. Crystallization fouling may occur with some gas ows that contain small
quantities of organic compounds that would form crystals on the cold surface. If the
deposited layer is hard and tenacious (as often found with inverse solubility salts such as
cooling water containing hardness salts), it is often referred to as scaling. If it is porous
and mushy, it is called sludge, softscale, or powdery deposit. The most important phenomena involved with precipitation or crystallization fouling include the following.
Crystal growth during precipitation require formation of a primary nucleus. The
mechanism controlling that process is nucleation, as a rule heterogeneous in the presence
868
of impurities and on the heat transfer surface. Transfer of particulate solids to the fouled
surface is accomplished by diusion. Simultaneously with deposition, removal phenomena caused by shear stress are always present. Deposit mechanical integrity changes over
time either by strengthening or by weakening it due to crystalization/recrystalization,
temperature change, and so on. All these phenomena are controlled by numerous factors,
the most dominant being local temperature and temperature gradient levels, composition
of the uid including concentration of soluble species.
Particulate fouling refers to the deposition of solids suspended in a uid onto a heat
transfer surface. If the settling occurs due to gravity, the resulting particulate fouling is
called sedimentation fouling. Hence, particulate fouling may be dened as the accumulation of particles from heat exchanger working uids (liquids and/or gaseous suspensions)
on the heat transfer surface. Most often, this type of fouling involves deposition of
corrosion products dispersed in uids, clay and mineral particles in river water, suspended solids in cooling water, soot particles of incomplete combustion, magnetic particles in economizers, deposition of salts in desalination systems, deposition of dust
particles in air coolers, particulates partially present in re-side (gas-side) fouling of
boilers, and so on. The particulate fouling caused by deposition of, for example, corrosion products is inuenced by the following factors: metal corrosion process factors (at
heat transfer surface), release and deposition of the corrosion products on the surface{;
concentration of suspended particles, temperature conditions on the fouled surface
(heated or nonheated), and heat ux at the heat transfer surface.
Chemical reaction fouling is referred to as the deposition of material (fouling precursors) produced by chemical reactions within the process uid, in the thermal boundary
layer, or at the uidsurface (foulinglm) interface in which the heat transfer surface
material is not a reactant or participant. However, the heat transfer surface may act as a
catalyst as in cracking, coking, polymerization, and autoxidation. Thermal instabilities
of chemical species, such as asphaltenes and proteins, can also induce fouling precursors.
Usually, this fouling occurs at local hot spots in a heat exchanger, although the deposits
are formed all over the heat transfer surface in crude oil units and dairy plants. It can
occur over a wide temperature range from ambient to over 10008C (18008F) but is more
pronounced at higher temperatures. Foulant deposits are usually organic compounds,
but inorganic materials may be needed to promote the chemical reaction. This fouling
mechanism is a consequence of an unwanted chemical reaction that takes place during
the heat transfer process. Examples of chemical fouling include deposition of coke in
petrochemical industries in cracking furnaces where thermal cracking of hydrocarbons is
realized. This fouling mechanism is found in many applications of process industry, such
as oil rening, vapor-phase pyrolysis, cooling of gas and oils, polymerization of process
monomers, and so on. Furthermore, fouling of heat transfer surface by biological uids
may involve complex heterogeneous chemical reactions and physicochemical processes.
The deposits from chemical reaction fouling may promote corrosion at the surface if the
formation of the protective oxide layer is inhibited. All fouling deposits may promote
corrosion.
In corrosion fouling (in situ), the heat transfer surface itself reacts with the process
uid or chemicals present in the process uid. Its constituents or trace materials are
carried by the uid in the exchanger, and it produces corrosion products that deposit
on the surface. Hence, corrosion fouling could be considered as chemical reaction fouling
y
It should be borne in mind that corrosion products may be soluble in a working uid, and hence both precipitation and particulate fouling would usually occur concurrently.
869
in which heat transfer fouling aects the exchanger mechanical integrity, and the corrosion products add thermal resistance to heat ow from the hot uid to the cold uid. If
corrosion products are formed upstream of the exchanger and then deposited on the heat
transfer surface, the fouling mechanism refers to particulate or precipitation fouling,
depending on whether the corrosion products are insoluble or soluble at the bulk uid
conditions. The interaction of corrosion and other types of fouling is the major concern
for many industrial applications. Corrosion fouling is dependent on the selection of
exchanger surface material and can be avoided with the right choice of materials (such
as expensive alloys) if the high cost is warranted. Corrosion fouling is prevalent in many
applications where chemical reaction fouling takes place and the protective oxide layer is
not formed on the surface. Corrosion fouling is of signicant importance in the design of
the boiler and condenser of a fossil fuelred power plant. The important factors for
corrosion fouling are the chemical properties of the uids and heat transfer surface,
oxidizing potential and alkalinity, local temperature and heat ux magnitude, and
mass ow rate of the working uid. It should be noted that although growth of corrosion
inuenced deposit has a detrimental eect on heat transfer, this inuence is less important than fouling caused by particulate fouling of corrosion products formed elsewhere
within the system. For example, fouling on the water side of boilers may be caused by
corrosion products that originate in the condenser or feedtrain.
Biological fouling or biofouling results from the deposition, attachment, and growth of
macro- or microorganisms to the heat transfer surface; it is generally a problem in water
streams. In general, biological fouling can be divided into two main subtypes of fouling:
microbial and macrobial. Microbial fouling is accumulation of microorganisms such as
algae, fungi, yeasts, bacteria, and molds, and macrobial fouling represents accumulation
of macroorganisms such as clams, barnacles, mussels, and vegetation as found in seawater or estuarine cooling water. Microbial fouling precedes macrobial deposition as a
rule and may be considered of primary interest. Biological fouling is generally in the form
of a biolm or a slime layer on the surface that is uneven, lamentous, and deformable
but dicult to remove. Although biological fouling could occur in suitable liquid
streams, it is generally associated with open recirculation or once-through systems
with cooling water. Since this fouling is associated with living organisms, they can
exist primarily in the temperature range 0 to 908C (32 to 1948F) and thrive in the
temperature range 20 to 508C (68 to 1228F). Biological fouling may promote corrosion
fouling under the slime layer. Transport of microbial nutrients, inorganic salts, and
viable microorganisms from the bulk uid to the heat transfer surface is accomplished
through molecular diusion or turbulent eddy transport, including organic adsorption at
the surface.
Freezing or solidication fouling is due to freezing of a liquid or some of its constituents, or deposition of solids on a subcooled heat transfer surface as a consequence of
liquidsolid or vaporsolid phase change in a gas stream. Formation of ice on a heat
transfer surface during chilled water production or cooling of moist air, deposits formed
in phenol coolers, and deposits formed during cooling of mixtures of substances such as
paran are some examples of solidication fouling (Bott, 1981). This fouling mechanism
occurs at low temperatures, usually ambient and below depending on local pressure
conditions. The main factors aecting solidication fouling are mass ow rate of the
working uid, temperature and crystallization conditions, surface conditions, and concentration of the solid precursor in the uid.
Combined fouling occurs in many applications, where more than one fouling mechanism is present and the fouling problem becomes very complex with their synergistic
870
13.2.2
871
Precipitation
Freezing
Particulate
Chemical
Corrosion
Biological
"#
#$
"
"
"
"
$
$
#
"#
"
#
"
"
$
$
"#$
#
"#
"
"$
"#
#
"
"
"#
"#$
"#
"$
"
"
"#$
"#
"#
"
"#
"#
is the combined eect of more than one fouling mechanism present. The qualitative
eects of some of the operating variables on these fouling mechanisms are shown in
Table 13.1.
The quantitative eect of fouling on heat transfer can be estimated by utilizing the
concept of fouling resistance and calculating the overall heat transfer coecient under
both fouling and clean conditions (see Section 13.3). An additional parameter for determining this inuence, used frequently in practice, is the cleanliness factor. It is dened as a
ratio of an overall heat transfer coecient determined for fouling conditions to that
determined for clean (fouling-free) operating conditions. The eect of fouling on the
pressure drop can be determined by the reduced free-ow area due to fouling and the
change in the friction factor, if any, due to fouling.
13.2.3
Gas-side fouling may be caused by precipitation (scaling), particulate deposition, corrosion, chemical reaction, and freezing. Formation of hard scale from the gas ow occurs if
a suciently low temperature of the heat transfer surface forces salt compounds to
solidication. Acid vapors, high-temperature removal of an oxide layer by molten ash,
or salty air at low temperatures may promote corrosion fouling. An example of particulate deposition is accumulation of plant residues. An excess of various chemical substances, such as sulfur, vanadium, and sodium, initiates various chemical reaction fouling
problems. Formation of frost and various cryo-deposits are typical examples of freezing
fouling on the gas side. An excellent overview of gas-side fouling of heat transfer surfaces
is given by Marner (1990, 1996). Qualitative eects of some of the operating variables on
gas-side fouling mechanisms are presented in Table 13.2.
13.2.4
Small channels associated with compact heat exchangers have very high shear rates,
perhaps three to four times higher in a plate heat exchanger than in a shell-and-tube
872
Particulate
Freezing
Chemical
Corrosion
"#
"#$
"
"
"$
$
#
#
#
"
"
"#$
"
"
"
"#$
"$
"
"$
"
exchanger. This reduces fouling signicantly. However, small channel size creates a
problem of plugging the passages. To avoid plugging, the particle size must be restricted
by ltering or other means to less than one-third the smallest opening of heat exchanger
passages. Even with this guideline, particulate fouling can occur and agglomerate, such
as with waxy substances.
13.2.5
From the empirical evidence involving various fouling mechanisms discussed in Section
13.2.1, it is clear that virtually all these mechanisms are characterized by a similar
sequence of events. The successive events occurring in most cases are the following: (1)
initiation, (2) transport, (3) attachment, (4) removal, and (5) aging, as conceptualized by
Epstein (1978). These events govern the overall fouling process and determine its ultimate
impact on heat exchanger performance. In some cases, certain events dominate the
fouling process, and they have a direct eect on the type of fouling to be sustained.
Let us summarize these events briey (Cannas, 1986).
Initiation of the fouling, the rst event in the fouling process, is preceded by a delay
period or induction period d as shown in Fig. 13.1. The basic mechanism involved during
this period is heterogeneous nucleation, and d is shorter with a higher nucleation rate.
The factors aecting d are temperature, uid velocity, composition of the fouling
stream, and nature and condition of the heat exchanger surface. Low-energy surfaces
(unwettable) exhibit longer induction periods than those of high-energy surfaces (wettable). In crystallization fouling, d tends to decrease with increasing degree of supersaturation. In chemical reaction fouling, d appears to decrease with increasing surface
temperature. In all fouling mechanisms, d decreases as the surface roughness increases
due to available suitable sites for nucleation, adsorption, and adhesion.
Transport of species means transfer of a key component (such as oxygen), a crucial
reactant, or the fouling species itself from the bulk of the uid to the heat transfer surface.
Transport of species is the best understood of all sequential events. Transport of species
takes place through the action of one or more of the following mechanisms:
873
. Diusion: involves mass transfer of the fouling constituents from the owing uid
toward the heat transfer surface due to the concentration dierence between the
bulk of the uid and the uid adjacent to the surface.
. Electrophoresis: under the action of electric forces, fouling particles carrying an
electric charge may move toward or away from a charged surface depending on the
polarity of the surface and the particles. Deposition due to electrophoresis increases
with decreasing electrical conductivity of the uid, increasing uid temperature,
and increasing uid velocity. It also depends on the pH of the solution. Surface
forces such as Londonvan der Waals and electric double layer interaction forces
are usually responsible for electrophoretic eects.
. Thermophoresis: a phenomenon whereby a thermal force moves ne particles in
the direction of negative temperature gradient, from a hot zone to a cold zone.
Thus, a high-temperature gradient near a hot wall will prevent particles from
depositing, but the same absolute value of the gradient near a cold wall will promote particle deposition. The thermophoretic eect is larger for gases than for
liquids.
. Diusiophoresis: involves condensation of gaseous streams onto a surface.
. Sedimentation: involves the deposition of particulate matters such as rust particles,
clay, and dust on the surface due to the action of gravity. For sedimentation to
occur, the downward gravitational force must be greater than the upward drag
force. Sedimentation is important for large particles and low uid velocities. It is
frequently observed in cooling tower waters and other industrial processes where
rust and dust particles may act as catalysts and/or enter complex reactions.
. Inertial impaction: a phenomenon whereby large particles can have sucient
inertia that they are unable to follow uid streamlines and as a result, deposit on
the surface.
. Turbulent downsweeps: since the viscous sublayer in a turbulent boundary layer is
not truly steady, the uid is being transported toward the surface by turbulent
downsweeps. These may be thought of as suction areas of measurable strength
distributed randomly all over the surface.
Attachment of the fouling species to the surface involves both physical and chemical
processes, and it is not well understood. Three interrelated factors play a crucial role in
the attachment process: surface conditions, surface forces, and sticking probability. It is
the combined and simultaneous action of these factors that largely accounts for the event
of attachment.
. The properties of surface conditions important for attachment are the surface free
energy, wettability (contact angle, spreadability), and heat of immersion. Wettability
and heat of immersion increase as the dierence between the surface free energy of
the wall and the adjacent uid layer increases. Unwettable or low-energy surfaces
have longer induction periods than wettable or high-energy surfaces, and suer less
from deposition (such as polymer and ceramic coatings). Surface roughness
increases the eective contact area of a surface and provides suitable sites for nucleation and promotes initiation of fouling. Hence, roughness increases the wettability of
wettable surfaces and decreases the unwettability of the unwettable ones.
. There are several surface forces. The most important one is the Londonvan der
Waals force, which describes the intermolecular attraction between nonpolar mole-
874
cules and is always attractive. The electric double layer interaction force can be
attractive or repulsive. Viscous hydrodynamic force inuences the attachment of a
particle moving to the wall, which increases as it moves normal to the plain surface.
. Sticking probability represents the fraction of particles that reach the wall and stay
there before any reentrainment occurs. It is a useful statistical concept devised to
analyze and explain the complicated event of attachment.
Removal of the fouling deposits from the surface may or may not occur simultaneously with deposition. Removal occurs due to the single or simultaneous action of
the following mechanisms: shear forces, turbulent bursts, re-solution, and erosion.
. Shear forces result from the action of the shear stress exerted by the owing uid on
the depositing layer. As the fouling deposit builds up, the cross-sectional area for
ow decreases, thus causing an increase in the average velocity of the uid for a
constant mass ow rate and increasing the shear stress. Fresh deposits will form
only if the deposit bond resistance is greater than the prevailing shear forces at the
soliduid interface.
. Randomly distributed (about less than 0.5% at any instant of time) periodic turbulent bursts act as miniature tornadoes lifting deposited material from the surface.
By continuity, these uid bursts are compensated for by gentler uid back sweeps,
which promote deposition.
. The removal of the deposits by re-solution is related directly to the solubility of the
material deposited. Since the fouling deposit is presumably insoluble at the time of
its formation, dissolution will occur only if there is a change in the properties of the
deposit, or in the owing uid, or in both, due to local changes in temperature,
velocity, alkalinity, and other operational variables. For example, suciently high
or low temperatures could kill a biological deposit, thus weakening its attachment
to a surface and causing sloughing or re-solution. The removal of corrosion deposits in power-generating systems is done by re-solution at low alkalinity. Re-solution
is associated with the removal of material in ionic or molecular form.
. Erosion is closely identied with the overall removal process. It is highly dependent
on the shear strength of the foulant and on the steepness and length of the sloping
heat exchanger surfaces, if any. Erosion is associated with the removal of material
in particulate form. The removal mechanism becomes largely ineective if the
fouling layer is composed of well-crystallized pure material (strong formations);
but it is very eective if it is composed of a large variety of salts each having
dierent crystal properties.
Aging of deposits begins with attachment on the heat transfer surface, and refers to
any changes the fouling material undergoes as time elapses. The aging process includes
both physical and chemical transformations, such as further degradation to a more
carbonaceous material in organic fouling, and dehydration and/or crystal phase transformations in inorganic fouling. A direct consequence of aging is change in the thermal
conductivity of the deposits.{ Aging may strengthen or weaken the fouling deposits.
y
A common nonfouling example of aging is the transformation of fresh, soft, uy snow in an open eld into
hard, crystalline, yellowish ice after a week or so of exposure to the sun resulting dierences in its material
properties.
13.2.6
875
Regardless of the type of fouling process, the principal characteristic feature of any type
of fouling is that the net mass fouling rate (i.e., the change of the mass m of foulant
deposited on the heat transfer surface for a given time, dm=d, is a consequence of a net
dierence between the foulant deposit rate m_ d and the foulant reentrainment rate m_ r :
@ms;
m_ d s; m_ r s;
@
13:6
In Eq. (13.6), s denotes symbolically the spatial dependence (say, x, y, and z) of the mass
of foulant. Note that the mass m of the foulant deposited uniformly is given as a simple
equation:
m f Af
13:7
where f represents foulant mass density, A denotes heat transfer surface area covered
with the foulant, and f is the thickness of the foulant layer. In general, all three terms of
Eq. (13.6) are spatially nonuniform and dependent on time. Equation (13.6) can conveniently be reformulated in terms of mass per unit heat transfer surface area,
MA m=A, and for a uniform spatial distribution of deposit, it is
dMA
M_ A;d M_ A;r
d
13:8
Equation (13.8) is a direct consequence of Eq. (13.6) after idealizing a uniform distribution of the fouling deposit over the surface A. Furthermore, mass per unit heat transfer
surface (uniformly distributed along the heat transfer surface) can be written as
^f
MA f f f kf R
13:9
^ f f =kf , the fouling factor, represents fouling unit thermal resistance; it reprewhere R
sents the thermal resistance of the layer of foulant deposited for a unit area of heat
transfer surface. Concisely, we refer to this entity as fouling resistance. From the fouling
^ f . Consequently,
factor denition, we obtain f kf R
^f
df
dR
dMA
f
f kf
d
d
d
13:10
In Eq. (13.10), it is assumed that both mass density and thermal conductivity of the
deposited layer are invariant with time. Combining Eqs. (13.8) and (13.10), we obtain
^f
dR
^_
^_ R
R
d
r
d
13:11
876
models for determining these variables have been developed; some of them are summarized in Table 13.3. Let us consider the model of Taborek et al. enlisted in that table as an
illustration.
According to Taborek et al. [as reported by Epstein (1978)], the deposition and
removal mass rates have the form
E
M_ A;d c1 }1 n exp
<Ts
mi
M_ A;r b1 s
13:12
^_ b f b kf R
^f
R
r
2 s
2 s
13:13
In Eq. (13.13), c2 and b2 are constants. It should be noted that both sets of equations
[Eqs. (13.12) and (13.13)], are semiempirical, to include the variables that govern fouling.
Introducing the expressions for mass per unit heat transfer surface area for both deposit
and reentrainment processes (or their thermal resistance) into Eq. (13.8) or (13.11), we
could integrate these governing equations and subsequently determine either deposited
mass or their thermal resistance. These solutions have to t empirical evidence that can
be generalized as presented in Fig. 13.1.
In Fig. 13.1, four characteristic scenarios for the growth of the fouling resistance are
presented (Knudsen, 1998). In this gure, d is the delay period for the onset of fouling
^f .
deposits for non-negligible R
^ f is linearly dependent on time) indicate that the
1. Linear characteristics (i.e., R
deposition rate is constant and there is no reentrainment rate (or at least their
difference is invariant in time). A linear fouling behavior is generally associated
with the crystallization of a well-formed deposit consisting of a substantially pure
salt that is largely uncontaminated by the presence of coprecipitated impurities.
The strong bonds characterizing the structure of such deposits make removal
mechanisms somewhat ineffective. If heat duty is kept constant linear fouling
behavior is also often observed for reaction fouling.
2. Falling rate fouling normally occurs in situations where the deposition rate is
decreasing but always greater than the removal rate. This type of fouling mechanism has been observed in crystallization fouling in a plate exchanger and also in
particulate fouling.
3. The curve characterized by asymptotic behavior reects the situation represented
^_ of Eq. (13.13), which corresponds to fragile deposits
by the expression for R
d
exposed to shear stress of the owing uid. The asymptotic fouling growth
model is often observed in cooling water heat exchangers. In these heat exchangers,
the conditions leading to the formation of a scale layer of a weak, less coherent
877
Commentsc
a
Specic symbols used in this table are all local and refer to the source cited for the units and detailed physical meaning. The list of models in this table illustrates the variety of models
reported in the literature.
b
D. Q. Kern and R. E. Seaton (1959a), Brit. Chem. Eng., Vol. 4, p. 258; D. Q. Kern and R. E. Seaton (1959b), Chem. Eng. Progr., Vol. 55, p. 71; J. W. Bartlett (1968), BNWL-676,
UC-80, Reactor Technology, Battelle Northwest, Richland, Wash.; D. H. Charlesworth (1970), Chem. Eng. Prog. Symp. Ser., Vol. 66, No. 104, p. 21; S. K. Beal (1970), Nucl. Sci.
Eng., Vol. 40, p. 1; S. K. Beal (1972), Paper 76-C, 65th Annual Meeting AIChE; S. K. Beal (1973), Trans. Am. Nucl. Soc., Vol. 17, p. 163; K. A. Burrill (1977), Can. J. Chem. Eng., Vol.
55, p. 54; J. Taborek et al. (1972), Chem. Eng. Progr., Vol. 68, pp. 59, 69; J. Taborek and R. B. Ritter (1972), Paper 76-A, 65th Annual Meeting, AIChE, New York; W. G. Characklis
(1990) Biolm Process, in Biolms, W. G. Characklis and K. C. Marshall, eds., Wiley, New York, pp. 195231; C. B. Panchal et al. (1997), in Fouling Mitigation of Industrial HeatExchange Equipment, C. Panchal, ed., Begell House, New York, pp. 201212.
c
_ d and M_ r .
Specic fouling/deposition/removal types are mentioned for each model for M
k1 k2 m2
b6 s m
k1 m
Km FCm km
b1 s
Taborek et al. (1972),
Taborek and Ritter (1972)
Characklis (1990)
Panchal et al. (1997)
Burrill (1977)
b5 Cs Cb m
b4 mloose
c1 }1 n eE=<Ts
Cb
1=km 1=Suw
bm
K d Cb
K d Cb
Bartlett (BNWL)
b3 M_ d
Kd Cb km Cb
Charlesworth (1970)
b0 s m
aub Cb
Sourceb
Reentrainment
Flux
M_ r
Deposition Flux
M_ d
878
Linear dependence
Decreasing
deposition rate
Rf
Asymptotic
Delay
Sawtooth pattern
d
Time
879
Assumptions: The assumptions invoked by the model of Taborek et al. (see Table 13.3),
as presented in Section 13.2.6, are valid. That includes the fact that all parameters and
variables for the problem are invariant in time.
Analysis: According to Eqs. (13.11) and (13.13), the model of fouling process is
^f
dR
kf ^
E
b2 s R
c2 }1 n exp
f
d
<Ts
at
0
The initial condition dened by Eq. (2) deserves an additional comment. In most fouling
cases, fouling resistance is often noticed after a certain delay period (i.e., for 0 d ,
where d represents the delay period for the onset of fouling deposits or a buildup of the
fouling resistance; see Fig. 13.1). This is attributed to simultaneous inuence of both
initial nucleation of the deposited material on the heat transfer surface and its inuence
on heat transfer reduction due to lower thermal conductivity of the foulant material.
Consequently, d does not represent the delay of an actual fouling process, but it signies
a delay in reduction of the heat transfer rate due to fouling. In our analysis we treat
d 0.
A solution of the problem dened by Eqs. (1) and (2) can easily be found by using any
of techniques for solving this linear, rst-order ordinary dierential equation. Let us
introduce the set of new parameters dened as follows:
a b2 s
kf
b c2 }1 n eE=<Ts
Integrating this linear rst-order ordinary dierential equation and simplifying, we get
^ f b Cea
R
a
The integration constant C in Eq. (5) can be determined by applying the initial condition
of Eq. (2) to Eq. (5):
C
c2 }1 n eE=<Ts
b2 s kf =
880
Equation (7) represents the time history of the fouling thermal resistance. For
^ f ;!1 is
large enough time ( ! 1), an asymptotic value of thermal resistance, R
obtained:
n E=<Ts
^ f ;!1 lim R
^ f c2 }1 e
R
!1
b2 s kf =
Finally, Eq. (7) can be rearranged by using Eq. (8) in a more convenient form as
follows:
^f
R
1 e=c
^
Rf ;!1
therefore;
447 h
Ans:
Discussion and Comments: In this example, we did not emphasize the inuence of
physical variables that are inherent in the original model [all the variables and
constants introduced in Eq. (13.12)] since the values of these variables and constants
are actually not known. Still, the model based on Eq. (9) describes quite well some
fouling processes (e.g., particulate or crystallization fouling) having an asymptotic thermal resistance represented by a time constant. Note that when c has a large value, Eq. (9)
reduces to
^f
R
^
Rf ;!1
c
This equation represents the limiting value; here higher-order terms are neglected. Thus,
^ f depends linearly on . For all cases for which the
in this case, the fouling resistance R
parameters of the dierential equation given by Eq. (1) are constant (as implied by the set
of assumptions in this example), the fouling process assumes the known deposit, dened
uid quality, and xed-ow conditions. It should be added that many other models of
fouling processes are available, as summarized by Epstein (1978). Most of them provide
^_ and R
^_ of Eq. (13.11). If these are known, the procedure for
the expressions for R
d
r
obtaining the time history of a fouling process will be similar to that demonstrated in
this example after utilizing the expressions specied for deposition and removal fouling
resistance rates and integrating the resulting dierential equation. All these solutions, as
a rule, can be treated at best as indicators of the fouling trend. However, the complexity
of the process and involved nonlinearities (not included in the simple model discussed)
prevent a reliable prediction.
881
13.3.1
As we have introduced in Section 3.2.4, the overall thermal resistance for a heat exchanger involves a series of thermal resistances from the hot uid to the cold uid, including
thermal resistances due to fouling on both uid sides, as shown in Fig. 3.4. If the overall
heat transfer coecient is based on the uid 1-side heat transfer surface area A1 , the
following relation holds in the absence of ns on both uid sides:
1
U1
1
^ f ;1
R
h1
w A1
1
^ f ;2 A1
R
kw A w
A2
h2
13:14
In Eq. (13.14), it is assumed that the wall thermal resistance is for a at plate wall [see Eq.
(3.31) for a tubular surface]. For a more general case, see Eq. (3.30), which includes the
n eects on both uid sides. Equation (13.14) is further rearranged and simplied as
1
1
^ f ;1 R
^ f ;2 A1 w A1 1 A1 1 R
^f R
^ w A1 1 A1
R
A2 kw Aw h2 A2 h1
Aw h2 A2
U1 h1
13:15
^ f ;1 R
^ f ;2 A1 =A2 represents the total fouling resistance, a sum of
^f R
Note that R
fouling resistances on both sides of the heat transfer surface, as shown. It should again
be reiterated that the aforementioned reduction in the overall heat transfer coecient due
to fouling does not take into consideration the transient nature of the fouling process.
According to Chenoweth (1990), use of the fouling resistance concept as represented by
Eqs. (13.14) and (13.15) must be based on the following recommendations:
. Fouling resistances should reect fouling alone and not uncertainties in the design
of the heat exchanger.
. Appropriate values of fouling resistances should be based on operating experience
and modied by economic considerations where possible.
. The buyer/user, not the manufacturer, should be responsible for selecting the fouling resistances because he or she may know his or her own application better.
882
. The eects of corrosion fouling and biofouling due to their complexity and a
questionable predictability should always be controlled externally. That means
that this control should be a system based on reducing or preventing fouling.
The current practice is to assume a value for the fouling resistance on one or both uid
sides as appropriate and to design a heat exchanger accordingly by providing extra
surface area for fouling, together with a cleaning strategy (see Section 13.4.3).
The complexity in controlling a large number of internal and external factors of a
given process makes it very dicult to predict the fouling growth as a function of time
using deterministic (well-known kinetic) models. A more realistic fouling growth model
can be devised by postulating fouling as a time-dependent random process{ and analyzing using the probabilistic approaches (Zubair et al., 1997) in conjunction with the
cleaning strategies as discussed in Section 13.4.3.
A note of caution is warranted at this point. There is an ongoing discussion among
scholars and engineers from industry as to whether either fouling resistance or fouling
rate concepts should be used as the most appropriate tool in resolving design problems
incurred by fouling. One suggestion in resolving this dilemma would be that the design
fouling-resistance values used for sizing heat exchangers be based on fouling-rate data
and estimated cleaning-time intervals (Rabas and Panchal, 2000).
13.3.2
In current practice, based on application and need, the inuence of fouling on exchanger
heat transfer performance can be evaluated in terms of either (1) required increased
surface area for the same q and Tm , (2) required increased mean temperature dierence
for the same q and A, or (3) reduced heat tranfer rate for the same A and Tm .{ For these
approaches, we now determine expressions for Af =Ac , Tm; f =Tm;c and qf =qc as
follows.} In the rst two cases, the heat transfer rate in a heat exchanger under clean
and fouled conditions are the same. Hence,
q Uc Ac Tm Uf Af Tm
13:16
Therefore,
Af
U
c
A c Uf
13:17
According to Eq. (13.15), the relationships between overall heat transfer coecients
(based on tube outside surface area) and thermal resistances for clean and fouled conditions are dened as follows. For a clean heat transfer surface,
1
1
^ w Ao 1 Ao
R
Aw hi;c Ai
Uc ho;c
13:18
y
The randomness in fouling process is due to time-dependent scatter in fouling resistance from a replicate to a
replicate (repeat tests).
z
The rst case is the design of an exchanger where an allowance for fouling can be made at the design stage by
increasing surface area. The other two cases are for an already designed exchanger in operation, and the purpose is
to determine the impact of fouling on exchanger performance.
}
Throughout this chapter, the subscript c denotes a clean surface and f the fouled surface.
883
R
Aw hi; f Ai ho;c
Aw hi;c Ai
Uf ho;f
13:19
Note that we have idealized that ho; f ho;c ; hi; f hi;c ; Ai; f Ai;c Ai , and
Ao; f Ao;c Ao . Here Ao represents the tube outside surface area and not free-ow
area in the exchanger. The dierence between Eqs. (13.18) and (13.19) is
^f 1 1
R
Uf Uc
13:20
It should be added that Eq. (13.20) is valid as long as clean overall heat transfer coecients are constant. If this assumption is not satised, the right-hand side in Eq. (13.20)
does not represent only the overall fouling resistance but a quantity that includes other
inuences on overall heat transfer coecients in addition to fouling. In that case, the
fouling assessment will be incorrect. Combining Eqs. (13.17) and (13.20), we get
Af
^f 1
Uc R
Ac
13:21
Similarly, when q and A are the same and Tm is dierent for clean and fouled
exchangers, we have
q Uc Ac Tm;c Uf Ac Tm; f
for constant A
13:22
Hence,
Tm; f
U
c
Tm; c Uf
13:23
13:24
Finally, if one assumes that heat transfer area and mean temperature dierences are
xed, heat transfer rates for the same heat exchanger under fouled and clean conditions
are given by qf Uf A Tm and qc Uc A Tm , respectively. Combining these two relationships with Eq. (13.20), we get
qf
1
^f 1
qc Uc R
13:25
13:26
1000
1000
100
100
Uc, (W/m2 K)
10,800
10
10
3,600
1,800
900
600
300
A or T lines
q lines
884
0.1
0.1
1.0 105
1.0 104
1.0 103
FIGURE 13.2 Percent change in heat transfer area, mean temperature dierence, or heat duty vs.
fouling unit thermal resistance for a fouled exchanger.
We nd that the right-hand sides of Eqs. (13.21), (13.24), and (13.26) are the same.
Equations (13.21), (13.24), and (13.25) are shown in Fig. 13.2 in terms of the percentage
increase in A and Tm and the percentage reduction in q for the fouled exchanger over
that for the clean exchanger. From this gure, it is clear that fouling has a signicant
^ f and/or Uc .
impact on the exchanger performance for high values of R
^ f as
The cleanliness factor CF is related to the fouling resistance R
CF
Uf
1
^ f Uc
Uc 1 R
13:27
Example 13.2 Overall heat transfer coecient of a heat exchanger operating under
clean conditions is calculated as 800 W/m2 K. Following industrial experience, the
cleanliness factor for this exchanger is 0.7. Determine the magnitude of the corresponding fouling resistance.
SOLUTION
Problem Data: The following data are given: Uc 800 W/m2 K and CF 0:7.
^ f of the deposit formed by this heat exchanger.
Determine: The fouling resistance R
Assumptions: The convective heat transfer coecients on the hot- and cold-uid sides are
the same for both fouled and clean heat transfer surfaces. Thermal resistance of the wall
is unchanged under fouled conditions. The change in heat transfer surface areas due to
fouling deposit formation is negligible. The n eciency is equal to unity. All idealizations adopted for heat exchanger design theory are valid (see Section 3.2.1).
Analysis: The relationship between fouling resistances and overall heat transfer coecients for clean and fouled conditions is given by Eq. (13.20):
^f 1 1
R
Uf Uc
885
1
1
1 1 CF
CF Uc Uc Uc CF
Ans:
Discussion and Comments: In some industries (such as power industry), use of the cleanliness factor has been prevalent for assessing the inuence of fouling. The reason for this
is the practice of the industry and the diculties associated with experimental determination of fouling thermal resistances (Somerscales, 1990). Equation (2) can be used to
calculate fouling resistance or unit thermal resistance if the cleanliness factor is known
(or vice versa) under the conditions governed by the above-mentioned assumptions.
Example 13.3 Determine how much will change the required heat transfer area of an
exchanger under fouling conditions if the fouling resistance changes from 104 m2 K=W
to 103 m2 K=W. The heat transfer rate and mean temperature dierence remain the
same, and Uc 1000 W=m2 K. Consider no extended surface on either uid side of the
exchanger.
SOLUTION
Problem Data: The following data are given:
Uc 1000 W=m2 K
qc qf
^ f ;1 104 m2 K=W
R
Tm;c Tm; f
^ f ;2 103 m2 K=W
R
o;1 o;2 1
Determine: The change in heat transfer surface area required if the fouling resistance
changes from 104 m2 K=W to 103 m2 K=W.
Assumptions: The convection heat transfer coecients are the same for fouled and clean
heat transfer surfaces. The thermal resistance of the wall is unchanged under fouled
conditions. Change in heat transfer surface areas due to deposit formation is negligible.
All assumptions adopted for heat exchanger design theory are valid (see Section 3.2.1).
Analysis: The heat transfer rate and mean temperature dierences in this exchanger
under clean and fouled conditions are the same. Hence, from Eq. (13.21), we get
Af
^f 1
Uc R
Ac
^ f ;2 causes
^ f ;1 to R
From Eq. (1), it follows that the change in total fouling resistance from R
a change in heat transfer area from Af ;1 to Af ;2 as follows:
886
Af ;1
^ f ;1 1 and
Uc R
Ac
Af ;2
^ f ;2 1
Uc R
Ac
Ans:
Thus an increase in the fouling resistance by a factor of 10 requires a surface area increase
for this exchanger of 82%.
Discussion and Comments: This example clearly shows a signicant increase in the surface
area requirement for this exchanger when the total fouling resistance is increased by an
order of magnitude. Inversely, a signicant reduction in surface area can be achieved (by
about one-half) if the total fouling resistance is reduced by an order of magnitude. Note
that the result provides a direct information on how large percent change in heat transfer
area would be compared to that for a clean heat exchanger for a given fouling resistance,
as shown in Fig. 13.2.
13.3.3
Empirical data for fouling resistances have been obtained over many decades by industry
since its rst compilation by TEMA in 1941 for shell-and-tube heat exchangers. Selected
data are summarized in Table 9.4 and hence are not repeated here. Many of the original
values of TEMA fouling factors or fouling resistances established in 1941 for a typical
exchanger service length of three months are still in use for a current typical service length
of ve years (Chenoweth, 1990)! TEMA fouling resistances are supposed to be representative values, asymptotic values, or those manifested just before cleaning to be performed. Chenoweth (1990) analyzed the current practice of customers specifying fouling
resistances on their specication sheets to manufacturers. He compiled the combined
shell- and tube-side fouling resistances (by summing each side entry) of over 700 shelland-tube heat exchangers and divided them into nine combinations of liquid, two-phase,
and gas on each uid side regardless of the applications. He then simply took the arith^ f for each two-uid combination value and plotted dimensional
metic average of total R
values in his Fig. I-3. His results are presented in Fig. 13.3 after normalizing with respect
^ f of liquidliquid applications so that
to the maximum combined shell- and tube-side R
the ordinate ranges between 0 and 1. For gases on both sides, the relative fouling resistance can be as high as 0.5 (the lowest value in Fig. 13.3) compared to liquids on both
sides (relative total fouling resistance represented as 1.0 in Fig. 13.3). If liquid is on the
shell side and gas on the tube side, the relative fouling resistance is 0.65. However, if
liquid is on the tube side and gas on the shell side, it is 0.75. Since many process industry
applications deal with liquids that are dirtier than gases, the general practice is to specify
larger fouling resistances for liquids compared to those for the gases. Also, if fouling is
anticipated on the liquid side of a liquidgas exchanger, it is generally placed in the tubes
for cleaning purposes and a larger fouling resistance is specied. These trends are clear
from Fig. 13.3. It should again be emphasized that Fig. 13.3 indicates the current practice
and has no scientic basis. Specication of larger fouling resistances for liquids (which
887
Relative 1
resistance
0.8
0.6
0.4
Liquid
0.2
0
Liquid
Two-phase
Shell side
Two-phase
Tube side
Gas
Gas
FIGURE 13.3 Combined tube- and shell-side relative fouling resistances. (Based on data from
Chenoweth, 1990.)
have higher heat transfer coecients than those of gases) has even more impact on the
surface area requirement for liquidliquid exchangers than for gasgas exchangers.
It should be reiterated that the recommended fouling resistances are believed to
represent typical fouling resistances for design. Consequently, sound engineering judgment has to be made for each selection of fouling resistances, keeping in mind that actual
values of fouling resistances in any application can be either higher or lower than the
resistances calculated. Finally, it must be clear that fouling resistances, although recommended following the empirical data and a sound model, are still constant, independent
^ f selected repreof time, while fouling is a transient phenomenon. Hence, the value of R
sents a correct value only at one specic time in the exchanger operation. As indicated by
Chenoweth (1990): . . . the new proposed (constant, independent of time) values reect a
careful review and the application of good engineering judgment by a group of knowledgeable engineers involved with the design and operation of shell-and-tube heat exchangers. . . . It needs to be emphasized that the tables may not provide the applicable values
for a particular design. They are only intended to provide guidance when values from
direct experience are unavailable. With the use of nite fouling resistance, the overall U
value is reduced, resulting in a larger surface area requirement, larger ow area, and
reduced ow velocity which inevitably results in increased fouling. Thus, allowing more
surface area for fouling in a clean exchanger may accelerate fouling initially.
Typical fouling resistances are roughly 10 times lower in plate heat exchangers than in
shell-and-tube heat exchangers (Zubair and Shah, 2001). Some fouling resistances for
PHEs are compared with those for shell-and-tube heat exchangers in Table 13.4.
TEMA (1999) presents fouling resistances for some gases used in process and petrochemical industries and Marner and Suitor (1987) summarize the literature data for gases
used in many industries, as reported in Table 13.5.
Example 13.4 A heat exchanger with water-to-phase change uid is designed keeping in
mind that the uid that changes its phase must be on the outside of the tube. The
empirical data available reveal that an average heat transfer coecient on the water side
is 2715 W/m2 K. On the tube outside, the heat transfer coecient is 3200 W/m2 K.
888
PHEs
^ f -TEMA
R
0.018
0.044
0.026
0.044
0.053
0.0180.053
0.009
0.180.35
0.180.35
0.180.35
0.350.53
0.36
0.36
0.18
^ f m2 K=kW)a
TABLE 13.5 Gas-Side Fouling Resistances R
Clean gas
Natural gas
Propane
Butane
Gas turbine
Average gas
No. 2 oil
Dirty gas
No. 6 oil
Crude oil
Residual oil
Coal
Miscellaneous
Sodium-bearing waste
Metallic oxides
FCCU catalyst nes
Weierman
(1982)
Zink
(1981)
TEMA
(1978)
Rogalski
(1979)
Henslee and
Bouge (1983)
0.08810.528
0.1760.528
0.1760.528
0.176
0.176
0.3520.704
0.264
0.528
0.528
1.76
0.5286.69
21.124.7
0.5281.23
0.7042.64
0.8813.52
0.8818.81
0.881
1.76
5.28
1.76
1.41
Tubes are made of steel with thermal conductivity of 40 W/m K. The tube outside
diameter is 19 mm with 1.6 mm wall thickness. The asymptotic value of the fouling
resistance on the water side is 4 104 m2 K=W. There is no fouling on the tube outside.
Based on past experience, the fouling phenomenon is of asymptotic nature for this
exchanger, and the time constant for the fouling process is 280 hours for the Kern
Seaton model. Determine percentage unit thermal resistance distribution contributing
to the overall unit thermal resistance for the following two cases: (a) after 280 hours of
fouling initiation, and (b) for the asymptotic fouling condition.
889
SOLUTION
Problem Data: The following data are provided:
hi 2715 W=m2 K
ho 3200 W=m2 K
^ f ;i 4 10
R
4
w 1:6 mm
kw 40 W=m K
m K=W for ! 1
do 19 mm
c 280 h
Determine: The distribution of thermal resistances for two cases: (1) ! 1, and (2)
c , where c is the time constant for asymptotic fouling.
Assumptions: The set of assumptions introduced for heat exchanger analysis (see Section
3.1.2) and the KernSeaton model (see Section 13.2.5) is valid; the tube wall is thin.
Analysis: The overall unit thermal resistance for the exchanger, from Eq. (13.14), is
1
Uo
1
^ f ;o
R
ho
w Ao
1 ^
Ao
1 A
1 ^
Ao
Rf ;i
w o
Rf ;i
kw A w
Ai ho kw Aw
Ai
hi
hi
1:203
Ai
di L 15:8 mm
Ao
do L
19 mm
1:092
Aw do di =2L 19 15:8=2 mm
3
^ w w 1:6 10 m 4 105 m2 K=W
R
kw
40 W=mK
Since the asymptotic fouling resistance is given, the actual fouling resistance on the water
side for c 280 hours can be determined from Eq. (9) of Example 13.1 as follows:
^ f ;i R
^ f ;!1 1 e=c 4 104 m2 K=W 1 e1 2:53 104 m2 K=W
R
Now the individual unit thermal resistances of the last equality of Eq. (1), in absolute
values and in percentages, are computed and summarized for this problem for 1
and c :
^ c;o 1
R
ho
(m2 K/W)
^ w Ao
R
Aw
(m2 K/W)
^ f ;i Ao
R
Ai
(m2 K/W)
^ c;i 1 Ao
R
hi Ai
(m2 K/W)
Overall
1=Uo
(m2 K=W)
!1
c
3.125 104
3.125 104
4.368 105
4.368 105
4.810 104
3.042 104
4.429 104
4.429 104
12.801 104
11.033 104
!1
c
24
28
3
4
Percentages
38
28
35
40
100
100
Fouling Time
890
Discussion and Comments: From the results of this problem, it is obvious that fouling has
quite a signicant inuence on the total unit thermal resistance 1=Uo . If cleaning is not
going to be performed, the fouling resistance will ultimately reach 38% of the total
resistance, more than any other contribution. Note that the thermal resistance of the
tube wall is an order of magnitude smaller, and hence our approximation of considering
^ w determination is reasonable. Distribution of thermal
the thin wall of the tube for R
resistances would be dierent if the foulant deposition is allowed to continue only up to
the deposition time that equals the time constant. In that case, the fouling thermal
resistance would be smaller compared to both the outside and inside tube convective
resistances. Still, it would have the same order of magnitude.
Among the most frequently used techniques for control of liquid-side fouling is the online
utilization of chemical inhibitors/additives. The list of additives includes (1) dispersants
to maintain particles in suspension; (2) various compounds to prevent polymerization
891
The standard techniques for control and/or prevention of fouling on the gas side are (1)
techniques for removal of potential residues from the gas, (2) additives for the gas-side
uid, (3) surface cleaning techniques, and (4) adjusting design up front to minimize
fouling. Details regarding various techniques for gas-side fouling prevention, mitigation,
and accommodation are given by Marner and Suitor (1987).
Control of gas (or liquid)-side fouling should be attempted before any cleaning
method is tried. The fouling control procedure should be preceded by (1) verication
892
of the existence of fouling, (2) identication of the feature that dominates the foulant
accumulation, and (3) characterization of the deposit.
Some of the methods for mitigation of gas-side fouling are as follows:
. Crystallization fouling can be prevented if the surface temperature is kept above the
freezing of vapors from the gaseous stream; the solidication can be minimized by
keeping a high velocity of freezable species, having some impurities in the gas
stream, and decreasing the foulant concentration, if possible.
. Particulate fouling can be minimized (1) by increasing the velocity of the gas stream
if it ows parallel to the surface and decreasing the velocity if the gas ow impinges
on the surface, (2) by increasing the outlet temperature of the exhaust gases from
the exchanger above the melting point of the particulates, (3) by minimizing the
lead content in gasoline or unburned hydrocarbons in diesel fuel, (4) by reducing
the fuelair ratio for a given combustion eciency, and (5) by minimizing ow
impact (e.g., ow over a staggered tube bank) or ensuring the narrowest dimension
in the ow cross section, to three to four times the largest particle size anticipated.
. Chemical reaction fouling can be minimized (1) by maintaining the right temperature range in the exhaust gas within the exchanger, (2) by increasing or decreasing
the velocity of the gaseous stream, depending on the application, (3) by reducing
the oxygen concentration in the gaseous stream, (4) by replacing the coal with fuel
oil and natural gas (in that order), and (5) by decreasing the fuelair ratio.
. Corrosion fouling is strongly dependent on the temperature of the exhaust stream in
the exchanger. The outlet temperature of the exhaust gas stream from the exchanger should be maintained in a very narrow range: above the acid dew point [above
1508C (3008F)] for sulfuric or hydrochloric acid condensation or below 2008C
(4008F) for attack by sulfur, chlorine, and hydrogen in the exhaust gas stream.
Since sulfur is present in all fossil fuels and some natural gas, the dew point of sulfur
must be avoided in the exchanger, which is dependent on the sulfur content in the
fuel (Shah, 1985). From the electrochemical condition of the metal surface,
the corrosion rate increases with velocity up to a maximum value for an active
surface and no sizable eect for a passive surface. The pH value has a considerable
role in the corrosion fouling rate; the corrosion rate is minimum at a pH of 11 to 12
for steel surfaces. Low oxygen concentrations in the ue gases promote the re-side
corrosion of mild steel tubes in coal-red boilers. Stainless steel, glass, plastic, and
silicon are highly resistant to low-temperature corrosion [Tgas < 2608C (5008F)],
stainless steel and superalloys to medium-temperature corrosion [2608C
5008F < Tgas < 8158C (15008F)], and superalloys and ceramic materials to
high-temperature corrosion [Tgas > 8158C (15008F)]. Chrome alloys are suitable
for high-temperature sulfur and chlorine corrosion, and molybdenum and chrome
alloys protect against hydrogen corrosion.
13.4.3
Cleaning Strategies
893
erally no cleaning is necessary. If fouling rate data are available, ideally the exchanger
can be optimized based on the life-cycle cost, and accordingly, the cleaning schedule can
be established.
The cleaning-cycle period may also be determined based on a regular maintenance
schedule during process shutdowns. In any case, the functional relationship between the
operation time and fouling resistance (see Section 13.2.6) should be known at least
partially. The importance of rational cleaning schedules based on such an understanding
is critical when (Somerscales, 1990) (1) the allowable deviation from the process stream
temperature is small compared to absolute values (steam power plant condensers), and
(2) the cost of cleaning is a signicant fraction of the operating cost.
Depending on the fouling process, the cleaning strategies for preventing maintenance
are of two types: reliability-based and cost-based. There are three scenarios for reliability-based cleaning strategies: (1) maintenance restores exchanger performance (this is an
idealized maintenance scheme, taking place at equal time intervals); (2) by decreasing
the preventive maintenance time interval gradually (due to xed degradation of performance after each maintenance interval), the exchanger performance is restored; (3) for the
preceding case of a xed degradation of performance after each preventive maintenance
time interval, if the maintenance takes place at equal time intervals, it reduces the
exchanger performance. The cost-based cleaning strategy includes the costs associated
with online chemical cleaning, oine cleaning, additional fuel/power consumption due
to fouling, and severity of the nancial penalty associated with exchanger performance
due to fouling.
It should be noted that operating a heat exchanger at the critical risk level of a system
or component is important in some applications, such as in a heat exchanger network in a
renery. In this situation, an acceptable level of heat exchanger overall heat transfer
coecient will primarily govern the maintenance strategy. However, in some situations,
heat exchangers are not in a network, or in a critical system; here, maintaining the
exchanger at a higher reliability level r (or at a lower risk level p) implies more frequent
maintenance intervals, which can often result in increasing operation and maintenance
costs. It is thus important to note that in situations in which the cost of operation and
maintenance is an important factor, along with exchanger reliability (r 1 p), maintenance decisions can be optimized by developing cost as a function of reliability (or risk
level) and then searching for a minimum cost-based solution. This cost-optimized maintenance solution will also result in an optimal level of heat exchanger reliability (Zubair et
al., 1997).
For further details on cleaning strategies, refer to Zubair et al. (1997) for shell-andtube heat exchangers and Zubair and Shah (2001) for plate-and-frame heat exchangers.
894
Environment factors
Environment
Stress corrosion
cracking
- Impurities
- Tempeature and concentration
of corrodent
- Degree of aeration
- Flow velocity
- pH
Corrosion
Stress
Design factors
- Geometry of
the joints
- Crevices
- Stagnant areas
- Ubends
Design factors
Metallurgical
factors
Fabrication
techniques
Material factors
Material
- Composition
- Alloying elements
- Heat treatment
- Effects of fabrication
- Microstructure
- Surface conditions
- Possivity
- Tendency for fouling
described in Section 13.2.1, thus adding thermal resistance in the heat ow path and
reducing heat transfer, increasing uid pressure drop and pumping power, and increasing
cost due to overdesign of the exchanger. The loss of material due to corrosion may result
in crevices, holes, and/or partial removal of heat transfer surfaces, resulting in loss
(leakage) of heat transfer uids, some of which may be costly. If the uid leaks outside,
it may harm the environment if the uid is corrosive or poisonous. If it leaks to the other
uid side, it may contaminate the other uid and deteriorate its quality. Corrosion may
add extra cost to the exchanger, due to the use of expensive material, maintenance,
warranty, inventory of parts, and so on. Corrosion products carried downstream of
the exchanger may corrode downstream components. Finally, corrosion may result in
complete failure of an exchanger or partial failure in a tube-n exchanger, due to corroding away ns, as in an automotive radiator.
There is no a single cause of corrosion and/or associated corrosion mechanisms.
However, corrosion in general has clear electrochemical roots. Namely, dierent parts
of a heat exchanger exposed to working uids easily become polarized. The role of an
electrolyte is usually taken by a working uid (or sometimes by solid deposits or thick
metal oxide scale) in the vicinity of or between parts made of dierent metals. If an
external electrical circuit is established, metal surfaces involved take the role of either
anode or cathode. Appearance of an electric current forces electrical particles (say,
positively charged metal ions) to leave the metal on the anode end and enter the surrounding electrolyte. On the other end, a metal surface that plays the role of cathode
serves as a site where electrical current escapes from the electrolyte. The presence of this
mechanism opens the way for metal dissolution at the anode end of the established
electrical circuit. This dissolution can be interpreted as a corrosion eect (if all other
mechanisms are suppressed). This very simplied picture provides a background for
many corrosion problems. Refer to Kuppan (2000) for further details.
Detailed study of corrosion phenomena is beyond the scope of this book. Due to the
importance of corrosion in heat exchanger design and operation, we will address only the
most important topics. A brief description of the main corrosion types is given rst
(keeping in mind their importance from a heat exchanger design point of view).
895
Corrosion Types
Corrosion types, important for heat exchanger design and operation, are as follows: (1)
uniform attack corrosion, (2) galvanic corrosion, (3) pitting corrosion, (4) stress corrosion cracking, (5) erosion corrosion, (6) deposit corrosion, and (7) selective leaching, as
categorized by Fontana and Greene (1978). Let us dene each corrosion type briey.
Uniform corrosion is a form of corrosion caused by a chemical or electrochemical
reaction between the metal and the uid in contact with it over the entire exposed metal
surface. It occurs when the metal and uid (e.g., water, acid, alkali) system and operating
variables are reasonably homogeneous. It is usually easy to notice corroded areas
attacked by uniform corrosion. All other forms of corrosion mechanisms discussed
below cause localized corrosion.
Galvanic corrosion is caused by an electric potential dierence between two electrically
dissimilar metals in the system in the presence of an electrolyte (such as water in a heat
exchanger). It occurs on the anode and does not aect the cathode (referred to as a noble
metal).
Pitting corrosion is a form of localized autocatalytic corrosion due to pitting that
results in holes in the metal. If anodes and cathodes rapidly interchange the sites
randomly, uniform corrosion occurs, as in rusting of iron. If the anode becomes xed
on the surface, pitting corrosion takes place.
Stress corrosion is a form of corrosion that involves cracks on susceptible metals
caused by the simultaneous presence of the tensile stress and a corrosive uid medium.
Erosion corrosion is a form of surface corrosion due to erosion of the heat transfer
surface due to a high-velocity uid with or without particulates (e.g., uid velocity
greater than 2 m/s or 6 ft/sec for water ow over an aluminum surface) and subsequent
corrosion of the exposed surface.
Crevice corrosion is a form of localized physical deterioration of a metal surface in
crevices or under deposits in shielded areas (i.e., in stagnant uid ow regions), often
caused by deposits of dirt and corrosion products.
Selective leaching or dealloying is the selective removal of one metal constituent from
an alloy by corrosion that leaves behind a weak structure.
13.5.2
Uniform (general) corrosion is not localized, and a surrounding corrosive medium aects
the surface of the exposed metal prone to corrosion. Temperature, concentrations, oxidation, acidity, and so on, have a signicant inuence on the extent of this type of
corrosion. Atmospheric corrosion and high-temperature gaseous corrosion are most
probable in heat exchangers. This corrosion usually thins the heat transfer surface.
Metals having 0.1 mm/yr surface thinning are considered excellent, those having 0.1 to
0.5 mm/yr satisfactory, and those having above about 1 mm/yr unsatisfactory for shelland-tube heat exchangers (Kuppan, 2000).
As opposed to general corrosion, galvanic corrosion often attacks interfaces/contacts
between tubes and baes and/or tubesheets, contact between the bae and shell, and
joint areas (either welded, brazed, soldered, or mechanically joined). The likelihood of
896
galvanic corrosion can easily be assessed knowing the position of the materials involved
in the galvanic series summarized in Table 13.6. The metals next to each other in the
galvanic series have little tendency to galvanic corrosion. In addition, metals closer to
the anodic end of the galvanic series are more prone to corrosion, and the materials at the
cathodic end are more stable. It can also occur in compact and other exchangers with
water and other electrolytes in the circuit in which the exchanger is one of the components.
Pitting corrosion takes place when a protective surface lm breaks down; these surface
lms are formed on the metal surface by reaction with an environment or during the
surface treatment. The common metals exposed to this type of corrosion in descending
order of nobility are aluminum, stainless steels, nickel, titanium, and their alloys. It is a
very aggressive type of corrosion. Pitting corrosion is inuenced by metallurgical and
environmental factors, such as breakdown of protective coatings, inhomogeneities in the
alloys, and inhomogeneities caused by joining processes. Consequently, the appearance
of pitting corrosion is possible whenever such conditions are present. Pits caused by
pitting corrosion are usually at places where the metal surface has surface deformities
and scratches.
Stress corrosion cracking may be present at locations within the construction where
the joint interaction of stress and a corrosive medium causes material deterioration.
The presence of higher stress levels, increased temperature and concentration of a
corrosive medium, and crack geometry may accelerate corrosion. For example, tubeto-tubesheet expanded joints may be prone to residual stresses, as well as thin-walled
expansion joints and/or U-bends. Cold working parts and U-bends in shell-and-tube
heat exchangers are locations where corrosion may take place in combination with
existing stress.
TABLE 13.6 Galvani Seriesa
Mg (anodic; least noble)
Zn
Fe (galvanized)
Al 3004
Al 3003
Cast iron
SS 430 (active)
SS 304 (active)
Admiralty brass
Monel 400
SS 430 (passive)
SS 304 (passive)
Lead
Copper
Nickel
Inconel 825
Hastelloy C
Titanium
Graphite
Platinum (cathodic; most noble)
a
897
Erosion corrosion involves solid particle or liquid droplet impingement and cavitation.
In shell-and-tube heat exchangers, impingement plates must be designed to prevent
this type of erosion corrosion in tubes exposed to nozzle inlet ow. Erosion corrosion
is more common at the inlet end of a heat exchanger ow passage or on the tube
side.
Crevice corrosion is localized corrosion and may occur at metal-to-metal or metal-tononmetal joints (e.g., gasketed joints), or underneath biological growth or fouling
deposits. In particular, areas prone to this type of corrosion are stagnant areas and
complex geometric designs with sharp edges. This type of corrosion usually starts with
an inltration of a corrosive substance into a crack and/or small opening, such as
clearances between rolled tubes and tubesheets, open welds, bolt holes, nut adjacent
areas, gasket areas, or contacts between plates in a plate heat exchanger. Fouling and
various deposits inuence corrosion at shielded areas if the combination of uid and heat
exchanger surface material is inappropriate.
Selective leaching (parting) takes place depending on the combination of alloys
selected and the presence of a corrosive substance in the surrounding medium. Some
typical problems encountered in heat exchanger operation are related to (1) removal of
Zn from brass in stagnant waters, (2) removal of Al from aluminum brass in acidic
solutions, and (3) removal of Ni in CuNi alloys under conditions of high heat ux.
Such removal processes are referred to as dezincication, dealuminumication, denickelication, and so on.
13.5.3
Corrosion Control
898
REVIEW QUESTIONS
899
Bott, T. R., 1990, Fouling Notebook, Institution of Chemical Engineers, Rugby, UK.
Cannas, F. C., 1986, Fouling in heat exchanger design, M.S. Thesis, Mech. Eng. Dept., State Univ.
New York, Bualo.
Chenoweth, J. M., 1990, Final report of the HTRI/TEMA joint committee to review the fouling
section of the TEMA standards, Heat Transfer Eng., Vol. 11, No. 1, pp. 73107.
Epstein, N., 1978, Fouling in heat exchangers, Heat Transfer 1978, Proc. 6th Int. Heat Transfer
Conf., Vol. 6, pp. 235253; also in Fouling of Heat Transfer Equipment, E. F. C. Somerscales and
J. G. Knudsen, eds., Hemisphere Publishing, Washington, DC, 1981, pp. 701734.
Fontana, M. G., and N. D. Greene, 1978, Corrosion Engineering, McGraw-Hill, New York.
Knudsen, J. G., 1998, Fouling in heat exchangers, in Heat Exchanger Design Handbook, G. F.
Hewitt, ed., Begell House, New York, Sec. 3.17.
Kuppan, T., 2000, Heat Exchanger Design Handbook, Marcel Dekker, New York.
Marner, W. J., 1990, Progress in gas-side fouling of heat transfer surfaces, Appl. Mech. Rev., Vol. 43,
No. 3, pp. 3566.
Marner, W. J., 1996, Progress in gas-side fouling of heat transfer surfaces, Appl. Mech. Rev., Vol. 49,
No. 10, Pt. 2, pp. S161S166.
Marner, W. J., and J. W. Suitor, 1987, Fouling with convective heat transfer, in Handbook of SinglePhase Convective Heat Transfer, S. Kakac, R. K. Shah, and W. Aung, eds., Wiley, New York,
Chap. 21.
Melo, L. F., T. R. Bott, and C. A. Bernardo, eds., 1988, Advances in Fouling Science and Technology,
Kluwer Academic Publishers, Dordrecht, The Netherlands.
Panchal, C. B., 1999, Review of fouling mechanisms, mitigation of heat exchanger fouling and its
economic and environmental impacts, Proc. Engineering Foundation Conf., Ban, Alberta,
Canada, July.
Panchal, C. B., and J. G. Knudsen, 1998, Mitigation of water fouling: technology status and
challenges, in Advances in Heat Transfer, Vol. 31, J. P. Hartnett, T. F. Irvine, Jr., Y. I. Cho,
and G. A. Greene, eds., Academic Press, San Diego, CA, pp. 437474.
Rabas, T. J., and C. B. Panchal, 2000, Fouling rates, not fouling resistances, Heat Transfer Eng.,
Vol. 21, No. 2, pp. 12.
Shah, R. K., 1985, Compact heat exchangers, in Handbook of Heat Transfer Applications, 2nd ed.,
W. M. Rohsenow, J. P. Hartnett, and E. N. Ganic, eds., McGraw-Hill, New York, p. 4-284.
Somerscales, E. F. C., 1990, Fouling of heat transfer surfaces: a historical review, Heat Transfer
Eng., Vol. 11, No. 1, pp. 1936.
TEMA, 1999, Standard of the Tubular Exchanger Manufacturers Association, 8th ed., Tubular
Exchanger Manufacturers Asscociation, New York.
Zubair, S. M., and R. K. Shah, 2001, Fouling in plate-and-frame heat exchangers and cleaning
strategies, in Compact Heat Exchangers and Enhancement Technology for the Process Industries
2001, R. K. Shah, A. Deakin, H. Honda and T. M. Rudy, eds., Begell House, New York, pp.
553565.
Zubair, S. M., A. K. Sheikh, M. A. Budair, and M. A. Badar, 1997, A maintenance strategy
for heat-transfer equipment subject to fouling: a probabilistic approach, ASME J. Heat
Transfer, Vol. 119, pp. 575580.
REVIEW QUESTIONS
For each question circle one or more correct answers. Explain your answers briey.
900
13.1
13.2
13.3
13.4
When thermal resistances are in series, the fouling has a signicant inuence on
heat transfer in an exchanger having:
(a) high U
(b) low U
(c) both of these
(d) cant tell
13.5
A change in the hydraulic diameter due to fouling inuences the local heat transfer
coecient h in a heat exchanger with constant m_ , L, A, and uid properties as
follows:
(a) h increases linearly with an increase in Dh .
(c) h is inversely proportional to D3h .
(b) h is inversely proportional to Dh .
13.6
A change in the hydraulic diameter due to fouling inuences the pressure drop p
in a heat exchanger with constant m_ , L, A, and uid properties as follows:
(a) p is inversely proportional to D5h .
(b) p is inversely proportional to D3h .
(d) p does not depend on Dh .
(c) p is proportional to D3h .
13.7
13.8
13.9
13.10 On the air side of a compact heat exchanger, 50% of the ow passages are large
and 50% are small. Which passages would have more change in thermal perfor-
REVIEW QUESTIONS
901
mance when the owing air is dirty and fully developed laminar ows are expected
in both passages?
(a) passages with larger ow area
(b) passages with smaller ow area
(c) both of these
13.11 In general, deposit formed on the heat transfer surface due to fouling shows the
time history character as follows:
(a) Deposit thermal resistance is hyperbolic with respect to time.
(b) Deposit thermal resistance changes linearly with time.
(c) Deposit thermal resistance changes exponentially with time.
13.12 The time constant in the KernSeaton correlation for time dependence of the
fouling resistance has the following physical meaning:
(a) time required for the fouling resistance to reach 50% of its asymptotic value
(b) time required for the fouling resistance to reach its asymptotic value
(c) time required for the fouling resistance to reach 63% of the asymptotic value
(d) time required for the fouling resistance to reach 99% of the asymptotic value
13.13 Water-side fouling can be most eciently mitigated by:
(a) utilizing chemical additives
(b) utilizing cleaning devices
(c) designing heat transfer surfaces with highly augmented heat transfer
characteristics
13.14 Combined tube- and shell-side fouling resistance is the largest for the following
combination of working uids:
(a) vapor on both sides of a heat transfer surface
(b) vapor on one side and phase-change uid on the other
(c) liquid on both sides
(d) phase-change uids on both sides
13.15 Fouling process models are based on the concept of:
(a) exponential decrease of the deposition rate and linear increase of the removal
rate
(b) deposition rate minus removal rate
(c) initiation plus aging minus removal
(d) initiation plus transport minus removal
13.16 Increasing the liquid-side velocity in a heat exchanger will denitely reduce the
likelihood of fouling due to:
(a) precipitation
(b) freezing
(c) particulate fouling
(d) biological fouling
(e) chemical fouling
13.17 Increasing the liquid-side temperature will denitely reduce the likelihood of
fouling due to:
(a) biological fouling
(b) freezing
(c) precipitation
(d) particulate fouling
902
13.18 In the case of gas-side chemical reaction fouling, the following inuences are
registered:
(a) Increase in the gas-side temperature decreases the likelihood of fouling.
(b) Increase in the oxygen level increases the likelihood of fouling
(c) Increase in the velocity may increase or decrease fouling.
13.19 Increased presence of sulfur on the gas side will denitely increase the likelihood of
fouling due to:
(a) chemical reaction fouling
(b) corrosion fouling
(c) freezing fouling
(d) particulate fouling
13.20 If fouling takes place, h is going to be higher, and for the same heat transfer, the
surface area should be:
(a) increased
(b) decreased
(c) unchanged
13.21 Corrosion fouling in a phosphoric acid condensation process may be prevented by
the use of a:
(a) counterow arrangement
(b) plate heat exchanger
(c) parallelow arrangement
(d) Teon coating over low-temperature exchanger surfaces
13.22 Corrosion fouling in a sulfuric acid condensation process may be prevented by use
of a:
(a) counterow arrangement
(b) plate heat exchanger
(c) parallelow arrangement
(d) Teon coating over low-temperature exchanger surfaces
13.23 Which materials are ideally prone to galvanic corrosion for contacting copper in a
seawater solution?
(a) nickel
(b) lead
(c) zinc
(d) cast iron
13.24 Circle the following statements as true or false.
(a) T F Fouling resistance is a time-dependent quantity and has some nite
value at time equal to zero.
(b) T F Attachment is associated with the delay period d .
13.25 The delay period d associated with initiation decreases with a(n):
(a) increase in the degree of supersatuartion in crystallization fouling
(b) decrease in surface roughness
(c) increase in the uid viscosity
13.26 In a plate-n heat exchanger with the n spacing of 3 mm and gas velocity of 10 m/s,
which of the following fuels would be likely to produce fouling exhaust gases?
(a) natural gas
(b) propane
(c) #2 oil
(d) coal
13.27 Chemical cleaning of foulant can be accomplished by:
(a) dilute acids
(b) steam
(c) chlorinated hydrocarbons
(d) hot air
(e) dispersants
PROBLEMS
903
13.2
Consider a simplied model of a heat exchanger with heat transfer areas equal on
the hot and cold sides and having no ns. Assume that this heat exchanger has to
be used alternately as (1) a gas-to-gas heat exchanger (fouling on the compressed
air side), (2) steam-to-heavy fuel oil heat exchanger (fouling on the heavy fuel oil
side), and (3) liquid-to-liquid heat exchanger (fouling on the ethylene glycol solution side). Determine what would be the required increase in heat transfer area for
the various working uids, comparing the fouled to the clean heat exchanger
operation. Assume that only the thermal resistance of the foulant would change
the overall heat transfer coecient (the other conditions remaining unchanged).
Assess the typical values of the overall heat transfer coecients and fouling resistances for described physical situations. The heat transfer rate and mean temperature dierence remain the same.
13.3
Using the same information as those given in Problem 13.2, determine what would
be the change of heat transfer rate between clean and fouled operations. Assume
the heat transfer area and the mean temperature dierence of the uid streams to
be invariant.
13.4
A waterwater gasketed plate heat exchanger has an overall heat transfer coecient under fouled conditions of 4200 W/m2 K. Hot- and cold-uid-side heat
transfer coecients are 15,000 and 14,000 W/m2 K, respectively. The plate thickness (stainless steel 316) is 0.6 mm, and the thermal conductivity is 17 W/m K.
Calculate the cleanliness factor CF and the total fouling resistance for this
exchanger.
13.5
In the cement industry, large gas-to-air heat exchangers are used to cool hot
exhaust gases leaving suspension preheaters. These gases are cooled prior to
being vented through a baghouse to the environment. In such heat exchangers,
the hot dirty gases are placed on the tube side. In one application, owing to static
electricity, very small particles of cement dust are deposited on the inside surface
of the tubes, thus reducing the eectiveness of the heat exchanger. In this
particular case, the tubes are 6.45 m long, 76 mm in inside diameter, with a wall
904
thickness of 3.2 mm. The tubes are made of carbon steel with a thermal conductivity of 43 W/m K. The average outside tube wall temperature is 1208C and the
average gas temperature is 2808C. For the clean exchanger, the gas-side velocity
is 15.6 m/s, resulting in a gas-side convective heat transfer coecient of
39.3 W/m2 K. If a 3.2 mm thick layer of cement dust (k 0:299 W/m K) is
^ f , the gas-side unit
deposited on the inner surface of the tubes, determine (a) R
2
fouling resistance (m K=W), (b) the reduction in heat transfer due to the
fouling layer, and (c) the increase in the pressure drop due to the fouling
layer. Consider the mean density of the air and gas as 0.64 kg/m3 and the
dynamic viscosity as 2.85 Pas. Assume the mass ow rate to be constant for
clean and fouled exchangers. Use the following correlation for turbulent ow
through the tube, and ignore entrance and exit losses as well as the momentum
eect for pressure drop calculation.
f 0:0014 0:125Re0:32
Explicitly mention any additional assumptions that you need to make.
13.6
13.7
PROBLEMS
905
side, and the tube heat transfer surface area was the same under clean and fouled
conditions. Explain what additional information you would need to compute the
water velocity in the fouled tubes. Assume the condensing heat transfer coecient
to be 4000 W/m2 K. Assume the following properties for water: 1000 kg=m3 ,
cp 4180 J=kg K, k 0:59 W=m K, and 0:001 Pa s. Hint: Use the Dittus
Boelter correlation for the tube-side heat transfer coecient.
13.8
A shell-side condenser with cooling tower water on the tube side is not performing
satisfactorily, due to tube-side fouling. There is negligible shell-side fouling.
Contemplated is replacement of the plain tubes with low-nned tubing having
1.18 ns/mm. This will increase the tube outside area by a factor of 2.9 over that of
the plain tubes. The following are some design data provided.
Quantity
Plain
Tube
Low-Finned
Tube
25.4
22.9
0.00018
8517
8517
17.3
2.9
3.22
906
903
875
385
420
380
447
420
439
444
460
381
2702
2770
8933
8800
8530
7870
7272
8510
8900
8314
9240
Specic
heat
c (J/kg K)
80.2
51
11.7
91
13
12.2
401
52
111
237
174
Thermal
Conductivity
k (W/m K
Steel
AISI 1010
AISI 1042 (annealed)
AISI 4130 (hardened)
AISI 302
AISI 304
AISI 316
AISI 347
AISI 410
Titanium
Pure
Ti-6Al-4V
Ti-21Al-2Mn
Metal
434
460
460
480
477
468
480
460
522
610
466
4500
4420
4510
Specic
heat
c (J/kg K)
7832
7840
7840
8055
7900
8238
7978
7770
Density
(kg/m3 )
21.9
5.8
8.4
63.9
50
43
15.1
14.9
13.4
14.2
25
Thermal
Conductivity
k (W/m K
Sources: Data from A. F. Mills, Heat and Mass Transfer, Richard D. Irwin, Burr Ridge, IL, 1995; F. P. Incropera and D. P. DeWitt,
Introduction to Heat Transfer, Wiley, New York, 2002; Y. S. Touloukian and C. Y. Ho, Thermophysical Properties of Matter, Vols. 19,
Plenum Press, New York, 1972; American Society for Metals, Metals Handbook, Vol. 1, ASM, Metals Park, OH, 1961.
a
4.4% Cu, 1.0% Mg, 0.75% Mn, 0.4% Si, Al balance.
Aluminum
Pure
Duralumina
Coper
Pure
Bronze (90% Cu, 10% Al)
Brass (70% Cu, 30% Zn)
Iron
Pure
4 C cast
Inconel X-750
Nickel, pure
Nichrome (80% Ni, 20% Cr)
Hasteloy B
Metal
Density
(kg/m3 )
APPENDIX A
Thermophysical Properties
907
THERMOPHYSICAL PROPERTIES
TABLE A.2 Thermal Conductivity [k (W/m E K)] and Specic Heat [c (J/kg E K)] of Metals as a
Function of Temperature
Temperature (K)
200
Metal
Al pure
Cu pure
Bronze
Brass
Iron Armco
Iron cast
AISI 1010
AISI 1042
AISI 4130
AISI 302
AISI 304
AISI 316
AISI 410
Inconel
Nichrome
Ti-pure
c
798
356
785
360
384
300
k
237
413
42
74
81
400
500
600
800
903
385
420
380
447
420
434
237
401
52
111
73
51
64
52
43
15
15
13
25
11.7
13
22
949
397
460
395
490
240
393
52
134
66
44
59
50
42
17
17
15
26
13.5
14
20
996
412
500
410
530
236
386
55
143
59
39
54
48
41
19
18
17
27
15.1
16
20
1033
417
231
379
1146
433
425
574
146
53
36
49
45
40
20
20
18
27
17
17
19
402
13
372
25
10.3
480
477
468
460
439
405
25
522
487
500
500
512
515
504
473
480
551
520
530
530
531
539
528
490
500
572
559
570
570
559
557
550
510
525
591
680
685
700
690
585
582
576
546
545
633
218
366
150
42
27
39
37
37
23
23
21
29
20.5
21
19
Sources: Data from A. F. Mills, Heat and Mass Transfer, Richard D. Irwin, Burr Ridge, IL, 1995; F. P. Incropera
and D. P. DeWitt, Introduction to Heat Transfer, Wiley, New York, 2002; Y. S. Touloukian and C. Y. Ho,
Thermophysical Properties of Matter, Vols. 19, Plenum Press, New York, 1972; American Society for Metals,
Metals Handbook, Vol. 1, ASM, Metals Park, OH, 1961.
(kg/m3 )
k (W/m K
cp J=kg K
104 (Pa s)
400
500
600
800
814
790
765
717
45.5
43.6
41.6
36.8
800
790
780
750
4.9
2.8
2.1
1.6
Sodium
500
600
800
1000
900
868
813
772
79.2
74.7
65.7
59.3
1335
1310
1260
1255
4.2
3.1
2.2
1.8
Lithium
500
600
800
900
514
503
483
473
43.7
46.1
50.7
55.9
4340
4230
4170
4160
5.31
4.26
3.10
2.47
Liquid Metal
Potassium
Source: Data from A. F. Mills, Heat and Mass Transfer, Richard D. Irwin, Burr Ridge, IL, 1995.
908
THERMOPHYSICAL PROPERTIES
106 K1
4217
4198
4184
4179
4178
4180
4184
4188
4195
4256
4660
7000
1
1.750
1.422
1.080
0.855
0.695
0.577
0.489
0.420
0.365
0.217
0.118
0.081
0.045
68.05
46.04
174.0
276.1
361.9
436.7
504.0
566.0
624.2
896
0.147
0.144
0.145
0.143
0.139
0.138
0.136
0.134
0.133
1796
1827
1909
1993
2076
2161
2250
2337
2427
3850
2170
486
141
53.1
25.2
14.1
8.74
5.64
700
700
700
700
700
700
700
700
700
1130.8
1125.8
1118.8
1114.4
1103.7
1096.2
1089.5
1083.8
1079.0
1074.0
1066.7
1058.5
0.242
0.244
0.248
0.252
0.255
0.258
0.260
0.261
0.261
0.261
0.262
0.263
2294
2323
2368
2415
2460
2505
2549
2592
2637
2682
2728
2742
65.1
42.0
24.7
15.7
10.7
7.57
5.61
4.31
3.42
2.78
2.28
2.15
650
650
650
650
650
650
650
650
650
650
650
650
1528
1498
1470
1439
1407
1374
1341
1306
1269
1229
0.068
0.069
0.070
0.073
0.073
0.073
0.073
0.072
0.069
0.068
881.6
892.3
903.7
916.3
930.1
945.0
960.9
978.1
996.3
1015.5
0.457
0.385
0.354
0.322
0.304
0.283
0.265
0.254
0.244
0.233
1850
1900
2000
2100
2250
2350
2550
2750
3050
3500
Liquid
T (K)
(kg/m3 )
k (W/m K
Water
273.15
280
290
300
310
320
330
340
350
400
500
600
647.3
1000
1000
999.0
997.0
993.0
989.1
984.3
979.4
973.7
937.2
831.3
648.9
315.5
0.569
0.582
0.598
0.613
0.628
0.640
0.650
0.660
0.668
0.688
0.642
0.497
0.238
Engine oil
(unused)
273
280
300
320
340
360
380
400
420
899.1
895.3
884.1
871.8
859.9
847.8
836.0
825.1
812.1
Ethylene
glycol
273
290
290
300
310
320
330
340
350
360
370
373
Freon R-12
230
240
250
260
270
280
290
300
310
320
cp (J/kg K
Source: Data from F. P. Incropera and D. P. DeWitt, Introduction to Heat Transfer, 5th ed., Wiley, New York,
2002.
THERMOPHYSICAL PROPERTIES
909
T (K)
(kg/m3 )
k (W/m K
cp (J/kg K
103 (Pa s)
Air
100
150
200
250
300
350
400
450
500
600
700
800
900
1000
3.5562
2.3364
1.7458
1.3947
1.1614
0.9950
0.8711
0.7740
0.6964
0.5804
0.4975
0.4354
0.3868
0.3482
0.00934
0.0138
0.0181
0.0223
0.0263
0.0300
0.0338
0.0373
0.0407
0.0469
0.0524
0.0573
0.0620
0.0667
1032
1012
1007
1006
1007
1009
1014
1021
1030
1051
1075
1099
1121
1141
71.1
103.4
132.5
159.6
184.6
208.2
230.1
250.7
270.1
305.8
338.8
369.8
398.1
424.2
Ammonia
300
320
340
360
380
400
420
440
460
480
500
0.6894
0.6448
0.6059
0.5716
0.5410
0.5136
0.4888
0.4664
0.4460
0.4273
0.4101
0.0247
0.0272
0.0293
0.0316
0.0340
0.0370
0.0404
0.0435
0.0463
0.0492
0.0525
2158
2170
2192
2221
2254
2287
2322
2357
2393
2430
2467
101.5
109
116.5
124
131
138
145
152.5
159
166.5
173
Steam
380
400
450
500
550
600
650
700
750
800
850
0.5863
0.5542
0.4902
0.4405
0.4005
0.3652
0.3380
0.3140
0.2931
0.2739
0.2579
0.0246
0.0261
0.0299
0.0339
0.0379
0.0422
0.0464
0.0505
0.0549
0.0592
0.0637
2060
2014
1980
1985
1997
2026
2056
2085
2119
2152
2186
127.1
134.4
152.5
170.4
188.4
206.7
224.7
242.6
260.4
278.6
296.9
Source: Data adapted and modied from F. P. Incropera, and D. P. DeWitt, Introduction to Heat Transfer, 5th
ed., Wiley, New York, 2002.
910
0.293
0.313
0.373
0.441
0.518
0.605
0.703
0.812
0.933
1.068
1.217
1.380
1.560
1.757
1.972
2.206
2.461
2.739
3.040
3.369
3.728
4.042
273.2
275
280
285
290
295
300
305
310
315
320
325
330
335
340
345
350
355
360
365
370
374
14.43
15.39
18.23
21.48
25.19
29.4
34.19
39.63
45.79
52.77
60.71
69.76
80.09
91.97
105.7
121.8
141
164.3
193.6
232.9
293.9
434.1
(kg/m3 )
0.90
0.91
0.93
0.96
0.98
1.01
1.04
1.08
1.12
1.16
1.21
1.27
1.34
1.42
1.52
1.66
1.85
2.13
2.61
3.58
6.86
137.20
cp
kJ=kg K
10.7
10.8
11.0
11.2
11.4
11.7
11.9
12.1
12.4
12.6
12.9
13.2
13.6
13.9
14.4
14.8
15.4
16.1
17.0
18.4
20.7
27.5
10
Pa s
0.0115
0.0117
0.0121
0.0126
0.0130
0.0135
0.0140
0.0145
0.0151
0.0157
0.0163
0.0170
0.0178
0.0187
0.0197
0.0209
0.0225
0.0246
0.0274
0.0318
0.0407
0.1018
k
(W/m K)
1295
1289
1272
1255
1237
1219
1200
1180
1160
1139
1117
1094
1069
1043
1015
984.7
951.3
913.8
870.1
816.3
740.3
587.9
(kg/m3 )
1.34
1.35
1.36
1.38
1.39
1.41
1.43
1.46
1.48
1.51
1.54
1.58
1.63
1.68
1.75
1.84
1.96
2.14
2.44
3.04
5.11
101.70
cp
kJ=kg K
Liquid
271.1
264.8
248.5
233.3
219.2
205.9
193.3
181.5
170.2
159.5
149.3
139.5
130.0
120.9
111.9
103.2
94.5
85.8
76.8
67.2
55.8
38.9
103
Pa s
0.0920
0.0912
0.0890
0.0868
0.0846
0.0825
0.0803
0.0782
0.0761
0.0739
0.0718
0.0696
0.0675
0.0653
0.0631
0.0609
0.0586
0.0563
0.0541
0.0522
0.0521
0.0881
k
(W/m K)
Source: Data from M. O. McLinden, S. A. Klein, E. W. Lemmon, and A. P. Peskin, NIST Thermodyamic and Transport Properties of Refrigrants and Refrigerant Mixtures, NIST
Standard Reference Database 23, REFPROP Version 6.01, U.S. Department of Commerce, Technology Administration, National Institute of Standards and Technology, Physical
and Chemical Properties Division, Boulder, CO, 1998.
P
(MPa)
Vapor
Thermophysical Properties of R134a as Liquid and Vapour along the Saturation Line
T
K
TABLE A.6
APPENDIX B
e-NTU Relationships for
Liquid-Coupled Exchangers
The liquid-coupled indirect-transfer type exchanger system, also referred to as a runaround coil system, connects two direct-transfer type exchangers (recuperators) usually
located apart by a circulating liquid as shown in Fig. B.1. Thus this system allows heat
transfer between source and sink which are not closely located and/or must be separated.
Such a system is commonly used in waste heat recovery applications such as HVAC and
low-temperature process waste heat recovery, industrial dryers having inlet and exhaust
ducts at the opposite end of the plant, drying of grains, and so on (Reay, 1979). It can
also be used in waste heat recovery from a hot gas to cold air, where the inlet density
dierence could be very high (such as over vefold). In such a case, complex gas ducting
can be simplied through the use of a circulating liquid. This is also the case for a gas-togas exchanger, where the circulating liquid can simplify the gas ducting arrangement.
Such a system would require two exchangers and theoretically would require a higher
surface area for heat transfer from the hot to the cold uid, about 10 to 20% (Kays and
London, 1998). In addition, it will require additional components for the circulating
liquid, adding cost and complexity.
FIGURE B.1
Fundamentals of Heat Exchanger Design. Ramesh K. Shah and Duan P. Sekulic
Copyright 2003 John Wiley & Sons, Inc.
911
912
Heat Capacity
Rate Criteria
" Formula
" Formula
CL > Cc > Ch
"
1
1 Ch =Cc Ch
"c
CL
"h
Ch > CL > Cc
CL > Ch > Cc
"
1
1 Cc =Ch Cc
"h
CL
"c
Ch Cc C > CL
Cc > Ch > CL
"
1
1
1
1
"c "h
Ch Cc C < CL
Ch > Cc > CL
"
1
1
1
1
"c "h
Cc Ch C CL
Cc > CL > Ch
"
Ch
CL
Cc
CL
1 Ch
" h CL
1
"
1 Cc
"c CL
"
"
1
1
1
"h
CL =C
1
1
1
"c "h
1
1
1
C
"c "h CL
"
1
1
1
1
"c "h
1
1
"c
Cc Tc;o Tc;i
Cc Th;i Tc;i
"h
Ch Th;i Th;o
Ch Th;i TL;i
"c
Cc Tc;o Tc;i
Cc TL;o Tc;i
B:1
Note that "h and "c here are the exchanger eectiveness values of the hot and cold uids,
and they are not the temperature eectiveness values dened by Eqs. (3.51) and (3.52).
The "h and "c are related to the overall eectiveness " of the liquid-coupled exchangers as
shown in Table B.1, where dierent formulas are presented depending on the relationships among the heat capacity rates of the hot uid, cold uid, and the circulating liquid.
Note that in most applications, CL is larger than Ch and Cc .
REFERENCES
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd ed., Krieger Publishing, Malabar, FL.
Reay, D. A., 1979, Heat Recovery Systems, E.&F.N. Spon, London.
APPENDIX C
Two-Phase Heat Transfer and Pressure
Drop Correlations
Although the focus in this book is on single-phase ow heat exchanger design and
analysis, there are situations when phase-change (condensation or vaporizing) uid
having negligible thermal resistance is on one uid side of a two-uid heat exchanger;
the design and analysis for such an exchanger can be done using the slightly modied
single-phase theory outlined in this book. However, we need to compute the heat transfer
coecient on the phase-change side even for this situation. Additionally, if one would
like to estimate approximately the performance or size of the phase-change exchanger, it
can be treated as a single-phase exchanger once the average heat transfer coecient on
the phase-change side is determined. Hence, in this appendix we provide some correlations for condensation and convective boiling. For the detailed information on the phasechange correlations and related phenomena, a comprehensive source is the handbook by
Kandilkar et al. (1999). For completeness, we also provide a method to compute the
pressure drop on the phase-change side and present it before the heat transfer correlations. Of course, many important topics of phase-change exchangers, such as the phasechange side not having the negligible thermal resistance, rating and sizing of the exchanger when phase change occurs on both uid sides, ow maldistribution, and so on, are
beyond the scope of this appendix and the book.
C.1 TWO-PHASE PRESSURE DROP CORRELATIONS
Due to the phase change during condensation or vaporization, the pressure gradient
within the uid changes along the ow path or axial length. The pressure drop in the
phase-change uid can then be computed by integrating the nonlinear pressure gradient
along the ow path. In contrast, the pressure gradient is linear along the ow length (axial
direction) in many single-phase ow applications, and hence we generally work directly
with the pressure drop since there is no need to compute the pressure gradient in singlephase ow.
The total local pressure gradient in two-phase ow through a one-dimensional duct
can be calculated as follows{:
dp dpfr dpmo dpgr
C:1
dz
dz
dz
dz
{
Additional symbols used in this appendix are all dened here and are not included in the main nomenclature
section.
913
914
where the three terms on the right-hand side correspond to the contributions by friction,
momentum rate change, and gravity denoted by the subscripts fr, mo, and gr, respectively. The analysis that follows is based on a homogeneous model. The entrance and exit
pressure loss terms of single-phase ow [see Eq. (6.28)] are lumped into the pfr term
since the information about these contributions is not available, due to the diculty in
measurements. The in-tube two-phase frictional pressure drop is computed from the
corresponding pressure drop for single-phase ow as follows using the two-phase friction
multiplier denoted as 2 :
dp
dz
flo
fr
4 G2
2
Dh 2gc l lo
where 2lo
dp=dzfr
dp=dzfr;lo
C:2
where flo is the single-phase Fanning friction factor (see Tables 7.3 through 7.8) based on
the total mass ow rate as liquid and G is also based on the total mass ow rate as liquid;
this means that the subscript lo indicates the two-phase ow considered as all liquid
ow. The subscripts l and g in Eqs. (C.2) and (C.3) denote liquid and gas/vapor phases,
respectively, and the subscript lo stands for entire two-phase ow as liquid ow.
Alternatively, dp=dzfr is determined using the liquid or vapor-phase pressure drop
multiplier as follows.
dp
dz
fr
dp
dz
2l
fr;l
dp
dz
2g
C:3
fr;g
where
2l
dp=dzfr
dp=dzfr;l
2g
dp=dzfr
dp=dzfr;g
dp
4fl G2
dz fr;l 2gc l Dh
dp
dz
fr;g
4fg G2
C:4
2gc g Dh
where the subscripts l and g denote liquid and gas/vapor phases. 2lo and 2l or 2g are
functions of the parameter X (Martinelli parameter). 2go [dened similar to 2lo of Eq.
(C.2), with the subscript lo replaced by go] is a function of Y (Chisholm parameter). The
X and Y are dened as follows:
dp=dzfr;l
dp=dzfr;g
X2
Y2
dp=dzfr;go
dp=dzfr;lo
C:5
Here the subscript go means the total two-phase ow considered as all gas ow. The
correlations to determine the two-phase frictional pressure gradient are presented in
Table C.1 for various ranges of G and l =g (Kandlikar et al., 1999, p. 228).
The momentum pressure gradient can be calculated integrating the momentum
balance equation (Collier and Thome, 1994), thus obtaining
dp
dz
mo
d G2 x2
1 x2
dz gc g 1 l
C:6
where represents the void fraction of the gas (vapor) phase (a ratio of volumetric ow
rate of the gas/vapor phase divided by the total volumetric ow rate of the two-phase
mixture), and x is the mass quality (a ratio of the mass ow rate of the vapor/gas phase
915
TABLE C.1 Frictional Multiplier Correlations Used for Determining the Two-Phase Frictional
Pressure Gradient in Eq. (C.2)
Correlation
Parameters
3:24FH
Fr0:045 We0:035
fgo
E 1 x x l
g flo
2
F x0:78 1 x0:24
0:91
l
g
Fr
G2
gdi 2hom
g
l
0:19
1
We
1
x 1x
hom g
l
Chisholm correlation (1973) for l =g > 1000 and
G > 100 kg/m2 s:
2n
2
g
l
0:7
G 2 di
hom
(
B
G < 500
500 G 1900
G 1900
520=YG1=2
G 600
21=G
G > 600
9
>
=
>
;
)
for 9:5 < Y 28
d=dzfr
c
1
1 2
dp=dzl
X X
2g
dp=dzfr
1 cX X 2
dp=dzg
divided by the total mass ow rate of the two-phase mixture). Equation (C.6) is valid for
constant cross-sectional (ow) area along the ow length. For the homogeneous model,
the two-phase ow behaves like a single phase and the vapor and liquid velocities are
equal. A number of correlations for the void fraction are given by Carey (1992) and
Kandlikar et al. (1999). An empirical correlation for the void fraction whose general
form is valid for several frequently used models is given by Butterworth (Carey, 1992) as
1 x p g q l r 1
1A
l
g
x
C:7
916
where the constants A, p, q, and r depend on the two-phase model and/or empirical data
chosen. These constants for a nonhomogeneous model, based on steamwater data, are
A 1, p 1, q 0:89, and r 0:18. For the homogeneous model, A p q 1 and
r 0. For the Lockhart and Martinelli model, A 0:28, p 0:64, q 0:36, and
r 0:07. For engineering design calculations, the homogeneous model yields the best
results when the slip velocity between the gas and liquid phases is small (for bubbly or
mist ows).
Finally, the pressure gradient due to the gravity (hydrostatic) eect is
dp
dz
gr
g
sin g 1 l
gc
C:8
Note that the negative sign (i.e., the pressure recovery) stands for downward ow in
inclined or vertical tubes/channels, and the positive sign (i.e., pressure drop) represents
upward ow in inclined or vertical tubes/channels. And represents the angle of tube/
channel inclination measured from the horizontal axis.
C.2
C:9
where hcon is the condensation heat transfer coecient, Tsat is the saturation temperature
of the condensing uid at a given pressure, and Tw is the wall temperature. We summarize
here the correlations for lmwise in-tube condensation, a common condensation mode in
TABLE C.2 Heat Transfer Correlations for Internal Condensation in Horizontal Tubes
Stratication Conditions
Annular owa
(lm condensation) (Shah, 1977),
accuracy 14:4%
(Kandlikar et al., 1999)
Stratied ow
(Carey, 1992),
accuracy: 18% (Ould
Didi et al., 2002)
a
Correlation
hloc 0:023
Rel
Gdi
;
l
kl
3:8x0:76 1 x0:04
0:4
1 x0:8
Re0:8
l Prl
di
psat =pcr 0:38
G total mass velocity kg=m2 s
11 G 1599 kg=m2 s
0:002 psat =pcr 0:44
Prl > 0:5
21 Tsat 3108C; 0 x 1;
3 uvap 300 m=s; no limit on q
Rel > 350 for circular tubes
7 di 40 mm
1 x g 2=3 3=4 k3l l l g ghlg0 1=4
hm 0:728 1
l
1 Tsat Tw di
x
where hlg0 hlg 0:68cp;l Tsat Tw
917
most industrial applications. The two most common ow patterns for convective condensation are annular lm ow in horizontal and vertical tubes and stratied ow in
horizontal tubes. For annular lm ow, the correlation for the local heat transfer coecient hloc hcon hloc in Eq. (C.9)] is given in Table C.2; and also for stratied ow, the
correlation for mean condensation heat transfer coecient hcon hm is given in Table
C.2. Shah et al. (1999) provide condensation correlations for a number of noncircular
ow passage geometries.
C:10
where htp is the two-phase heat transfer coecient during the vaporization process. We
present here a most general intube forced convective boiling correlation proposed by
Kandlikar (1991). It is based on empirical data for water, refrigerants and cryogens. The
correlation consists of two parts, the convective and nucleate boiling terms, and utilizes a
uidsurface parameter. The Kandlikar correlation for the two-phase heat transfer
coecient is as follows:
8
< 0:6683Co0:2 f2 Frlo 1058 Bo0:7 Ffl 1 x0:8
htp
larger of
:
hlo
1:136Co0:9 f Fr 667:2Bo0:7 F 1 x0:8
2
C:11
fl
lo
where
hlo
(
f2 Frlo
8
>
>
>
<
>
>
>
:
2300 Relo 10
25Frlo 0:3
1
C:12
4
C:14
Here hlo is the single-phase heat transfer coecient for the entire ow as liquid ow. Also,
the convection number Co, the nucleate boiling number Bo, and the Froude number Fr
for the entire ow as liquid are dened as follows:
Co
g
l
0:5
1x
x
0:8
Bo
q 00
Ghg
Fr
G2
2l gdi
C:15
918
Ffl
Fluid
Ffl
1.00
1.30
1.50
1.31
2.20
1.30
R-114
R-134a
R-152a
R-32/R-132 (60%40% wt.)
Kerosene
1.24
1.63
1.10
3.30
0.488
Ffl is a uidsurface parameter and depends on the uid and the heat transfer surface. Ffl
values for several uids in copper tubes are presented in Table C.3. Ffl should be taken as
1.0 for stainless tubes. This correlation is valid for either vertical (upward and downward) or horizontal intube ow. A mean deviation of slightly less than 16% with water
and 19% with refrigerants has been reported by Kandlikar (1991).
Note that being uid specic, Ffl cannot be used for other uids (new refrigerants) and
mixtures. It is also not accurate for stratied wavy ows and at high vapor qualities since
it is not based on the onset of dryout. The Thome model (Kattan et al., 1998; Zrcher et al.,
1999), based on a ow pattern map, is recommended for those cases.
REFERENCES
Carey, V. P., 1992, Liquid-Vapor Phase Change Phenomena, Taylor & Francis, Bristol, PA.
Chisholm, D., 1967, A theoretical basis for the LockhartMartinelli correlation for two-phase ow,
Int. J. Heat Mass Transfer, Vol. 10, pp. 17671778.
Chisholm, D., 1973, Pressure gradients due to friction during the ow of evaporating two-phase
mixtures in smooth tubes and channels, Int. J. Heat Mass Transfer, Vol. 16, pp. 347358.
Collier, J. G., and J. R. Thome, 1994, Convective Boiling and Condensation, 3rd ed., McGraw-Hill,
New York.
Friedel, L., 1979, Improved friction pressure drop correlations for horizontal and vertical two-phase
pipe ow, European Two-Phase Flow Group Meeting, Ispra, Italy, Paper E2.
Hewitt, G. F., 1998, Gasliquid ow, in Handbook of Heat Exchanger Design, G. F. Hewitt, ed.,
Begell House, New York, Sect. 2.3.2.
Kandlikar, S. G., 1991, Development of a ow boiling map for subcooled and saturated ow boiling
of dierent uids in circular tubes, ASME J. Heat Transfer, Vol. 113, pp. 190200.
Kandlikar, S. G., M. Shoji, and V. K. Dhir, eds., 1999, Handbook of Phase Change: Boiling and
Condensation, Taylor & Francis, New York.
Kattan, N., J. R. Thome, and D. Favrat, 1998, Flow boiling in horizontal tubes, Part 1; Development of a diabatic two-phase ow pattern map, ASME J. Heat Transfer, Vol. 120, pp. 140147;
Part 2; New heat transfer data for ve refrigerants, ASME J. Heat Transfer, Vol. 120, pp. 148
155; Part 3; Development of a new heat transfer model based on ow patterns, ASME J. Heat
Transfer, Vol. 120, pp. 156165.
Ould Didi, M. B., N. Kattan, and J. R. Thome, 2002, Prediction of two-phase pressure gradients of
refrigerants in horizontal tubes, Int. J. Refrig., Vol. 25, pp. 935947.
Shah, M. M., 1977, A general correlation for heat transfer during subcooled boiling in pipes and
annuli, ASHRAE Trans., Vol. 83, No. 1, pp. 205215; also, M. M. Shah, 1982, Chart correlation
REFERENCES
919
for saturated boiling heat transfer: equations and further study, ASHRAE Trans., Vol. 88, No. 1,
pp. 185196.
Shah, R. K., S. Q. Zhou, and K. Tagavi, 1999, The role of surface tension in lm condensation in
extended surface passages, J. Enhanced Heat Transfer, Vol. 6, pp. 179216.
Zrcher, O., J. R. Thome, and D. Favrat, 1999, Evaporation of ammonia in a smooth horizontal
tube: heat transfer measurements and predictions, ASME J. Heat Transfer, Vol. 121, pp. 89101.
APPENDIX D
U and CUA Values for Various Heat
Exchangers
920
921
5,000
1,000
q=T
(W/K)
Boiling organic
liquid
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Low-viscosity
organic liquid
High-viscosity
liquid
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Low-viscosity
organic liquid
High-viscosity
liquid
Boiling water
Cold-Side Fluid
LowPressure
Gas
(< 1 bar)
55
5.70
93
5.02
120
5.51
105
4.89
99
4.96
68
5.39
105
4.89
99
4.96
55
2.11
93
1.63
120
2.26
105
1.56
99
1.59
68
1.86
Parameter
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)}
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
93
5.02
300
4.18
350
4.81
484
3.98
375
4.09
138
4.61
467
3.99
375
4.09
93
1.63
300
1.11
350
1.89
484
1.00
375
1.05
138
1.43
MediumPressure
Gas
(20 bar)
120
5.51
350
4.81
400
6.25
600
4.56
450
4.38
200
5.50
550
4.91
450
4.38
120
2.26
350
1.89
400
2.25
600
1.10
450
1.46
200
1.93
HighPressure
Gas
(150 bar)
102
4.93
429
4.03
600
4.56
938
3.77
600
3.91
161
4.46
875
3.79
600
3.91
102
1.58
429
1.02
600
1.10
938
0.88
600
0.95
161
1.36
Process
Water
99
4.96
375
4.09
450
4.38
714
3.85
500
3.97
153
4.51
677
3.87
500
3.97
99
1.59
375
1.05
450
1.46
720
0.91
500
0.99
153
1.38
LowViscosity
Organic
Liquid
63
5.50
120
4.76
200
5.50
142
4.59
130
4.67
82
5.16
140
4.60
130
4.67
63
1.95
120
1.49
200
1.93
142
1.41
130
1.46
82
1.71
HighViscosity
Fluid
Hot-Side Fluid
107
4.87
530
3.95
600
4.56
1607
3.61
818
3.81
173
4.42
1432
3.64
818
3.81
107
1.55
530
0.98
600
1.10
1607
0.83
818
0.89
173
1.32
Condensing
Steam
100
4.95
388
4.07
400
4.82
764
3.83
524
3.95
155
4.50
722
3.85
524
3.95
100
1.59
388
1.05
400
1.45
764
0.90
524
0.98
155
1.37
Condensing
Hydrocarbon
86
5.11
240
4.28
300
4.81
345
4.12
286
4.20
214
4.33
336
4.13
286
4.20
86
1.68
240
1.18
300
1.45
345
1.07
286
1.13
124
1.48
Condensing
Hydrocarbon
with
Inert Gas
922
100,000
30,000
q=T
(W/K)
105
1.56
99
1.59
55
1.11
93
0.76
120
1.06
105
0.71
99
0.74
68
0.94
105
0.71
99
0.74
55
0.95
93
0.58
120
0.93
105
0.52
Parameter
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
Cold-Side Fluid
Boiling water
Boiling organic
liquid
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Boiling organic
liquid
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Low-viscosity
organic liquid
High-viscosity
liquid
Boiling water
LowPressure
Gas
(< 1 bar)
467
1.00
375
1.05
93
0.76
300
0.37
350
0.62
484
0.29
375
0.33
138
0.57
467
0.29
375
0.33
93
0.58
300
0.23
350
0.35
484
0.17
MediumPressure
Gas
(20 bar)
550
1.20
450
1.46
120
1.06
350
0.62
400
0.94
600
0.40
450
0.53
200
0.73
550
0.49
450
0.53
120
0.93
350
0.35
400
0.58
600
0.24
HighPressure
Gas
(150 bar)
875
0.88
600
0.95
102
0.73
500
0.28
600
0.40
938
0.23
600
0.27
161
0.52
875
0.23
600
0.27
102
0.54
429
0.18
600
0.24
938
0.116
Process
Water
677
0.93
500
0.99
99
0.74
375
0.33
450
0.53
714
0.25
500
0.38
153
0.53
677
0.25
500
0.28
99
0.55
375
0.20
450
0.28
714
0.134
LowViscosity
Organic
Liquid
140
1.42
130
1.46
63
0.99
120
0.63
200
0.73
142
0.56
130
0.59
82
0.83
140
0.56
130
0.59
63
0.83
120
0.47
200
0.64
142
0.41
HighViscosity
Fluid
Hot-Side Fluid
1432
0.84
818
0.89
107
0.71
530
0.28
600
0.40
1607
0.19
818
0.24
173
0.50
1432
0.20
818
0.24
107
0.52
530
0.16
600
0.24
1607
0.086
Condensing
Steam
722
0.91
524
0.98
100
0.73
388
0.32
400
0.62
764
0.24
524
0.28
155
0.53
722
0.25
524
0.28
100
0.55
388
0.19
400
0.32
764
0.129
Condensing
Hydrocarbon
336
1.08
286
1.13
86
0.80
240
0.42
300
0.62
345
0.34
286
0.38
124
0.60
336
0.35
286
0.38
86
0.62
240
0.27
300
0.39
345
0.21
Condensing
Hydrocarbon
with
Inert Gas
923
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
99
0.55
68
0.77
105
0.52
99
0.55
375
0.20
138
0.42
467
0.168
375
0.20
450
0.28
200
0.64
550
0.26
450
0.28
609
0.145
161
0.37
875
0.121
600
0.146
500
0.162
153
0.38
677
0.137
500
0.162
130
0.44
82
0.65
140
0.41
130
0.44
Source: Selection and Costing of Heat Exchangers, ESDU Engineering Data 92013, ESDU International, London, 1994.
A CUA value for a q=Tm between values of q=Tm j , j 1, 2 should be calculated by logarithmic interpolation:
lnCUA;1 =CUA;2 lnq=Tm =q=Tm 1
CUA exp ln CUA;1
lnq=Tm 1 =q=Tm 2
Boiling organic
liquid
Low-viscosity
organic liquid
High-viscosity
liquid
Boiling water
818
0.125
173
0.35
1432
0.091
818
0.125
524
0.158
155
0.38
722
0.133
524
0.158
286
0.24
124
0.45
336
0.22
286
0.24
924
10,000
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Low-viscosity
hydrocarbon
High-viscosity
hydrocarbon
Boiling water
5,000
Boiling
hydrocarbon
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Low-viscosity
hydrocarbon
High-viscosity
Cold-Side Fluid
q=T
(W/K)
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
Parameter
491
1.55
NUS
NUS
NUS
NA
NA
315
1.55
NUS
NUS
NUS
402
3.10
217
1.55
325
1.55
NA
491
3.10
NUS
315
3.10
NUS
270
3.10
163
1.57
217
1.55
NA
217
3.10
325
3.10
NA
MediumPressure
Gas
(20 bar)
163
3.10
217
3.10
NA
LowPressure
Gas
(< 1 bar)
NA
NUS
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NUS
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Process
Water
NA
HighPressure
Gas
(150 bar)
TABLE D.2 U and CUA (C) Values for Plate-Fin Heat Exchangersa
NUS
NUS
NUS
453
3.10
264
1.55
377
1.55
NA
NA
NUS
NUS
NUS
264
3.10
377
3.10
NA
LowViscosity
Hydrocarbon
Liquid
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
HighViscosity
Hydrocarbon
Liquid
Hot-Side Fluid
NUS
NUS
NUS
NA
NUS
NUS
NA
NA
NUS
NUS
NUS
NA
NUS
NUS
Condensing
Steam
NUS
NUS
NUS
530
3.10
270
1.55
402
1.55
NA
NA
NUS
NUS
NUS
270
3.10
402
3.10
NA
Condensing
Hydrocarbon
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
Condensing
Hydrocarbon
with
Inert Gas
925
Boiling
hydrocarbon
Boiling
hydrocarbon
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Low-viscosity
hydrocarbon
High-viscosity
hydrocarbon
Boiling water
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
UW=m2 K
C[/(W/K)]
NA
NA
402
0.227
NUS
NUS
270
0.273
491
0.210
NUS
NA
NA
315
0.250
NUS
NUS
NUS
402
0.532
217
0.301
325
0.245
NA
491
0.513
NUS
315
0.560
NUS
270
0.579
163
0.336
217
0.301
NA
402
1.55
217
0.607
325
0.551
NA
NA
270
1.55
163
0.677
217
0.607
NA
NA
NA
NA
NA
NA
NA
NUS
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NUS
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
453
0.216
NA
NUS
NUS
NUS
453
0.527
264
0.280
377
0.231
NA
NA
NUS
NUS
NUS
453
1.55
264
0.574
377
0.537
NA
NA
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
Source: Selection and Costing of Heat Exchangers, ESDU Engineering Data 92013, ESDU International, London, 1994.
a
A CUA value for a q=Tm between values of q=Tm j , j 1, 2 should be calculated by logarithmic interpolation:
lnCUA;1 =CUA;2 lnq=Tm =q=Tm 1
CUA exp ln CUA;1
lnq=Tm 1 =q=Tm 2
100,000
30,000
Boiling
hydrocarbon
Low-pressure
gas (< 1 bar)
Medium-pressure
gas (20 bar)
High-pressure
gas (150 bar)
Treated cooling
water
Low-viscosity
hydrocarbon
High-viscosity
hydrocarbon
Boiling water
hydrocarbon
Boiling water
NA
NA
NUS
NUS
NUS
NA
NUS
NUS
NA
NA
NUS
NUS
NUS
NA
NUS
NUS
NA
NA
530
0.205
NA
NUS
NUS
NUS
530
0.518
270
0.273
402
0.227
NA
NA
NUS
NUS
NUS
530
1.55
270
0.579
402
0.532
NA
NA
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
NUS
NUS
NUS
NA
927
Foster, B. D., and J. B. Patton, eds., 1985, Ceramic Heat Exchangers, American Ceramic Society,
Columbus, OH.
Foumeny, E. A., and P. J. Heggs, eds., 1991, Heat Exchange Engineering, Vol. 1; Design of Heat
Exchangers, Ellis Horwood, London.
Foumeny, E. A., and P. J. Heggs, eds., 1991, Heat Exchange Engineering, Vol. 2; Compact Heat
Exchangers: Techniques for Size Reduction, Ellis Horwood, London.
Fraas, A. P., and M. N. Ozisik, 1989, Heat Exchanger Design, 2nd ed., Wiley, New York.
Ganapathy, V., 1982, Applied Heat Transfer, PennWell Publishing, Tulsa, OK.
Ganapathy, V., 2002, Industrial Boilers and Heat Recovery Steam Generators Design, Applications,
and Calculations, Marcel Dekker, New York.
Garrett-Price, B. A., S. A. Smith, R. L. Watts, J. G. Knudsen, W. J. Marner, and J. W. Suitor, 1985,
Fouling of Heat Exchangers, Noyes Publications, Park Ridge, NJ.
Gupta, J. P., 1986, Fundamentals of Heat Exchanger and Pressure Vessel Technology, Hemisphere
Publishing, Washington, DC; also as Working with Heat Exchangers, in soft cover, Hemisphere
Publishing, Washington, DC, 1990.
Hausen, H., 1983, Heat Transfer in Counterow, Parallel Flow and Cross Flow, McGraw-Hill, New
York.
Hayes, A. J., W. W. Liang, S. L. Richlen, and E. S. Tabb, eds., 1985, Industrial Heat Exchangers,
American Society for Metals, Metals Park, OH.
Hesselgreaves, J. E., 2001, Compact Heat Exchangers: Selection, Design, and Operation, Elsevier
Science, Oxford.
Hewitt, G. F., exec. ed., 1998, Heat Exchanger Design Handbook, three vols. (ve parts), Begell
House, New York; former publication: G. F. Hewitt, coord. ed., 1989, Hemisphere Handbook of
Heat Exchanger Design, Hemisphere Publishing, New York.
Hewitt, G. F., G. L. Shires, and T. R. Bott, 1994, Process Heat Transfer, CRC Press and Begell
House, Boca Raton, FL.
Hewitt, G. F., and P. B. Whalley, 1989, Handbook of Heat Exchanger Calculations, Hemisphere
Publishing, Washington, DC.
Hryniszak, W., 1958, Heat Exchangers: Applications to Gas Turbines, Butterworth Scientic
Publications, London.
Idelchik, I. E., 1994, Handbook of Hydraulic Resistance, 3rd ed., CRC Press, Boca Raton, FL.
Ievlev, V. M., ed., 1990, Analysis and Design of Swirl-Augmented Heat Exchangers, Hemisphere
Publishing, Washington, DC.
Jakob, M., 1957, Heat Transfer, Vol. II, Wiley, New York.
Kakac, S., ed., 1991, Boilers, Evaporators, and Condensers, Wiley, New York.
Kakac, S., ed., 1999, Heat Transfer Enhancement of Heat Exchangers, Kluwer Academic Publishers,
Dordrecht, The Netherlands.
Kakac, S., A. E. Bergles, and E. O. Fernandes, eds., 1988, Two-Phase Flow Heat Exchangers:
Thermal Hydraulic Fundamentals and Design, Kluwer Academic Publishers, Dordrecht, The
Netherlands.
Kakac, S., A. E. Bergles, and F. Mayinger, eds., 1981, Heat Exchangers: Thermal-Hydraulic Fundamentals and Design, Hemisphere Publishing, Washington, DC.
Kakac, S., and H. Liu, 1998, Heat Exchangers: Selection, Rating, and Thermal Design, CRC Press,
Boca Raton, FL.
Kakac, S., R. K. Shah, and W. Aung, eds., 1987, Handbook of Single-Phase Convective Heat
Transfer, Wiley, New York.
Kakac, S., R. K. Shah, and A. E. Bergles, eds., 1983, Low Reynolds Number Flow Heat Exchangers,
Hemisphere Publishing, Washington, DC.
928
Katinas, V., and A. Zukauskas, 1997, Vibrations of Tubes in Heat Exchangers, Begell House, New
York.
Kays, W. M., and A. L. London, 1998, Compact Heat Exchangers, reprint 3rd edn., Krieger Publishing, Malabar, FL; rst ed., National Press, Palo Alto, CA (1955); 2nd ed., (1964), 3rd ed.,
McGraw-Hill, New York (1984).
Kern, D. Q., 1950, Process Heat Transfer, McGraw-Hill, New York.
Kern, D. W., and A. D. Kraus, 1972, Extended Surface Heat Transfer, McGraw-Hill, Chaps. 912,
pp. 439641.
King, R., ed., 1987, Flow Induced Vibrations, BHRA Publication, London.
Kraus, A. D., 1982, Analysis and Evaluation of Extended Surface Thermal Systems, Hemisphere
Publishing, Washington, DC.
Kraus, A. D., A. Aziz, and J. R. Welty, 2001, Extended Surface Heat Transfer, Wiley, New York.
Kroger, D. G., 1998, Air-Cooled Heat Exchangers and Cooling Towers, Tecpress, Uniedal, South
Africa; also, Begell House, New York.
Kuppan, T., 2000, Heat Exchanger Design Handbook, Marcel Dekker, New York.
Lokshin, V. A., D. F. Peterson, and A. L. Schwarz, 1988, Standard Handbook of Hydraulic Design
for Power Boilers, Hemisphere Publishing, Washington, DC.
Ludwig, E. E., 1965, Applied Process Design for Chemical and Petrochemical Plants, Vol. III, Gulf
Publishing, Houston, TX, Chap. 10.
Manzoor, M., 1984, Heat Flow through Extended Surface Heat Exchangers, Springer-Verlag, Berlin.
Martin, M., 1992, Heat Exchangers, Hemisphere Publishing, Washington, DC.
Marto, P. J., and R. H. Nunn, eds., 1981, Power Condenser Heat Transfer Technology, Hemisphere
Publishing, Washington, DC.
Marvillet, Ch., gen. ed., 1994, Recent Developments in Finned Tube Heat Exchangers: Theoretical
and Practical Aspects, DTI Energy Technology, Danish Technological Institute, Taastrup,
Denmark.
Marvillet, C., and R. Vidil, eds., 1993, Heat Exchanger Technology: Recent Developments, Eurotherm Seminar 33, Editions Europeennes Thermique et Industrie, Paris.
McNaughton, K. J., ed., 1986, The Chemical Engineering Guide to Heat Transfer; Vol. 1; Plant
Principles, Vol. 2; Equipment, Hemisphere Publishing, Washington, DC.
Melo, L. F., T. R. Bott, and C. A. Bernardo, eds., 1988, Advances in Fouling Science and Technology,
Kluwer Academic Publishers, Dordrecht, The Netherlands.
Miller, D. S., 1990, Internal Flow Systems, 2nd ed., BHRA Fluid Engineering Series, Vol. 5, BHRA,
Craneld, UK.
Minton, P. E., 1986, Handbook of Evaporator Technology, Noyes Publications, Park Ridge, NJ.
Mori, Y., A. E. Sheindlin, and N. H. Afgan, eds., 1986, High Temperature Heat Exchangers,
Hemisphere Publishing, Washington, DC.
Muller-Steinhagen, H., ed., 2000, Heat Exchanger Fouling: Mitigation and Cleaning Technologies,
Publico Publications, Essen, Germany.
Palen, J. W., ed., 1987, Heat Exchanger Sourcebook, Hemisphere Publishing, Washington, DC.
Panchal, C. B., T. R. Bott, E. F. C. Somerscales, and S. Toyama, 1997, Fouling Mitigation of
Industrial Heat Exchange Equipment, Begell House, New York.
Podhorsky, M., and H. Krips, 1998, Heat Exchangers: A Practical Approach to Mechanical
Construction, Design and Calculations, Begell House, New York.
Putman, R. E., 2001, Steam Surface Condensers: Basic Principles, Performance Monitoring and
Maintenance, ASME Press, New York.
Reay, D. A., 1979, Heat Recovery Systems, E&FN Spon, London.
929
Reay, D. A., 1999, Learning from Experiences with Compact Heat Exchangers, CADDET Analyses
Series 25, Centre for the Analysis and Dissemination of Demonstrated Energy Technologies,
Sittard, The Netherlands.
Rifert, V. G., 1998, Condensation Heat Transfer Enhancement, Computational Mechanics Publications, WIT Press, Southampton, UK.
Roetzel, W., P. J. Heggs, and D. Butterworth, eds., 1991, Design and Operation of Heat Exchangers,
Springer-Verlag, Berlin.
Roetzel, W., and Y. Xuan, 1998, Dynamic Behaviour of Heat Exchangers, Vol. 3, Computational
Mechanics Publications, WIT Press, Southampton, UK.
Saunders, E. A. D., 1989, Heat Exchangers: Selection, Design and Construction, Wiley, New York.
Schlunder, E. U., ed.-in-chief, 1982, Heat Exchanger Design Handbook, 5 vols., Hemisphere Publishing, Washington, DC.
Schmidt, F. W., and A. J. Willmott, 1981, Thermal Energy Storage and Regeneration, Hemisphere/
McGraw-Hill, Washington, DC.
Shah, R. K., K. J. Bell, H. Honda, and B. Thonon, eds., 1999, Compact Heat Exchangers and
Enhancement Technology for the Process Industries, Begell House, New York.
Shah, R. K., K. J. Bell, S. Mochizuki, and V. V. Wadekar, eds., 1997, Compact Heat Exchangers for
the Process Industries, Begell House, New York.
Shah, R. K., A. W. Deakin, H. Honda, and T. M. Rudy, eds., 2001, Compact Heat Exchangers and
Enhancement Technology for the Process Industries 2001, Begell House, New York.
Shah, R. K., and A. Hashemi, eds., 1993, Aerospace Heat Exchanger Technology, 1993, Elsevier
Science, Amsterdam, The Netherlands.
Shah, R. K., A. D. Kraus, and D. Metzger, eds., 1990, Compact Heat Exchangers: A Festschrift for
A.L. London, Hemisphere Publishing, Washington, DC.
Shah, R. K., and A. L. London, 1978, Laminar Flow Forced Convection in Ducts, Supplement 1 to
Advances in Heat Transfer Series, Academic Press, New York.
Shah, R. K., and A. C. Mueller, 1985, Heat exchangers, in Handbook of Heat Transfer Applications,
W. M. Rohsenow, J. P. Hartnett, and E. N. Ganic, eds., McGraw-Hill, New York, Chap. 4, pp.
1312.
Shah, R. K., and A. C. Mueller, 1989, Heat exchange, in Ullmans Encyclopedia of Industrial
Chemistry, Unit Operations II, Vol. B3, Chap. 2, VCH Publishers, Weinheim, Germany.
Shah, R. K., and D. P. Sekulic, 1998, Heat exchangers, in Handbook of Heat Transfer, W. M.
Rohsenow, J. P. Hartnett, and Y. I. Cho, eds., McGraw-Hill, New York, Chap. 17.
Shah, R. K., E. C. Subbarao, and R. A. Mashelkar, eds., 1988, Heat Transfer Equipment Design,
Hemisphere Publishing, Washington, DC.
Sheindlin, A. E., ed., 1986, High Temperature Equipment, Hemisphere Publishing, Washington, DC.
Singh, K. P., and A. I. Soler, 1984, Mechanical Design of Heat Exchangers and Pressure Vessel
Components, Arcturus Publishers, Cherry Hill, NJ.
Smith, E. M., 1997, Thermal Design of Heat Exchangers: A Numerical Approach: Direct Sizing and
Stepwise Rating, Wiley, New York.
Smith, R. A., 1987, Vaporisers: Selection, Design, Operation (Designing for Heat Transfer), Wiley,
New York.
Somerscales, E. F. C., and J. G. Knudsen, eds., 1981, Fouling of Heat Transfer Equipment,
Hemisphere/McGraw-Hill, Washington, DC.
Soumerai, H., 1987, Practical Thermodynamic Tools for Heat Exchanger Design Engineers, Wiley,
New York.
Stasiulevicius, J., and A. Skrinska, 1987, Heat Transfer of Finned Tube Bundles in Crossow,
Hemisphere Publishing, Washington, DC.
930
Sukhotin, A. M., and G. Tereshchenko, 1998, Corrosion Resistance of Equipment for Chemical
Industry Handbook, Begell House, New York.
Sunden, B., and M. Faghri, eds., 1998, Computer Simulation in Compact Heat Exchangers, Computational Mechanics Publications, WIT Press, Southampton, UK.
Sunden, B., and P. J. Heggs, eds., 1998, Recent Advances in Analysis of Heat Transfer for Fin Type
Surfaces, Computational Mechanics Publications, WIT Press, Southampton, UK.
Sunden, B., and R. M. Manglik, eds., 2001, Plate and Frame Heat Exchangers, Computational
Mechanics Publication, WIT Press, Southampton, UK.
Taborek, J., G. F. Hewitt, and N. Afgan, eds., 1983, Heat Exchangers: Theory and Practice,
Hemisphere/McGraw-Hill, Washington, DC.
Taylor, M. A., 1987, Plate-Fin Heat Exchangers: Guide to Their Specication and Use, HTFS,
Harwell Laboratory, Oxon, UK.
Walker, G., 1990, Industrial Heat Exchangers: A Basic Guide, 2nd edn., Hemisphere Publishing,
Washington, DC.
Webb, R. L., 1994, Principles of Enhanced Heat Transfer, Wiley, New York.
Willmott, A. J., 2001, Dynamics of Regenerative Heat Transfer, Taylor & Francis, New York.
Yokell, S., 1990, A Working Guide to Shell-and-Tube Heat Exchangers, McGraw-Hill, New York.
Zukauskas, A. A., 1989, High Performance Single-Phase Heat Exchangers, Hemisphere Publishing,
Washington, DC. [This book has a misleading title. It should be Forced Convection Heat
Transfer.]
Zukauskas, A., and R. Ulinskas, 1988, Heat Transfer in Tube Banks in Crossow, Hemisphere
Publishing, Washington, DC.
Zukauskas, A. A., R. Ulinskas, and V. Katinas, 1988, Fluid Dynamics and Flow Induced Vibrations of
Tube Banks, Hemisphere Publishing, Washington, DC.
Index
Absorptivity, 539
gas, 542
Advection, 439
ASME code(s), 13
Analogy between uid ow and electric
entities, 9899
Analytical correlations, 473. See also
Correlations, and Heat transfer
coecient
fully developed ows, 475
hydrodynamically developing ows, 499
laminar ow, 475
simultaneously developing ow, 507
thermally developing ows, 502
Annular ow, 916
Arithmetic mean, 187
Bae(s):
disk-and-doughnut, 683
grid, 682
impingement, 684
plate, 682
rod, 684
segmental, 682, 683
strip, 683
Bae geometry, 588
Balances, 776
cost, 783
energy, 779
exergy, 786
Balance equations, 102, 115, 260, 269, 314,
739, 750
Bavex welded-plate, 30
Bell-Delaware method, 294, 647. See also Heat
exchanger design methodology
correction factors, 648, 650
Bend losses:
circular cross section, 405
miter bends, 409
rectangular cross section, 409
Bhatti-Shah correlation, 482
Biological fouling, 869
931
932
INDEX
INDEX
933
934
INDEX
INDEX
935
uidized-bed, 6
gas-liquid, 8
gas-to-uid, 11
heat transfer elements, 3
immiscible uid, 8
indirect transfer type, 1
indirect-contact, 3
irreversibilities, 755
laminar ow, 9
liquid-coupled
liquid-to-liquid, 12
liquid-vapor, 8
meso heat exchanger, 9
micro heat exchanger, 9
modeling, 738
MTD method, 209
multipass cross-counterow, 168
multipass crossow exchangers, 65
multipass cross-parallel ow, 66, 170
multipass, 64
number of shells in series, 163
operating condition variables, 104
overall counterow, 65
over-and-under passes, 65
P1 -P2 method, 211
paralleow, 58
parallel coupling, 65
performance, 787
phase-change, 12
P-NTU method, 209
psi ( )-P method, 210
principal features, 676
recuperators, 1, 4
sensible, 1
series coupling, 65
side-by-side passes, 65
single-pass, 57, 122
storage type, 5
surface compactness, 8
surface geometrical characteristics, 563
surface heat exchanger, 3
train, 164
tubular, 13
two-pass, 57
Heat Exchanger Arrays, 201
Heat exchanger design methodology, 78. See
also Heat exchanger
costing, 90
exchanger specication, 81
manufacturing considerations, 90
mechanical design, 87
optimum design, 93
overview, 78
problem specications, 79
936
INDEX
INDEX
937
938
INDEX
Passage-to-passage ow maldistribution
(continued)
Other eects, 833
two-passage model, 822
Peclet number (Pe), 443, 448
Performance (eectiveness) deterioration
factor, 813
Performance evaluation criteria, 699, 713,
714
algebraic formulas, 717
direct comparisons of j and f, 700
xed ow area, 714
xed geometry, 714
uid pumping power, 700
reference surface, 700
variable geometry, 714
Induction period, 872
Periodic ow, 437
Periodic ow regenerator, 47
Petukhov-Popov correlation, 482, 484
Pinch analysis, 776, 779
Pipe losses, 399
Plate-n heat exchanger, 37, 584, 605
Plate heat exchanger, 185, 597, 632. See also
Heat exchanger
heat transfer-limited design, 635
limiting cases for the design, 633
mixed channels, 635
multipass, 185
pressure drop-limited design, 635
rating a PHE, 637
rating and sizing, 635
sizing, 645
Plate pack, 23
rating problem, 605
sizing problem, 617
super elastically deformed diusion bonded,
40
Plate-type heat exchangers, 22, 693
advantages and limitations, 28
channel, 25
ow arrangements, 27
gasket materials, 26
geometrical and operating condition
characteristics, 27
hard or soft plates, 25
looped patterns, 71
major applications, 29
multipass, 64, 71
pass, 25
series ow, 71
thermal plates, 27
U-arrangement, 72
Z-arrangement, 72
INDEX
939
transition, 497
Run-around coil system, 911
Sand-grain roughness, 497
Schmidt number (Sc), 509
Second law eciency, 787
Second law of thermodynamics, 723, 735, 776,
796
evaluation, 723, 796
performance evaluation criteria, 796
Sedimentation fouling, 868
Selection guidelines for major exchanger
types, 680
extended-surface exchangers, 694
plate heat exchangers, 693
plate-n exchanger surfaces, 694
regenerator surfaces, 699
shell-and-tube exchangers 680
Shell-and-tube exchangers, 13, 68, 183,
291, 646, 766. See also Flow
arrangements
additional considerations, 291
approximate design method, 658
baes, 18, 682
bundle-to-shell bypass stream, 292
comparison of various types, 21
correction factor pressure drop, 649
crossow section, 591
crossow stream, 292
design features, 689
disk-and-doughnut bae, 683
divided-ow exchanger, 71
external low-nned tubes, 648
nite number of baes, 297
front and rear end heads, 18, 688
grid baes, 18, 682
heat transfer calculation, 646
helical bae, 18
impingement baes, 684
increase heat transfer, 693
leakage and bypass streams, 292
low ns, 17
multipass, 183
no-tubes-in-window design, 648
nozzles, 17
parallel counterow exchanger 68
plate baes, 18, 682
preliminary design, 646
pressure drop calculation, 646
rating, 646
rear-end heads, 688
reduce pressure drop, 693
rigorous thermal design method, 663
rod baes, 18, 684
940
INDEX
INDEX
941