Professional Documents
Culture Documents
GIFT OP
SABRE FOUNDATION U M
NOT FOR RESAf pr
McGraw-Hill
A Division o f The M cG row -H ill Companies
Copyright © 2002 by The McGraw-Hill Companies, Inc. All rights reserved. Printed in the United
States of America. Except as permitted under the United States Copyright Act of 1976, no part of this
publication may be reproduced or distributed in any form or by any means, or stored in a data base or
retrieval system, without the prior written permission of the publisher.
1 2 3 4 5 6 7 8 9 0 DOC/DOC 0 9 8 7 6 5 4 3 2
ISBN 0-07-136372-6
The sponsoring editor fo r this book was Larry S. Hager and the production
supervisor was Sherri Souffrance. It was set in Times Roman by Lone Wolf
Enterprises, Ltd.
McGraw-Hill books are available at special quantity discounts to use as premiums and sales
promotions, or for use in corporate training programs. For more information, please write to the
Director of Special Sales, McGraw-Hill, Professional Publishing, Two Penn Plaza, New York, NY
10121-2298. Or contact your local bookstore.
Information contained in this work has been obtained by The McGraw-Hill Companies, Inc. (“McGraw-
Hill” ) from sources believed to be reliable. However, neither McGraw-Hill nor its authors guarantee the
accuracy or completeness of any information published herein, and neither McGraw-Hill nor its authors
shall be responsible for any errors, omissions, or damages arising out of use of this information. This
work is published with the understanding that McGraw-Hill and its authors are supplying information
but are not attempting to render engineering or other professional services. If such services are required,
the assistance of an appropriate professional should be sought.
CONTENTS
Contributors ix
Preface xi
Index 1.1
CONTRIBUTORS
Fraser King, Ph.D. NOVA Research and Technology Corporation. (Chapter 25)
John Tuzson, Ph.D. Pump Consultant, John Tuzson and Associates. (Chapter 17)
This handbook required expertise in many areas of fluid flow, and I would like
to acknowledge each of the 26 professionals who contributed their time and tal
ent to the project. I have been involved in software development of fluid flow
computer simulation for the oil and chemical industries for a few years, with a
few more years as a flow assurance engineer for the deepwater oil and gas pro
duction and transportation industry, but I couldn't have completed a task of this
size without the help of these contributors. I am very glad I could get profes
sionals in many of the fluid flow disciplines to summarize some of their experi
ence in this handbook. I admit that I have learned a lot while proofreading the
various chapters, and I am sure readers of different backgrounds will enjoy and
learn while reading this handbook.
Special thanks to the staff of Lone W olf Enterprises, including Roger Woodson,
Barb Karg, and Rick Sutherland for their patience and guidance throughout the
production process of the book. My family has sacrificed priceless moments to
see this book come into existence. When I started this project, my daughter Huda
was a week old and I was distracted many times, days and nights, by her smiles
and cries. I am writing this preface after 18 months, and I am distracted again, but
this time by her crayon. My other kids, Sondoss, Yahya, and Heba, have also
grown used to see me working on the book and I appreciate their patience and sac
rifice of free time ever since the project began. My wife has suffered the most and
has helped type many pages and tables of the book. She was the shining and guid
ing star when things appeared dark. My parents deserve my deepest appreciation
for their continuous support and encouragement.
1.1
1.2 FL U ID FLO W HANDROOK
have continued to our current time. The reader is encouraged to explore the refer
ences at the end of this introduction for a detailed history of fluid mechanics [1,2].
Fluid flow may be classified into many pairs of categories, some of which are:
In this handbook, the reader will find chapters that discuss all of the above
categories, and much more. The purpose o f this book is to provide the reader with
a quick reference to many areas in fluid flow. In addition to covering traditional
areas in fluid flow, such as pressure drop in pipes and networks, pumps, flow
meters, incompressible, compressible, and multiphase flows, minor losses due to
pipe fittings, and flow in open channels, I have strived to add non-conventional
and new developing areas in fluid flow, such as: flow in nano- and micro- chan
nels, deepwater flow assurance for gas and oil pipelines, flow of fluids under vac
uums, dynamics of blood flow, corrosion and erosion associated with fluid flow
in pipes, transient flow, and many more.
The reader will be able to gain a general understanding of particular topics
through definitions, classifications, and industrial applications. Also, equations,
tables, and graphs are included to enable the user to solve, analyze, and assess
fluid flow problems. It is not intended to cover each topic in full detail, but rather
to provide a well-documented entry point with balanced theory and estimation
methods so that reader may enjoy, learn, and apply solutions to practical prob
lems. The reference list at the end of each chapter is an excellent resource for full
coverage of the particular topic.
While a few chapters deal with derivation of fluid flow equations, such as
Chapter 6, “The General Equations,” Chapter 7, “Application of the Equation of
Change,” and Chapter 31, “Turbulent Flow,” the bulk of the handbook provides
equations, charts, figures, and recommendations to solve and analyze fluid flow
problems. It was necessary to add the “Fluid Statics” chapter (Chapter 4) to clar
ify many concepts, such as pressure head. Chapters 2, 3, and 27 discuss the topics
o f “Fluid Properties,” “Fluid Flow and Thermodynamics,” and “Heat Transfer in
Pipe Flow,” respectively, which is necessary since the flow of fluids is directly
affected by the fluid properties, the thermodynamics of the flow process, and the
heat transfer between the fluid and the surroundings.
Chapters 8, 9, 11, 12, 13, and 14 cover the principles and characteristics of
different fluid flows inside pipes. When considering flow of fluid in pipes, it is
essential to include the effect of piping and other elements (elements to control
the flow of fluids) on flow behavior. Topics on fluid flow, piping, and How com
INTRODUCTION 1.3
ponents such as fittings, flow meters, flow control, and fluid machinery are dis
cussed in Chapter 10, “Flow Minor Losses,’' Chapter 15, “Flow Metering," Chap
ter 16, “Flow Control,” and Chapter 17, “Fluid M achines.” In real life, flow of
fluids is encountered in complex networks of pipes. Chapter 18 deals with analy
sis and design of fluid flow networks.
Flow in open channels, flow past immersed objects, and flow through porous
materials have wide applications in urban development and industrial sectors, and
is therefore included as Chapters 19, 20, and 21, respectively. Numerous fluid
flow applications must be treated in a transient, rather than steady state to account
for the time trends of pressure, temperature, flow rates, and other variables that
may be severe enough to cause failure in the transport systems. Chapter 22 deals
with transient flow. Chapter 23 deals with flow hydrodynamics in processing
units such as mixers, reactors, contactors, and heat exchangers. The efficiency of
processing units is a strong function of the fluid hydrodynamics. Serious m al
functions, such as channeling or dead zones, may result due to improper design
of the hydrodynamic aspects of the processing unit. Chapter 24, “Computational
Fluid Dynamics,” covers the ever-growing science of computational fluid
dynamics. This chapter discusses numerical approximation to the equations that
govern fluid motion, and algorithms to solve the system equation by direct or iter
ative procedures. Flow-induced corrosion and erosion in pipes are common prob
lems in industry and can damage operating systems. Chapter 25 covers
flow-induced erosion-corrosion problems.
C hapter 26, “ Blood Flow Dynamics,” discusses biofluid mechanics or fluid
flow inside our bodies, which is an important subject to the medical community.
Micro- and nano-scale devices are emerging technologies and have many appli
cations in the areas of biomedical engineering, information technology, materials
engineering, and energy and environmental engineering. Chapter 28 discusses
fluid flow in micro and nano channels.
The production and transportation of oil and gas from deepwater reservoirs
offshore west Africa and in deep sections of the Gulf of Mexico faces many chal
lenges, such as hydrate formation and wax deposition. Chapter 29 discusses top
ics in oil and gas transportation. Energy conservation has and will continue to be
a concern to world civilization because fluid transport requires pumping power.
C hapter 30 discusses the fluid flow and drag reduction topic.
REFERENCES
Knowledge and good understanding of the fluid physical and transport proper
ties are essential to accomplish sound engineering design and analysis of fluid
flow problems. In fluid flow analysis and design, density and viscosity are the
most pertinent physical and transport properties. Most fluid flow hydrodynamic
and pressure-drop functions are, in some way, functions of fluid density and vis
cosity. The role of surface tension, a liquid phase transport property, comes into
play for two-phase flow. Accurate fluid flow design requires precise methods to
predict fluid density, viscosity, and surface tension. Although, the subject of
physical and transport properties is a complex and lengthy topic that might
require a book of this book size to cover, a brief description of fluid properties,
which have a significant effect on hydrodynamic calculation, is necessary. The
following sections summarize the definitions, units, and methods to predict (or
sources for reliable) fluid properties pertinent to fluid flow.
2.1
2.2 FLUID FLO W HANDBOOK
Fluid density is defined as the fluid mass per unit volume of the fluid. The property
o f fluid density plays a sensitive role in flow hydrodynamics and pressure-drop
calculations. For a given mass flow rate, the fluid velocity, and hence pressure-
drop calculation, is a function of the density property. The following example illus
trates the effect of density on gas frictional pressure-drop calculation.
Case 1
V = — (2.1)
pA
where
Using Equation 2.1, velocity is calculated to be: 41.16 ft/s (12.54 m/s)
Using Darcy equation to calculate the frictional pressure drop in feet per unit
length with the assumption of a constant frictional factor of 0.016
h = & -— ( 2 .2 )
D 2g
where
From Equation 2.2, the frictional head loss is calculated as /? = 1.255 ft (0.382 m)
Case 2
Using Equation 2.2, frictional head loss, assuming a constant frictional factor
of 0.016, is calculated to be: 13.52 ft (4.123 m).
The above example illustrates the effect of density on the frictional pressure-
drop calculations. The only difference between Case 1 and Case 2 is in the den
sity property. The change in density value affects the fluid velocity and hence the
final pressure-drop value. Many of the flow hydrodynamics parameters such as
Reynolds Number, pressure drop, and flow regime maps (for 2-phase flow) are
nonlinear functions of the velocity. For example, Equation 2.2, which is the Darcy
equation for the prediction of frictional pressure drop of a homogeneous fluid, is
nonlinear in velocity. For incompressible fluids (such as liquids), density depen
dency on the temperature is strong, while the liquid density is a weak function of
pressure. However, when the fluid is gas, the dependency of density on tempera
ture and pressure is strong. Liquid density decreases with increasing temperature,
while gas density decreases with increasing temperature and increases with increas
ing pressure. Figure 2.1 shows the air density as function of temperature and pres
sure. Density has the units of mass per unit volume, such as pounds per cubic
foot, kilograms per cubic meter, or grams per cubic centimeter. Methods and
sources to predict the gas and liquid densities for pure components and mixtures
will follow.
Experimental saturated liquid density data for hundreds of organic and inorganic
pure components have been compiled by many chemical and research organiza
tions. The following references or sources provide an excellent source for pure
component saturated liquid densities.
1. The Design Institute for Physical Property Data (DIPPR), which is an orga
nization of the American Institute of Chemical Engineers (AICHE), pro
vides a large compilation of experimental liquid densities for more than
2 .4 FL U ID FLO W HANDBOOK
FIGURE 2.1 Air density as function of temperature and pressure. (To change chart
values to kg/m3, multiply chart value by 16.018.)
1,500 components. Data has been curve-fitted into polynomials with tem
perature dependency coefficients. Data are available commercially via
many software interfaces.
2. Chemical Properties Handbook by Yaws [32] lists more than 20 temperature
dependents liquid and vapor property such as viscosity, density, vapor pres
sure, and heat capacity for more than 1,700 organic and inorganic components.
3. The Thermodynamic Research Center at Texas A & M University has also
com piled data for thousands of components over wide range of tem pera
ture. More information may be obtained from their Web site at
http://www.trcweb.tamu.edu.
Liquid Density
Temp.F (C)
Saturated liquid density may also be estimated by the modified Rackett equa
tion [22, 26). The modified Rackett equation is summarized as
where
Tc = Critical temperature
Pc = Critical pressure
R = Universal gas constant
Z ra = A unique constant for each compound
Vs = Saturated liquid molar volume
Values for Z RA may be found in Reid [24], Rackett [22] and Spencer and
Danner [26], or estimated by the following correlation which was suggested by
Yamada [30]:
Solution
T c= 1165.14 R (647.3 K)
Pc = 3208.23 psia (22120.15 kPa)
w = 0.344 (Acentric factor)
ZRA for W ater (Table 3.10, Reid [24]) = 0.2338
W ater density is 63.027 lb/ft3 (1009.56 kg/m3), which is about 3.89% higher
than the reported measured value. Compressed liquid density, liquid density at a
FLUID PROPERTIES 2.7
Liquid density o f mixtures is calculated using the following mixing rule over the
pure com ponent’s molar volume. The mixing rule will give a good approxima
tion when m ixture’s components do not differ greatly.
I n
— = v!: = X • <2 -5 )
2.8 F LU ID FLO W HANDBOOK
where
EXAMPLE 2.3
Estimate the liquid density o f a hydrocarbon mixture o f n-Hexane and n-Hep-
tane (50% mole composition) at 77° F (25° C) and 14.7psia (101.326 kPa),
and compare it with a reported value o f 41.565 lb/ft? (665.807kg/m3).
Solution
Using Equation 2.5 with pure components liquid densities of n-Hexane and
n-Heptane of 40.957 lb/ft3 (656.068 kg/m3) and 42.54 lb/ft3 (681.425
kg/m3) respectively, the mixture density is calculated as
Pi Mw, yt
lb/ft3 (kg/m3) ftVlbmol (mVkmol)
The calculated mixture liquid density is 0.54% higher than the reported value.
The modified Rackett equation, Equation 2.3, may also be used to find mix
ture liquid density. Few of the equation terms must be changed to accommodate
the mixture critical data rather than the pure component critical data. The modi
fied Rackett equation for mixtures is
( 2 .6 )
Many mixing rules have been suggested for calculating the mixture reduced
temperature. A simple one is
/ * i l ci
FLU ID PR O PERTIES 2.9
Solution
ZRA values for n-Hexane and n-Heptane (Table 3.10, Reid [24]) are 0.2635
and 0.2604 respectively.
PM xv
(2.7)
ZTR
where
p = Absolute pressure
Mw = Gas molecular weight
T = Absolute temperature
Z = Compressibility factor
R = Universal gas constant
Any system o f consistent units may be used. Table 2.2 lists values of the Uni
versal Gas Constant for different units. The compressibility factor Z refers to the
deviation from ideal gas behavior at high pressure and temperature. The Z factor
approaches l.O at standard conditions (1 atmosphere and 60° F). As gas temper
ature and pressure increase, the compressibility factor will decrease. However, at
very high pressure, Pr > 5, the compressibility factor might exceed 1.0. There are
2 .1 0 F LU ID FLOW HANDBOOK
many correlations and methods to predict the gas compressibility factor. These
methods can be categorized as:
A brief discussion with examples of the above categories will be explained below.
( 2 .8 )
(2.9)
FLU ID PRO PERTIES 2.11
Pr
FIGURE 2.3 Compressibility factor Z, as function of reduced temperature and reduced pressure.
where
Pure components critical data may be obtained from critical data tables such
as those available in Reid [24], and Yaws [32]. Table 2 .1 lists critical data and
molecular weight for some of the most common industrial gases.
For a gas mixture of known composition, the m ixture's reduced temperature
and reduced pressure are calculated by Equations 2.8 and 2.9 with the m ixture’s
pseudocritical data rather than the true critical data. The pseudocritical data are
found to be the following set of equations:
N
Tc = Y x J ci (2.10)
i= I
N
Pc = l X,pci (2.I1)
i= I
2 .1 2 FLU ID FLO W HANDBOOK
where
When the specific gravity or m olecular weight is the only known parameter
o f the gas, i.e.gas composition is unknown, such as the case with many natural
gas wells, thepseudocritical data may be estimated bythe following formula of
the American Gas Association (AGA)
where
The gas specific gravity at standard conditions is defined as the gas density at
standard condition divided by the air density at standard conditions
£ — PcJus _ M W q os ^2 |4)
pAir M w Air
Equations 2.12 and 2.13 apply hydrocarbon gases with specific gravity of less
than 0.75. Hankinson et al [ 12] have developed the following set of equations to esti
mate the pseudocritical data for natural gases over the gravity range of 0.55 - 1.7
Pc = 7 0 9 .6 4 -5 8 .7 1 8 5 , (2.15)
where
Equations 2.15 and 2.16 are applicable with impurity contents less than 7%.
FLU ID PRO PERTIES 2 .13
EXAMPLE 2.5
Estimate the compressibility fa cto r and gas density fo r a natural gas at / 10° F
(43.3° C) and 270 Psia (1861.6 kPa). The natural gas average molecular
weight is 31.8.
Solution
The gas specific gravity is estimated using Equation 2.14
5,, = 3 1 .8 / 2 9 = 1.09
Gas critical pressure and critical temperature are estimated using Equations
2.15 and 2.16 respectively
Gas-reduced pressure and temperature are calculated using Equations 2.8 and
2.9 respectively
where
A summary o f values for Z (0) and Z U) recommended by Lee and Kesler [17]
are listed in Table 2.3 and Table 2.4. For the complete tabulation of Z (0) and Z (1),
the reader is referred to Reid [24] and Lee and Kesler [17].
2 .1 4 F LU ID FLO W HANDBOOK
T rl P r—» 0.010 0.10 0.40 0.80 1.20 1.50 2.00 5.00 10.0
0.30 0.0029 0.0290 0.1158 0.2315 0.3470 0.4335 0.5775 1.4366 2.8507
0.55 0.9804 0.0195 0.0778 0.1553 0.2323 0.2899 0.3853 0.9475 1.8520
0.80 0.9935 0.9319 0.0661 0.1307 0.1942 0.2411 0.3182 0.7598 1.4456
0.97 0.9963 0.9625 0.8338 0.5580 0.2055 0.2474 0.3164 0.7052 1.2968
1.02 0.9969 0.9679 0.8610 0.6710 0.2629 0.2715 0.3297 0.6980 1.2650
1.30 0.9985 0.9852 0.9396 0.8764 0.8111 0.7624 0.6908 0.7358 1.1580
1.80 0.9995 0.9955 0.9823 0.9659 0.951 1 0.9413 0.9275 0.9297 1.1391
2.00 0.9997 0.9972 0.9892 0.9796 0.9715 0.9664 0.9599 0.9772 1.1516
2.20 0.9998 0.9983 0.9937 0.9886 0.9847 0.9826 0.9806 1.0094 1.1635
2.40 0.9999 0.9991 0.9969 0.9948 0.9936 0.9935 0.9945 1.0313 1.1728
2.80 1.0000 1.0001 1.0007 1.0021 1.0042 1.0063 1.0106 1.0565 1.1830
3.00 1.0000 1.0004 1.0018 1.0043 1.0074 1.0101 1.0153 1.0635 1.1848
3.50 1.0001 1.0008 1.0035 1.0075 1.0120 1.0156 1.0221 1.0723 1.1834
4.00 1.0001 1.0010 1.0043 1.0090 1.0140 1.0179 1.0249 1.0747 1.1773
T rl
P r—> 0.010 0.10 0.40 0.80 1.20 1.50 2.00 5.00 10.0
0.30 -.0008 - 0081 -.0323 - .0645 .0966 -.1207 -.1608 -.3996 -.7915
0.55 -.0 314 - 0086 -.0343 .0682 - .1015 -.1263 -.1669 -.3991 -.7521
0.80 -.0044 0487 -.0272 .0526 - .0767 -.0940 -.12 17 -.2682 -.4740
0.97 -.0010 - 0101 -.0450 .1647 - .0669 -.0759 -.0921 -.18 3 7 -.3163
1.02 -.0005 0051 -.0198 .0303 .0227 -.0524 -.0722 -.1 1 5 6 -.2731
1.30 .0006 0061 .0267 .0612 .1048 .1420 .1991 .0875 -.0423
1.80 .0008 .0081 .0325 .0652 .0978 .1216 .1593 .2846 .2576
2.00 .0008 .0078 .0310 .0617 .0916 .1133 .1476 .2819 .3096
2.20 .0007 .0074 .0293 .0579 .0857 .1057 .1374 .2720 .3355
2.40 .0007 .0070 .0276 .0544 .0803 .0989 .1285 .2602 .3459
2.80 .0006 .0062 .0245 .0483 .0711 .0876 .1138 .2372 .3443
3.00 .0006 .0059 .0232 .0456 .0672 .0828 .1076 .2268 .3385
3.50 .0005 .0052 .0204 .0401 .0591 .0728 .0946 .2042 .3194
4.00 .0005 .0046 .0182 .0357 .0527 .0651 .0849 .1857 .2994
FLU ID PRO PERTIES 2 .1 5
Solution
The critical data and acentric factor may be obtained from critical data
tables such as Table 2.2.
Tc= -2 2 1 .2 6 ° F ( - 140.7° C)
Pc = 547.3724 psia (3774.02 kPa)
w (acentric factor) = 0.0
M\v = 28.95
Tr = (210 + 4 5 9 .6 9 )/( -2 2 1 .2 6 + 459/69) = 2.808
Pr = (500 / 547.3724 ) = 0.913
Values for Z (()) and Z (l) are read from Tables 2.3 and 2.4 as
Z(0) = 1.0025
Z(,) = 0.053
Z = 1.0025 + 0.053 X (0 .0 )= 1.0025
Air density = 500 X 28.95 / ((210 + 459.69) X 10.73 X 1.0025)
= 2.009 lb/ft3 (32.186 kg/m 3)
A2 , A^ A*
Z= 1 + M, + — + + + “S f" ) P r + + 4 r + ~ 2 ~ ) Pr
Tr T? Tr4 7/
A~! An \ s ................... 2 \ ( P r2
-* 9 l- y + Pr + A . o d + W ) ( y j ExP < - A n P r )
(2.18)
3595
4 0
2 .1 6 FLUID FLO W HANDBOOK
where
A\ = 0.3265 A 2 = -1 .0 7 0 0 A t, = -0 .5 3 3 9 A 4 - 0.01569
^ 5 ~ -0 .0 5 1 6 ^6 = -0 .5 4 7 5 A-, = -0.7361 Ax = 0.1844
Ay = 0.1056 ^10 - 0.6134 A n = 0.721
and
The comparison performed by Takas showed that Equation 2.18 has an aver
age error of less than 0.6%. A simple trial-and error algorithm should be used to
solve Equation 2.18 since the equation is implicit in terms of Z. One should guess
a value for Z, solve for the reduced density, Equation 2.19, and then solve Equa
tion 2.18 for Z again. The process should be repeated until thechange in Z is less
than a set tolerance value. The following example should illustrate the algorithm.
EXAMPLE 2.7
Estimate the compressibility factor and gas density o f methane at 670° F
(354.4° C) and 600 psia (4136.88 kPa) using the Dranchuk-Kassem method.
Solution
Tr = ( T / T c)= 1.83
Pr = (P / Pc) = 0.9
EXAMPLE 2.8
Repeat Example 2.7 using the P apay’s correlation.
Solution
Tr = 1.83
Pr = 0.9
Applying Equation 2.20 with the above reduced temperature and pressure
gives
Z = 0.956
p = PMw / (TRZ) = 600 X 16.043 / ((670 4- 459.69) X 10.73 X 0.956)
= 0.83 lb/ft3 (13.3 kg/m3)
( 2 .21)
u= (2.25)
P.
2 .1 8 FLU ID FLO W HANDBOOK
Solution
T r= 1.83
Pr = 0.9
Z ' = 0.9208(From Equation 2.24)
P ' = 3.77 (From Equation 2.22)
u = 0.238 (From Equation 2.25)
N = 0.835 (From Equation 2.23)
Z = 0.965
p = PMw (TRZ)=600
/ X 16.043 / ((670 + 459.69) X 10.73 X 0.965)
P= RT <*(T)
(2.26)
v - b v(v + b) = b{v - b)
FLU ID PROPERTIES 2.19
Term s a and b are dependent on the critical data and the acentric factor. Equa
tion 2.26 may he rearranged as
, (, R T \ 2 (a RT R T l2 ab\ _ „ ~
v' + [ b -------- + -------------3b --------- 2b I v + \ b' H-----------------------------b ------ 0.0
\ P \P P \ P P
(2.27)
Pv
Z = ---- (2.28)
RT
where
M anipulation o f Equations 2.27 and 2.28 results in the following cubical form
of compressibility factor, which may be solved iteratively for the gas and liquid
Z factors
where
N
'N
a =X " X y i yA j (2.30)
./
N
B - Y Y/Bi (2.31)
b P ( P \ ( PI P)
Bt ^ T 0 7 7 8 0 = 0 0 7 7 8 0 a 3 3 )
( / 0.45721^ ( 2 J 4 )
Pc
2 .2 0 FLU ID FLO W HANDBOOK
M , = 0 .3 7 4 6 4 + 1.5 4 2 2 6 w, - 0 .2 6 9 9 2 w ,2 ( 2 .3 7 )
a = { a ca ) = XX Y ’ YJ ( a <a h j ( 2 .3 8 )
a, ~ ( a ca ) i = ( a c)i X a i ( 2 .4 0 )
A jj = ( 1- k tj) ( A A / '5 ( 2 .4 1 )
Although the P-R solution equations 2.29 to 2.41 appear complex, once pro
gram med, they will prove to be very powerful in predicting the liquid and gas
densities and other thermodynamic calculations such as flash and enthalpy cal
culations. In the following example, step-by-step calculations of liquid and gas
densities by P-R are shown.
S olution
1. Find the pure com ponent critical data, m olecular weight, and acentric factor data.
R eferences 24 and 31 are good source for pure com ponent data.
2. V apor and liquid phase com position must be known prior to using the P-R equation
for density calculation. Since the com position given is the m ixture com position, a
tem perature-pressure (T-P) flash should be perform ed to find the vapor and liquid
content and com position. The flash calculation algorithm will be explained in the
next chapter. To find the vapor and m ixture com position, a T -P flash calculation
was perform ed using PPDS2™ . The follow ing results were obtained:
The gas to feed mole ratio was calculated as .997432, which means liquid phase
exists in trace amount.
FLU ID PRO PERTIES 2.21
3. Applying Equations 2.29 to 2 .4 1 with the vapor com position found in step 2 to find
the vapor phase com pressibility factor, the following data were calculated:
Tr Pr rrij ai bi Ai Bi
A jj is then evaluated by Equation 2.41 assum ing that interaction param eters
equal 0.0
A = Y\ Yx X A u + K, X Y2 X A X2 + Y2 X Yx X A lx + Y2 X Y2 X A 22
Finally, Equation 2.29 must be solved for Z. Tw o m ethods may be applied— a sim
ple trial and error m ethod or the more advanced m ethod o f New ton-Raphson. Both
m ethods are explained below.
0.9 9.7124e —2
0.8 2.550e—2
0.75 9.660e—4
0.73 —7.00e —3
where
M ixture gas density is then calculated by Equation 2.7. The average mixture
m olecular weight should be used
The gas density calculated by the P-R equation o f state is 4.6% less than the
measured value. More accuracy could be achieved by taking into account the
interaction param eters in Equation 2.41. Liquid density calculation is sim ilar
to the gas density calculation steps except that the liquid m ole com position
should be used.
2.2 VISCOSITY
t = a — (2.44)
dY
where
t= Shear stress
dV/dy = Rate of shear strain
cK//dy
Sat. Vap cP
Sat. Liq. cP
Temp.Deg. F (C)
FIGURE 2.5 Viscosity of saturated liquid water and saturated water vapor. (Liquid viscos
ity decreases with temperature.)
2 .2 4 FLU ID FLO W HANDBOOK
'■y
velocity gradient N.S/m . The following units are also common: cP = 0.001 N.sec/m
and Piose = 100 cP = 0 .1 N.se/m2.
References listed in section 2.1.1 are also excellent sources for pure component
(vapor and liquid) viscosity. Examples of vapor and liquid pure component vis
cosity are listed in Tables 2.5 and 2.6.
A detailed review and discussion o f the pure component and mixture viscos
ity for liquids and gases is available in Reid [24]. For the purpose of providing
means for the reader to estimate liquid and gas viscosity in order to solve the fluid
flow problems in this book, few simple methods that require only the critical data
and molecular weight are selected.
Golubev [11] suggested the following correlation to estimate the vapor viscosity
from the component critical data
Degree F
(Degree C) Air Hydrogen Ethane
* (T a b le v alues are in c P units. T o convert to N .s/m 2, m ultiply table value by 1,000. T o c onvert to
lbm /ft/hr, m ultiply by 2.419)
FLU ID PRO PERTIES 2 .2 5
Degree F
( Degree C) Water Benzene n-Decane Sulfuric Acid
♦(T able v alues are in cP units. T o convert to N .s/m 2. m ultiply table value by 1,000. T o convert to
Ibm /ft/hr m ultiply by 2.419)
where
V? = (2-47)
*C
Mw = M olecular weight.
Pc = Critical pressure, atm.
Tc = Critical temperature, K.
p - Viscosity, micro P.
ft* = Viscosity at the critical temperature
EXAMPLE 2.11
Estimate the vapor viscosity o f n-Octane at 100° F (37.8° C) using the Golubev
equation. Compare to the reported value o f 58.2 fiP.
Solution
Tc = 1023.84R (568.8 K)
Pc = 360.15 psia (24.5 atm)
Mw = 114.232
Tr = 559.75/1023.84 = 0.5467
fi = 61.21 fiP = 6 .12e —3 cP, which is about 5.17% higher than the
reported value.
2.2 6 FLU ID FLO W HANDBOOK
The following relation, which is an extension to the Cham an-Enskog kinetic the
ory, may be applied to estimate the viscosity of vapor mixtures at low pressures
n
(2.48)
j= i
where
(2.49)
Equation 2.49 114] is an approximate estimation for the </>,-, parameters. Other,
more complex expressions have been suggested to estimate this parameter as well
[24].
EXAMPLE 2.12
Estimate the vapor viscosity o f a mixture o f methane (80% mole) and ethane at
130° F (54.4° C), 19 psia (131 kPa) and compare it with the value calculated
by PPDS2™ o f 0.0117 cP.
Solution
Pure Component
Mw Mole Fraction Vapor Viscosity, cP
_ yM yifk>
r~m . .
yi + y 2 <Pi2 y i + y\<t>2 \
= 0.01165 cP (0.36% lower than reported mixture value, however, higher than both of the pure
component’ s viscosity).
FLU ID PR O PERTIES 2.27
Liquid viscosity for mixtures may be estimated by the following simple mixing rule
Equation 2.50 should be used with care when system exhibits a maximum or
a minimum in mixture viscosity.
EXAMPLE 2.13
Estimate the liquid viscosity o f a benzene-toluene mixture (50% mole benzene)
at 89° F (31.6° C) and 14,7 psia (101.35 kPa). Compare with a reported value
o f 0.5333 cP.
Solution
Pure Component
Mole Fraction Liquid Viscosity, cP
Degree F
(Degree C) Water Benzene n-Decane Sulfuric Acid
where
S olution
Pi
lb/ft3 (kg/m3) N/m
REFERENCES
1. A m brose, D. Septem ber 1978. “C orrelation and Estim ation o f Vapour-Liquid Criti
cal Properties. 1. C ritical Tem perature o f O rganic C om pounds.” National Physical
Laboratory, Teddington, U.K. N P L Report. Chem . 92.
2. A m brose, D. Septem ber 1979. “C orrelation and Estim ation o f Vapour-Liquid Criti
cal Properties. I. Critical Pressure and V olum es o f Organic Com pounds.” National
Physical Laboratory, T eddington, U.K. N P L Report. Chem . 98.
FLU ID PRO PERTIES 2 .2 9
3. Am erican Petroleum Institute. 1992. Technical Data Book-Petroleum Refining, 5th ed.
4. Burnett, R. R. June I 1, 1979. “C alculator Give Com pressibility Factors.” The Oil
an d G as Journal. 70-72.
5. C oker, A. K. D ecem ber 11, 1989. The Oil and Gas Journal. 63.
6. C om pressed G as A ssociation, Inc. 1981. H andbook o f C om pressed G ases, 3rd ed.
Van N ostrand Reinhold, New York.
7. D ranchuk, P. M. and J. H. A bou-K assem . 1975. “Calculations of z-Factor of N at
ural G ases U sing Equations o f State.” Journal o f Canadian Petroleum Technology.
14: 3 4-36.
8. Prauntiz, J. M ., and P. L. Chueh. 1968. C om puter C alculations fo r H igh-Pressure
V apor-Liquid Equilibria. Prentice-H all, Inc., Englew ood Cliffs, New Jersey.
9. Gas Processors Suppliers A ssociation. 1994. Engineering Data Book, 10th ed. V 11.
10. G lanville, J. W ., B. H. Sage, and W. N. Lacey. 1950. Industrial and Engineering
C hem istry. 42: 508-513.
1 1. G olubev, I. F. 1959. Viscosity o f G ases and Gas M ixtures: A Handbook. Natl. Tech.
Inf. Serv. T T 70 50022.
12. H ankinson, R. W ., L. K. Thom as, and K. A. Phillips. April 1969. “Predict Natural
G as Properties.” H ydrocarbon Processing. 106-108.
13. H enley, E., and J. D. Seader. 1981. Equilibrium -Stage Separation O perations in
C hem ical Engineering. John W iley & Sons, Inc., New York.
14. H erning, F., and L. Zipperer. 1936. Gas W asserfach. 79: 49.
15. H ougen, O. A., K. M. W atson, and R. A. Ragatz. 1959. Chem ical Process P rinci
ples, Part II, Therm odynam ics, 2nd ed. New York: John, W iley & Sons, Inc.
16. Kudchadker, A. P., G. H. Alani, and B. J. Zwolinski. 1968. Chemical Reviews. 68: 659.
17. Lee, B. I., and M. G. Kesler. 1975. A IC H E Journal. 2 1 :5 1 0 .
18. N elson, L. C ., and E. B. Obert. 1954. Trans. ASM E. 76: 1057.
19. Papay, I. 1968. O G IL Musz. Tud. K ozl. Budapest. 267-73.
20. Peng, D. Y., and D. B. Robinson. 1976. “A New Tw o-C onstant Equation of State.”
Ind. Eng. Chem. Fundam . 15: 59-64.
21. Perry, R. H., D. W. Green, and J. O. M aloney. 1984. P erry's Chem ical Engineer's
H andbook, 6th ed. M cG raw -H ill, New York.
22. Rackett, H. G. 1970. Journal o f C hem ical Engineering Data. 15: 514.
23. Redlich, O., and J. N. S. Kwong. 1949. Chem ical Reviews. 44: 233-244.
24. Reid, R. C., J. M. Prausnitz, and B. E. Poling. 1987. The Properties o f Gases and
Liquids, 4th ed. M cG raw -H ill, New York.
25. Soave, G. 1972. C hem ical E ngineering Science. 27: 1197-1203.
26. Spencer, C. F. and R. P. Danner. 1972. Journal o f Chem ical Engineering Data. 17:
236.
27. Standing, M. B., and D. E. Katz. 1942. “ Density o f Natural G ases.” Trans. AIM E.
146: 140-49.
28. Takacs, G. 1989. “Com paring M ethods for C alculating Z-Factor.” The Oil and Gas
Journal. 43.
29. van der W aals, D. J. 1873. D octoral D issertation, Lieden.
30. Y am ada, T. and R. D. Gunn. 1973. Journal o f C hem ical Engineering Data. 18: 234.
31. Y arborough, L. February 18, 1974. “H ow to Solve Equation o f State for Z-Factors.”
The O il and G as Journal. 86-88.
32. Y aw s, C. L. 1999. C hem ical P roperties H andbook. M cG raw -H ill, New York.
CHAPTER 3
FLUID FLOW AND
THERMODYNAMICS
Internal energy, u, represents the energy stored in a unit mass of the fluid due to
the state of molecular activity. The higher the fluid internal energy, the more
activity the molecules possess. Units for internal energy are expressed in Btu/lb,
cal/g, or kJ/kg. Internal energy may be expressed for the entire mass of the fluid, U;
in this case, units should be Btu, cal, or kJ. There is no direct way to calculate or
measure the internal energy; however, for a stationary mass the internal energy
3.1
3.2 FLU ID FLO W HANDBOOK
is equal to the difference between the net amount of heat added to the system and
the net amount of work done by the system on the surroundings,
A U = A Q -A W (3.1)
du = dq - dw (3.2)
where
Pressure energy represents the work energy exerted on the fluid as it moves from
one point to another, P/p. It is known as the flow work. Pressure energy may be
expressed inmechanical units such as ft-lbf/lbm (N-m/kg), which may be con
verted to thermal units, P/Jp, with the use of the mechanical equivalent of heat, J
(J = 778.16 ft-lbf/Btu).
3.1.3 Enthalpy
Enthalpy is a term used to indicate the total o f the internal energy and the pres
sure energy.
h = u + — (3.3)
Jp
u = q - w (3.4)
h = q - w + Pv ( 3 .5 )
where
The conversion term J is dropped from the above equation, assuming that all
terms have the same thermal units.
Equation 3.3 may be expressed in a differential form as,
dh = dq - dw + d(P v) (3.6)
where
dw = pdv (3.8)
From Equation 3.7, assuming constant pressure system, the change in ent
halpy is equal to the energy added to the system as,
dh = dq (3.10)
Energy added to the system may be found by the following relation when the
initial and final temperature o f the fluid state are known:
dq = Cp dT (3.11)
where
From Equations 3.10 and 3.11, the change in fluid enthalpy at constant pres
sure is found by
dh = Cp dT (3.12)
Fluid specific heat capacity is a measure for the amount of heat required to
raise a unit mass o f the fluid by one temperature degree. The heat capacity con
stant refers to the ideal heat capacity at constant low pressure.
Internal energy may be expressed in an equation similar to Equation 3.12 by
following the same derivation steps,
du = dq - dw (3.13)
3 .4 FLUID FLO W HANDBOOK
Pressure work, dw, is equal to pdv. At constant volume, the pressure work
term drops and the internal energy is expressed as
where
The ratio o f specific heat at constant pressure to the specific heat at constant
volume, k = Cp/Cv (specific heat ratio), plays an important role in the calculation
of sonic velocity, and in isentropic processes. Table 3.3 lists specific heat ratios
for common industrial gases. For a perfect gas, with the assumption that enthalpy
is a function o f tem perature only, the following relation may be applied
Cp = Cv + R
where
AH = / ( T ,P ) = A T + & P (3.15)
dP dT
where
,dH_
(— )P = CPd T (3.16)
dT
At low pressure the change of fluid enthalpy, ideal enthalpy, may be esti
mated by Equation 3.16. As the pressure increases, a deviation term must be
included. The deviation term accounts for the change in fluid enthalpy due to
change in pressure at constant temperature. The deviation term is known as the
FLU ID FLO W AND TH ERM O D YNAM ICS 3.5
T A B L E 3.1 Vapor Specific Heat Capacity Coefficients for Common Industrial Chemicals.
C p = U| + a2 T + a? T “ + a4 T 3 + a5 T4, T in °F, Cp in Btu/lb-mole-°F. To convert C p to
metric units (kJ/kmole-°C) multiply by 4.1868
Component at ai a3 a4 a5
Carbon dioxide 8.398 -0.6475e-2 -0.3555e-5 0.1194e-8 -0.185 le-12
Hydrogen 6.647 0.2472e-2 -0.4557e-5 0.3117e-8 -0.6643e-12
Hydrogen Sulfide 8.031 0.9868e-3 0.2388e-5 -0.1593e-8 0.3203e-12
Water 7.985 0.4633e-3 0.1402e-5 -0.6578e-9 0.97952e-13
Ammonia 8.276 0.39006e-2 0.3524e-6 -.27402e-9 0.0
enthalpy departure function. The change in ideal fluid enthalpy over a given
range o f temperature is estimated by the following integral equation:
r
h ; = J c ;v dT (3 .17)
TO
where
Equation 3.17 is used to calculate the change in the ideal vapor enthalpy.
However, since the vapor enthalpy at the reference tem perature, 70, is assumed
to be 0.0, Equation 3.17 may be used to estimate the vapor enthalpy at a tem per
ature equal to T.
Table 3.1 lists the coefficients for the vapor specific heat capacity o f selected
gaseous components as a function of temperature. An extensive coverage of the
3.6 FLUID FLO W HANDBOOK
where
For a vapor mixture, Equation 3.18 is modified with the following mixing rule,
where
i Component index
n Total number of gaseous components in mixture
j Polynomial coefficient index
Yj W eight fraction of component i in mixture
Units for specific heat capacity data tabulation may be expressed per unit
mass or unit mole of the fluid; K, should match the specific heat capacity units.
Mole %
M ethane 79.1
Ethane 6.8
Propane 14.1
S olution
The vapor-phase specific-heat capacity coefficients for the above pure compo
nents may be obtained from Table 3.1. Equation 3.19 may be used to calculate
the mixture ideal vapor enthalpy. The reference temperature is taken as 77.0°F
(25.0°C).
FLU ID FLO W AND TH ERM O D YN A M IC S 3.7
(TJ - T( V) { V -T0>)
r , l “,
j j =l j
The mixture ideal vapor enthalpy is very comparable with most available com
mercial software.
Similar approach may be applied to find the liquid-phase ideal enthalpy.
Table 3.2 shows the liquid-phase specific heat capacity coefficients for selected
components.
3.1.3.2 Enthalpy Departure Function. Fluid Enthalpy, vapor and liquid, is a func
tion o f pressure. The ideal fluid enthalpy, which is discussed above, must be cor
rected for high pressure. A correction term to the ideal fluid enthalpy will yield the
real enthalpy at the fluid temperature and pressure. Fluid real enthalpy is described as
Hv = + ( H v - H° v ) (3.20)
h l = X x ,w ;;. + (.h l - h ;) 0 .2 1 )
where
A // „ A D , Z + 2.4145
----- — Z —1 + — j=—(1 H— ) ln(---------------------------------)(3.22)
RT 2a/2B a Z - 0.414/?
where
AH Departure function.
Z Liquid or vapor-phase compressibility factor.
T Fluid temperature.
R Universal gas constant, units of R and T must match.
A, B y a See Equations (2.26-2.41)
D - II W M o (3.23)
Chapter 2 has a complete definition for the terms used in the above equation.
W hen applying Equations 3.22 and 3.23, the vapor mole composition and com
pressibility factor should be used to find the vapor-phase enthalpy deviation and
similarly for the liquid-phase.
Solution
The mixture data and compressibility results may be found in Example 2.9. The
following data will be used to estimate the vapor enthalpy:
AH A _. __ 0.266 36107.11 ,
-----= 0.7477 - 1 + — j=------------ ( 1 + -------------- ) + In (------------------------------ )
RT 2s[2 x 0.046 53672.36 0.7477 - 0.414x0.046
AH
= -0 .8 1 1
RT
R = 1.987 Btu/lbmole-R
A moving mass o f fluid possesses kinetic energy equals to V 2/2gc where V is the
fluid velocity. The fluid kinetic energy changes as the fluid velocity decreases or
increases. Thus, for an incompressible fluid flowing in a constant area conduit,
the fluid kinetic energy remains constant.
3 .1 0 FL U ID FLO W HANDBOOK
Potential energy represents the energy possessed by the fluid due to its position
to a reference elevation. The potential energy of a fluid is given by
PotentialEnergy = — Z (3.24)
gc
where
The fluid total energy is the sum of the kinetic energy, potential energy, pressure
energy, and internal energy,
P V2 g
Fluid total energy = u + — + ----- + — Z (3.25)
P 2 gc gc
Fluid total energy remains constant as long as the fluid is not losing or gain
ing energy. Examples of energy gain are heat transfer from the surroundings to
the fluid, and the addition of a pumping head to the fluid. Other than work done
by the fluid or a heat transfer to the surroundings, the fluid total energy remains
constant. However, the individual energy terms may change. A reduction in the
fluid velocity at constant elevation and constant temperature will decrease the
kinetic energy; however, since total energy is constant, there will be a gain in the
pressure energy. The following example illustrates.
S olution
±
+ -----
^ +
^ * Z|
7 — 112 +
-u P*
— +
+ V22
----- +
4. * 7Z-,
P 2 gt. gcp 2g, gc
(3.26)
The fluid total energy is constant, i.e. fluid total energy at point l equals the fluid
total energy at point 2.
Since there is no change in elevation and no change in internal energy. Equa
tion 3.26 is simplified to
P vx2 p; v2
-j- + = — + -2 - (3.27)
P Sc i P 2 gc
p, = Pl + {V' - V^
2gr l44
where
A gain of 2.676 psig (18.45 kPag) in the pressure energy will occur due to the
reduction in kinetic energy.
Similarly, Equation 3.26 may be applied to fluids at static. The derivations and
examples shown in Chapter 4 are direct derivation of Equation 3.26. The inlet and
outlet velocity terms are dropped since the fluid is at static. At constant temperature,
the internal energy is constant. For fluid statics application Equation 3.26 is reduced
to the following form:
^ + -£-Z, = + — Z, (3.28)
P 8C P Sc
P, v x = P2 v2 (3.29)
where
P v = n (3.30)
where
n Constant
Solution
From the steam tables [15]
At P = 114.7 psia (790.83 kpaa) and T = 400°F (204.4°C),
v = 77.014 ftVlbmol (4.802 m Vkmole)
Application o f equation 3.29 yields:
V2 = P, v, /P 2
V2 = 114.7 X 77.014 / 94.7 = 93.27 ft3/lbm ole (5.81 nrVkmole)
The steam table gives a value of 94.04 ftVlbmole (5.864 m3/kmole) for the
specific volume at pressure o f 80.0 psig (551.58 kpag). The error in using
the ideal gas law is -0.81 percent.
dq = 0
For simplicity, fluid enthalpy may be assumed to be constant under this assump
tion (In reality, enthalpy will increase since heat of friction will be added to the
fluid). Recalling that the fluid enthalpy is the summation of the internal energy
and the pressure energy, any change in the pressure energy will cause a change
in the fluid internal energy, hence a change in fluid temperature. This is known
as the Joule-Thomson effect o f expandable gases. The following relation may be
used to find the temperature change due to a reversible adiabatic pressure change.
The relation is used for perfect gases.
II 2 \(k-l)/k
(3.31)
T
where
Absolute temperature and pressure values must be used in the above equation.
EXAMPLE 3.5
Find the exit temperature o f air flow ing in pipe. The pipe is well insulated; adi
abatic flo w may be assumed. The total pressure drop is 100 psig (689.48 kpag).
3 .1 4 F L U ID FLO W HANDBOOK
S olution
The specific heat ratio for air is l .4 (Table 3.3)
The air exit temperature is estimated from Equation 3 .3 1 as
The above calculation is true for perfect gases. Large deviation may be
obtained when compared with the rigorous flash calculation. The way to solve the
above example rigorously is by performing a pressure-enthalpy (PH) flash calcu
lation using an equation of state such as Peng-Robinson. Since the inlet condi
tions are given, a pressure-temperature is performed to find the enthalpy at the
inlet conditions. The exit enthalpy is set equal to the inlet enthalpy (adiabatic
process, ignoring the heat of friction). The exit temperature o f the fluid is found
by a PH flash. Example 3.5 is repeated for illustration.
Components K = C,,/Cv
Water 1.32
Air 1.4
Oxygen 1.4
Carbon dioxide 1.3
Sulfur dioxide 1.3
Helium 1.66
Argon 1.66
Nitrogen 1.4
Carbon Dioxide 1.4
Chlorine 1.36
Hydrogen sulfide 1.32
Methane 1.31
Ethane 1.2
Propane 1.1
n-Butane 1.1
Ethylene 1.26
FLU ID FLO W AND TH ERM O D YN A M ICS 3 .1 5
Solution
W hen heat crosses the boundary o f the flowing fluid, the process is called non-
adiabatic. The analysis shown in Example 3.6 (PH flash calculation) is applica
ble to a non-adiabatic process. The heat transferred per unit mole of the fluid must
be added to (or subtracted from) the fluid inlet enthalpy before performing a PH
flash. A positive heat sign indicates an energy added to the fluid, while a nega
tive sign indicates heat lost from the fluid to the surroundings
S olution
Processes that occur at a high speed such as a throttling process undergo an isenthal
pic process, constant enthalpy process. Isenthalpic process must not be confused
3 .1 6 FLU ID FLO W HANDBOOK
with adiabatic process; in the latter, enthalpy may increase due to the heat ot fric
tion, while in the former, the friction force does not add heat to the fluid and no
heat crosses the flow boundary. Flow of gas through nozzles and constriction are
exam ples of constant enthalpy processes.
The sound velocity in a fluid medium is known as the sonic velocity. Sonic veloc
ity is the rate o f propagation of a pressure pulse of infinitesimal strength through
a fluid or a solid. The sound speed is defined as
a (3.32)
where
a (3.33)
where
a (3.34)
FLU ID FLO W AND TH ERM O D YNAM ICS 3.17
Substituting the gas density, which is defined in equation 2.7, into the above
equation yields,
where
where
where
EXAMPLE 3.8
Calculate the speed o f sound in air at I50°F (65.55°C).
Solution
Using Equation 3.36, the speed of sound in air at 150°F is
Sonic velocity in liquids and solids is calculated by Equation 3.33, which may
be rewritten as
/ \,/2
144 K g /
a = (3.38)
where
dP .
K = p -f- (3.39)
dp
Values for the bulk modulus of elasticity for selected liquids are given in
Table 3.5.
EXAMPLE 3.9
Compare the speed o f sound in water and sea water at 77°F (25°C).
Solution
Component K, psi K, M pa
From Table 3.5, the bulk modulus of elasticity of water is 3.25e+5 psi
(2240.8 M Pa) and for sea water is 3.51e+5 psi.(32420.0 M Pa)
For sea water.
Mach number is defined as the ratio of the fluid velocity to the velocity of sound
at the local temperature and pressure conditions,
Ma = — (3.40)
a
where
Substituting the equation of the speed of sound, Equation 3.35, into the above
equations yields the following gas-phase mach number relation:
Ma = 7 V -- (3.41)
y]gck Z R T / Mw
Mach number is a local quantity that will change as a function of the local
temperature and pressure. Mach number changes inversely, at constant flow area,
with the square root of the fluid density. Higher pressure reduces the fluid veloc
ity and increases the fluid density thus reducing the mach number. Higher tem
perature increases the fluid mach number. Increasing the flow cross sectional area
will decrease fluid velocity, hence, the mach number will be decreased. Mach
number is a measure of the fluid incompressibility effects. High mach numbers,
3 .2 0 FLU ID FLO W HANDBOOK
Ma > 0.3, is an indication of the How compressibility effects. Mach number is « 0.3
for liquid-phase flow and compressibility effects may be ignored. However, com
pressibility effects, and hence mach number, may be very large in gas-phase flow.
Sonic velocity in liquids is in the order of 4500-5000 ft/sec (1370-1500 m/sec),
much higher than liquid velocities in the industrial practice. Water flow will reach
a mach number of 0.3 when the fluid velocity equals 0.3 of the sonic velocity on
water (4920 ft/sec at 77°F (1499.6 m/sec at 25°C)). A water velocity in this range,
1446 ft/sec (440.7 m/sec), requires a mass flow of 112. le +6 Ibm/hr (50.91 kg/hr)
in a 6 inch schedule 40 pipe. This flow is beyond any industrial practice and the
assum ption of the absence of compressibility effects on liquid flow is valid.
Unlike liquid-flows, gas-flow may easily reach sonic conditions. The following
exam ple illustrates the mach num ber calculation for gas-phase flow.
Estim ate the mach number at the inlet o f the pipe fo r the following air Jlow
conditions:
Air mass flow rate is 60000 lb/hr (27243.67 kg/hr), pipe I.D. is 7.891 in
(200.43 mm)
S olution
Pipe cross sectional area = (6.065/12/2)2 X 3.14 = 0.196 ft 2 (0.0182 m2)
Air molecular weight = 29.0
a. Air density at 325°F (162.77°C) and 55 psig (379.2 kpag) = 0.2391 lbm/ft3
(3.83 kg/m3) (Z = 0.99974 using P-R equation o f state)
Air velocity = 60000 Ibm/hr X (hr/3600 sec) / (0.2391 lbm/ft3) / 0.196 ft 2
= 355.65 ft /sec (108.4 m/sec)
Mach number is calculated using Equation 3.41
Ma = -------------------------- 3 5 5 65 - . = = 0.26
V 32.174x1.4x0.9997x1545.4x(325 + 460) / 29
Pressure drop due to acceleration per unit foot of pipe length at these con
ditions = 9.1e-3 psig/ft (0.205 kpag/m) which is about 8.9% of the total
pressure drop.
Air velocity = 60000 lbm/hr X (hr/3600 sec) / (0 .1365 lbm/ft3) / 0.196 ft2
= 608.7 ft /sec (185.5 m/sec)
Mach number is calculated using Equation 3.41
Ma = , - - 608-7- = = 0.443
^32.174x1 .4 jc0.9998 jc1545.4 jc(325 + 460) / 29
Pressure drop due to acceleration per unit foot of pipe length at these con
ditions = 0.061 psig/ft (1.379 kpag/m), which is about 26% of the total
pressure drop (frictional plus acceleration).
c. Air density at 425°F (218.33 C) and 25 psig (172.37 kpag) = 0.121 lbm/ft 3
(1.938 kg/m 3) (Z = 0.9999 using P-R equation of state)
Air velocity = 60000 lbm/hr X (hr/3600 sec) / (0.121 lbm/ft3) / 0.196 ft 2
= 686.27 ft /sec (209.2 m/sec)
Mach number is calculated using Equation 3.41
Ma = - 6 8 6 -7 - —■ = 0.471 nyrri
V 32.174x1.4jc0.9999.rl 545.4^(325 + 460) / 29
Pressure drop due to acceleration per unit foot of pipe length at this con
ditions = 0.082 psig/ft (1.855 kpag/m), which is about 30 percent of the
total pressure drop (frictional plus acceleration).
From the above example, it is clear that pressure drop due to acceleration
(compressibility effects) increases as the mach number increases. It is recom
mended to add the pressure drop due to acceleration effects to the total pressure
drop (frictional + elevation) for mach number > 0.3.
Isothermal Process
T2 = T 7I
P'L _ Ma^_
Px ~ M a2 (3.42)
p2 _ M a{
p, Ma2
3 .2 2 FLUID FLO W HANDBOOK
Adiabatic Process
t2 Xl
T, y2
P2 M at [ X
(3.43)
Pt V
M a2 '
p2_ Max [ X
P, M a2
where
S olution
Since the pressure and temperature are specified at the pipe inlet, the steam enthalpy
is found from the steam tables.
FLU ID FLOW AND THERM O D YNAM ICS 3.23
Steam enthalpy, H = X X HG + (1 - X) X HL
where
Hg Gas-phase enthalpy, Btu/lbm (kJ/kg).
Hl Liquid-phase enthalpy, Btu/lbm (kJ/kg).
X Vapor mass / Total mass
Vapor and liquid enthalpies at the inlet conditions are [ 15]
H g = 1180.2 Btu/lbm (2745.4 kJ/kg)
H l = 269.7 Btu/ lbm (627.32 kJ/kg)
Steam enthalpy = 0.05 X 910.4 + (1-0.05) X 269.7 = 315.225 Btu/lbm
(733.2 kJ/kg).
Exit pressure = 66.98 - 9.45 = 57.53 psia (396.65 kpag)
Since the pipe is perfectly insulated, there is no change in fluid enthalpy (assum
ing no frictional heat is added to the fluid), the exit enthalpy is 315.225 Btu/lbm
(733.2 kJ/kg).
From the steam tables, at P = 57.53 psia (396.65 kpag), the vapor and liquid
enthalpies are
H g = 1177.2 Btu/lbm (2738.16 kJ/kg)
H l = 259.4 Btu/lbm (603.36 kJ/kg)
Since the exit fluid enthalpy, 315.225 Btu/lbm (733.2 kJ/kg), is less than the
saturated steam vapor enthalpy and greater than the saturated steam liquid enthalpy,
the exit condition is a saturated steam at the saturated temperature of 290.0 F
(143.33).
The mass fraction o f vapor is found by solving the following equation for X,
H = X X HG + (1-X) X H l
315.225 = X X 1177.2 + (1-X) X 259.4
X = 0.0608 (M ass fraction of vapor phase)
The im portance of thermodynamic calculations is evident in the above exam
ple, the therm odynam ics calculations, not hydrodynamics, enables one to find the
exit temperature and gas-liquid phase ratio.
Non-adiabatic conditions may also be handled by adding (or subtracting) the
heat from the fluid enthalpy, and then find the exit temperature and vapor to liq
uid ratio at the exit pressure and new fluid enthalpy. For example, assuming there
is a net heat transfer (gain) of 25 Btu/lbm (58.15 kJ/kg) of steam, this will make
the exit steam enthalpy equal to 315.225 + 25.0 = 340.225 Btu/lbm (791.4 kJ/kg).
The exit tem perature remains as the one calculated above, however, the vapor
quality changes,
H = X X HG + (1-X) X HL
340.225 = X X 1177.2 + (1-X) X 259.4
X = 0.08806
3 .2 4 FLU ID FLO W HANDBOOK
(3.44)
P
where
Table 3.6 lists the Antonie vapor pressure coefficients for selected com po
nents. Figure 3.2 shows the solution algorithm to the following set of PT flash
equations:
0.0 (3.45)
» N
/v
ffv) = I (3.46)
U + i / / ( K . - l )]2
y/ new (3.47)
FLU ID FLO W AND TH ER M O D YN A M ICS 3 .25
z
X. =
i + y/(Ki - (3.48)
Y = X. K:
X.
X. = — —
l
1 IX (3.49)
Y
Y =
ly.
where
EXAMPLE 3.12
Find the vapor-phase mole ratio and the vapor and liquid phase composition o f
a mixture consisting o f a 50 percent mole benzene-toluene at 205°F (96.11°C)
and 14.7 psia (101.353 kpaa)
>
T + A,
T in Degree F
P- Vapor pressure, psia
P,c Critical Pressure, psia
To convert psia to k pa multiply by 6.8948
Component Pc Ai a2 A3
S olution
At 205°F (9 6 .1°C), the ideal values for the equilibrium constants using equation
3.44 and the Antonie coefficients in Table 3.6:
For sim plicity, vapor and liquid phases are assumed to be ideal, the Ki values
are independent of the vapor and liquid composition.
The solution is iterative and requires an initial estim ate of the vapor and liq
uid mole fractions such as
Xi Yi
An initial estim ate o f the vapor-phase mole ratio is also required to start the
algorithm, let ijj= 0.5, the follow ing iterative steps are calculated:
FLUID FLOW AND THERMODYNAMICS 3 .27
Calculation is terminated after 3 iterations since the change in i//(k) is less than
le- 6 , the vapor and liquid composition may be calculated using Equations 3.48.
REFERENCES
Although the subject of this handbook is fluid flow, many of the hydrodynamic
concepts such as fluid pressure, pressure measuring concepts and devices, forces
acting on submerged bodies, and buoyant forces may be illustrated with a fluid at
static. A flowing fluid starts with a fluid at static, pumping fluids from tanks (fluid
at static) is a simple example. The pump suction head is a function of the fluid
hydrostatic pressure at the pump suction. Fluid statics provide means to measure
pressure or pressure difference of flowing or static fluids. Barometer and manome
ter devices are examples where a static fluid is used to measure the pressure or
pressure difference. In this chapter, the following concepts will be illustrated:
• Fluid pressure
• Pressure measuring concepts and devices
• Forces acting on submerged bodies
• Buoyant forces
4.1
4.2 FLUID FLO W HANDBOOK
where
Fluid pressure has units of force (weight) per area such as lbf/ft 2 (psf), lbf/in 2
(psi), and N/m (pa). Equation 4.1 is applicable when the acting force is uniform
and constant over the entire area. A differential form of Equation 4.1must be used
when the acting force is not uniform, such as the case for fluid weight acting on
a submerged inclined slab (see Figure 4.1),
cIF
dP = — (4.2)
dA
For fluids, the force term in Equations 4.1 and 4.2 is replaced by the fluid
weight. Fluid weight is related the mass of fluid as,
Vy _ (43)
gc
where
M = p V
where
When area in Equations 4.1 is taken as a unit area, 1.0 in 2 or 1 ft 2 (m2), the
pressure is defined as,
P =p h — (4.4)
gc
FLU ID STATICS 4.3
w here
h H e ig h t, d e p t h o f fluid, ft
Units for Equation 4.4 are lbf/ft 2 (psf). For SI unit system. Equation 4.4 is
expressed as
P =p hg (4.4a)
where
h Height in m
g G ravitational constant, 9.81 m/s 2
p Fluid density, kg/m 3
P Hydrostatic Pressure, Pa = kg/m-s
Equation 4.4 and 4.4a is a basic fundamental equation to convert between the
fluid hydrostatic pressure and the fluid pressure head. Fluid pressure head is
defined as the height of the liquid that will give the same fluid pressure. A given
fluid pressure may be converted into an equivalent pressure head (fluid height) by
rearranging Equation 4.4 as
(4.5)
(4.5a)
x z
w eight
inc rea se s
w ith depth
Solution
(The pressure calculated above is the gauge pressure; more discussion will fol
low about gauge and absolute pressures.)
A plot of fluid pressure, P, versus the fluid height (depth, positive downward)
is shown in Figure 4.3. The fluid pressure increases linearly with the depth o f the
fluid.
Solution
Fluid mass density = 55.5 lbm /ft 3 (888.999 kg/m3)
Required fluid pressure = 15.0 psig (pound force per squared inch, gauge)
15.0 psig = 15.00 x 144 in 2/ft 2 = 2160 lbf/ft2 (103.41 k Pa)
Use Equation 4.5 to the find the fluid pressure head as h = 2160 (lbf/ft2) /
55.5 (lbm /ft3) / 32.172 (ft/s2) x 32.174 (lbm-ft/lbf-s2).
h = 38.9 ft (11.86 m) of fluid is required to produce a 15.00 psig at the
pump suction.
Atmospheric pressure refers to the weight of a column of air per unit area of
the earth surface. Atmospheric pressure decreases with altitude. At sea level, the
atm ospheric pressure has an average value of:
4.6 F L U ID FLOW HANDBOOK
Gauge pressure has a minim um value of -14.70 psig (-101.35 k Pa), which
refers to the absolute vacuum. A gauge pressure of 0.0 value is equivalent to
atm ospheric pressure. Absolute pressure lowest possible value is 0.0, which
refers to the perfect vacuum. Any pressure value below atmospheric pressure is
known as vacuum pressure. Therefore, when the pressure is reported in gauge,
any value below 0.0 indicates a vacuum pressure. When the pressure is reported
in absolute terms, any value below the atmospheric pressure (14.7 psia or the
local atm ospheric pressure) indicates a vacuum pressure. Figure 4.4 illustrates the
concept of gauge and absolute pressure. A suffix (g for gauge or a for absolute)
is added the end o f the pressure unit to differentiate between the two systems.
A static fluid in tanks and tubes (that are connected) tends to level its surface in a
way such that the static pressure is uniform at similar depths. For example, in Fig
ure 4.5 points A, B, and C will have the same hydrostatic pressure since all have
/ ' / \
Gage
Absolute
pressure
Pressure
Atmospheric Pressure
0.0 psig
Vacuum . 14.7 psia
/
Pressure
Absolute (Perfect) Vacuum
-1 4 7 psig
0.0 psia
the same depth. If the external pressure over the fluid surface in one of the tubes
increases (see Figure 4.6), the fluid levels in all connected tubes will be readjusted.
However, the fact that points at similar depths have similar static pressure will hold
to be true. In Figure 4.6, points A ', B \ and C ' have the same hydrostatic pressure.
To find the value o f the external pressure, P, one may use the concept that pressure
at point C ' is the same as the pressure at point A ' since both points are at the same
horizontal level. Pressure at point A ' is found by converting the liquid static head
into pressure as explained earlier. Pressure measuring devices are direct application
of this phenomenon. Many types of pressure measuring devices exist. Barometers,
manometers, and piezometers are among the most commonly used devices.
4.2.1 Barometers
A barometer is a device made to measure the atmospheric pressure. Figure 4.7 shows
a barom eter consisting of a long hollow tube with an opening at one end only. To
remove air from the tube, the tube is filled with mercury (or any other fluid) and
then is im m ersed in a container of mercury. Liquid mercury will rise in the tube
due to the effect of atmospheric pressure on the mercury surface. The closed end
o f the tube is near perfect vacuum, except for the mercury vapor pressure which
may be neglected. The mercury rise above the container surface is equivalent to
the pressure head of the atmospheric pressure which may be converted into pres
sure using Equation 4.4. When mercury is used in a barom eter, the following for
mula may be used to convert the mercury head into pressure
P = 0 .4 9 15 x hHg (4.6)
where
P = 0.03625 x Sg x h + Pv (4.7)
where
Sg Specific gravity
Pv Vapor pressure, atm
P Pressure, atm
Vacuum
Atmospheric
Pressure.
4.2.2 Piezometers
A piezom eter is a device used to measure the pressure difference between the
atmospheric and the fluid hydrostatic pressure in a tank. It is basically a tube con
nected to the side o f a container (see Figure 4.8); the other end of the tube is open
to the atmosphere. The liquid rises in the tube due to the fluid hydrostatic pres
sure. The height o f the liquid in the piezometer is the equivalent pressure head of
the fluid hydrostatic pressure. Equation 4.4 is used to convert the liquid height
into pressure.
A manometer is a bent tube (U tube) filled with a liquid of different specific grav
ity. When the pressure at one of the tube ends is known, such as an end open to
the atmospheric pressure, the pressure at other end is calculated by converting the
manometer liquid head into pressure. When pressures at both ends are unknown,
the manometer will measure the pressure difference. In this case the manometer
is called a differential manometer. Figure 4.9 shows a manometer with one end
open to the atmosphere, while Figure 4.10 shows a differential manometer. The
deflection in height of the manometer fluid is converted into pressure using the
following equation
AP = pm AH l \ 44 (4.8)
where
P re s s u re P r e s s u re
h ea d at p o in t h ea d at p o in t
B A
u
FIGURE 4.9 A manometer with one
end open to the atmosphere.
CH j
Manometers may also be used to measure pressure drop of fluids flowing in pipes
and conducts (see Figure 4.11). Examples to illustrate these concepts will follow.
EXAMPLE 4.3
A barom eter reads 29.5 in (749.3 mm) o f mercury. What is the atmospheric
pressure?
Solution
M ercury specific gravity = 13.56
M ercury mass density = 13.56 X 62.4 = 846.144 lbm/ft 3 (13553.53 kg/m3)
Applying Equation 4.4,
P = h X p X g / g c = 2080.104 psfa = 14.445 psia (99.59 kPaa)
FLU ID STATICS 4.11
EXAMPLE 4.4
Find the pressure (psia) at the bottom o f the pressurized vessel. The piezom eter
attached reads 50 inches (1270 mm). The flu id specific gravity is 1.5. (see
Figure 4.12).
Solution
Pressure at Point B = Pressure at Point A+ Liquid head between points A-B
Pressure at Point A = Pressure at Point C
Pressure at Point C (psia) = Atmospheric Pressure + h X SG X 62.4 / 144
Pressure at C = 14.7 psia + 50/12 X 1.5 X 62.4 /144 = 17.4 psia
(120.27 kPaa)
Pressure at Point B = 17.4 psia + 1 X 1.5 X 62.4 / 144 = 18.05 psia
(124.45 k Pa)
Atmospheric
Pressure
S olution
Fluid height in peizometer, h = 144 X PA / p vt.
where
S olution
Pressure at B = Pressure at A + Liquid static head (11 inches,SG = 1.1)
Pressure at B = Pressure at C
Atmospheric
Pressure
/ ......... \
20 psig
S G -1 .
5ft
,8
Solution
Since the flowing fluid is gas, the static head of gases is very small and may
be ignored.
Dp (psi) = h (ft) X p (lb/ft3) / 144
Dp = 15 /12 X 1.65 X 6 2 .4 /1 4 4 = 0.89 psi (6.616 k Pa)
S olution
Pressure at point B = Pressure at point A = Atmospheric pressure
= 14.7 psia (101.35 kpa)
Pressure at point B = Pressure at E + Pressure head (5 inches, SG = 0.7)
15
Submerged structures such water dams are subjected forces caused by the hydro
static water pressure. The acting resultant hydrostatic force must be accounted for
in the static design of the submerged objects. The hydrostatic force acting on a
submerged object is defined as
F = P A (4.9)
where
Equation 4.9 is applied when the pressure is uniform over the entire surface
area, such as the case for submerged horizontal planes. It may also be applied to
small area of inclined surfaces, the pressure is assumed to be constant over the
small area. The total hydrostatic force, acting on inclined surfaces, is the sum
mation of all forces acting on small areas. The total hydrostatic force is repre-
5 in, SG = 0.7
4 in, Sg = 0.9
sented by a resultant force acting at a single point of the surface. The point at
which the resultant forces is acting on is known as the pressure center. For hori
zontal planes, the pressure center is located exactly at the centroid of the area.
Figure 4 . 17 shows the area centers for different common shapes. The resultant
hydrostatic force, acting on submerged horizontal planes, is calculated by multi
plying the pressure by the area.
EXAMPLE 4.17
Find the fo rce acting on a submarine circular door (2.5 f t in diamete
(0.762 m)). The submarine is 1 mile (1.609 km) below the surface o f the sea.
Assume the sea water density is 67.0 lbm/ft3 (1073.2 kg/m3).
Solution
1 mile = 5280 ft (1609.34 m)
The submarine door may be assumed a perfect horizontal circle. The pres
sure center is located on the centroid of the area (see circle shape in
Figure 4.17)
Circle area = 3.14 X (2.5/2 )2 = 4.906 ft2 (0.4557 m2)
Pressure at 5280 ft depth
P = 14.7 ■+— —— p = ,4 ,7 ,^ 0 ^ 1 7 4 ^
144 gc K 144x32.174
For vertical surfaces, the pressure center is below the centroid of the area.
Figure 4.18 shows the pressure centers for common vertical planes. The resultant
force may be calculated using the following equation
F = H„— P (4.10)
8C
where
For inclined surfaces, Figure 4.18 must be modified for the plane inclination.
Figure 4.19 shows the same shapes when inclined with an angle 0. The resultant
force on inclined surfaces is calculated by Equation 4.10 after the depth of the
pressure center is found by equations on Figure 4.19.
4.16 FLU ID FLO W HANDBOOK
X= D /2
Y =D/2
X= B /2
Y=A/2
X= B /2
Y=H/2
X= b / 2
Y = h /3
EXAMPLE 4.18
A rectangular gate located 10 ft (3.048 m) below' the surface o f water (see Fig
ure 4.20). Find the following:
Solution
a. Since the gate is vertical, equations in Figure 4.18 may be used to find the
depth of the pressure center
where
A = 6 ft
d = 10 ft
3x10 + 2*6 6
H a= 10 + 13.23 f t (4.033 m)
2x10 + 6 3
b. Resultant pressure
EXAMPLE 4.19
Resolve Example 4.18 i f the gate is inclined with an angle o f 60° fro m the water
surface level (see Figure 4.21).
Solution
a. Equations in Figure 4.19 will be used since object is inclined, for rectangu-
lar shape
3x10 + 2x6 6
Ha = 10 + sin(60)= 12.79/r (3.9 m)
2x10 + 6 3
4 .1 8 F LU ID FLO W HANDBOOK
8J + 5D
Ha = d + —
8 2d + D
Circle
\7
3d + 2A A
H a= d+
2 d +A 3
Rectangle
h(2d + h
H a- d+
2(3 d + h)
Triangle
u j , D &d + 5 D
Ha = d +—x ---------- Sin 6
8 2d + D
3d + 2A A
H a- d + ---------- — Sin 0
2d +A 3
Sin 6
3d+ 2 A A
H a= d +
2d + A 3
Rectangle
r7 3d + 2A A
Ha= d + ---------- — Sin U
2d + A 3
Inclined Rectangle
, „ Haxg 12.79*32.174
b. P = 14.7 + ------ -- p = 14.7 + -------------------62.4
144 gc y 144x32.174
= 20.24 psia (139.56 kpaa)
Buoyant force is the upward force that acts in an opposite direction of weight force
(against gravity) on all bodies submerging (partially or com pletely) or floating in
a fluid. The buoyant force may is calculated by Archim edes’ principle: the buoy
ant force on a submerged object is equal to the weight o f the displaced fluid.
FLU ID STATICS 4.21
F„=— V„ (4 .11)
where
The buoyant force acts at the center of the displaced mass. The object will
submerge to a point where the weight of the displaced fluid equals the entire
object weight.
EXAMPLE 4.20
Find the height o f a cubical solid that will submerge in water (62.4 lb/ft3,
999.53 kg/m 3). The solid has a density o f 50 Ibm/ft3 (800.9 kg/m3). The dim en
sions o f the cubical solid are 2 X 3 X 4 f t (0.609 X 0.914 X 1.22 m) the largest
dimension is the height o f the cubic.
Solution
Volume o f the cubic = 2 X 3 X 4 = 24 ft2 (2.23 m3)
Total mass of the solid = 24 X 50 = 1200 Ibm (544.87 kg)
W eight o f solid = 1200 lbf (5337.6 N )
Buoyant force - 1200 lbf (5337.6 N )
From Equation 4.11, the displaced volume of the water
= 19.23 ft3 (0.5445 in3)
Height o f solid that will be submerged in water
= V/A = 19.23 / (2 X 3) = 3.2 ft (0.976 m)
REFERENCES
1. Giles, R. V., J. B. Evett, and C. Liu. 1994. F luid M echanics and Hydraulics. 3rd ed.
McGraw-Hill, Inc., N ew York.
2. Lindeburg, M. R. 1992. Engineer-In-Training Reference Manual. 8th ed. Profes
sional Publication, Inc., Belmont, CA.
3. Sabersky, R. H., Acosta, A. J., and E. G. Hauptmann. 1989. Fluid Flow: A First
Course in Fluid M echanics. 3rd ed. Macmillan Publishing Co., New York.
4. Simon, A. L., and S. F. Korom. 1997. H ydraulics. 4th ed., Prentice Hall, Colum bus,
Ohio.
____________ CHAPTER 5____________
FLUID FLOW: FUNDAMENTAL
CONCEPTS
Basic hydraulic concepts such the flow continuity, Bernoulli equation, hydraulic
and total energy lines are presented in this chapter. It is very essential for the engi
neer to classify the fluid flow problem prior to choose the set of equations to solve
it. Fluid flow problems may be classified according to many properties or phenom
ena such as Newtonian or non-Newtonian, compressible or non-compressible, etc.
A summary of fluid flow classification is covered in this chapter as well.
Fluid flow may be classified into many pairs of categories according to the fol
lowing conditions:
5.1
5.2 F L U ID FLO W HANDBOOK
A given flow may be described by more than one of the above classifications. For
example, the flow of a large quantity of water in a small pipe diameter where water
tem perature is identical to ambient temperature is described as incompressible,
Newtonian, turbulent, isothermal and homogeneous flow' process. If the quantity
o f w ater is independent of time, it is also a steady state flow. The different flow
categories require different solution assumptions and will yield different results.
A brief description of the above categories is followed.
v = 2 K v ,(1 - ^ t) (5.1)
where
W hen r = 0, pipe center, Equation 5.1 gives the maximum velocity which equals
twice the average velocity. When r = /?, pipe wall, the velocity reduces to 0.0.
Average fluid velocity at steady state may be found by dividing the fluid volu
metric flow rate by the cross sectional area of the pipe
where
Lam inar flow is a result of the predominant viscous forces at low velocities
between the fluid layers. Each layer resists the movement o f the adjacent layers.
The largest resistance is at the wall of the pipe where the adjacent fluid layer is
stationary. As the fluid flow increases, the layers will start to macro-mix.
FLU ID FLO W : FU N D A M EN TA L CONCEPTS
<
L a m in a r F lo w
r
T u r b u le n t f lo w
— —
R
---------------- m a x
-----------
V e lo c ity p r o file in la m in a r f lo w
R - »
----------► —
------------------- ►_
— »-
EXAMPLE 5.1
Estimate the local velocity o f water at the follow ing radial locations:
a. Pipe center, r = 0
b. r = 0.1/?
c. r = 0.5R
d. r = R
Pipe inside diameter is 6 inch (152.4 mm), and water flo w rate is 20 GPM
(4.54 m3/hr).
5.4 FLU ID FLO W HANDBOOK
Solution
Flow rate = 20 GPM = 2.67 ft3/m in (4.54 m Vhr)
Pipe cross sectional area = 0.19635 ft2 (0.01824 n r)
Average velocity (Equation 5.2) = 2.67/0.19635 = 13.59 ft/min = 0.22 ft/sec
(0.06907 m/sec)
Applying Equation 5.1
V = V max (5.3)
where
Equation 5.3 is applicable at high Reynolds numbers where the laminar sub-layer
is very small. Equation 5.3 may be expressed as a function of the average veloc
ity (average velocity is by Equation 5.2)
1/7
(5.4)
FIGURE 5.3 Velocity profiles for the laminar and turbulent flows.
where
y =R - r
R = Pipe radius
n = 1/V f
f = Friction factor
Solution
Volumetric flow rate = 300 GPM = 40.1 ft3/min = 0.668 ftVsec (68.13 mVhr)
Average velocity = 3.4 ft/sec (1.037 m/sec)
Friction factor of fully developed turbulent flow in 6 inch pipe = 0.016
(Crane technical paper No. 410)
From the above example it is clear that the velocity gradient, in turbulent flow, is
very sharp near the pipe wall and flat in most of the pipe cross sectional area.
5.1.1.3 Transition Flow an d Reynolds Num ber. The transition in flow from lami
nar into turbulent depends on many factors such as the fluid physical properties,
specifically the fluid viscosity and density, fluid velocity, and the dimension and a
characteristic length of the flow element such as a diameter of pipe or hydraulic
radius of a non-circular duct. The criterion which distinguishes laminar flow from
5.6 FLU ID FLO W HANDBOOK
turbulent is the Reynolds number. For pipes and circular ducts with full flow,
Reynolds number is defined as
DVp DV _ 2 RV _ D Qp _ 4Q p _ DG
Re =
V v v jjA 3.14 p D fiA
4G
3.14 pD
( 5 .6 )
where
D Pipe Diameter
R Pipe radius
V Fluid velocity
p Fluid density
p Fluid dynamic viscosity
v Fluid kinematics viscosity, p ip
Q Volumetric flow rate
G Mass flow rate
A Pipe cross sectional area
Re Reynolds number
Volume Mass
Dynamic Kinematics How How
Velocity Diameter Density viscosity viscosity rate rate
An hydraulic diam eter is defined as four times the cross sectional area divided by
the wetted perimeter of the duct. Reynolds numbers for different flow character
istics (such 2-phase flow, flow over bank of tubes, and flow in ducts and open-
channels) will be discussed at appropriate chapters through out the handbook.
S olution
2 /1 2 x 1 x 6 2 .1 8
Re = ----------------------- = 16947.6
0 .9 1 x 6 .7 1 9 7 ^ -4
S olution
Re = ..5 ^ 7 .4 0 7 ,0 .0 7 0 8 = 6 ]2 8 9 ]5
0.0191x6.7197^-4
The fluid viscosity behavior determines the fluid classification in the Newtonian
and non-Newtonian category. Fluid viscosity has been defined in Chapter 2 for
gases and liquids. Newtonian fluids are characterized by a constant viscosity at
constant temperature and pressure and by the independence o f viscosity on time.
The viscosity of a Newtonian fluid is equal to the shear stress divided by the rate
of deformation. The relation between the stress and the rate of deformation is a
straight line (see Figure 2 .1). Non-Newtonian flow poses a more complex viscos
ity behavior since the relation between the shear stress and the rate of deformation
is nonlinear and is a function of time. Non-Newtonian fluids are incompressible in
nature. Examples of non-Newtonian fluids are:
In compressible flow, fluid density changes as the pressure decreases. If the flow
is steady and the flow cross sectional area is uniform, the velocity will increase and
the change in the kinetic energy term becomes substantial. Unlike compressible flow,
the incompressible flow has a negligible density change. Velocity is merely the same
and the kinetic term may be dropped unless there is a decrease or increase in the
flow cross sectional area. Flow of gases with high pressure drop falls under the
com pressible flow category. Gas relief networks are good example were the fluid
compressibility must be taken into account.
Homogeneous flow refers to the flow of a fluid in a distinct flow-phase such as liq
uid or vapor (pure components or mixtures). When two or more distinct phases flow
FLU ID FLOW: FU N D AM EN TAL CONCEPTS 5.9
concurrently, the tlow is known as heterogeneous. Flow of liquid and gas slurries,
paper pulps, and oil-gas are a few examples. The hydrodynamics of a heteroge
neous flow is more complex and depends mainly on the flow regime. More dis
cussion of the heterogeneous flow will be covered in later chapters of the handbook.
In isothermal flow, the fluid has a constant temperature which means properties
(density and viscosity) are constants. Though isothermal flow is an ideal assump
tion, a few flow processes may approach this state such as the firewater network
case. In non-isothermal flow, the temperature changes drastically due to heat
transfer with the surroundings, which reflects on the fluid properties and hence
on the flow hydrodynamics.
Flow o f fluids in pipes and ducts are examples of a closed-conduit flow. Closed-
conduit indicates that the fluid is totally enclosed by a solid boundary (pipe or
duct). Open-channel flow, on the contrary, is only partially enclosed by a solid
boundary such as rivers and creeks. Only liquids or liquids-like, such as slurries,
fluids may undergo open-channel flow. In open-channel flow, the liquid is under
atmospheric pressure and has a free surface. If a flow in a closed-conduit has a
free surface, it also may be treated as open-channel flow.
The principles of mass conservation are applicable to fluid flow. Mass cannot be
created or destroyed. Therefore, the mass balance holds true between any two
5 .1 0 FLU ID FLO W HANDBOOK
points in a flow element such as a pipe or a venture meter. Referring to Figure 5.4,
a mass balance between points 1 and 2 yields
C, = G2
Q\P\ = Q2P2
(5.7)
VAPi = V2A2p2
VtD?Pl = KD2P2
where
G, = g2
<2, = Q2 (5.8)
II
VA
N>
VtD? =
Rl
D,
(5.9)
\
The velocity at point 2 is expressed as
V2 = (5.I0)
D
EXAMPLE 5.5
Find the velocity? o f an incompressible fluid in a pipe contraction. The upstream
flu id velocity is 5 ft/se c (1.524 m/sec) and the diameter ratio, B, is 0.5.
Solution
£ = 0.5
B 2 = 0.25
For unsteady state application, such as the flow shown in Figure 5.5, Equation 5.7
is modified to include the time-rate term
M, - M„ = (5.10)
dt
where
Since
M' = Q p
M T= V p
5 .12 FLUID F LO W HANDBOOK
where
a a - dp* = (5 .ii)
dt
Pi - Po - P r
q, - a. = dt
' 5. 12)
EXAMPLE 5.6
Determine the rate at which the liquid height in Figure 5.6 changes with time.
H ow long it will be before the tank reaches a flooding condition. Tank fluid is
water at 77.0°F (25°C). Tank has a cylindrical shape with a radius o f 5 f t
(1.524 m).
Solution
The problem may be solved with a direct substitution into Equation 5.12
dVIdt = 5 -2 = 3 GPM = 0.401 ft3/min (0.01135 m 3/min), where V is the liq-
uid volume in tank
Tank cross sectional area, A = 3.14 X (5/2)2 = 19.625 ft 2 (1.823m2)
O u tle t
Energy in fluid flow is also conserved. Fluid energy between any two points in a
flow element is identical unless there is an energy transfer from or to the sur
roundings or energy lost due to friction. Two forms of energy equations are of
practical application in fluid flow, the mechanical energy equation which also
known as Bernoulli equation and the thermal energy equation.
Both energy equations are direct application of the following energy conser
vation balance
5 GPM
0.0189 m 3 /m in
10*
3.05r l 5
|l.S 24m f
2 GPM
0.0075 m3Amin
Total fluid energy at any point is equal to the summation of pressure energy,
kinetic energy, potential energy and internal energy as
P V2
Total Fluid Energy = — I------- \- Z + u (5.14)
P 2g
Units must he consistent in the above equation (internal energy units must be
expressed in ft of fluid to be consistent with other terms). Total energy balance
for a fluid between two points becomes
P V2 P V2
+ ~z~ + Z, + h - hf = — + - ^ + Z2 (5.15)
P 2g p 2g
where
± Q + Qf = H2 (5.16)
where
H Fluid enthalpy
Q Heat transfer from/or to the surroundings
Qf Heat added to fluid due to friction
Equations 5.15 and 5.16 are the bases to solve fluid flow problems. While Equa
tion 5.15 is used to determine the pressure change, Equation 5.16 is used to track
the change in fluid enthalpy and hence fluid temperature, examples are shown in
chapter 3. A simultaneous solution of both equations isessential to accurately
predict the fluid temperature and pressure change.
EXAMPLE 5.7
Determine the pressure at point 2 in Figure 5.7. The flu id is incompressible and
has mass density o f 61.5 lbf/ft? (985.107 kg/m3). The follow ing data simplify the
solution process.
Total head loss due to friction = 1.251 psi (8.624 kpa) = 2.93 ft (0.893 m)
Pump head = 55 ft (16.764 m)
FLU ID FLOW : FU N D AM EN TAL CONCEPTS 5.15
Solution
The Bernoulli equation is solved for the pressure at point 2. Velocity at the suc
tion and discharge of the pump is found by dividing the volumetric flow rate by
the pipe cross sectional area
EXAMPLE 5.8
I f point 2 in Example 5.7 discharges to an atmospheric tank, find the pump
head required to deliver a flow rate o f 300 GPM (68.13 m3/hr).
Point 2
20* (6.096 m)
1Qa(3.408m) 4“ (101.6 m m
3" (76.2 m m )
P o in t 1
Solution
P ] = p2 = 0.0 psig (since both tanks are atmospheric)
VI = 13.62 ft/sec (4.151 m/sec)
V2 = 7.663 ft/sec (2.33 m/sec)
The frictional pressure drop is calculated by Equation 5.17. The friction factor
and frictional losses calculations will be discussed in the next chapter. For this
example a friction factor of 0.017 is used
fL V 2
= ^ — (5.17)
2 Dg
A graphical representation of the fluid total energy line versus the hydraulic line
is helpful in solving and understanding the pipe flow analysis. Both lines are
expressed in terms of fluid head (feet or meter). Total fluid energy line has been
discussed before and may be calculated by following equation
P V2
Total Energy Line = ----- 1---------- 1- Z (5.18)
gp
At any point in the flow, the hydraulic line is the total energy at that point less the
kinetic energy
P
H ydraulic Energy Line = ---------v Z (5.19)
gp
FLU ID FLOW : FU N D AM EN TAL CONCEPTS 5.17
From the above two equations, the hydraulic line is always less or equal to the
total energy line. Both lines coincide when the velocity term is zero. As the fluid
velocity increases, the hydraulic line drops below the energy line. To avoid the
risk of collapsing in large pipelines, the hydraulic line should always be below the
pipe elevation. Figure 5.8 shows both lines (total and hydraulic) for a How due to
gravity between two atmospheric tanks. Note how the two lines coincide when
fluid velocity approach zero at the inlet and outlet of atmospheric tanks.
Many of the fluid flow phenomena are characterized by dimensionless group num
bers and not by a single flow variable or fluid property. For example, Reynolds num
ber, which is a dimensionless group of fluid velocity, pipe inside diameter, and fluid
viscosity and density, distinguishes the laminar and turbulent flow boundary. Flow
ing fluids with identical Reynolds number, even if the fluids have different physi
cal properties and flow velocities, are known to be under similar hydrodynamic flow
as long as there is a geometrical similarity in the flowing boundary. Reynolds num
ber defines the ratio of the inertial force to the viscose force. When two fluids flow
with similar Reynolds number, the inertia to viscose force ratio are similar. The
dimensionless groups are very useful for scaling experim ent models into real life
application as long as geometrical similarity exists.
The following is a partial list and definition of dimensionless groups that pertiin
to fluid flow problems and applications.
5 .5 .1.1 Reynolds Num ber, Re. Reynolds number defines the ratio of ineitia
force to the viscose force
5.20)
where
Reynolds number characterizes the flow type (laminar or turbulent) and is lseful
parameter for correlating the friction factor coefficient.
5.5.1.2 Weber Num ber, We. W eber number defines the ratio of the inertia force
to the surface tension
5.21)
or
where
5 .5 .1.3 E uler Num ber, Eu. Euler number defines the ratio of inertia (or liffer-
ence in static pressure) to the dynamic pressure
FLU ID FLOW : FU N D AM EN TAL CONCEPTS 5.19
where
P Fluid pressure
Eu Euler Number, dimensionless
Static pressure of a flowing fluid is the measured fluid pressure less the kinetic
energy term. Euler number is important in predicting fluid flow cavitations char
acteristics. Euler number is also called Cavitations Number and may be defined as
(5.23)
where
Fluid cavitations are phenomena that cause vapor bubbles to form when the fluid
pressure falls below the fluid vapor pressure at the local fluid pressure and tem
perature. Cavitations Number is important in the design of rotary power equipment.
5 .5 .1.4 Froude N um ber, Fr. Froude number defines the ratio of inertia force to
the gravitational force
(5.24)
where
5 .5 .1.5 M ach N um ber, Ma. Mach number defines the ratio of the fluid velocity
to the velocity of sound in the same fluid
(5.25)
where
In incompressible flow, Ma number is very small (Ma « 1.00) since the sonic
velocity in liquids is much higher than it is in gases and the liquid velocity is
5 .20 FLU ID FLO W HANDBOOK
much smaller than gas velocities. Gas flow with Mach number larger than 0.3
indicates that gas compressibility affects the pressure drop calculation (accelera
tion term) and must be accounted for.
5.5.7.6 O ther Geometric Ratios. Relative pipe (or wall) roughness, the fineness
ratio, and the contraction ratio are some o f the important geometry ratios which
define the flow boundary of the fluid. The relative roughness is defined as the
ratio of the pipe surface roughness factor to the pipe diameter:
e/D
where
L /W
The area contraction ratio, /3, is the ratio o f the small flow diameter to the larger
one.
where
W hen a number o f flow variables and/or fluid properties are known to affect and
contribute to a certain flow characteristic or phenomenon, it is more convenient
to express this characteristic as a function o f a dimensionless group. To formu
late the dimensionless group, the following steps may be taken:
1. Variable and fluid properties are expressed in terms of the basic units of
length (L), time (T) and mass (M). Table 5.2 has a list of variables defined
in the basic units of time, length, and mass.
FLU ID FLOW: FU N D A M EN TA L CONCEPTS 5.21
2. The dimensionless group is set equal to the product of all variable each
raised into unknown exponent variable as:
where
1. Express all variables in the basic units o f time, length, and mass as:
Velocity L/T
Density M /L 3
Viscosity M/L-T
Diameter L
2. Define the dim ensionless group as a product o f the above variable. Each
variable is raised into unknown exponent:
Re = V". D h .p (
(L) 0 = a +b - 3 c - d (5.28)
Re = V '.D 1 .p 1 .fi
Sim ilar results will be obtained if other exponent was set, say b = 1, and then the
system of equation solved for other variables.
FLU ID FLO W : FU N D AM EN TAL CONCEPTS 5.23
REFERENCES
6 .1 1NTRODUCTION
In developing the equations of change of a pure fluid, the Euler point-of-view will
be used. Here a control volume (system) fixed in space will be selected having the
fluid passing through it. Although we speak of a pure fluid we can apply the results
to a mixture as long as the fluid properties are continuous and do not change sig
nificantly if the composition changes.
The equations o f change are the result o f applying three physical laws to an
open system. The laws are the law of conversation of mass, Newton’s second law
of motion and the first law of thermodynamics. The resulting partial differential
equations are referred to as the continuity equation, the equations of motion or the
Navier-Stokes equations and the general energy equation. To now we have referred
to the corresponding one- and two-dimensional simplifications of the above as
material balances, momentum balances and energy balances. An open system such
as we refer to here is a finite control volume fixed in space with material flow
ing through it.
6.1
6.2 FL U ID FLOW H A N D B O O K
The law of conservation of mass will be applied to the control volume shown in
Figure 6.1. The dimensions of the control volume (system) are Ax, Av, and Az,
and is orientated in the flow field such that all o f the flow enters through three
faces and leaves through three faces.*
The law o f conservation of mass can be expressed in the following mnemonic
representation:
The mass flowrate into the control volume consists of the sum of the mater
ial entering the three faces AxAy, AxAz and AyAz. The mass flow rate entering
face AyAz is the product of the area, AvAz; the density of the fluid, p; and the
component o f the velocity vector perpendicular to the face, vv. This is p v xA y A z \ x
as shown in Figure 6 .1. Similar terms can be obtained for the mass flow rates enter
ing and leaving the other faces.
FIGURE 6.1 Control volume fixed in space with fluid passing through.
*This derivation after Bird, Stewart, and Lightfoot is preferred to others employing a control vol
ume of differential dimensions or to the Lagrangian point-of-view of following a fixed mass as it
moves with the fluid [1]. It is felt that this is the easiest to comprehend initially, although it is not
the most concise presentation.
THE G EN ER A L EQUATIONS 6.3
Since AxAyAz are fixed in space and not a function of time the above can be
written as
The terms on the right hand side of the above equation are in the form of the defi
nition o f partial derivatives. The result of letting Ax, Ay and Az approach zero is
which can be written in vector notation in terms of the vector differential opera
tor V as
dp
=V-pv (6.5)
at
The term V • pv can be thought of as the net mass flux per unit volume through
the control volume. Thus, the decrease in density with time of a control volume
fixed in space is equal to the net rate o f mass efflux per unit volume.
If the Lagrangian point-of-view had been used, the continuity equation would
have been obtained in terms of the substantial derivative. Expanding Equation 6.4
and rearranging gives:
6.4 FLU ID FLO W HANDBOOK
The substantial derivative of the density is written Dp/Dt and is the term in the
brackets of the above equation. Written in vector notation the above is: (See Bird,
et.al. [ 1 ] for the expanded form in other coordinate systems.)
The next chapter will illustrate the solution of the above equation for various
situations, but one simplification is worth mentioning here. This is the incom
pressible fluid simplification (p = constant) which is applicable to the flow of
most liquids. The previous equation becomes
V •v = 0 (6 .8)
The result of applying N ewton’s second law of motion to an open system— a con
trol volume fixed in space with fluid flowing through— is referred to as the equa
tions of motion.* To derive these equations, the same control volume will be used
and Newton’s second law for an open system can be written mnemonically as
(6.9)
There are three components of the above equation, the x , y, and z components.
The x component can be indicated by the following equation
* Again. Bird, Stewart, and Lightfoot’s approach2 appears best for the reasons previously stated.
THE G E N E R A L EQUATIONS 6.5
Before being able to apply the above equation, it is necessary to review the forces
and stresses that act on a body.
The forces that act on a body are either surface forces or body forces. Body
forces are ones that are proportional to the volume or mass o f the body and com
prise those forces that involve action at a distance. Examples are gravitational attrac
tion, magnetic forces and electrodynamic forces. Surface forces are ones that exert
force on the control surface by material outside the control volume. Such forces are
exerted in the form of surface stresses (force per unit area) and are of two types—
normal or pressure stresses and shear or viscous stresses. Normal stresses are usu
ally considered to consist of the sum of two terms. These are stress due to static
pressure, P, and stress due to linear deformation of the fluid element, a x.
For the fluids at rest or in motion such that the velocity is everywhere the
same each normal stress is equal to the pressure.
A stress Tacts on each of the control volumes and has one component normal
to the surface and two tangential to the surface as shown in Figure 6.2 for the face
in the y-z plane. The sign conventions to be used are that normal stresses are con
sidered positive in compression and shear stresses are considered positive on
faces in contact with the point (x,y,z). As shown in Figure 6.2, the first subscript
on a stress gives the face on which the stress acts by giving the axis normal to the
face and the second subscript gives the direction that the stress acts. Therefore
stresses with repeated subscripts are normal stresses, and those with mixed sub
scripts are shear stresses.
Derivation o f the x-com ponent: In Figure 6.3 the control volume is shown with
the shear and normal stresses that action the ^-direction only. Before evaluating
the first term of Equation 6.10 a body force term should be included. Let x be the
component o f this force per unit volume in the x direction.
The sum o f the forces in the x direction is the sum of the body, shear and nor
mal forces.
X Fx = X p A x A y k z + TvvA y A z | v + t vvA a A v | v + t . vA a A ;y ,
( 6 . 12)
- ', , - V v A q , + a.v - tvvA, vA j | v + Av - T-rA v A ; | - + A,
Rearranging gives
Xp- Txx I .v + Ax
Ax
Tx.x x
+
Tyx y + Av
Ay
^yx v
(6.13)
^zx\z + Az ~ Tz x L
+
6.6 FLU ID FLO W HANDBOOK
Similar results can be obtained for the y and z directions. In Figure 6.4 the con
trol volume is shown with the convective momentum entering and leaving the cm-
trol volume in the x-direction only. Rate of momentum is defined as the proactt
o f the mass flow' rate and the velocity in the direction of the mass flow rate.
FIGURE 6.3 Shear and normal stresses acting on the control volume in the ^-direction only.
6.8 FLU ID FLO W HANDBOOK
The rate of change of the momentum in the control volume in the x-direction is
given by
Substituting Equations 6.13, 6.16 and 6.17 into Equation 6.10 and taking the
limit as A x, Ay and Az go to zero the result is
(6.18)
The above equation is the jc-component of the equation of motion. In the con
cise form of vector notation the equation of motion Bird, et al. [3] is:
pB - V • t = V • Vvv + — pv (6.19)
dt
The first term represents the body force acting per unit volume; the second term
represents the shear and normal forces acting per unit volume, the third represents
the rate of change of convective momentum per unit volume and the last term rep
resents the rate of change of momentum with time per unit volume. The term vv
is a dyadic product, the terms V • Tand V • Vvv are single dot products of a vec
tor V and tensors t and vv.
TH E G E N E R A L EQUATIONS 6.9
Dv
p -— = -V • t + pB (6.20)
This would be the form obtained if the Lagrangian approach were used to
derive the equation. (See Bird et al., [1] for the expanded form.)
To this point the above equations are very general and apply to any fluid from
a gummy mass to a gas at a low pressure. The fluid must be continuous however,
and the equations will not hold for the case where the flow must be described by
the movement of discrete molecules.
To apply the equations of motion to the flow of a fluid, a relation between the
shear stress r a n d the velocity v is needed to eliminate rfro m the equations. For
many fluids, the stress is proportional to the velocity gradient and relatively sim
ple relationships result. These are referred to as Newtonian fluids. However, for
some fluids such as polymer solutions and melts, the stress is not proportional to
the velocity gradient; these are referred to as non-Newtonian fluids. The devel
opment of mathematical models relating the stress and velocity gradients in a
non-Newtonian fluid is a subject in itself.
Our next effort will be devoted to deriving the relations between the shear
stress and velocity gradient for a Newtonian fluid. This will give some insight to
the understanding of the other models. The results will just be stated for the rela
tion for the normal stress as this derivation is somewhat more detailed. Also, some
o f the more widely used non-Newtonian relations will be discussed.
A one dimensional flow field is shown in Figure 6.5a and a differential element in
this How field, Bennett and Myers [4]. At time t the dimensions of the fluid ele
ment are dx - dy ' 1, and at time t + dt it is deformed by shear stress ryx as shown in
Figure 6 .6 b. The arc length, dl, is related to the velocity gradient by the following:
d l= ° ^ -d y d t (6 .2 1 )
ay
For small angles the arc length is equal to the product of radius and the angle, i.e.
dl = dyd(j) (6 .22 )
Combining Equations 6.21 and 6.22 gives the relation between the rate of
deformation of a fluid element (in one dimension) d(f)/dt and the velocity gradient.
dvx _d(j)
(6.23)
dy dt
6.10 F LU ID FLO W HANDBOOK
ryx = - p - ~ - (6.24)
dt
The proportionality constant is the viscosity. Combining Equations 6.23 and 6.24
gives the relation for one dimensional flow.
= (6.25)
ay
Extending the above analysis for a two-dimensional flow field, this is shown
in Figure 6 .6 . The rate of deformation d(f) is the sum of the changes in the .v and y
directions and is given by:
It can be shown (Bird et. al. [5]) that ryx = r ^ . The other shear stresses can be
obtained similarly and are
6.12 F L U ID FLO W HANDBOOK
The normal stresses are obtained by considering the normal deformation of the
fluid element (Bird, et. al. | 6 ]). The normal stress includes the pressure and is given
by the following:
(6.32)
(6.33)
, \ 2 /x
r „ = P - 2H - S - + X V ' V (6.32)
Substituting Equations 6.29,6.30, 6 .3 1, 6.32, 6.33, and 6.34 into 6.20, the result
is the three components of the general equation of motion for a Newtonian fluid
with varying viscosity and density.
6.14 FLU ID FLO W HANDBOOK
Dy
- - V P + /xV2v + pB (6.38)
Dt
These equations are given in expanded form in rectangular, cylindrical and spher
ical coordinates in Bird et.al. [7]
For the case of t = 0, an ideal fluid,
D\
p — = - V P + pB (6.39)
Dt
dvx
Tyx = T0 - f l ' (6.40)
dy
Polymer solutions and melts have been described by the pseudoplastic model for
one-dimensional flow (Bird et al., [ 8]):
n- 1
dvt dvx
T ry = m n< 1 (6.41)
dy dy
One of the more complicated models is that of Bird and Meter describing the flow
of polymer solutions [9].
a- 1
dvx _ z L (6.40)
dy Vo 1 + {Tx y l T m) a 1T]Jr)„
The reader is referred to Bird et al. [10] for more information on models now in
use. The multidimensional generalization of Equations 6.40 and 6.41 are given
in reference [ 1 1 ].
TH E GENERAL. EQUATIONS 6.15
The result o f applying the first law of thermodynamics to a control volume fixed
in space is referred to as the general energy equation. To derive this equation, the
same control volume will be used, and the first law of thermodynamics for an open
system can be written mnemonically as*
By the usual sign convention, work done by the control volume (system) on the sur
roundings is positive. Also energy transport by thermal, nuclear and other electro
magnetic radiation is not specifically included. The internal energy per unit mass, U,
is the energy associated with molecular motion and is a function of the material and
the temperature and pressure. The kinetic energy per unit mass, v 2/2 is the energy
associated with bulk fluid motion (v 2 = | v |2).
To evaluate each term in Equation 6.43, consider the fluxes shown on the con
trol volume of Figure 6.7. For the first term, the rate of accumulation of internal
and kinetic energy in the control volume is given by the product of the mass of
the control volume and the internal and kinetic energy per unit mass in the con
trol volume.
The control volume is orientated in the flow field such that the internal and
kinetic energy associated with the flow entering passes through the faces touching
point U,y,z), and leaving through faces touching point (x + Ajc, y + Ay y z + Az). The
components entering and leaving through faces AyAz are shown. The combined
*Again, the approach of Bird, Stewart, and Lightfoot appears best for the reasons stated previously [1].
6 .1 6 FLU ID F LO W HANDBOOK
t)VAxAzvJ ^
FIGURE 6.7 Typical components of the flux of internal and kinetic energy, heat by conduction and
work done by the control volume.
net flux of internal and kinetic energy (second term of Equation 6.43) can be
written as
37
= -k
dx
dr
Q\ =
—- k (6.47)
dy
dT
<lz = ~k
dz
and
q = -k V T (6.48)
Com bining the above with Equation 6.47 gives the net rate of heat addition in
terms o f the tem perature gradient and the thermal conductivity.
k dT/dz Iz + Az k dT /dz\z
(6.50)
Az
If qg is the rate of heat generation per unit volume by resistance heating or some
other means, the total rate of heat generation in the control volume is given by
qgA x A y A z (6.51)
The evaluation of the work term of Equation 6.43 consist of two parts— work
done by the control volume against volume forces and work done against surface
6 .18 FLUID FLO W HANDBOOK
forces. The only volume force to be considered will be gravity, and surface forces
to be considered will be the shear and normal forces. Actually, in Equation 6.42
it is necessary to include the rate of work done which is the product of the force
and the velocity in the direction of the force. The rate of work done against grav
ity can be written in terms of the three components of the gravitational force g
(g x> gy> g z ) and is
The negative sign arises because work is done against gravity (i.e. on the sur
roundings) when v and g are opposed. This can also be thought of as the change in
potential energy where the potential energy increases when v and g are opposed.
In Figure 6 .8, the shear and normal forces in the x-direction of the control vol
ume are shown that oppose those forces from the surroundings. The net rate of
work done associated with v* for the flow field shown is
+ VxTzx k x A y I, + - vxTyx Ax A y | z
7 XX Vx Ix + Ajc T xx
A xA yA z =
Ax
~Ay
vx TZX I z + Az V v T-zx I z
(6.54)
Az
There are corresponding terms associated with vy and vz which will be left to the
reader to develop.
All of the terms for Equation 6.43, the first law of thermodynamics, have been
developed. Each is multiplied by A xA yA z and has been put in the form of the def
inition of a derivative. Canceling the AjcAjAz coefficients and taking the limit as
Ajc, Ay, A z go to zero, and performing some rearranging, results in the general
energy equation.
TH E G EN ER A L EQUATIONS 6 .19
FIGURE 6.8 Shear and normal force reactions by the control volume in the .v-di recti on.
d d
^ - ( T x x V x + Tx y Vy + TXZVZ) + — ( T yj[VX + TyyVy + T y ^ )
ox oy
+ ^ r ( Tz x vx + r z y Vy + Tzzvz) (6.55)
az
(6.56)
This is the most general form of the general energy equations. It is cumbersome
to manipulate since it is so general. Consequently, it is convenient to have sev
eral specific forms to use. An extensive tabulation of various forms is given by
Bird et. al., [12] along with the manipulations illustrating the methods for obtaining
some of these forms. The reader is referred to this reference for cases not covered
in the following discussion.
6.20 FLU ID FLO W HANDBOOK
If the expressions for the shear and normal stresses from Equations 6.32, 6.33, and
6.34 are substituted into Equation 6.56, the result is the general equation for a New
tonian fluid. An important form o f this equation is obtained by subtracting from
Equation 6.56 the equation obtained by forming the dot product of the velocity with
the equation of motion and substituting for U in terms of cv (Bird et al. [13]).
(6.57)
The above equation written in terms of the substantial derivative for an ideal gas is
DT
p cv — =V -(k V T )-P V \ (6.58)
An incompressible fluid is one in which the density is constant. The energy equa
tion neglecting the viscous dissipation term is
p c v ----- = V • (k V T) (6.59)
Dt
6.4.4 Solid
pcv — = V ■( k V T ) (6.60)
Dt
TH E G EN ER A L EQUATIONS 6.21
6.5 SUMMARY
The general equations of change have been derived using the Euler point-of-view
of a control volume fixed in space. These equations include the continuity equation,
the equation of motion, and the general energy equation. Each one was written
in several convenient forms for subsequent use. The next chapter will illustrate
their use in developing the mathematical models for a variety of applications.
For convenience the various forms of the general energy equation are given
in Bird et al. [ 14]. These tables include an expanded form of the energy equation
is given in rectangular, cylindrical and spherical coordinates in terms of the
energy and momentum fluxes. Also, the corresponding set is given in terms of
the transport properties. In addition, the general energy equation is given in a
variety of forms along with the equations of motion and continuity.
REFERENCES
1. Bird, R.B., W.E. Stewart and E.N. Lightfoot. 1962. Transport Phenomena. John
W iley, Inc., New York, p. 74.
2. Ibid. p. 76.
3. Ibid. p. 78.
4. Bennett, C.O. and J.E. M yers. 1962. M om entum , H eat and M ass Transfer.
M cG raw -H ill, Inc., New York, p. 92.
5. Ibid. p. 90.
6. Ibid. p. 94.
7. Bird, R.B. et. al. op. cit. p. 11.
8. Bird, R.B. et. al. op. cit. p. 11.
9. M eter, D.M ., 1964. “Tube Flow o f non-N ew tonian Polym er Solutions: Part II T u r
bulent Flow .” A.I.Ch.E. Journal. V olum e 10(6): 8 8 1 -4 .
10. Bird, R. B., R. C. Arm strong and O le Hassager, 1987, D ynam ics o f P olym er L iq
uids, Vol. 1 Fluid D ynam ics, 2nd ed., John W iley and Sons, Inc., New York, NY.
11. Bird, R.B. et. al. op. cit. p. 103.
12. Ibid. p. 311.
13. Ibid. p. 322.
14. Ibid. p. 318ff.
CHAPTER 7
APPLICATIONS
OF THE EQUATIONS
OF CHANGE
7.7 INTRODUCTION
In the previous chapter the equations of change were developed in detail and in this
chapter the use of these equations to obtain mathematical models will be illus
trated. The approach will be first to show how the general equations can be sim
plified to obtain the same equations that were derived from first principles. In these
relatively simple problems, analytical solutions were obtained. Next, some more
complicated illustrations not previously discussed will be presented. This will be
followed by a discussion o f the use of the equations of change for dimensional
analysis. Then classical results from various branches of fluid mechanics will
conclude the chapter.
The equations of change will be simplified to obtain the same equations for the
problems of laminar flow in circular tubes, cooling a sphere and the “start-up” of
laminar flow.
7.1
7.2 FLU ID FLO W HANDBOOK
r dr r dO dz
If the z axis is selected as the direction of flow, then ve = 0 and vr = 0, since there
is only flow in the axial direction. The above equation becomes
4 ^ = 0 (7.2)
dz
This says that can be a function of r and 6 only. By axial symmetry v, = vz(r)
only, a pertinent result.
dp dvz i a2
+ /x + + + PZz (7.3)
dz ~a7 r 2 dO2 dz
Simplifying the above equation, the first term on the left hand side is zero since
there is steady flow; the second and third terms are zero since v,. = 0 ; and ve = 0 ;
and the fourth is zero by Equation 7.2. On the right hand side the third, fourth and
fifth terms are zero since v. = j{r) only. For the gravitational term, gr = g cos /3
when (3 is the angle between the axis and the direction which gravity acts. Here
(3 = FI and cos IT = —1. Therefore Equation 7.3 simplifies to give
A dP i± d I dv~
(7.4)
° ‘ - & + 7 3 7 r1 7
rdvz
= — — (P + pgz)r (7.5)
d r \ V dr /jl dz.
dv~ 1 dp C\
r+ — (7.6)
dr 2/jl dz
where p = P + pgz. The constant c, is zero since the velocity gradient is finite at
the centerline, r = 0. Integrating again gives
1 dp 2
v, = ------- —r + c 2 (7.7)
4/x dz
Applying the boundary condition of no slip at the wall = 0 at r - /?, the con
stant of integration is evaluated, and the equation for the velocity profile is
7.4 F L U ID FLO W HANDBOOK
0 = ^ + pgr (7.8b)
or
n dP (7.8c)
Both go and gr are zero since the angle (3 = 11/2 in each case. These equations
show that the pressure is uniform in these directions.
1 d ( ■ a 'd T 1 d2r
+ —:— — I s i n ft — - +
dr r “ sin0 d6\ d0 r sin 0 3</>
(7.9)
dT 1 d / 2 dT (7.10)
dt pcP
“S ta rt-u p ” o f Lam inar Flow. This flow has a stagnant fluid in a horizontal pipe
suddenly be subjected to a pressure gradient and movement begins. The Navier-
Stokes equations contain a term to describe this transient flow.
The continuity equation with constant properties in cylindrical coordinates is
Equation 7.1. Selecting the z-axis as the direction of flow, the same simplifica
tions are obtained (ve = vr = 0), and the result is Equation 7.7 again.
A PPLIC A TIO N S OF THE EQUATIONS OF CHANGE 7.5
(7.11)
The solution of this partial differential equation is obtained by the method of sep
aration of variables and is given by Bird et al. [ 1].
Several additional examples of the use of the equations of change will be given.
These include illustrations of non-Newtonian flow and the combined solution of
the equation of motion and general energy equation. Some of these and others are
given by Bird et al. [ 1].
Non-N ew tonian Flow in a Tube. To illustrate the use of the equations of change
for the case where the shear stress is not proportional to the velocity gradient,
consider the case of the isothermal flow of a power law fluid in a horizontal tube.
The shear stress-shear rate relation from Equation 6.41 can be written as
(7.12)
For one-dimensional flow m is referred to as the apparent viscosity. For n < 1 the
behavior is referred to as pseudoplastic, and for n < 1, dilatant.
To determine the volumetric flow rate as a function of the pressure gradient,
fluid properties and tube radius, the velocity profile must be determined. Proceed
ing, the continuity equation holds for any fluid and results in the same as that of
Equation 7.2.
For the equation o f motion, it is necessary to start with the form in terms of
the shear and normal stresses. For the z-component in cylindrical coordinates from
Equation 6.20, the result is
(7.13)
7.6 FLU ID FLO W HANDBOOK
The terms containing t 0z and rzz do not appear as they are functions of the gadi-
ents d v J d d , d v jd z, dvr/d and d v j d z which are zero. The above equation cai be
integrated once since the pressure gradient is a constant.
d t
— \ (rTrz)
V dP / rdr
=-\ — C-14)
rrrz = 0 dr *'() dz
For the lower limit on the integral the shear stress at the center line is finite and
the product of r and Trz is zero. Integrating gives
dP_ r_
(M 5 )
dz 2
Mn
^1 + 1In
-1 dP
V- = 1- f - f (7.17)
2m dz 1 + Mn
A sketch is given in Figure 7.2 illustrating the effect of n on the velocity pnfile.
To obtain the volumetric flow rate Q, the following integral must be ^alu-
ated which sums all of the differential elements of dimensions (rd6)(dr) ov* the
tube cross-section for 6 from 0 to 2FI and r from 0 to R.
rR r2ir
Q=\ v^rdddr (7.18)
Jo Jo
By axial symmetry vzr is not a function of ft and the first integration give.^
rR
Q = 2 n R 2 \ vz(r/R)d(r/R) (7.19>
Jo
7tR 7,
-d P /d z - R
Q= 7.20)i
(3 + Mn) 2m
pseudoplastic (n<l)
r
— Newtonian (n=l)
dilatant (n> l)
FIGURE 7.2 A sketch of the velocity profiles comparing pseudoplastic, newtonian and dilatant fluids.
since = vy = 0 .
Examining the z-component of the equation of motion, Equation 6.37, it sim
plifies to
(7.22)
<7-24>
The second term on the right is from the viscous dissipation portion of the gen
eral energy equation. Thus, it is necessary to use the velocity profile, Equation 7.23,
to evaluate the velocity gradient to use in Equation 7.24. Combining Equations 7.24
and the derivative of 7.23 with respect to x gives
"\2'T'
0 = k - ~ ~ - + i x ( ( - B 2/2/j.)(dp/dz)(\ ~ 2,x/B 2) Y (7.25)
dx
The temperature profile is obtained by integrating twice using the boundary con
ditions T = T0 at x = ± B
. 4
rp _ T* 2 I M '; max X
/-/() + (7.26)
3k B
Flow in the Kidney. The flow in the kidney can be idealized to the flow between
two horizontal flat plates where the plates are made of a porous material. Fluid flows
into the bulk flow on one side and out through the porous wall on the other side.
F IG U R E 7.3 Sketch of velocity and temperature profiles of a viscous fluid in laminar flow.
APPLICATIONS OF THE EQUATIONS OF CHANGE 7.9
In this case there is two-dimensional flow as shown in the sketch of Figure 7.4.
Selecting the v-axis as the direction of hulk flow, fluid is injected normal to the
bulk flow with a velocity V at x = B and is removed at the same rate at x = -B .
The continuity equation in rectangular coordinates for steady incompressible
flow o f a Newtonian fluid. Equation 6 .8, is
(7.28)
dx
(7.29)
d 2vy pv dvy 1 dP
(7.30)
dx2 dy dx p dz
x
A y
-B
F IG U R E 7.4 Sketch of flow between flat porous walls with fluid entering and leaving.
7.10 FLU ID FLO W HANDBOOK
(D 2 - m D )vy = k (7.31)
vv = C, + c 2e mx - — (7.32)
m
B2 dP 2{ 1 - exp\-pV (B - x)/i± ]}
[1 - x / B \ (7.33)
pV dz - exp\-pVB//i]}
The velocity can be written in terms of the sum of the two components as
where i and j are the unit vectors of the x and y axis. Substituting the above becomes
\ = V\ + vvj (7.35)
e — = -V • pv0 (7.36)
ot
APPLICATIO N S OF TH E EQUATIONS OF CHANGE 7.11
Integrating gives
a
v0 = ------ (VP - pg) (7.39)
where v0 is the superficial velocity (= v,e). This simplifies to the following for
one-dimensional horizontal flow
a dP
v0 = --------- — = v,e (7.40)
M dz
To obtain the pressure distribution for steady flow (W = constant), Equations 7.35
and 7.37 are combined with the ideal gas law v = WRT/PM and rearrangement gives
P pdp = ^ ™ \ p T dz (7 .4 1)
P0
1/2
2 WeR f T pi
P = P<t - dt (7.42)
aM J t0 (dt - dz)
_ dT d ( dT \ dP tn a a \
pCpv~— = — — + V -— (7.44
1 p z dz dz\ dz I ' dz
At this point one must resort to a numerical solution of Equations 7.42 and 7.44
simultaneously as they are coupled.
To continue and obtain an approximate but analytical solution, constant ther
mal conductivity and viscosity are assumed, and the work term is neglected
which is small compared to the convective term. The energy equation for the gas
combined with W = pvz, mass flux of gas in the pores from Equation 7.38 gives
C„W— = (7.45)
' dz ' d z2
The energy equation for the porous medium with constant thermal conductivity.
Equation 6.60 simplifies to
0 =k S (7.46)
dz
Assuming the gas and solid are in intimate contact such that at any cross-section
both are at the same temperature, Equations 7.45 and 7.46 can be combined to give
e C p W ^ ~ = [ * „ + ( 1 - e ) y 4 t ( 7 4 7 )
dz dz
d ^ _ m c J L dL __ o
dz2 ke dz
To integrate the above for variable properties, Cp and ke, a numerical solu
tion would be required. For constant properties the solution o f the above is
T -T n \ - e NZIL
^ - - — ------ (7 49)
T,L -~ Ti n0 '1 - e N
where N = WC,,eL/ke. Using the above with Equation 7.42 the pressure distribu
tion could be obtained. These manipulations will be left for the reader.
APPLICA TIO N S OF TH E EQUATIONS OF CHANG E 7.13
FIGURE 7.5 Sketch of flow through a porous medium subjected to a temperature gradient.
Many times the situation to be modeled is so complicated that the equations are nearly
impossible to solve and dimensional analysis has to be applied. The problem then
becomes one of determining the important variables of the system and formulating
the dimensionless groups. This can be accomplished also by non-dimensionalizing
the equations of change, and confidence can be placed in the results obtained by this
method that only these variables need be considered.
The approach consists of selecting characteristic dimensions o f the system and
then using these to non-dimensionalize the equations of change [2]. This will be
illustrated for several cases using the equations of change.
the equations of continuity and motion for constant fluid properties, the results in
terms of the dimensionless variables are
V :* * v* = 0 (7.50)
Dv* - _ v * ^ ^ y*
• p* + -------- + g/g (7.51)
Dr* Re Fr
V tV
y p* = (P ~ P0) t* =
“ ~V pv2 D
X
y* = y -7* z
x* <. —-
~ D D D
D
v * = DV V * 2 = D 2V 2
Dr* { -
Reynolds Number, Re =
V2
Froude Number, Fr =
ZD
General Energy Equation. For problems involving heat transfer, the general energy
equation can be non-dimensionalized as was done in the case of the equations of
continuity and motion. Employing the same dimensionless variables only a dimen
sionless temperature, 7*, need be added.
7-* _ T~ T0
T t-T o
where T\ and T0 are characteristic temperatures of the system. T0 might be the tern-
perature of the air far from the wing of a plane in supersonic flight having a leading
APPLICA TIO N S OF THE EQUATIONS OF CHANGE 7.15
edge temperature T {. In this case V could be the velocity of the airplane and L the
width o f the wing. Using these characteristic dimensions to non-dimensionalize the
general energy equation, the result in terms of the dimensionless variables for a fluid
with constant properties is
where is given by the last term of Equation 6.56 written in terms of the dimen
sionless variables. The additional dimensionless groups are
Prandtl Number, Pr =
k
la V 2
Brinkman Number, Br =
Ty~T<0
DT*
R eP r------ = V *27^ + Br<J>* (7.53)
Dt*
Thus, if the Brinkman Number is small, the viscous dissipation term will not be
important. If the Reynolds Num ber is small also, then heat transfer by conduction
will predominate as in a slowly flowing liquid of high thermal conductivity, such
as a liquid metal.
A more realistic case for the flow past a wing is the non-dimensional form of
the energy equation for the steady flow of an ideal gas. The general energy equa
tion for no viscous dissipation simplifies to
pC v\ • V 7 = kV 2T - P V • v (7.54)
V* 2r*
v* • V*T* = ------------- (8 - 1)Ma (v* • V*/?* —p*V* • v*) (7.55)
RePr
7 .1 6 FLU ID FLO W HANDBOOK
where 8 = Cp/Cv and M(l is the Mach Number defined as the ratio of the fluid
velocity to that of the velocity of sound in the fluid. For an ideal gas the velocity
of sound Vy = (■8ZRT/M )W2. For large values of Mach Number the term involving
the pressure (second on right) becomes significant. This is demonstrated each
time a space vehicle re-enters the atmosphere.
In this chapter the uses of the equations of change have only been introduced. The
applications shown are only a minute fraction of the wealth ot technology that is
available. In fact, many books have been written on the topics discussed, and in
this brief discourse, only a glimpse can be given of the material that is available.
Consequently, it was the purpose here to present the derivation of equations of
change in a relatively general form so that an understanding could be obtained for
their intelligent use. Then, having a variety of forms of the equations from which
to select, illustrations were given to show how to simplify the general form and
obtain meaningful results for the particular system. To put the topic in the proper
perspective, the following outlines the general areas that subdivide the solutions
o f the general equations of change.
L am inar Flow. The equations of change are applicable to laminar flow. This is
the case where fluid layers slide over adjacent layers with transport of momentum
and energy by molecular rather than bulk movement of fluid. There are many
outstanding references that deal with this topic, as Leal (3) for example.
Turbulent Flow. In laminar flow, if a small disturbance is introduced into the flow,
it will decay. However if the flow velocity is increased past a certain value of the
Reynolds Number, a small disturbance will grow and the flow passage will be filled
with eddies in random motion. The equations o f change apply at any instant to the
flow but none but the most simple solutions to the equations o f change can be
attained. The difficulty lies in the fact that for one-dimensional mean flow, for exam
ple, the instantaneous velocity at any point has three non-zero components that are
time varying. Two approaches are being used to describe turbulence, one involves
measuring the correlation between two points in a turbulent field and relating this
with the equations of change. The other involves using average properties of the
flow that are not only a function of the material flowing but position in the flow.
These approaches are the subject of the book by Pope and have been discussed in
detail by Deissler and Holmes [4, 5, 6 J. A summary is given by Bird et al. [7].
N on-N ew tonian Flow. This subject has been mentioned briefly and typical exam
ples of this type of flow were given. A two volume series on this subject by Bird
et al. is comprehensive, and polymer processing is emphasized by Grulke [8,9 , 10|.
APPLICATIO N S O F TH E EQUATIONS OF CHANGE 7.17
Boundary Ixiyer Flow. The classical reference in this field is by Schlichting, and
boundary layer flow is concerned with flow where the change in velocity is con
fined to a very thin layer near the wall [11]. This occurs commonly in flow around
streamlined objects, particularly airfoils. By an order analysis, considering each
term in the equations of change simplifications can be obtained to give a set of
partial differential equations that are solvable and describe the flow [11]. The solu
tions to these equations have found wide applications in describing not only fluid
dynamics but heat and mass transfer also.
Flow Through Porous Media. Darcy’s law which corresponds to the equation of
motion for flow in a porous medium was presented in a previous example. The appli
cation o f D arcy’s law finds wide use in modeling oil and gas reservoirs and exten
sive numerical solutions of three dimensional, nonisothermal flow have been made.
The standard text for this subject are those by Greenckorn, Bear, and Holzbecker
[13, 14, 15].
M ultiphase Flow. This subject can be subdivided into gas-liquid and gas-solid
flow. Gas-solid flow is usually thought of in terms of fluidized beds which have
found extensive use as chemical reactors. This field is covered by Fan [16]. Gas-
liquid flow finds applications from flow of crude-gas mixtures from off-shore
wells to cooling nuclear reactors. The lack of an accurate description of the gas-
liquid interface has prevented progress in the description past one-dimensional
models. The subject has been summarized by Levy [ 17].
7.6 SUMMARY
In this chapter the use of the equations of change were illustrated by obtaining the
differential equations for several flow systems. Then the solutions were obtained
for some more complex systems such as ones that involved simultaneous solu
tions of the equations of motion and energy and ones that involve non-Newtonian
flow and two-dimensional flow. The chapter was closed with a brief description
of the important topics in fluid dynamics giving current texts on these subjects.
7 .18 FLU ID FLO W HANDBOOK
REFERENCES
1. Bird, R.B., W.E. Stewart, and E.N. Lightfoot. I960. Transport Phenomena. John
W iley and Sons Inc., New York.
2. Ibid. p. 107 and 339.
3. Leal, G. L., 1992. Lam inar Flow and C onvective Transport Processes. Butterworth-
Heinem ann, Boston.
4. Pope, S. B., 2000. Turbulent Flows. C am bridge U niversity Press, New York.
5. D eissler, R. G., 1998, Turbulent Fluid Motion. T aylor and Francis, Philadelphia.
6. H olm es, P, J. L. Lumley and G. Berkosz, 1998. Turbulence, Coherent Structures,
D ynam ical Systems and Symmetry. C am bridge University Press, New York.
7. Bird, R.B. et al. op. cit. Chapters 5 and 12.
8. Bird, R. B., R. C. Arm strong, and Ole H assager, 1987. D ynam ics o f Polymer Liq
uids, Fluid M echanics, Vol. I, 2nd ed. John W iley and Sons Inc., New York.
9. Bird, R. B., R. C. Arm strong, and Ole Hassager, 1987. D ynam ics o f Polymer Liq
uids, K inetic Theory, Vol. II, 2nd ed. John W iley and Sons Inc., New York.
10. G rulke, E. A., 1994. Polym er Process Engineering. Prentice-H all, Englewood
C liffs, NJ.
11. Schlichting, H.5., 1955. Boundary Layer Theory. M cG raw -H ill Inc., New York.
12. Anderson, J. D., 1990. M odem Compressible Flow , 2nd ed. M cGraw-Hill, New York.
13. G reenkorn,R. A., 1983. Flow Phenom ena in Porous M edia. M arcel Dekker.
14. Bear, J. 1988. Dynamics o f Fluids in Porous Media. Dover Publishing Co., New' York.
15. Holzbecker, E. O. and E. Holzbecher, 1998. M odeling D ensity-D raw n Flow in
Porous Media. Springer Verlag, New York.
16. Fan, L-S., 1989. G as-Solid Fluidization Engineering. B utterw orth, Stoneham, MA.
17. Levy, Salomon, 1999, Two-Phase Flow in C om plex System s. John W iley and Sons
Inc., New York.
_________ CHAPTER 8_________
INCOMPRESSIBLE FLOW
The Bernoulli’s equation derived in Chapter 4 is the most useful single equation in
fluid mechanics especially for incompressible flow calculations. Here the incom
pressible flow refers to a flow in which the changes in density are unimportant,
rather than o f an incompressible fluid. In general, all steady flows of liquids and
most steady flows of gases at low velocities (v < 300 ft/s) may be considered incom
pressible, whereas some unsteady flows o f liquids, and all flows of gases at high
velocities (v > 300 ft/s) may not be considered incompressible.
In incompressible flow calculations, the Bernoulli’s equation has the follow
ing form:
( 8 . 1)
1. The system is under steady state and steady one-dimensional flow opera
tion (e.g., flow in a pipe);
2. The density of the fluid is constant during flow;
3. Electrostatic, magnetic, and surface energies are negligible; and
4. The acceleration of gravity is constant.
8.1
8.2 FLU ID FLO W HANDBOOK
Note that the term “F ’ in Equation 8.1 represents the friction loss which, accord
ing to the second law of thermodynamics, has a value o f zero for frictionless flows
and is positive for all real flows. In many occasions, it is convenient to divide both
sides of Equation 8.1 by the gravitational acceleration (g) to have
( 8 .2 )
When a pump is involved and the friction loss inside the pump (Fpump) is con
sidered, then Equation 8.3a may be expressed as
The fluid-wall friction loss (Fw) appearing in Equation 8.3 can be estimated based
on the following equations depending on the flow regimes, i.e., laminar or turbu
lent. They are described below.
_ 128qL(/j/p) (R
w" (/rD4) 1 ^
, _ 32vL(jU/p)
VV . 2\ (8.5)
( ttD")
IN C O M P R ESSIB LE FLOW 8.3
( 8 .6 )
Besides Figure 8.1, the Moody friction factor can also be estim ated from any
of the two equations proposed by Colebrook [2] and Wood [3], respectively. The
two equations are given below:
Colebrook Equation:
(£/D) 1.255
0.5 = - 4 log + (8.7)
(f) 3.7 (SKf05)
Wood's Approximation:
f = a + bCSty* ( 8 .8 )
Surface roughness
Material e, ft e\ in
Drawn tubing (brass, lead, glass, etc.) 0.000005 0.00006
Commercial steel or wrought iron 0.00015 0.0018
Asphalted cast iron 0.0004 0.0048
Galvanized iron 0.0005 0.006
Cast iron 0.00085 0.010
Wood stave 0.0006-0.(X)3 0.0072-0.036
Concrete 0.001-0.01 0.012-0.12
Riveted steels 0.003-0.03 0.036-0.36
From Moody [ I /.
(j/i ssouqSnoj 3Ai^joy £
T3
<D
o
3
XJ
£
Q.
<u
OC
5
O'
<N
r-
vo
vb
sO
I
<0
£
o
c=
8.
$ £
I J
£3 .a
,C
c U.
>>
"O
E
I
8.
Ou
u
a
o
Q.
O u.
*52 £
c
•C a
U- W
^3 o
00 C
Wo
CC
3 .2
iz/rAY.aM\) a
/ JOJO BJ U O I^O IJJ NM ret)
U. D.
8 .4
IN C O M P R ES SIB LE FLOW 8.5
where
+ 0.1325
£ \ °. 134
It is worth pointing out that, the Colebrook equation (Equation 8.7) predicts
exactly the values o f Moody friction factor displayed in Figure 8.1. The W ood’s
approximation (Equation 8.8) tends to overestimate the Moody friction factor by
approximately 5 percent.
It should be pointed out that Equation 8.6 , although designed for estimating
friction loss in turbulent flows, can also be used in laminar flow calculations. In
such applications, the roughness factor is usually ignored and the friction factor
can be calculated by:
Note that the coupling of Equations 8.6 and 8.9 is exactly identical to either Equa
tion 8.4 or Equation 8.5.
It should also be pointed out that an alternative expression for the estimation
o f friction loss (Fw) especially popular in civil engineering practice is the use of
Fanning friction factor ( f ) defined as: [1,4]
f' = 4f ( 8. 10)
With the Fanning friction factor, Equation 8.5 for friction loss estimation becomes:
( 8 . 11)
The corresponding Fanning friction factor for laminar flow applications, i.e.,
Equation 8.9, is then
( 8 .1 2 )
8.6 FLU ID FLOW HANDBOOK
The fluid-wall friction loss discussed above, i.e., Fw, is normally the most signif
icant contributor to the overall friction loss in fluid flows. However, in many flow
systems we must also take into account the losses due to fittings such as valves,
elbows, etc. One popular way to estimate the friction loss due to fittings, i.e., F f,
is to estimate the “equivalent length” of the fitting and to insert the estimated
equivalent length into Equation 8.6 to calculate the friction loss. This practice can
be expressed as:
(8.13)
In the practice, the equivalent length (Leq) can be estimated by the following
expression:
/\v + ^ = 4 f ( M f c ) (8.15)
where
8.5 CONTRACTION/ENLARGEMENT
FRICTION LOSS (FC/FE)
The contraction/enlargement loss refers to the flow conditions where there is a sud
den change in pipe diam eter or where the fluid enters a pipe from a tank or leaves
a pipe into a tank. The losses due to contraction and enlargement mostly come
from the changes in fluid kinetic energy and the creation of turbulent eddies. In
the practice, the losses are calculated by the following two expressions:
IN C O M PRESSIBLE FLOW 8.7
Equivalent length
Type of fitting L/D (dimensionless)
and
* = (8 I 8 )
(L adj '
F = 4f + Kc + Kc (8.19)
D
8.8 FLU ID FLO W HANDBOOK
Inserting this equation into Equations 8.1 and 8.2, the two equations
become:
Conventional Form:
W
A |E + gZ + - ( 8 .2 0 )
m
Head Form:
W
Al — + Z + (8.21)
2 g, mg
The two equations, Equations 8.20 and 8.21, represent the complete work
ing form of the Bernoulli’s equation for incompressible, turbulent, one
dimensional flow-in-pipes calculations.
Conventional Form:
w 32vL(/i/p)
( 8 . 22 )
m ( ttD2)
Head Form:
1L 32vL(jU/p)
AI-E- + Z + — (8.23)
mg (*D 2g)
The Darcy’s formula given below represents the simplified versions of the Bernoulli’s
equation. They only calculate the pressure loss (A'P) due to the fluid-wall friction
loss based on Fanning friction factor ( f ). Note that this pressure loss is the same in
IN C O M P R ESSIB LE FLOW 8.9
a sloping, vertical, or horizontal pipe. Depending on the variables used, the Darcy’s
formula have the following forms with various units involved [5 ]:
The corresponding head loss forms for the above D arcy’s Equations IP
through 8P can be expressed as: (where the head loss hL is defined to be
A 'p/pg)
(jU L Q)
Darcy 3PL. A 'P = 0.000 273 d4 (8.26c)
The corresponding head loss forms for the above Darcy 1PL through
5PL formula can be expressed as: (Again, the head loss hL is defined to be
A'p/pg.)
For the flow in non-circular pipe, the above equations for flow in circular pipes
can still be used using the concept of Hydraulic Radius (HR), defined as
Therefore, for non-circular pipe applications, an equivalent diam eter (Deq) ex
pressed as
Deq = 4 HR (8.30)
is used to substitute the diameter (D) appearing in all the equations for circular pipes.
The rest of the calculations remain the same.
The economic pipe diameter is the one that minimizes the total annual costs for
transporting a given amount of a specific fluid. Based on a simple economic analy
sis, the following equation has been derived for the economic pipe diam eter [6 ]
(8 .3 1)
This equation indicates that the economic pipe diameter is mainly a function of mass
flow rate (m), fluid density (p), and the economic terms (PC, CC and PP). The diam
eter is independent of the length of transportation.
In practice, a popular alternative for the determination of economic diameter
(Dec0n) is to calculate economic velocity (vecon) instead. It is defined as
Substituting for the economic diam eter from Equation 8.31 yields
Note that, since the friction factor does not vary significantly, the above equation
indicates that the economic velocity is mainly a function of fluid density.
Boucher and Alves [7] have provided a table (Table 8.3) to summarize the eco
nomic velocity as a function of fluid density under turbulent flow in schedule 40, car
bon steel pipes. The values in Table 8.3 indicate that for a fluid with p = 100 lbm/ft3,
the economic velocity is 5 .1 ft/s; however, for p = 0 .1 lbm/ft3, the economic veloc
ity is 39.0 ft/s. Since the density of water is around 62.3 lbm/ft3, the economic veloc
ity is expected to be around 6 ft/s. It is worth pointing out that, although the
economic parameters in Equations 8 .3 1 and 8.33 affect the economic velocity, the
effect is not expected to be significant due to the 1/6 power effect appearing in
Equation 8 .31.
The complete Bernoulli's equations, i.e., Equations 8.20 through 8.23, are nor
mally the equations for solving fluid flow problems. However, when the kinetic
and potential energy terms are not significant and when pump work is not
involved, the Darcy’s formula, i.e.. Equations 8.24 through 8.27, can be used. In
the applications, most of the fluid flow problems can be classified into three types:
The Type l problems do not require a trial and error procedure and can be solved in
a relatively straightforward manner. The Type 2 and Type 3 problems, however,
cannot be solved directly and require a trial and error procedure simply because the
friction factor cannot be calculated without the knowing o f velocity (v) and pipe
diam eter (D). The example problems given below will demonstrate all three types
o f problem solving procedures.
100 1 5.1
50 1 6.2
10 1 10.1
1 0.02 19.5
0.1 0.02 39
0.01 0.02 78
This section contains example calculations to demonstrate the use of the Bernoulli's
equations and D arcy’s formula for fluid mechanics calculations.
Data: For water at 68°F, p = 62.3 lbm/fr ? and p = 6.73 x 10~4 lbm/( ft sec)
Solution
The D arcy’s formula are perfect choices for this application. First is to
determine if the flow is in laminar or turbulent condition through the calcu
lation of flow rate and the Reynolds number:
q ^ [0.5/(7.48 x 60)]
= 0.0217 (ft/sec)
A " [3.1416 x (3.068/12)2/4]
^ _ (p D v) _ [62.3 x (3.068/12) x 0.02171
n (6.73 x 10“4)
= 513.6 < 2000 (Lam inar Flow)
Since the flow is under laminar flow condition, Equations 8.26 and 8.27
can be used to estimate pressure loss (A' P) and head loss (hL):
Using Equation 8.26a
tuLQ )
A P = 0.000 2 7 3 — -:.—
(6.73 x 10~4 x 1 0 0 0 x 0 .5 )
= 0.000 273
(3.068)4
= 1.037 x 10“6 (psi)
8 .14 FLU ID FLO W HANDBOOK
hL = 0.0962 (A/ V)
d p
. (6.73 x 1 0 4 x 1000 x 0.0217)
— u .u y o z -------------------- ^
[(3.068) x 62.3]
= 2.396 x 10“6 (ft)
O iL Q )
h, = 0.0393
d 4p
(6.73 x 10“4 x 1 0 0 0 x 0 .5 )
= 00393 [(3.068)4 x 62.3]
Data: For water at 68<)F, p = 62.3 lbm/ft3 and fj = 6.73 x 10 4 lbm/(ft sec)
Solution
The Darcy formula again are perfect choices for this application. First is to
determine if the flow is in laminar or turbulent condition through the calcu
lation o f flow rate and the Reynolds number:
v = — = [ 5 0 0 /^ 4 8 ^ 6 0 ) ] ^ — = 21.7 (ft/sec)
-
A [3.1416 x (3.068/12)2/4J
^ (p D v) __ [62.3 x (3.068/12) x 21.7]
li (6.73 x lO " 4)
= 5.14 x 105 > 2000 (Laminar Flow)
Since the flow is under turbulent flow condition, Equations 8.24 and 8.25
can be used to estimate pressure loss (A' P) and head loss (hL):
A'P = 0 .0 0 1 2 9 4 — L p v r )
d
(0 .0 184 x 1 0 0 0 x 6 2 .3 x 21.72)
= 0 .0 0 1 294
3.068
= 227 (psi)
(f' L p Q 2)
A'P = 0.000 216
d5
(0.0184 x 1000 x 62.3 x 5002)
= 0.000 216
(3.068)-
= 227 (psi)
(f' L v2)
hL = 0.1863
d
(0.0184 x 1000 x 21.72)
= 0.1863
(3.068)
= 526 (ft)
Data: For the oil, Specific G ravity (SG) = 0.9; kinematic viscosity = 10 cSt
(assuming commercial steel with s ' = 0.0018 in)
Solution
The solution procedure is very similar to that demonstrated in Example 8.2
shown above. First is to determ ine if the flow is in laminar or turbulent con
dition through the calculation of flow rate and the Reynolds number: ^
8.1 6 FLU ID FLO W HANDBOOK
Given:
Given:
fi - p x v
= 62.43 x 10 x 1.073 x 10“s
=
= 0.0067 [lbm/(ft s)]
Given:
Q = 120 x — = 84 (gal/min)
60
u 184/(7.48 x 6 0 )]
v = — = — 1----- --------------- 5— = 3.65 (ft/sec)
A [3.1416(3.068/1
(p D v) [56.2 x (3.068/12) x 3.65]
p “ (0.0067)
= 7826 > 2000 (Turbulent Flow)
Since the flow is under turbulent flow condition, Equations 8.24 and 8.25
can be used to estimate pressure loss (A'P) and head loss (hL):
Using Equation 8.24a
(f' L v2)
hL = 0.1863
d
(0.034 x 1000 x 3.652)
= 0.1863
(3.068)
= 27.5 (ft)
Solution
Because the problem does not involve the changes in kinetic and potential
energy, the Darcy’s formula are again the choices for determining the pres
sure loss. The first step is to determine the Reynolds number that requires
the air density that can be estimated by the ideal gas law, i.e.,
_ m _ MP
P ~ V “ RT
[(28.84)(14.7)]
[(10.73)(68 + 460)]
= 0.075 (lbm/ft3)
Then
v= q = [(80/60)]
A {3.1416 x [(3.068/12)2/4 ]}
= 26.0 (ft/sec)
8 .18 F L U ID FLO W HANDBOOK
Since the flow is under turbulent flow condition, Equation 8.24 can be u;ed
to estimate pressure loss (A'P).
(f' L p q 2)
A'P = 43.5
d5
[0.027 x 1000 x 0.075 x 80/60)2]
= 43.5
3.068s
= 0.58 (psi)
= 15.9 (in H 20 )
Data: For air at 68°F, p = 0.018 cP = 1.21 x 1 0 '5 lbm/(ft sec) and
p = 0.075 lbm(ft3
Solution
There are two considerations in solving the problem: (1) The D arcy’s fcr-
mula can be used since the flow neither involves the changes in kinetic ind
potential energy nor the compressor work; and (2) The method of hydralic
radius (HR) applies since the flow is in a non-circular duct. Therefore, tie
IN C O M P R ESSIB LE FLOW 8 .19
HR = ----- - - - -L5)-----
[2 x (1.5 + 1.0)]
= 0.3 (ft)
Since the flow is under turbulent flow condition, Equation 8.24 can be used
to estimate pressure loss (A'P).
Solution
1. Power Requirement
Unlike the previous examples, the complete Bernoulli's equation, i.e.,
Equation 8.20 (if turbulent flow) or Equation 8.22 (if laminar flow), must
be used since this problem involves a pump and a change in potential
energy. We first calculate the flow rate and the Reynolds number to deter
mine the flow regime:
[200/(7.48 x 60)]
A {3.1416 x [(3.068/12) 2/41)
= 8.68 (ft/sec)
( p D v) _ [62.3 x (3.068/12) x 8.68]
li (6.73 x 10"5)
= 2.06 x 10 5 > 2000 (Turbulent Flow)
Since the flow is turbulent, Equation 8.20 is used. It has the form
w
y | (8.20)
m 4 f (~ ~ D / + K c + K e
At this point, it is worth pointing out that the friction loss term, i.e.,
[4f(Ladj/D ) + Kc + ATe](v 2/2) appearing in Equation 8.20, should include all
three segments of flow between “surface 1” and “surface 2 ,” namely, flow
in the first reservoir, flow in the 3" pipe, and flow in the second reservoir.
However, since the fluid velocities in the reservoirs are near zero, the fric
tion loss term in those two segments of the flow is extremely small and can
be neglected. Therefore, only the friction loss in the 3' pipe segment is
IN C O M PR ESSIBLE FLOW 8.21
W adj v
g( 20) = -------- 4f + K c + Kc
m D T
(8.34)
W 3000 8 . 68 -
4f + 0.5 + 1.0
m (3.068/12)
0.0018
= 0.0006
l 3.068 ,
w (200/7.48) ^ n/11
m = p x V = 62.3 x -------------- = 27.8 (lbm/s)
60
Inserting f and m into Equation 8.31 and with g = 32.174 ft/s2, Equation
8.34 becomes
W 3000 8.68‘
32.174(20) = - 4(0.0048) + 0.5 + 1.0
m (3.068/12)
or
The above expression points out an interesting observation that, for this
type of problem, the fluid-wall friction loss term dominates the total fric
tion loss for the process, e.g., 11734 for Fw as compared to 0.5 for Fc and
1.0 for Fe.
After simplification, the above equation yields
F L U ID FLO W HANDBOOK
That is
-W = 27.8(9131)
= 253842 (lbmft 2/s3)
, = = 789Q
32.174
7890= 14.3
= ------ 1 ^ (hp) x
550 F
Note that since friction loss inside the pump was not involved in the above
calculation, the obtained result represents the pump power required for an
ideal pump, i.e., 100% pump efficiency. However, if a pump efficiency of
rf = 80% is assumed, then the real pump power requirement will be
—W re a l = ( - W ' d cu l/ m _ 9 ^ 1 _ j ( f t 2^2x
m r] 0.8
Note that the difference between (-W 'eaI) and (-W [deai) is indeed the fric
tion loss inside the pump (Fpump), i.e.,
A ^ + gZ + y J = - ^ - F pump (8.35)
IN C O M P R ES SIB LE FLOW 8 .23
In the application, the terms in Equation 8.35 will have the following
expressions:
AI —| is to be determined
\P)
A(gZ) = g (z 2 - Zi) = 0 (assuming the pump inlet and outlet are at the
same elevation)
/ v2\ ( \ \ — v2)
A I — I = ----- ------ = 0 (since the velocities at pump inlet and outlet are the
same)
A ( E \ = _Wideai (g36)
\P m
or
A(p) = p ( - ^ i )
This is
or
Note that if W rea| is used in Equation 8.35, then Fpump = 2291 ft 2/s 2 and,
instead o f Equation 8.36, the following equation will be yielded from
Equation 8.32:
This gives
A (p )= p ( - ^ f - ^ p)
This gives
= 17665 (lb,/ft2)
or
17665
A(P) = = 122.7 (psi)
144
This is the same result as that calculated above based on - W idea| and with
no pump loss, i.e., F pump = 0 .
Solution
W
m
or
8 .682
14873 - 6435 = - —— - 4f - — + 0.5 + 1.0 (8.38)
m (3.068/12) 2
That is
W
----- = 8438 + 8488 = 16926 (ft2/s2)
m
or
-W = 27.8(16926)
= 470543 (lbmft2/s 3)
In terms of other units
470543
- W' = = 14625 (ft ibf/s)
32.174 1
= ^ = 2 6 .6 ( h p )
550 H
Note that the above result is the ideal pump power requirement.
a (p) =“ ^i r 1 (8,36)
or
A(p)=p(~ ^ ir)
= 62.3(16926) = 1.054 x 106 ( l b j f t s2)
This gives
_ 1 05 4 x 106 = 2
(32.174) 1
or
32775
A(P) = ------- = 227.6 (psi)
144 F
Solution
Unlike all the example problems described previously, this is a typical
Type 2 problem and requires a trial and error procedure. Assuming the
flow is in the turbulent flow regime, Equation 8.20 given below is used:
W V“
m 4 ffr? + kc+ Kc
T
( 8 .20 )
In this application, “surface 1” is the top surface of the storage tank, and
“surface 2” is the top surface o f the truck tank. With the selection, the fol
lowing expressions are obtained
= 100 + 6 x 20 x + 2 x 13 x
= 112.8 (ft)
K c = 0.5 (flow from the storage tank into the pipe)
Kc = 1.0 (flow from the pipe into the truck tank)
With all the above expressions, Equation 8.20 has the form
112.8
32.174(-30) = - 4f + 0.5 + 1.0
(1.049/12)
Note that the equation has two unknowns, f and v. After simplification, the above
equation becomes
IN C O M P R ESSIB LE FLOW 8.27
A “Trial & Error" procedure must be involved to solve the equation. With (6*'/d)
for the pipe being
(£ \ M 018
\ d / 1.049
we may first assume the flow is highly turbulent, i.e., 9^ > 2 x 106, then accord
ing to Figure 8.1, the friction factor will have a value f=0.0057. Inserting this value
to Equation 8.39 yields
Solving for v
v = 7.9 (ft/s)
Note that the Reynolds number is less than the assumed value and therefore the
first trial is not successful. Next try f = 0.0058, Equation 8.39 yields
Solving for v
V = 7.8 (ft/s)
Since at this Reynolds number and with (e/D ) = 0.0017, the friction factor is approx
imately equal to the assumed value, i.e., f = 0.0058, the trial is successful and the
8.28 FLU ID FLO W HANDBOOK
obtained v = 7.8 ft/s is the result. At this velocity, the volumetric flow rate is
Q = (v)(A)(7.48)(60)
(1.049/12)2 (7.48)(60)
= 7.86 x (3.1416)
= 21.0 (gpm)
Data: For water at 68°F, p = 62.3 lbm/ft3 and fi = 6.73 x 10 4 lbm/(ft sec)
Solution
This is another Type 2 problem requiring a “Trial & Error” procedure to
solve the flow rate. Equation 8.20 again is used to solve the problem
La d j' v
A ( £ + gZ + y = - - 4f + Kc + Ke (8.20)
ip 2 m D ~2
In this application, “surface 1” is the top surface of the water, and “surface 2” is
the outlet o f the pipe. With that, the following expressions are obtained
v2 11.6
+ 32.174 (-5 0 ) = 4f + 0.5
(6.065/12)
Note that the two velocity terms, i.e., v2 and v, are the same and can be combined.
The above equation thus has two unknowns, f and v. After simplification, the above
equation becomes
or
A trial and error procedure, again, must be involved to solve the equation. With
(£'/d) for the pipe being
£ = ^ 0 1 8 = 0 .000264
d / 6.065
we again assume the flow is highly turbulent, i.e., 9? > 1 x 107, then according to
Figure 8.1, the friction factor will have a value f = 0.0036. Inserting this value to
Equation 8.40 yields
-1 6 0 9 = - ( 9 1 .8f x 0.0036 + l - 5 ) ( y
Solving for v
v = 41.9 (ft/s)
Since the Reynolds number is larger than 1 x 107, the assumption is correct and
therefore the trial is successful. At this velocity, the volumetric flow rate is
Q = (v)(A)(7.48)(60)
Data: For water at 68°F, p = 62.3 lbm/ftJ and p = 6.73 x 10~4 lbm/(ft sec)
Solution
The problem is to estimate the pipe diameter and, therefore, is a Type 3
problem that requires a “Trial & Error” procedure. Again, Equation 8.20 is
the starting equation, i.e.,
(8.20)
In this application, “surface 1” is the inlet surface, and “surface 2” is the out
let surface. With the conditions given, the following expressions are obtained
= -2231 (ft2/s2)
A(gZ) = g(z2 - Z|) = g(0) = 0 (horizontal flow)
In addition
Q 500 .42
v=
[ A(7.48)(60) ] [3. l4 l6 (D /2 )2(7.48)(60)] D;
or
v2 =
2.02/D
-2231 = - 4f
(?:
Note that the equation has two unknowns, f and D. After simplification, the above
equation becomes
2020f
-2231 = - (8.41)
D5
A “Trial & Error” procedure, again, must be involved to solve the equation. Since
(£/D ) for the pipe is a function of D also, i.e.,
I ± \ _ 0.00015
d )" D
we will first try a 3" schedule 40 pipe, i.e., D = (3.068/12) = 0.256 ft. With the
assumed D
e \ _ 0.00015
= 0.0006
, D / ~ 0.256
and
1.42 _ 1.42
v= = 21.7 (ft/s)
D2 (0.256Y
8.32 FLU ID FLOW HANDBOOK
f = 0.0044
_2^ 3 | _ (-2020)(0.0044)
D5
Solving for D
D = 0.33 (ft)
The calculated D (0.33 ft) apparently is greater than the assumed D, which is
0.256 ft. The trial is not successful.
A 4 in schedule 40 pipe is then tried following the same procedure where
I e\ 0.00015
= 0.00045
,D / 0.336
and
f = 0.0042
2231 = (-2020X0.0044)
D5
Solving for D
D = 0.33 (ft)
The calculated D is close enough to the assumed D and the trial is successful.
The minimum pipe size for the design is therefore a 4 in schedule 40 commercial
steel pipe.
VARIABLES
Reynolds number
P density, lbm/ft3
n pump efficiency, %
e pipe roughness, ft
£' pipe roughness, in
viscosity, lbm/(ft s)
V kinematic viscosity, cSt
REFERENCES
1. M oody, L. W. 1944. “ Friction factors for Pipe Flow ," Trcms. A SM E , 66.
2. C olebrook, C. F. 1938. “Turbulent Flow in Pipes, with Particular Reference to the
T ransition Region between the Sm ooth and Rough Pipe Law s,” J. o f Inst. Civ. Eng.
1 1 ,1 3 3 -1 5 5 .
3. W ood, D. J. 1966. “An Explicit Friction Factor R elationship,” Civil E ngineering,
3 6 ,6 0 - 6 1 .
4. B oucher, D. F. and G. E. Alves. 1973. F luid an d Particle M echanics, in Chem ical
E ngineer's H andbook, 5th ed. R. H. Perry, C. H. C hilton, and S. D. Kirkpatrick,
eds., M cG raw -H ill, Inc., New York, pp 5 -2 2 .
5. C rane Technical Paper. No. 410, 1957, Flow o f Fluids through Valves, Fittings, and
Pipes. Crane Com pany, 475 N. G ary Ave., Carol Stream , 111.
6. Noel de Nevers. 1991. Fluid M echanics f o r C hem ical E ngineers, 2nd ed. M cGraw-
Hill, Inc., New York.
7. Boucher, D. F. and G. E. Alves. 1973. F luid and Particle M echanics, in Chem ical
E n g in eer's H andbook, 5th ed., R. H. Perry, C. H. C hilton, and S. D. Kirkpatrick,
eds., M cG raw -H ill, Inc., New York, pp 5-32.
__________CHAPTER 9__________
FLOW OF GASES IN PIPES
AND DUCTS
The flow of compressible fluids in pipes and ductlines is discussed in this chap
ter. The density of compressible fluids is a function of temperature and pressure,
as was detailed in Chapter 2. The gas density decreases as the pressure decreases—
at constant temperature— along the pipe flow path. The change in density will
change the fluid velocity and hence the acceleration term in the Bernoulli’s equa
tion has to be taken into account. Many methods have been developed to predict
the gas pressure drop in transmission and plant pipelines. A summary o f these
methods will be covered in this chapter. An accurate incremental computer algo
rithm, which is applicable to any gas flow case, is also detailed in this chapter.
The concept of the conservation of total energy, the Bernoulli equation, is also
applicable to the gas flow.
(9.1)
P 2 g
where
9.1
9.2 FLU ID FLOW HANDBOOK
Since the gas density is much smaller than liquid density, the Bernoulli equation
elevation term is relatively small and is ignored (at low pressure) in many of the gas
total pressure drop correlations. The following example illustrates the effect of the
elevation term in gas flow problems. Only the elevation term effect on the total pres
sure change will be considered, i.e. the frictional and acceleration terms are ignored
for the time being.
EXAMPLE 9.1
An air stream flo w s upward 4 inches (101.6 mm) schedule-40 pipe. The pres
sure at the discharge o f the blower is 18 psig (124.1 kpag). The pipe net eleva
tion change is 2 5 0 f t (76.2 m). Find the change in air pressure. Ignore the
frictional and acceleration effects. Temperature is constant at 77°F (25°C).
Solution
l80yl44 = ^ + 2 5 0 (9.2)
0.1646 0.1646
The pressure change due to elevation is (18.00 - 17.41) = 0.59 psig (4.06 kpag)
To compare the effect o f the elevation term in liquid flow, the 250 ft (76.2 m)
net elevation in a sim ilar water (density 61.4 lbm/ft3 (983.5 kg/m3)) flow case
will yield a pressure change of 106.6 psig (734.98 kpag). This what we meant by
declaring that the pressure change due to elevation in gas flow problem is rela
tively much smaller than it is in liquid flow.
FLO W O F GASES IN PIPES AND DUCTS 9.3
Many methods have been proposed to estimate the pressure change in plant and
transmission gas pipelines. A simplification assumption (or assumptions) is usu
ally included to avoid the complication of iterative and tedious solution steps.
The assumptions are mainly:
In many cases, one (or more) of the above assumptions may apply well. For
example, at low pressure, the compressibility factor approaches one. When the pres
sure drop is very small, the change in density is minimal and acceleration effects
may be dropped. The following is a summary of the proposed simplified methods; a
detailed discussion may be found in the Crane Technical Manual num ber 410 and
in section 17 o f the Engineering Reports of the Gas Processor Association [2, 5].
These equations were proposed by the Panhandle Eastern Pipe Line Company in
the early 1940s to predict the gas flow in transmission pipelines. Two versions
are available. The Panhandle A Equation is used to predict the gas flow rate in
smooth pipes in partially turbulent flow, while the Panhandle B Equation approx
imates the fully developed turbulent flow. Both of the A and B forms o f the Pan
handle equations are independent of the roughness of the pipe surface. However,
a pipe efficiency factor may be adjusted to account for the pipe roughness. An
average gas com pressibility factor is used; the acceleration and elevation terms
are ignored.
9.1.L I P anhandle A Equation. The gas flow rate in smooth pipes at relatively
partially developed turbulent flow is estimated by:
Q = 435.87
f T„) P P\-Pi d 2.6182
(9.3)
n0.853 j nr 'y
UJ m avg avg
where
When E = 1.0, Equation 9.3 reflects a smooth pipe. A value of 0.9 to 0.92 of
E has been used to apply Equation 9.3 to partially turbulent flow. The partially
turbulent flow is defined by
where
Re = 1.934 Q S/d
d Pipe inside diameter, in.
Re Reynolds number
0.5392
P - P2
r \ /2.6I82
Q = 20407.13 E (9.4)
S°'853Lm tuvg z avf>
where
p = - Pl + P2 - PA
3
P>+P,;
Equations 9.3 and 9.4 are best solved for the gas flow rate given pressure drop.
Solving the equations for the pressure drop given the flow rate is an iterative pro
cedure since the resulting rearrangement is implicit in terms of the compressibil
ity factor which is a function o f the outlet pressure.
Q 737 (9.5)
UJ
Again, Equation 9.5 does not account for the surface roughness, ignores the
elevation and acceleration effects, and assumes an average constant com pressibil
ity factor. Crane Technical Paper 410 suggests the following values for the effi
ciency factor:
E = 1.0 for brand new pipes without any fittings or change in pipe diameter
E = 0.95 for very good operating conditions
E = 0.92 for average operating conditions
E = 0.85 for unusually unfavorable operating conditions.
The equation was suggested by Weymouth [14] in 1912 for gas flow in short pipes.
V CL,
P ;-P ;
UJ S LmTavgZ avg ^
Unlike previous equations, the AGA equation accounts for the pipe roughness and
may be applied to both fully and partially turbulent flows. The AGA equation is
summarized as
0.5
Q (9.7)
where
Re
§TTf =4 log10 - 0.6 (9.9)
where
Crane Technical Manual number 410 [2] recommends using the Darcy equation to
calculate the pressure drop of a compressible How [2). The following conditions
should be observed:
1. When pressure drop is less than 10% of the inlet pressure, the gas density
calculated at the inlet or the outlet pipe conditions may be used and is
assumed to be constant over the entire pipe length.
2. For pressure drops larger than 10% but less than 40%, the average o f the
inlet and outlet conditions should be used.
3. For pressure drop larger than 40%, a correction factor, the net expansion
factor K, should be used to account for the change in fluid density due to
expansion. The Darcy equation becomes,
(9.10)
(9.11)
where
Table 9.1 lists the Y values recommended by the Crane Technical Paper as a func
tion of k = Cp/Cv (specific heat ratio) and K (pipe resistance), Table 9.1 applies
FLO W OF G A SES IN PIPES AND DUCTS 9.7
1. Find the total pipe resistance coefficient K. The resistance includes the pipe
length, any entrance or exit effects and any fittings or hand valves.
2. Calculate the pressure drop ratio &P/P\
3. If the calculated AP/Pi is below the table AP/P\ value for the calculated K,
flow is subsonic and Equations 9.10 and 9.11 may be used. However, if the
calculated pressure ratio exceeds the table value for the given pipe sonic
flow is encountered and the AP!P\ and Y should be adjusted to match the
table values. Table interpolation may be needed in this step.
Examples to illustrate the calculation procedure will follow at the end of this section.
Equations 9.10 and 9.11 are recommended for adiabatic flow condition, which
is more pertinent to fluid flow in short pipes. Crane Technical Paper recommends
the following equations for isothermal flow, which is more pertinent for flow in
long pipes:
i0.5
A'-p, rp ? -p r
w = 68.0 (9.12)
' L + 21og * i /
D P,
* =1.3 k =1.4
(C02, S 0 2,H20, H2S,NH3( (Air, H2, 0 2, N2, CO,
N20,C12,CH4,C2H2, C 2H4) NO, HCI)
K A/VP, Y K \P /P X Y
1.2 0.525 0.612 1.2 0.552 0.588
1.5 0.550 0.631 1.5 0.576 0.606
2.0 0.593 0.635 2.0 0.612 0.622
3.0 0.642 0.658 3.0 0.662 0.639
4.0 0.678 0.670 4.0 0.697 0.649
6.0 0.722 0.685 6.0 0.737 0.671
8.0 0.750 0.698 8.0 0.762 0.685
10.0 0.773 0.705 10.0 0.784 0.695
15.0 0.807 0.718 15.0 0.818 0.702
20.0 0.831 0.718 20.0 0.839 0.710
40.0 0.877 0.718 40.0 0.883 0.710
100 0.920 0.718 100 0.926 0.710
9.8 FLU ID FLO W HANDBOOK
or
0.5
/ \
(9.13)
where
Equations 9.12 and 9.13 assume isothermal and perfect gas flow.
EXAMPLE 9.2
Find the flo w rate o f a natural gas, average specific gravity o f 0.82, that could be
obtained if the pressure drop in the 14 miles (22.5 km) 12 inches (304.8 mm) I.D.
pipe is 420 psi (2895.8 kpa). The upstream pressure is 760 psia (5240.0 kpaa),
and the gas average temperature may be assumed to be 72°F (22.2°C).
Solution
1. Panhandle A Equation
Th = 520 R (288.88 K)
Ph = 14.73 psia (101.56 kpaa)
To find the average gas compressibility factor, we will need the gas critical
data. The AGA correlations may be used to find the natural gas critical
data. These correlations are explained in Chapter 2.
The AGA correlations are:
Pc = 6 9 0 - 3 1 .0 X S
Tc = 157.5 + 336.15 X 5
oo
d
o
k.
a,§•
CD "a
o
Q.
£
a 2
Q 0
1o
o c:
D.
CNJ
£
o
d
o
o
C
_o
"t/5
c
r3
a.
<u
Z
in cd in oo in m cd m ©V
cr> cd ^ m w
o °9 o o
o o as
O
9.9
10 FLU ID FLO W HANDBOOK
/ __ \
p =- P + P - PA
avg 3
Substituting the inlet and outlet pressure into the above equation
Natural gas reduced pressure and reduced temperature are then calculated
as:
For simplicity, the Papay’s correlation is used to estimate the gas average
compressibility factor. Any other compressibility factor correlation or any
other equation of state may also be used. Reader is referred to Chapter 2
for more discussion on estimation methods for the gas compressibility fac
tor. The Papay’s correlation is:
Substituting the gas reduced pressure and reduced temperature in the above
equation yields,
0.5392
( 520 V 760 - 340 2.6182
Q = 435.87 0.92 12
U 4 .7 3 ; 0.82°853;d4jc(72 + 460)^0.826
= 141 M M CFD (0.166 M M CM H)
The above flow rate is calculated at the base conditions of 14.73 psia
(101.56 kpaa) and 60°F (15.56°C).
FLO W OF GASES IN PIPES AND DUCTS 9.11
2. Panhandle B Equation
When Equation 9.5, the Panhandle B Equation, is used, the calculated How
rate is
0 .5 1
/ \l-02 760“ - 340‘
‘ 520 1 2.53
Q = 737 0.92 12
v
14.73 / J 0.82° 961 a 14.v(72 + 460)*0.826
= 137.77 M M CFD
3. AGA Equation
A pipe roughness, £, o f 0.0018 inches (0.0457 mm) is assumed. A fully tur
bulent flow is assum ed, hence, the transmission factor (1 //) is calculated
using Equation 9.8,
3.7*1
= 4\og 10 = 17.568
0.0018/12
The flow rate is then found by substituting the above into Equation 9.7,
0.5
520 7602 - 3 4 0 2 2.5
Q = 38.77 0.92 *17.568 12
14.73, 0.82 *14*(72 + 460)*0.826
= 1 14.78 M M CFD
If the flow is assumed for partially turbulent flow, then a trial and error
solution is used to find the transmission factor and the gas flow rate.
0.5
760 - 340‘ 2.667
Q = 433.51 10.92* 12
14.73 0.82*14*(72 + 460)*0.826
= 101.77 M MCFD
J
dd 55
/ -> \
q = 114.2 (9.15)
{f K T S )
9 .1 2 FLU ID F LO W HANDBOOK
where
/ Friction factor
L Pipe length in miles
q Gas flow rate in standard cubic feet per hour.
For simplicity, the full turbulent friction factor may be used in the calcula
tion. For 12 inch (304.8 mm) pipe, Crane recommends a friction factor of
0.013. The flow rate from the above equation is
7602 - 3 4 0 2
q = 114.2 12 = 4345520 SCFH
0.01 3jc 14 (72 + 460)* 0.82
EXAMPLE 9.3
Find the air flow rate in a 4 inch (101.6 mm) l.D. pipe with an upstream pressure
o f 150 psia (1034.22 kpaa) and downstream pressure o f 65 psia (448.16 kpaa).
The flo w may he assumed adiabatic at an average temperature of70°F
(21.1°C). The pipe length is 100 ft. Use the modified Darcy equation.
S olution
From Table 9.1, at K = 5.1, the maximum achievable pressure drop ratio and
gas expansion value are: AP/P\ = 0.7189 and Y = 0.661
The maximum pressure drop = 0.7189 X 150 = 107.835 psi
Since the given pressure drop is below the maximum, sonic flow is not encoun
tered and the mass flow at the given conditions is calculated using the Darcy m od
ified equation, Equation 9.11.
At the above pressure drop ratio and K = 5 .1. the gas expansion factor is Y ■
0.656 (Table 9 .1, table interpolation is required).
The maximum flow, or sonic flow, is found by using the largest possible pres
sure drop ratio for the given pipe resistance, K = 5.1
Calculate the pressure drop for natural gas pipe with 50 M M SCFD (0.059
M M SCM H) (75% Methane, 20% Ethane, 5% Propane). The pipe is 1 mile
(1.609 km) in length and has an inside diam eter o f 10 inches (254 mm). The
pipe inlet conditions are:
Solution
A trial and error procedure is needed to solve for the pressure drop. Solution steps
will be illustrated using the modified Darcy equation and the Panhandle B Equation.
Modified Darcy Equation
Mixture Density (P = 385 psig (2654.5 kpag) and T = 70°F (21.1 °C)) = 1.586
lbm /ft3 (25.4 kg/m3)
Friction factor for fully turbulent flow = 0.014 (Crane Technical Paper No. 410)
K - f U D = 0.014 X (1 X 5280)/( 10/12) = 88.7 (Pipe resistance coefficient)
k = 1.3, specific heat capacity ratio (approximate)
S = M w/29 = (0.75 X M W(^|-|4 + 0.2 x M w^ 2H6 + 0.05 X M w ^ ^g ) /29
= (0.75 X 16 + 0.2 x 30 + 0.05 X 44)/29 = 20.2/29 = 0.6965
Assume AP/P] = 0.5
AP = 0.5 X Pi = 0.5 x (350 + 14.7) = 182.35 psi
From Figure 9.1, at APIPX= 0.5 and K = 88.7, Y = 0.845
9 .1 4 FLU ID FLOW HANDBOOK
1 0.6965 \ 88.7
= 5 A \0 9 x \0 b SCFH (0.153x10 6 SCM H)
Solution
1. Panhandle B Equation
Assume a roughness factor Of 0.0018 inches (0.04572 mm).
? 614.7.v469.7
545.43 psia (3760.6 kpag)
614.7 + 469.7
„ , 3 .5 2 x 0 .8 2 2 0 .2 4 7 a O .8 2 2 2
Z = 1------------------------------------------ 0.8337
i.5l
614.72 - 469.72
100x10 737 0.92.r ------— ---------------------------------
V14.7 0.87 x78x(70 + 460)^0.8337
1672.1 = d 253
2. AGA Equation
Since the AGA equation requires the calculation of the transmission factor
( l / / ) 0 5 which is a function of the inside diameter and the flow turbulence,
Equations 9.7 and 9.8 are combined (assuming fully turbulent How) to give
0.5
3.7D PC ~ Pi 2.5
Q = 38.77 E 4 log. d
°S L m Tavg Z avg
where
D Inside diameter in ft
d Inside diameter in inches
A trial and error table is constructed below to solve for the diameter.
Q - ll 708.95.V logl0(24666.6*D) d 25
12 1 25.66
14 1.167 38.3
16 1.34 54.2
18 1.5 73.52
19 1.583 84.6
20 1.66 96.64
21 1.75 109.68
From the above table the inside diameter should be slightly larger than 20 inches.
A difference of about 7-8% is noticed in diameter calculation between the AGA
and Panhandle B Equation.
(9.17)
where
T Absolute temperature
P Absolute pressure
1,2 Inlet and outlet indexes, respectively
n Expansion factor
Isothermal flow n- 1
Adiabatic flow n = k (Cp/C v)
Polytropic flow I< n <k
(9.18)
T .. 2
i i +. k~\
m ci;
2 2
k -1
(9.19)
(9.20)
where
Ma Mach number
9 .18 FLU ID FLO W HANDBOOK
V V
Mci = - = - r - - (921>
Vs g,.kZRT
V Mw
where
V, Sonic velocity
V Gas velocity
R Universal gas constant, 1545 lbf-ft/lbmol-R
gc Gravitational constant, 32.174 lbm-ft/lbf-sec2
Choked flow conditions are reached when Mach number is equal to 1.0 (1 k°'
for isothermal flow), a condition that should always be avoided since very ligh
pressure drop (due mainly to acceleration effects) accompanies high speed fbw.
Implementation of Equations 9.18, 9.19, and 9.20 requires an iterative proce
dure (unless simplifications are assumed or mach number at exit conditions is spec
ified) since calculation of mach number at the exit conditions requires the p ior
knowledge of the temperature and pressure. The following examples illustrate the
solution steps.
EXAMPLE 9.6
Estimate the exit pressure and pressure o f a natural gas with specific gravity o f
0.87. The follow ing inlet and outlet information is provided:
Inlet Conditions
Solution
Since the exit Mach number is set, the solution is simplified and non-iterative. How
ever, the Mach number, at the inlet conditions, must be calculated first. The gas com
pressibility factor at the inlet condition is also required. The AG A and the Papay’s
correlations may be used to estimate Z as discussed in the previous examples:
215.56
Ma = = 0.1942
\g k Z R T \32x\ .32.v0.846.vl 545^(100 + 460)
Mw 25.23
Exit temperature and pressure are then estimated by Equations 9.18 and 9.19 as
1 32 — I
T 1+ — ---- A'0.19422
= 0.948
1 + l ' 2 - | ,0.62^
1.32
1 - 1 1.32-1
1+ jcO. 1942
2___________ = 0.802
.1.32-1 2
1+ ---------- *0.62
EXAMPLE 9.7
Repeat the above example fo r a pressure drop o f 200 psi (1378.96 kpa), fin d the
temperature and mach number at the exit conditions.
Solution
The solution for this example is more complex than the previous since the exit mach
number is unknown. The exit mach number is a function of the exit temperature
9.20 FLU ID FLO W HANDBOOK
and gas compressibility factor, which also have to be calculated. An iterative pro
cedure is required and the following steps are taken:
600
4. Find the gas compressibility factor at T2, use Equation 9.14; Z = 0.843;
Density = 2.19 lbm /ft3 (35.1 kg/m3); Velocity = 293 ft.sec (89.3 m/sec)
5. Calculate the mach number at exit conditions using the latest 72, Z, density
and velocity values, M a2 = 0.277
6. Calculate T2 with Equation 9.18 using the latest Ma2 value: T2 = 556.52 R
= 96.52°F (35.8°C)
7. Repeat steps 4-6, the second iteration yields the following at the exit condi
tions: Z = 0.88; Density = 1.92 lbm/ft3 (30.75 kg/m3); Velocity = 334.2 ft/sec
(101.86 m/sec); M a2 = 0.296; T2 = 555.7 R = 95.57°F (35.3°C)
To find the pipe length required to reach a given mach number the following
equations are used:
(9.22)
or
L+ L
f D
(9.23)
FLO W OF G A SES IN PIPES AND DUCTS 9.21
where
v l -M a 2 k +\ (k + \)M a2
Y = --------— + ------ Ln (9.24)
kM a2 2k
2(\ + ~ M a 2)
2
where
Solution
Equation 9.24 is used to find K, and Y2:
Y, = 16.575
Y2 = 5.865
L = 637.5 ft (194.31 m)
1. All terms of the Bernoulli equation, including the acceleration and eleva
tion terms.
2. Accurate prediction o f gas properties (gas compressibility, density and vis
cosity) at the local temperature and pressure values.
9 .22 FLU ID FLO W HANDBOOK
Two algorithms are presented to achieve the above features in a rigorous gas
model. Both algorithms require the use of incremental (integral) approach to
assure the accurate prediction of physical properties at the local temperature and
pressure conditions. Both models are best suited for a computer program rather
than for hand calculations. However, we will show few solution steps for each
algorithm and compare results and other calculations overhead.
In this algorithm, the pipe length is divided into a specified number o f small seg
ments; the Bernoulli equation is evaluated for each segment in a sequential order.
Two slightly different algorithms may be developed under this category. They
differ in the way to calculate the physical and thermodynamic properties for each
segment. A short description of each will follow:
1. Fluid properties are calculated at the segment inlet conditions. The fluid
properties such as density, viscosity, and gas compressibility factor are
evaluated at the segment inlet temperature and pressure and assumed to
stay constant through out the segment length. The accuracy of this algo
rithm depends on the number of segments, the higher number of segments
will make the segment length shorter and hence the likelihood of constant
physical and thermodynamic properties assumption is more pertinent.
Figure 9.2 shows a flowchart for the algorithm steps. The pressure drop is
calculated using the Bernoulli equation including the acceleration and ele
vation terms. The Bernoulli equation may be rearranged to the following
form to calculate the segment pressure drop:
(9.25)
where
h/?/f Frictional pressure drop, ft
V Fluid velocity
1,2 Subscripts for pipe inlet and outlet conditions, respectively
AZ seg Segment net elevation change.
FLO W O F G A SES IN PIPES AND DUCTS 9.23
I
Seg. Length= Pipe Length/N
Terms in the right hand-side of Equation 9.25 are expressed in feet, there
fore, the left hand-side is expressed in feet and must be converted to psi or
kpa. The segment net elevation term, AZ seg, is calculated by:
AZ
AZ =— (9. 26)
L
where
L Pipe length
AZpipe Pipe net elevation change, Z 2 - Z,.
Calculation and equations for the frictional pressure drop term were explained
in Chapter 8. The Darcy equation for frictional pressure drop term is used
in the examples below. The Darcy equation is summarized as
.. fL V 2
= D 2g <9 -27)
where
The acceleration term is not applicable to the first segment, since the
velocity is assumed constant in each segment. For other segments, the
acceleration term is calculated based on the difference of the segment inlet
square velocity and the previous segment square velocity.
The temperature change for each segment may be calculated by different
means. The most accurate and comprehensive method is by performing an
enthalpy balance over the segment under focus to find the fluid enthalpy at
the exit of the segment and then perform a PH flash at the exit pressure and
enthalpy to find fluid exit temperature. In the enthalpy balance, the sum of
heat of friction and the heat transfer from/to the surroundings must be
added to the fluid inlet enthalpy. Other simpler methods such as the ones
explained in Section 9.3 may be used with less accuracy.
Any equation of state or other gas compressibility calculation method may
be implemented in the solution algorithm. The discussion of equation of
state to calculate the gas Z factor, the gas density and then the gas velocity
has been explained in Chapter 2. Mach number is calculated at the inlet con
ditions of each segment. Calculation should stop if mach number exceeds
0.99. Such a condition indicates the need to change one of the following
flow parameters:
Fluid properties are calculated using the segment average temperature and
pressure. Average segment temperature and pressure may be calculated by
different means, the simplest is to add the inlet and outlet value and then
divide by two. Application of this algorithm requires an iterative solution
for each segment, thus increasing the complexity and computer overhead.
Since the segm ent’s outlet pressure and temperature are unknown (and this
is true for the calculation for each segment), an initial estimate is required
to start the solution. Once new values for the exit pressure and temperature
are found, the average values must be recalculated and the solution must be
repeated for a new estimate of exit temperature and pressure. Normally,
few iterations (4-6) are sufficient to converge to a stable solution. Since the
fluid properties are calculated at the average conditions, a larger segment
length may be used with comparable results to the algorithm discussed
above (A). Figure 9.3 shows a flowchart for this algorithm.
An incremental pressure drop value is set as a percentage of the segment inlet pres
sure (say 1-5% of Pl of each segment) or as an absolute fixed value (say 1 psi).
Therefore, segment outlet pressure is known, however, the length of the segment
is unknown. Once the outlet pressure is fixed, the temperature may be found as
described in the sections above, and physical properties at average conditions are
calculated. Fluid velocities at inlet and outlet conditions are also evaluated. Bernoulli
equation is then solved to the find the segment length that will give the set pressure
drop. Once the length of a segment is found, the algorithm is repeated for the next
segment until the pipe length is met. The actual number o f increments is unknown
prior to solving the case, only the case characteristics such as flow rate will deter
mine the length of the calculated segments and therefore the total number of incre
ments. A case with small flow rate will result in smaller number of increments and
larger segments lengths, unlike a case with high mass flow rate, which will result in
higher number of increments with smaller segments lengths. If the pressure drop
increment taken is small (less than 1% o f segment inlet pressure), physical prop
erties may be calculated at inlet conditions rather than at average conditions. Com
parable results are obtained. The flowchart in Figure 9.4 shows a detailed outline
of the algorithm.
The following example will be used to illustrate the above three algorithms.
EXAMPLE 9.9
45000 Ibm/h (99105.6 kg/hr) o f gas flo w s in a pipe with 6 (152.4 mm) inches
inside diameter. The pipe inlet pressure is 100 psia (689.48 kpaa) and the inlet
temperature 100 F (37.78 C). The pipe length is 5 0 0 f t (152.4 m) and the pipe
roughness may be assumed to be 0.0018 in (0.00658 m). Find the exit pressure,
9.26 FLU ID FLOW HANDBOOK
temperature, and mach number using the above explained rigorous gas algo
rithms. The following physical properties were used in the example
Solution
The following notation and assumptions will be used to solve this example.
Assumptions:
(9.28)
where
Nr Reynolds number
/ Moody friction factor approximation.
(.B - A ) 2
f= A - — ------------- (9.29)
(C - 2 B + A )
FLO W OF G A SE S IN PIPES AND DUCTS 9 .2 9
where
A = - 2 Logw\, £-------+
/D
—12
11,1 3.7 Nr
n „f , e l D 2.51 A
B = - 2 L o g .„ \-------+ ---------
1(1 3.7 Nr
(e/D 2.51 B
C = - 2 ^ Sl° — +—
1. Find Z, p, and velocity at the inlet pressure of 100 psia (689.48 kpaa) and
temperature o f 100°F (37.7°C): Z = 0.8638; p = 0.5586 lb/ft3 (8.87 kg/m3);
Velocity = 114.008 ft/sec (34.749 m/sec).
2. Calculate Reynolds number and friction factor: Nr = 5 0 4 3 9 0 8 ;/ = 0.015.
3. Calculate Mach Number: Ma = 0.15. If Mach Number > 0.99 stop calculation.
4. Calculate pressure drop for the first segment using Equation 9.25. The
acceleration term should be ignored for the first segment. DP = 1 . 1 8 psia
(8.135 kpaa). The pressure at the exit of segment 1 is set equal to 100 -
1.18 or 98.82 psia (681.344 kpaa).
5. Estimate the temperature at the exit of the first segment using Equation 9.16.
T2 = 98.1°F (36.722°C)
6. The calculation for increment one is complete. The second increment calcula
tion starts with the exit conditions of increment one. Steps 1-5 are repeated
for the rest of the increments; the acceleration term in Equation 9.25 is
accounted for by taking the difference in the square fluid velocity in the
current and preceding increment.
Table 9.2 shows the temperature, pressure, mach Number, gas com pressibil
ity factor and gas density profile for the 10 increments. The pressure at the exit of
the pipe (with 10 increments)is calculated to be: 87.47 psia (603.018 kpaa). More
9 .3 0 FLU ID FLOW HANDBOOK
accuracy is achieved when the total number of segments increases. Table 9.2a shows
the same pipe profiles when N = 50, and Table 9.2b shows the profiles when N =
100. Any further increase in the number of pipe increments will yield a negligi
ble improvement in profile calculation. The exit pressure with 100 increments is
87.39 psia (602.58 kpaa).
1. Assume the exit pressure and temperature for the first increment to 95% of
the inlet o f the inlet pressure and temperature.
2. Find Z, p, and velocity at the inlet and outlet segment conditions:
Inlet: Z = 0.8638, p = 0.5586 lbm/ft3 (8.947 kg/m3), V = 114.008 ft/sec
(34.75 m/sec)
Outlet: Z = 0.8687, p = 0.5325 lbm/ft3 (8.529 kg/m 3), V = 119.61 ft/sec
(36.45 m/sec)
3. Calculate Reynolds num ber and friction factor: Nr = 5 0 4 3 9 0 8 ;/= 0.015.
4. Calculate Mach N um ber using an average velocity and compressibility val
ues: Ma = 0.106. If M ach Number > 0.99 stop calculation.
5. Calculate pressure drop for the first segment using Equation 9.25. The
acceleration term should be ignored for the first segment. DP = 1.28 psia
(8.82 kpaa). The pressure at the exit o f segment 1 is set equal to 100 - 1.28
or 98.72 psia (680.654 kpaa).
FL O W O F G A SE S IN PIPES A N D D U C T S 9.31
(continued)
9.32 FLU ID FLO W HANDBOOK
6. Estimate the temperature at the exit of the first segment using Equation
9.16. T2 = 97.93°F (36.63°C).
7. Repeat steps 2-5. Five iterations for the inside loop (2-5 steps) should be
sufficient to provide a very good estim ate of the exit conditions for the
segment under focus.
8. Repeat steps 1-7 for the next increment.
Table 9.3 shows the data profile for the entire length. The results obtained for
N = \ 0 (number of length increment) is very comparable with the results obtained
with model A assuming N = 100. By comparing Tables 9.2, 9.2a, 9.2b and 9.3,
the advantage o f model B over model A is very clear.
The com puter code in BASIC to solve for the above example is shown in Fig
ures 9.6-9.8.
Gas flow at very low pressure, vacuum, is characterized to be molecular How and
not viscous flow. The resistance to the flow is dominated by the m olecules’ col
lision with the pipe wall. If we recall, in viscous flow, the resistance to the flow
is dominated by the m olecules’ collision with each other, which is represented by
the gas absolute viscosity. Figure 9.5 shows an illustration of viscous and mole
cular flows.
Knudsen Number, which is defined as the ratio of the molecule mean free path
(average) to the pipe diam eter, is used as a criterion to define the molecular and
viscous flow. Knudsen number is defined as
(9.30)
where
Kn Knudsen Number
D Pipe diameter, in.
A Molecule average mean free path, in.
where
When the mean free path is much smaller than the pipe diameter, it indicates
that m olecules’ collision with other molecules is dominant and flow is viscous.
The following Knudsen Number criteria are used to distinguish between the vis
cous and molecular flow:
Equations presented in Sections 9.2 and 9.3 can’t be used with molecular flow.
The viscosity in the molecular flow regime appears to be higher than viscous flow.
Also viscosity, in molecular flow, is very dependent on pressure and independent
of temperature, unlike viscous flow. The concepts of throughput and conductance
provide a better means of expressing vacuum flow problems. Throughput is defined
as the product of pumping speed and pumping suction pressure
Q = SPs (9.32)
where
Q = qPmg (9.33)
where
Conductance represents inverse of the pipe (or flow element such as elbow)
resistance to the flow. Conductance relates the throughput and pressure drop
across the pipe (flow element) as
Q = CAP (9.34)
where
Equations to calculate the conductance values for simple flow elements are
summarized in Table 9.5. Full coverage may be found in [9, 12].
Gas flow in ducts, such as ventilation and air conditioning ducts, normally occurs
at close to atmospheric pressure and temperature. The equations and algorithms
presented in the rigorous calculation model of gases may be applied. The diameter
term in all relations including the Reynolds number, Darcy equation, and veloc
ity calculation must be replaced by the equivalent diameter. The equivalent diam
eter is defined to be four times the hydraulic radius:
H (9.34)
where
Rh (9.35)
Short pipe
with end correction
Long Pipe
Duct (rectangular)
FLO W OF G A SE S IN PIPES AND DUCTS 9 .39
where
For a rectangular duct of width w and height /z, the hydraulic radius is
expressed as
= 2(w
2 + h) ( 9-36)
Deq = T — — (9-37)
(w + h)
REFERENCES
1. Arnold, E. 1995. Fluid Flow for Chem ical Engineers. 2nd ed. Holland Bragg,
London.
2. “Flow o f Fluids through Valves, Fittings, and Pipe.” 1980. Technical Paper 410.
Crane C om pany, New York.
3. C rocker S. 1945. Piping Handbook. M cG raw -H ill, New York.
4. D ushm an, S. 1962. Scientific Foundation o f Vacuum Technology. 2nd ed.
W iley, New York.
5. G as Processors Suppliers A ssociation. 1994. Engineering Data Book. 10th ed. V II.
Section 17.
6. Papay, I. 1968. O G IL Musz. Tud. Kozl. Budapest. 267-73.
7. Peng, D. Y., and D. B. Robinson. 1976. “A New Tw o-C onstant Equation o f State.”
Ind. Eng. Chem . Fundam. 15: 59-64.
8. Perry, R. H., D. W ., Green, and J. O. M aloney. 1984. Perry’s Chem ical Engineer’s
Handbook. 6th ed. M cGraw -H ill, New York.
9. Ryans, J. L. and D. L. Roper 1986. Process Vacuum System Design And Operation.
M cG raw -H ill, New York.
10. Serghides, T. K. March 5, 1984. “Estim ation o f Friction Factor A ccurately.” C hem i
cal Engineering. Pp 63-64.
11. Streeter, V. L. and E. B. W ylie. 1979. Fluid M echanics. 7th ed. M cG raw -H ill,
New York.
12. Vann Atta, C. M. 1965. Vacuum Science and Engineering. M cG raw -H ill,
New York.
13. “Steady Flow in Gas Pipelines.” IG T Technical Report 10. C hicago, 1965.
14. W eym outh, T. R. 1912. T ransactions o f the A m erican Society o f M echanical E ngi
neers. Vol. 34.
15. W hite, F. M. 1994. Fluid M echanics. 3rd ed. M cG raw -H ill, New York.
_______ CHAPTER 10_______
FLOW MINOR LOSSES
Piping systems in chemical plants, refineries, city water networks, oil and gas
transportations grids, and many other examples are very complex. Pipes have to
be laid out according to many geometrical and physical constraints. Also, flow
has to be measured, controlled, mixed (into a larger pipe diameter), split (into
smaller pipes diameters), or redirected to meet the needs of the process. Pipe
elbows and bends, expanders and reducers, diverging and converging tees, and
control valves are very common in any fluid flow piping system. One may find
all o f the above in a simple residential house piping system. These pipe elements
are known as pipe fittings and hand-valves. Figure 10.1 has a summary of the
most common pipe fittings. While in some flow cases, the fluid head loss due to
elbows, bends, valves, tees, crosses, and other fittings is minor, fittings may con
tribute significantly in the overall hydrodynamics of the flow system and their
effects should be taken into account for a proper design and analysis of flow sys
tems. Energy loss in fittings and valves is attributed to[ 1]:
1. Energy dissipation due to friction with the fittings’ interior wall. Just like
flow in pipe, the frictional loss is a function of fluid physical properties,
pipe roughness, and fluid velocity. Since the fittings’ length is normally
short with respect to the pipe length, this type of energy loss is small.
2. Energy loss due to change in flow direction. Fittings and valves tend to
change flow path and create more turbulence.
3. Obstructions in flow path will cause more energy to be dissipated.
4. Change in flow cross sectional area will decrease or increase the fluid sta
tic head. In this chapter, methods to represent and predict the resistance
from fittings and valves will be summarized.
10.1
10.2 FLU ID FLO W HANDBOOK
s
b c
Reducers/Expanders: a)Sharp Reducer b)Smooth
Reducer c) Sharp Expander d) Smooth Expander
The resistance o f pipes and fittings may be represented in many ways. The resis
tance factor (loss coefficient) K and the equivalent length (in pipe diameter or in
actual feet of pipe length) are among the most common ways to account for fittings
and valves flow resistance in fluid mechanics. A flow coefficient is also a com
mon way to represent resistance in valves; this will discussed later in this book.
FLO W MINOR LOSSES 10.3
* =
D
+
2g
00 . 1)
or
/( L + L ) V1
h = - --------- ^ ------ (10.2)
D 2g
or
where
The resistance factor K indicates the flow resistance in terms of the velocity
head. The K factor represents the resistance as a number (multiple or fraction) of
the local velocity head. The K factor is defined as
hr
K = — r L— (10.4)
V-12 g
where
The loss coefficient is independent of the friction factor and may be used for
any fitting or valve of a similar geometry. The loss coefficient is related to the fit
tings equivalent length (of pipe diam eters) as
K= 0 0 .5 )
10.4 F LU ID FLO W HANDBOOK
K P. = K c ( 10.6 )
where
Fluids passing bends develop more turbulence (eddies) due to the pipe curvature.
The additional turbulence appears as a result of a centrifugal force directed from
the center of the curvature to the outer wall. The flow velocity, in curvature, is
much lower near the outer wall and higher near the inner wall. A detailed dis
cussion of the flow theory in elbows and bends may be found in [3]. K factors in
bends and U turns are summarized in Tables 10.1, 10.3, 10.4 and 10.5. The fol
lowing sections will be present a short description of the methods classified
according to the data originators.
Crane Technical Paper 410 recommended the K value listed in Table 10.1 for pipe
bends and U returns. The K values are calculated as function of the rid (ratio of
bend curvature to pipe diameter) and pipe size d. The friction factor in the fully tur
bulent zone,/r , is used to compute the K value for all of the Crane fittings and valves.
Table 10.2 has a list of the friction factor in the fully turbulent zone as function
of the pipe nominal size, which may also be calculated using the Colebrook equa
tion with a very high Reynolds number.
The method was developed by Hooper [4, 5]. The advantage of the 2-K method
lies in the inclusion of the fitting size as an independent variable in the estimation
FLO W MINOR LO SSES 10.5
TABLE 10.1 Loss Coefficients for Elbows and Bends by Crane. See Table 10.2 fo r /T Value
E l b o w / Bend K
r / d = l .0 K=20/ t
r / d = l .5 K=14/ t
r / d = 2.0 K=12/ t
r/d=4.0 K=14/ t
r / d = 6 .0 K=17/ t
r / d = 8 .0 K=24/ t
r / d = 1 0 .0 K=30/ t
r / d = 1 2. 0 K=34 f T
r/ d=14.0 K=38/ t
r/ d=16.0 K=42/ t
r/d=20.0 K=50 f T
r
Kc = (n - 1) (0.257 if T — h 0.5AT) + K , Use above to find K
d
M i t r e B e nds (see figure 10.2 for dimensions)
Angle, a =0° K = 2/ t
Angle, a - 15°
ii
Angle, a = 30° K = 8 /t
Angle, a = 60° K = 2 5/ t
= 90°
*
Angle, a
o
>
o f the fittings resistance coefficient and in the method’s dependency upon Reynolds
number. A general form for the 2-K resistance loss coefficient is
10.6 F LU ID FLO W HANDBOOK
H" 0.027
Z" 0.025
I" 0.023
1%" 0.022
VA” 0.021
2" 0.019
2'A", 3" 0.018
4" 0.017
5" 0.016
6" 0.015
8-10" 0.014
12-16" 0.013
18-24" 0.012
where
Reynolds number is based on the fluid velocity in the fitting. If the cross sec
tional area o f the fitting changes, then the calculated K value is in reference to the
area at which the velocity was used to calculate Reynolds number. Values tor K\
and Kk appear in Table 10.3. K ] represents the elbow K value at very low Reynolds
number (Re = l, laminar flow) while represents the elbow K value at very high
Reynolds num ber (Re = oo, turbulent flow). The 2-K method is advantageous over
the Crane method in the laminar flow region. The 2-K method gives comparable
agreement with the Crane method for high Reynolds numbers.
Idelchik [31 had compiled the resistance coefficient data for a wider variety of bends
and elbows such as bends with 30° angle, elbows composed with separate elements
at different angles, Z-shaped elbows and many others [3]. Tables 10.4 and 10.5
show K factor for selected elbows and bends. For an extensive review and exper
imental data the reader is referred to [3].
FLO W MINOR LOSSES 10.7
TABLE 10.3 Values of /(j and K * for Various Elbows and Bends
(2-K Method: K = + K (1 + - )
Re d
TABLE 10.4 Resistance Coefficients for Pipe Elbows and Bends [3]
S olution
Crane method
Crane estim ation of the resistance coefficient is independent of the
Reynolds number. K value for a standard {rid = 1.0) screwed 90° elbow
may be found in Table 10.1. Values for the friction factorf T may be found
FLO W M INOR LO SSES 10.9
2-K Method
The resistance coefficient for a standard 90° elbow may be estimated from the
parameters available in Table 10.3:
K x = 800; Kx = 0.4
The following table gives the elbow resistance as function of the size and
Reynolds number:
Table 10.6 summarizes the K values recommended by the Crane paper and by the
2-K method. Table 10.7 sum m arizes the K values recommended by Idelchik for
straight inlets and inlets with a perforated area or with an orifice [3]. Many differ
ent inlet orientations and experim ental data for circular and rectangular inlets
may be found in Idelchik [3].
Tables 10.8 to 10.11 summarize the equations to predict the resistance coefficient
for pipe reducers and expanders. Common nomenclature is given for all tables.
10.10 F LU ID FLO W HANDBOOK
TABLE 10.5 Resistance Coefficients for Pipe Elbows and Bends [3]
D (in) 1/2 1 1% 2
L (mm) 38 102 102 127
e/D 0.02 0.01 0.0075 0.005
K 1.23 0.7 0.65 0.58
D1 is the small area diameter while D2 is the large area diameter. Care is needed
when using the 2-K method, K is always estimated based on the inlet diameter which
for reducers is D2 and for enlargers is D 1. The following example illustrates the use
of the three tables.
FLO W MINOR LOSSES 10.11
TABLE 10.6 Resistance Coefficients for Pipe Entrance and Pipe Exits
Sharp-Edged K= 1.0
Rounded K =1.0
2-K Method
K = ^ - + K„
Re
Kl=160
Ko. = 0.5 (Borda)
= 1.0 (Square-edged, r/d=0.0)
Flush
Crane
o
o
R/d = K = 0.5
R/d = 0.02 K = 0.28
R/d = 0.04 K = 0.24
R/d = 0.06 K = 0.15
R/d = 0.10 K = 0.09
R/d >= 0.15 K = 0.04
EXAMPLE 10.2
Find the resistance fa c to r K fo r a pipe expander 4-2 inches. The flo w may be
assumed fu lly turbulent and the tapered reduction angle is 10°. Compare the
results using the Crane, 2-K, and methods recommended by Idelchik [3].
10.12 FLU ID FLO W HANDBOOK
Pipe Inlet
a/Dh b/Dn
0 0. 005 0. 01 0. 05 0 .1 0 .2 0 .5 w
= # 0
0..004
0..008
0. 50
0 . 50
0. 50
0 .63
0 .58
0 .55
0 .68
0 .63
0 .58
0 .80
0 .74
0 .68
0. 86
0. 80
0. 74
0. 92
0. 86
0. 81
1. 00
0. 94
0. 88
1..00
0..94
0.,88
0.,012 0 . 50 0 .53 0 .55 0 .63 0. 68 0. 75 0. 83 0..83
0..016 o. 50 0 .51 0 .53 0 .58 0. 64 0. 70 0. 77 0..77
0..020 0 . 50 0 .51 0 .52 0 .55 0. 60 0. 66 0. 72 0..72
0..024 0 . 50 0 .50 0 .51 0 .53 0. 58 0. 62 0. 68 0..68
0..03 0. 50 0 .50 0 .51 0 .52 0. 54 0. 57 0. 61 0..61
0 .04 0. 50 0 .50 0 .51 0 .51 0. 51 0. 52 0. 54 0..54
0 .05 0. SO 0 . 50 0 .50 0 .50 0. 50 0. 50 0. 50 0..50
For Ducts: D= Dh
oo 0 .50 0 .50 0 . 50 0 . 50 0 .50 0 .50 0 .50 0 .50
Dh = Hydraulic Radius
Pipe Entrance
Circular Bellmouth K for the following cases
r/D
0 0 0 01 0 05 0 08 0 12 0 16 >0.2
A 1 0 0 87 0 40 0 20 0 10 0 06 0.03
B 1 0 0 65 0 27 0 18 0 10 0 06 0.03
C 0 5 0 44 0 22 0 15 0 09 0 06 0.03
Solution
e = io°
D2 = 4 in (Inlet)
D1 = 2 in (Outlet)
/T = 0 .0 1 7 (Table 10.2, for D = 4")
Crane method (Table 10.8)
(3 = D1/D2 = 2/4 = 0.5
FL O W M IN O R LO SSES 10.13
TABLE 10.7 (continued) Resistance Coefficients for Pipe and Duct Entrances
Pipe Inlet
Orifice or Perforated Inlet
B Resistance
0.05 1100
0.10 258
0.15 98
0.20 57
0 .25 38
0.30 24
0.35 15
0.40 11
0 .45 7.8
0.50 5.8
0.55 4.4
0.60 3.5
0.65 2.6
0.70 2.0
0.75 1.7
0.80 1.3
0.90 0.8
1.00 0.5
The above is for square reducer; to correct for the angle, multiply by
Q
1.6sin(—)
2
K 2 = 7.3 X 1.6 X sin(10/2) = 1.0179
K { = K 2 X (D ,/D 2)4 = 0.063
10.14 FLU ID FLO W HANDBOOK
Table 10.10
n = A ,/A 2 = D|2/D f = 0.25
K ] = 0.047
K 2 = 0.752
Tables 10 .12 to 10.16 summarize the Crane equations, 2-K method equations, and
a few selections of the compilation of Erwin and Idelchik experimental data and
formulas to estimate the flow resistance in converging and diverging tees [3]. Unlike
the Crane and 2-K method, values for K in Tables 10.14 to 10.16 are functions of
the volumetric flow ratio and the area ratio. Tables 10.14 to 10.16 provide data
for tees with branches at angles other than 90°, which may be useful for many
applications. The Crane and 2-K method K values represent the pressure drop due
to friction and turbulence; one should also add the pressure change due to the
FLO W MINOR LOSSES 10.15
TABLE 10.9 Resistance Coefficients for Pipe with Change in Diameters by Hopper 1988
(Re and/are based on the upstream diameter,/is the Moody friction factor)
K = [1.2 + ^ ] A * - l ]
Re Dl
Re > 2500
3) Pipe Reducer
^ , 1 + | 0 ] A 4 -1 ]
Re D,
4) Thin, sharp Re < 2500
orifice
D2 is the orifice K = [2.72 + A ) 2 - 1)][1 - A ) 2] [ A ) 4 -1 ]
diameter D2 Re D2 D,
Re > 2500
K = [2.72 + A ) 2( ^ ) ] [ l - A ) 2] A * -11
D2 Re D2 D,
TABLE 10.9 (continued) Resistance Coefficients for Pipe with Change in Diameters by Hopper
1988 (Re and/ are based on the upstream diameter,/is the Moody friction factor)
K = 2 [ l - 3 - ) 4]
2
Re > 2500
K = [ l + 0 . 8 / ] [ l - A ) 2]2
2
6 < 45°
change in velocity head. The flow static head may increase along the main branch
flow due to reduction in the flow velocity.
EXAMPLE 10.3
Estimate the K value the branch (2 inches) and the manifold (4 inches) o f a
diverging tee. Assume fu lly turbulent flo w with Re = 10 7.
Solution
Crane (Table 10.12)
Run through = 0.017 X 20 = 0.34 (0.017 is the friction factor for a 4 inch
diameter in the complete turbulent regime, see Table 10.2.)
Branch = 0.019 X 60 = 1.02 (0.019 friction factor for the 2 inch branch)
2-K method (Table 10.13)
K for run through = 0.1 X (1 + 1/4) = 0.125
K for branch through = 0.7 X (1 + 1/2) = 1.05
EXAMPLE 10.4
Estimate the pressure gain across the run through leg. Use Figure 10.3 fo r
velocity in inlet and outlet legs. Use the K values calculated above (Crane
fo r simplicity).
FLO W M IN O R LO SSES 10.17
F it t in g Resistance coefficient, K
n 3 5 10 15.-40 50-60
Of2
0. 64 0 .072 0 .067 0 .054 0. 040 0.058
A, 0 . 45 0 .076 0 .064 0 .052 0. 050 0.072
D2
0. 39 0 .098 0 .070 0 .051 0. 046 0.064
A2_ 0. 25 0 .100 0 .071 0 .047 0. 044 0.068
0. 16 0 .108 0 .084 0 .048 0. 044 0.074
0. 10 0 .118 0 .093 0 .053 0 . 050 0.079
D
n 76 90 105 120 150 180
0. 64 0 .076 0 .094 0 .112 0. 131 0.167 0 .190
0 . 45 0 .104 0 .138 0 .170 0. 202 0.246 0..255
n=Al/A2 (Ratio of cross
0 . 39 0 .110 0 .162 0 .210 0. 250 0.319 0 .364
section area)
0. 25 0 .127 0 .174 0 .220 0. 268 0.352 0..408
0r=O.017456
0. 16 0 .136 0 .184 0 .232 0. 278 0.362 0..420
0. 10 0 .142 0 .190 0 .237 0. 285 0.367 0..427
R/D
n 0.5 1. 0 1.5 2.0
0 .64 0.044 0.044 0.044 0.045
0.45 0.054 0.052 0.049 0.047
0.33 0.052 0.048 0.045 0.048
n=Al/A2
0 .35 0.062 0.053 0.052 0.052
10.18 FLU ID FLO W HANDBOOK
F it t in g Resistance coefficient, K
TA BLE 10.12 Resistance Coefficients for Pipe Tees, Crane Technical Paper 410
FLO W MINOR LOSSES 10.19
V = 10 ft/ s
( 3 . OSm/s )
Solution
For Pipe 1, the exit pressure Ph is estimated using the Bernoulli equation,
V2
P p - h - K .-L -
b 1/ 2g
where
v 2- V 2 v 2 V2- V 2
Pb~Pc=hff+ 2g
= Kt„—2 g + 2
2g
'
102 32 - 1 0 2
1 1.91895- P . = (0 .3 4 ---------------------------------------------------------------- + ------- )x62.4x2.1584e —
2 2
= -0 .4 9 5 psia
10.20 FLU ID FLO W HANDBOOK
TABLE 10.13 Valve and Tee Resistance Coefficients, 2-K Method. Hooper I98I.
K l
K = —L + K (l + — ) Re: Reynolds Number. ID: pipe inside diameter in inches.
Re ~ ID
Fittings Ki
V a lv e s
Gate, Ball, or Plug (3=1.0 300 0.10
3=0.9 500 0.15
1000 0.25
P = 0 .8
1500 4 .00
Globe, Standard nooV
1 VV 2 .00
1
Globe, angle or Y-type
1000 2 .00
Diaphragm, dam-type
800 0.25
Butterfly
2000 10.00
Check, Lift 1. 50
1500
Check, Swing 0.50
1000
Check, Tilting-Disc
Te e s
Run-through Tee
Screwed 200 0.10
Flanged or welded 150 0.50
Stop-in-type branch 100 0.00
Elbow branch
Standard, Screwed 500 0.70
Long-radius, screwed 800 0.40
Standard, flanged or welded 800 0.80
Stub-in-type branch 1000 1.00
Solving the above equation for Pc, Pc = 12 .4 14 psia (85.592 kpaa). The net
pressure gain is due to the change in velocity head.
Flow resistance in valves is summarized in Tables 10.17, 10.18, and 10.19 (Crane),
Table 10.13 (2-K method Hooper) and Table 10.20 lists K values for selected valve
types from the Idelchik Hydraulic Resistance book [3].
FLO W MINOR LOSSES 10.21
K3_2
90° Converging Tee (Flow r e s i s t a n c e l e g 2 t o 3, side
branch)
A1+A2 > A3 (Q V o l u m e t r i c f l o w r a t e )
A1=A3
(A i s c r o s s sectio n al area) Q2/ Q3 A2/ A3
0. 1 0. 2 0. 3 0.4 0.6 0.8 1. 0
-1.00 -1.00
H
0
0
0 -1.00 -1.00 -1.00 -1.00
1
0. 1 0.40 -0.37 -0.51 -0.54 -0.59 -0.60 -0.61
OJ
0.2 3 .80 0 . 72 0 . 17 -0.03 -0.17 -0.22
1
0
0
0.3 9 . 2 0 2 . 27 1 . 00 0.58 0.27 0.15 -0.11
0.4 16. 3 4 . 30 2 . 06 1. 30 0 . 7 5 0 . 5 5 0. 44
0.5 2 5. 5 6 . 75 3. 23 2.06 1.20 0. 89 0.77
0.6 3 6. 7 9 . 70 4 . 70 2.98 1 . 6 8 1 . 2 5 1. 04
0. 7 4 2. 9 13 .0 6 . 30 3 .90 2 20 1 . 6 0 1 . 3 0
0. 8 64 .9 1 6 . 9 7 . 92 4.92 2 . 7 0 1 . 92 1 . 5 6
0.9 8 2. 0 2 1 . 2 9. 70 6.10 3 .20 2 . 2 5 1 . 8 0
1.0 101. 26.0 11. 9 7.25 3 . 80 2 . 5 7 2 . 0 0
\2
K 3„2 = Al 1+
Q i\ -2 i - 2 l
Q,j
A is found from t h e follow ing ta b le
K3-1
(Flow resistance leg 1 to 3)
Q1/ Q3 K 3-1
0 0
0.1 0.16
0.2 0.27
0.3 0.38
0.4 46
0.5 53
0.6 57
0.7 59
0.8 60
0 .9 0.59
1. 0 0.55
\2
K, , = 1.55 — £2
G, QJ
10.22 F LU ID FLO W HANDBOOK
K3-2
45° Converging Tee (Flow resistance leg 2 to 3, side branch)
(Q Volumetric flow rate)
A1+A2 > A3
A1=A3 Q2/Q3 A2/A3
(A is cross sectional area) 0. 1 C.2 0. 3 0. 4 0.6 0.8 1.0
Y
1+
Q,A,
-2 -1 .4 ,A Qi
I CM: M < 2 3y
K3-1
(Flow resistance leg 1 to 3)
Q2/Q3 A2/A3
0. 1 0.2 0. 0.^ 0.6 0. 1 0
0 0. 00 0 00 0 00 0 00 0 00 0 00 0 00
0 1 0. 05 0 12 0 14 -0 54 0 17 0 17 0 17
0 2 - 0 . 20 0 17 0 22 -0 03 0 27 0 29 0 31
c 3 - 0 . 76 -0 13 0 08 0 58 0 28 0 32 0 40
0 4 - 1 . 65 -0 50 -0 12 1 30 0 26 0 36 0 41
0 5 - 2. 77 -1 00 -0 49 06 0 16 0 30 0 40
0 6 -4 . 30 -1 70 -0 87 2 98 -0 04 0 20 0 33
0 7 - 6. 05 -2 60 -1 40 3 90 -0 25 0 08 0 25
0 8 - 8. 10 -3 56 -2 10 4 92 -0 55 -c 17 0 06
0 9 -10 .0 -4 75 -2 80 6 10 -0 88 -0 40 -0 18
1 0 -18 -6 10 -3 70 n 25 _1 35 - c 77 -0 42
\2
*3-1=1- i - & - 1 . 4 1 ^ Cl
g 5. M e,
FLO W M IN O R L O SSES 10.23
Kx-2
90° D i v e r g i n g T e e (Flow r e s i s t a n c e l e g 1 t o 2, side branch)
T h read ed ; M alleable iro n (Q V o l u m e t r i c f l o w r a t e )
A1+A2 > A l
A l =A3 Q2/ Q1
0.1 0.2 0. 3 0.4 0.5
(A i s c r o s s s e c t i o n a l a r e a )
A2/A1
0.09 2.80 4.50 6.00 7.88 9 .40
0.19 1.41 2.00 2 . 50 3.20 3.97
0.27 1.37 1.81 2.30 2 .83 3 .40
0 .35 1.22 1.54 1.90 2.35 2.73
0 . 44 1.10 1.45 1.67 1.89 2.11
1 —► 0.55 1.09 1.20 1.40 1.59 1 . 65
1 ---------- *
1 1.00 0.90 1.00 1.13 1.20 1.40
i Q2/ Q1
•iL 0.6 0.7 0.8 0.9 1.0
A2/A1
0.09 11.1 13.0 15.8 20 . 0 24 .7
0.19 4.95 6.50 8.45 10 . 8 13.3
0.27 4.07 4.80 6.00 7 .18 8.90
0 . 35 3.22 3.80 4.32 5.28 6.53
0 . 44 2.38 2.58 3 .04 3 . 84 4 .75
0.55 1.77 1.94 2.20 2.68 3.30
1 .00 1.50 1.60 1 . 80 2 .06 2 . 80
TA B L E 10.17 Valve Resistance Coefficients. Crane Technical Paper 410 (K2 is Valve K
reference to pipe diameter. d2 is valve diameter)
V a lv e R e s is ta n c e C o e f f ic ie n t
G a te V a lv e s P = 1 .0;6=0
(Wedge Disc, Double Disc, or K = 8 /T
Plug)
P < 1.0; 0 <=45°
d2 pipe diameter 8S t + sin ^[0.8(1 - P 2) + 2.6(1 - p 2) 2 )
d x valve diameter
P=di/d2 K l ~ p*
0 Wall co wall contraction
a n gle.
P < 1.0; 45° < 0 <=180°
A n g le V a lv e s P=1
Welded K i=K2= 55/t
P < 1
,, 55fT+m-/32)+(l-/32)2'
K l ~ ..... f
P=1
Flanged Ki=K2= 150/t
P < 1
150/r +/7[(l-/?2) + ( l - ^ 2)21
2 p
FLO W M INOR LOSSES 10.25
TA BLE 1 0.18 Valve Resistance Coefficients, Crane Technical Paper 410. ( K2 is Valve K
reference to pipe diameter, d2 is valve diameter, d| pipe diameter; [3 = d,/d2; 0 Wall to wall
contraction angle.)
V a lv e R e s is ta n c e C o e f f ic ie n t
P lu g V a lv e s P=1 . 0;
Straight-Way K = 18/t
P < 1.0;
3 -W a y V a l v e s [5=1.0;
K = 30/t (
K = 90/t ( b r a n c h flow)
P < 1.0;
K + 0.5(sin - ) 05 (1 - P 1) + (1 - P 1) 2
K , = ------------------------ 7-----------------------
P
B a ll V a lv e s p = 1.0, 0 = 0 K = 3/t
P < 1.0, 0 < 45°
3 /r + 0 .5 (s in ^ )05( l - / ? 2) + ( l - / ? 2)2
K j = ------------ ---- ^---------------
2 P“
Foot v a lv e s w ith S tr a in e r
K= 420/ t
P o p p e t Disc
K= 7 5/t
Hinged Disc
S w in g C heck v a lv e
90° Check K = 50/t
TA B L E 10.19 Valve Resistance Coefficients, Crane Technical Paper 410. (K2 is Valve K
reference to pipe diameter. d2 is valve diameter, d, is pipe diameter; = d|/d2; 6 Wall to wall
contraction angle)
V a lv e R e s is t a n c e C o e ffic ie n t
L if t C h ock V a lv e s
90 ° Stem 3=1.0; K = 600/t
0 < 1.0;
600/r +/?[0.5(l-£2) + (l-/?2)2]
0=1.0; K = 55/t
45° Stem
0 < 1.0;
S to p -C h e c k V a lv e s 0=1.0;
,, K + /9 [0 .5 (l-/? 2) + ( l - / ? 2)2]
K > -
FLO W M IN O R LO SSES 10.27
0 5 10 20 3040 50 5567
Rectangular tube
0 .5 10 20 30 40 50 55 67
0 20 30 40 50 60 70 75
K 1.7 3.2 6.6 14 30 62 90
Dd' = Dd/Do
Do “__
a/b 0, Degree
0 10 20 30
0.5-1.0 0.04 0.30 1.10 3.00
1.5-2.0 0.04 0.35 1.25 3 .60
a/b 0, Degree
40 50 60 70
0.5-1.0 8.00 23.0 60 190
1.5-2.0 10.0 29.0 80.0 230
REFERENCES
1. Crane C om pany. 1980. Flow o f Fluids through V alves, Fittings, and Pipe. Technical
paper 410. C rane Com pany, New York.
2. D arby, R., 1986. Chem ical Engineering Fluid M echanics. 1st ed. M arcel Dekker,
Inc, New York.
3. Erwin F., and Idlechik, I. E., 1989. Flow Resistance: A Design Guide for Engineers.
H em isphere Publishing C orporation, New York.
4. H ooper, B. W . 1981. “The Tw o-K M ethod Predicts Head Losses in Pipe Fittings.”
Chem ical Engineering. (A ugust) pp. 96-100.
5. H ooper, B. W . 1988. “C alculate Head Loss C aused by C hange in Pipe Size.” C hem
ical Engineering. (N ovem ber) pp. 89-92.
6. Idelchik, I. E. 1954. “Hydraulic Resistance.” Physical and M echanical Fundamentals,
p. 316. G osenergoizdat, M oscow.
7. Idelchik, I. E. 1953. Determ ination o f Resistance C oefficients in Discharge Through
Orifice. G idrotekh. Stroit., no. 6, 31-36.
8. Idelchik, I. E. 1947. A erodynam ics o f the Flow and Pressure H ead Losses in D if
fusers. Prom. A erodin., no. 3, 132-209, BNT MAP.
9. Idelchik, I. E. 1944. Hydraulic Resistance During Flow in E ntrance in Channels and
Passage T hrough Orifice. Prom. Aerodin., no. 2, 27-57, BNT, NKAP.
____________ CHAPTER 11____________
HYDRODYNAMICS OF VAPOR-
LIQUID TWO-PHASE FLOW
Two-phase gas-liquid flow is a common flow in many areas such as gas and oil
production and transportation lines, heat exchangers, pressure relief systems and
steam utilities networks. The gas-liquid flow is more complex than the hom oge
neous flow. The coexistent flow of two phases complicates the theoretical and
empirical hydrodynamics analyses mainly due to dynamic fluctuation and inter
action of the two phases with each other and with inside wall of the pipe. Exten
sive research and experimental studies have been conducted on two-phase flow
during the last 60 years, yet it is rare to find two correlations with exact predic
tions. It will be impossible to cover all of the gas-liquid flow theoretical and
empirical studies in this chapter, however, a few selected methods to predict
pressure drop, flow maps, and liquid holdup will be covered.
Many concepts and nom enclature are used specifically in two-phase gas-liquid
flow hydrodynamics. A brief description of the two-phase flow concepts and
terms is necessary before discussing methods to predict pressure drop in two-
phase gas-liquid pipe flow.
11.1
11.2 FLU ID FLO W HANDBOOK
11.1.1 Velocity
Since gas and liquid phases flow concurrently in the pipe, different fluid velocity
terms apply.
A ( ii.D
vu = ^ r (11.2)
A
where
11.1.1.2 M ixture Velocity. Mixture velocity is defined as the total volumetric flow
(that is the liquid plus the gas) per unit area of the pipe inside cross sectional area:
(11.3)
A
where
Vm = QL±QcL (11.4)
K = ^ - + ^ - = v u + v os (11.5)
H YD RO D YN A M ICS OF VAPOR-LIQUID TWO-PHASE FLOW 11.3
From Equation 11.5, it can be seen that the mixture velocity is equal to the
summation of the liquid and gas superficial velocities.
11.1.1.3 Liquid and Gas True Velocities. In reality, the gas and liquid will flow
at different velocities, depending on the gas mass split fraction and physical prop
erties of each phase. Gas-phase travels faster in most cases due to the relatively
low gas density in comparison to the liquid phase density, thus
Vc > Vl
where
When a no-slip flow is assumed, this will indicate that the liquid travels at the
same gas velocity
Vb = Vi
Liquid holdup is defined as the volume fraction of the pipe occupied or filled by
liquid. Liquid holdup and gas holdup are related as
H l + H g = 1.0
where
V
vp
where
vl = ^ t (1L6)
h L
V V
VG = — = — — (11.7)
G Ha ] - H ,
When a no-slippage flow is assumed (VL = VG), Equations 11.6 and 11.7 may
be grouped as
( 1 , -8)
Equation 11.8 maybe rearranged to solve for the liquid holdup in terms of
superficial velocities as
V V O
/ / = ----- ±— = - ^ = (11.9)
1 vLs +v
T VGs Vm V^r
v o
When a flow slippage is assumed, which is the actual flow behavior in most
cases, one of the empirical correlations has to be used to predict the liquid holdup.
Under slippage conditions, the liquid velocity is less than liquid velocity at no
slip conditions, and liquid holdup is higher than that of the liquid holdup at no
slip conditions.
11.1.3 Quality
Quality in two-phase flow is defined as the mass ratio of the gas-phase to the total
fluid mass in a particular pipe segment; quality should not be confused with holdup
since it is a mass rather than a volume fraction.
H YD RO D YN A M IC S O F VAPOR-LIQUID TWO-PHASE FLOW 11.5
x —— ^ — ( l l . 10)
M g + M,
where
x Quality, dimensionless
Mg Gas-phase mass at particular position in the pipe, lbm (kg)
Ml Liquid-phase mass at particular position in the pipe, lbm (kg)
-i
Hc,= 1 + S R \------ (11.11)
where
A value for the slip ratio is not easily calculated, which makes the above equa
tion unusable unless measurements of gas and liquid actual velocities are available.
The concurrent flow of gas (vapor) and liquid in pipes may result in many phase
interactions (flow regimes) depending on many parameters. Authors and resear
chers have used different terminologies to describe the gas-liquid flow behavior.
The characteristics of two-phase flow hydrodynamics are strongly affected by the
flow patterns; many of the pressure drop and hold-up correlations are flow pat
tern dependent. The flow regime is a strong function of the following parameters
liquid flow rates where the gravity force acting on the liquid is much higher than
the drag forces acted on the liquid by the gas-phase. Depending on the smoothness
of the gas-liquid surface, many other terms may be used such as: 1) stratified smooth
2) stratified wavy. The wavy behavior results at higher gas flow rates. While strat
ified flow does not occur in pipe with an upward inclination of more that 1 degree,
it is the most common flow pattern in pipes with downward inclination. The strat
ified wavy flow regime may be referred to as wavy flow by many investigators.
Stratified flow may also be referred to as segregated flow.
11.1.4.2 Slug Flow. As the liquid flow rate increases in a stratified wavy flow,
the liquid waves will grow higher and fill the entire pipe section at irregular inter
vals forming a slug flow. The slug flow is characterized by the flow of alternating
slugs of liquid and large gas bubbles. The term plug flow may be used to charac
terize a slug flow with a smaller gas volume fraction. Slug flow is a common flow
regime in horizontal or near horizontal and upward vertical or near vertical flow. In
upward vertical flow, slugs are characterized by a smaller size than that in the hori
zontal flow. Slug flow may cause serious problems due to vibration and corrosion;
slug catchers are normally inserted to eliminate slug formation. Others terms used
by many investigators to describe this flow include: plug flow, intermittent flow,
churn flow. Slug flow may be classified into hydrodynamic slug or terrain slug
based on the initiation cause. In terrain slug, the gravity forces due to the topog
raphy of the hilly terrain cause the liquid to accumulate in the lower pipe sections;
as the gas pressure increases, liquid will be pushed and slug flow will form. More
on slug flow will follow in slug catcher design section of this chapter.
11.1.4.3 A n n u la r Flow. In annular flow, the liquid flows as a film around the inner
circumference of the pipe, while gas flows in the center. Liquid mist or droplets may
be carried in the pipe center due to the high speed of the gas. As the gas velocity
increases, the liquid film decreases and more liquid is carried by the gas phase as
mist. For a given flow rate, the liquid film thickness in vertical flow is uniform,
however, in horizontal flow, due to gravity the liquid film thickness is larger at
the bottom of the pipe. Annular flow is very common in most high-pressure gas
transfer lines and may exist for any angle of inclination. Depending on the ratio
of mist entrained and film thickness, this flow may be referred to as annular-mist
or mist flow.
11.1.4.4 D ispersed Bubble Flow. The dispersed bubble flow is the opposite of
the annular flow; liquid-phase is the predominated flow while gas flows as dis
persed bubbles. As the gas flow rate increases, bubbles start to increase in size and
plug or slug flow starts to form. Bubble flow is very common in upward vertical
flow lines and also in horizontal flow with high liquid flow rates. This flow may
be referred to as bubble flow, or forth flow.
Figure 11.1 shows an illustration of gas-liquid flow regimes in vertical and hori
zontal flow. Boundaries of flow patterns are not easily defined; rather the gradual
change between flow regimes may contain a combination of many flow forms flue-
H YD RO DYNAM ICS OF VAPOR LIQUID TWO-PHASE FLOW 11.7
Stratified Flow
Smooth
Wavy
0o
Annular Flow
(Annvlar-Uist)
Slug Flow
(Intermittent)
(Churn)
(Plug)
tuating until a stabilized distinct flow pattern characterizes the new flow pattern
once again. In general, the following summarizes the chances for the flow pattern
• Annular flow: High gas superficial velocity. Vapor velocity greater than
20 ft/s (6 m/s). Mist flow may occur at gas velocities higher than 195 ft/s
(60 m/s).
• Dispersed bubble flow: High liquid superficial velocity. Liquid velocities
range between 4.8-16 ft/s (1.5-5 m/s). Vapor velocity between 1-10 ft/s
(0.3-3.0 m/s).
• Slug flow: Moderate low gas and liquid superficial velocities. Occurs
over a wide range of velocities.
A general trend of the two-phase flow regimes as a function of gas and liquid
superficial velocities illustrated in Figure 11.2 and Figure 11.3 for horizontal and
vertical upward flow respectively.
o
■w
J
?
•H
O
O
H ANNULAR
•H
U
•H
SLUG
8
§■
S
STRATIFIED BUBBLE
F IG U R E 11.2 Effect of gas and liquid superficial velocities on flow regimes in horizontal flow.
H YD RO DYNAM ICS OF VAPOR-LIQUID TWO-PHASE FLO W 11.9
tn
o
■W
J
?
•H
U
0
H
?!
•H
U
•H
5
6
------------------------------------------ ►
L iq u id S u p e r f i c i a l v e l o c i t y (Log)
FIGURE 11.3 Effect of gas and liquid superficial velocities on flow regimes in vertical upward flow.
11.2.1 Baker
Among the earliest attempts is the map developed by Baker [1,2]. The Baker map,
shown in Figure 11.4, uses the superficial mass flux terms for the gas and liquid
phases. The gas and liquid superficial mass flux are defined as
Ml
(11.13)
where
For systems other than air-water, Baker used the following gas-liquid prop
erty correction factors:
{ x l/2
Pg Pl
A = (11.14)
V Pa P
and
( \ 2“
<yw Ml Pw
o L <P l y
1
where
o Surface tension
/x Viscosity
Subscripts G and L refer to the gas and liquid phases respectively, while sub
scripts W and A refer to physical properties of water and air respectively at 68°F #
and 14.7 psia (20°C and 1 atm). For air-water fluid system at 68°F and 14.7 psia
(20°C and 1 atm), Equations 11.14 and 11.15 reduce to unity.
Beggs and Brill suggested the following set of equations to predict the flow
regime in horizontal flow. The Beggs and Brill correlation is based on experi
mental data measured by the authors using acrylic pipes 90 feet long; pipe diam
eters of 1 and 1.5 inches were used. Data were collected for horizontal, vertical,
and many other angles of inclinations (-90°, +90°). Fluids used were water
(0-30 GPM ) and air (0-300 mmscfd). The flow regime map in the Beggs and Brill
correlation consists o f three areas as
1. Intermittent flow which includes the slug and plug flow regimes.
2. Segregated flow which includes the stratified, wavy, and annular flow
regimes.
3. Distributed flow which includes the bubble and mist flows.
The following criteria are suggested for the above flow regimes:
Intermittent: L| < N FR < L 2
Segregated: N FR < L\
Distributed: N FR > L x and N FR > L 2
where L x, L 2, and N FR are defined as:
A = — ——
Ql + Qc
11.12 F LU ID FLO W HANDBOOK
Figure 11.5 shows the Beggs and Brill gas-liquid flow map.
FIGURE 11.6 Mandhame, et al [21] flow map for horizontal gas-liquid flow.
The M ukherjee and Brill flow map for horizontal flow is shown in Figure 11.9.
The map isconstructed using air-lube oil or air-kerosene systems in a 1.5 inch
(39mm) pipe inside diameter. The map is expressed interms of the gas and liq
uid velocity number which are defined as:
N lv = VL,t] p L / g o (11.18)
n lv = Vu \ j p L I ga (11.19)
EXAMPLE 11.1
Use the follow ing air-water flo w data to predict the two-phase flo w regime
using all o f the above-m entioned methods.
Air viscosity = 0 .0 18 12 cP
W ater viscosity = 0 .8 1 cP
W ater surface tension = 70.2 dyne/cm
Pipe inside diameter = 2 inch (5.08 cm)
Gas flow rate = 8500 ft3/hr (240.7 m3/hr)
W ater flow rate = 25 GPM (5.677 m3/hr)
Solution
x= Qt = 2 5 /7 .4 8 1 /6 0 = 0.05569 =QQ23
Q l +Q g 2 5 /7 .4 8 1 /6 0 + 8 5 0 0 /3 6 0 0 0.05569 + 2.36
X = ln(0.023) = - 3.77
V2 1 10.8552
AL„ = -= - = ------------------ = 2292
™ dg 2/12x32.174
H YD RO D YN A M ICS OF VAPOR-LIQUID TWO-PHASE FLOW 11.15
Velocity
Superficial
Liquid
o
Downward, 5
Superficial
Liquid
._o
Downward,10
Velocity
Superficial
Liquid
Downward, 30°
Superficial
Liquid
Downward,70°
Velocity
Superficial
Liquid
_ _o
Downward, 90
Liquid
Horizontal Flow
Upward, 2
^ o
Upward ,10
^
Upward ,5 0o
O
Upward , 70
Velocity
Superficial
Liquid
FIGURE 11.9 Gas-liquid flow map for horizontal flow by Mukherjee and Brill [23].
A few selections of the liquid hold-up correlations will be presented and compared.
l-b P .
X = J - 7r f 0 l 20)
-A Pc;
where
where
For a fully turbulent gas and liquid flow the following estimation for X may
be used
M ( 11.22)
~o
, I /O
a
K V O sf
The Lockhart and Martinelli correlation is based on data for a flow in 1 inch pipe
diameter. Baker [2] reported acceptable results for the Lockhart and Martinelli
correlation when applied to pipes 8 inch and 10 inch I.D. diameters. Farooqui and
Richardson [15] reported an over predicted liquid hold-up values by the Lockhart
H YD RO DYNAM ICS OF VAPOR-LIQUID TWO-PHASE FLOW 11.23
X (Lockhart-Martinelli)
FIGURE 11.10 Comparison of liquid holdup predicted by the Lockhart-Martinelli and Farooqi-
Richardson correlations.
and Martinelli correlation and suggested the following correction bases on the X
parameter defined above
Beggs and Brill have developed the following formula to predict the liquid hold
up at any angle of inclination
where
0.845 A05351
H l (0) = —:— — (Intermittent Flow) (11.25)
N FR
r-d
1.065 A05824
H l (°) = * ^ 0.0609— (Distributed Flow)
i VFR
r-o
where
Prediction of flow regime by Beggs and Brill has been discussed in the per
vious section of this chapter. The inclination correction factor, y/, is estimated as
where
6 Angle of inclination
C is defined in Table 11.1 based on the flow pattern for upward and
downward inclinations
Eaton et al. [14] proposed the following empirical correlation for the liquid-phase
hold-up
o.io
(11.27)
H YD RO D YN A M ICS OF VAPOR-LIQUID TWO-PHASE FLOW 11.25
where
p
5.0 < ——— < 65.0 (11.30)
14.69 14.69
N L ^ / u Li j g / p Lo ' (11.31)
where
TA BLE 11.1 C Coefficient for the Angle Correction Factor in Beggs and Brill |6] Correlation
Flow Pattern
(Horizontal) Upward Inclination Downward Inclination
0.1244
4.7 N LV____
Distributed 0.0 C = (1 —A)ln 10.3692 i r 0.5056
t FR
11.26 FLU ID FLO W HANDBOOK
Figure 11. 11 may be used to find the Eaton's liquid holdup fraction as a func
tion o f the above dimensionless groups. The Eaton’s correlation is based on their
data on the flow of gas-water in 2 and 4 inch horizontal pipes.
Hoogendoom and Buitelaar have developed a liquid holdup correlation that is inde
pendent of pipe diameter and fluid physical properties. Their correlation is a func
tion of the liquid-gas superficial velocity ratio and the gas superficial velocity
(11.32)
Mukherjee and Brill developed the following correlation which may be applied
to any angle of inclination
1
c
o
o
0.8 -
<c
0.6 --
E
3
£ 0 .4 --
2
D
<75 0 —
- 0.C01 0.01 0.1 1 13
- 0.2 —
A ^ A C 77U 4 . 6 9 j V0.00266 J
where
f \ 0.9 / \ 0.5
1-JC - 1/6 Rc. Mg
(ReLF r L )
P l ) x Ml s
c
o
'■oH
CO
k-
LL
0
E
D
O
>
2
*5
cr
□
3
+*
(?)
when
, xO.9, x0.5
1-JC - 1/6 Pc j I Mg -3
( R e t Frt ) < 2.5*10 (11.35)
Pl ) \ M lj
where
jc Quality
R et = (11.36)
Vi
FrL = M l { \ - x f
(11.37)
PUD
The Hughmark correlation for the in-situ liquid volume fraction was developed
for vertical flow. But it may also be applied to horizontal flow as confirmed by
Dukler et al. The Hughmark correlation is
r
" 1 - jc “
Pg Hl
" ‘ -4
Hg X P l. Hg
Table 11.2 shows different values for factors /?, q , and r as suggested by many
investigators.
Chen and Spedding derived the following flow-based correlations for the liquid
holdup in horizontal idealized flow.
1. Stratified Flow
Gas-liquid turbulent-laminar flow
1.8 .0.2
3 Qc
0.8 (11.40)
l Q l VgO
0.8
1/MS
He Pg Qg VgD
. .0.2 (11.41)
Hl P l Q'l
where
D Pipe diameter
v Kinematics viscosity
Values for vvb W2, and w2 are dependent of the HG/HL ratio and are shown
in Table 11.3.
11.30 F LU ID FLO W HANDBOOK
Homogeneous 1 1 1 0
Zivi 1 1 0.67 0
Turner-Wallis 1 0.72 0.40 0.08
Lockhart-Martinelli 0.28 0.64 0.36 0.07
Thorn 1 1 0.89 0.18
Baroczy 1 0.74 0.65 0.13
HC]/ H l ratio W\ w1
2. Annular Flow
The following equation may be used for any range of HG/HL, however differ
ent values for the K , a , b, and c should be used for the different gas-liquid
turbulence combinations:
-a b
Q g_ Pc Hl (11.42)
Hl Ql Pl He
Table 11.4 shows the values for the above equation parameters for gas-liquid
turbulent-turbulent and laminar-laminar flows.
HSL 1
(11.43)
HL [0.2 +1.2 / (Qc I Ql )]
HYD RO DYNAM ICS OF VAPOR-LIQUID TWO-PHASE FLO W 11.31
TABLE 1 1.4 Parameters for the Chen-Spedding [ 10] Annular Flow Holdup
Correlation
EXAMPLE 11.2
Find the liquid holdup fraction fo r the air-water flo w system detailed in Exam
ple 11 . 1.
Solution
Hoogendoorn correlation
The Hoogendoorn correlation, Equation 11.32, is implicit in terms of gas
holdup fraction. A trial and error procedure may be applied to find the gas
holdup fraction
\ 0.85
/ V N
Ls
= 0.60 V,Gs 1 -
1~ E r 1~ E c V yV Gs
11.32 FLU ID FLO W HANDBOOK
0 . 8?
Eg
0 .60^ V* 1 -
1~E r 1- E , [vG
s
0.92 11.5 24.58
0.93 13.28 23.37
0.95 19.01 19.36
Since the left- and right-hand sides of Equation 11.32 are almost equal at
EG = 0.95; E l = 0.05
Hughmark correlation
108.3
H g = 0.82 j = 0.801
2.555 + 108.3,
Therefore, liquid holdup is equal to 0.1988
Hc
aPc bH
= K Qc l
Ql Pl He,
where K , a, b, and c given in Table 11.4. Solving the above equation for H L gives
H l = 0.2568
dP_ dP_
(11.45)
\d Z )T dZ. dZ m
The gravitational term is significant in vertical flow and flow through hilly
terrains. The gravitational term drops out when the flow is horizontal. The grav
itational term of the total pressure drop may be expressed as
d P ^ = p M g sin fl
(11.46)
d Z )g 144g(.
HYD RO DYNAM ICS OF VAPOR-LIQUID TWO-PHASE FLO W 11.33
where
pM Mixture density
6 Angle o f inclination
gc Gravitational constant
g Acceleration due to gravity
Mixture density is a function of the gas and liquid holdups, and the gas and
liquid densities and may be estimated as
P m = " l p L + ( l - 0 - H L) p G (11.47)
Different correlations have been proposed for the frictional and acceleration
terms of the pressure drop of gas-liquid flow systems.
Frictional pressure drop in two-phase flow is due to the friction between each
phase and the wall and also between the gas and liquid phases. Depending on the
area of the gas-liquid interface and its roughness, frictional loss due to gas-liquid
interface friction may be higher than the wall friction such as the wavy How or it
may be negligible as in dispersed bubble. Two-phase frictional pressure drop is
highly dependent on the void fraction and the flow regime. A large number of
mechanistic and empirical correlations have been proposed by many investiga
tors; selected ones are presented below.
or
where
(pG and 0L are plotted as a function of X (Equation 11.20) for various turbu
lence and laminar combinations of the gas and liquid phases which they defined as:
Baker [2] has suggested gas-phase multipliers based on the flow regime for
the turbulent-turbulent region as
1. Stratified flow
15400*
0c - (11.54)
H YD RO DYNAM ICS OF VAPOR-LIQUID TWO-PHASE FLOW 1 1.35
2. Elongated-bubble flow
27.315 X 0855
(11.55)
3. Dispersed-bubble flow
14.2 X075
(11.56)
4. Slug flow
1190 X 0815
(11.57)
5. Annular-mist flow
where
11.4.1.2 Beggs a nd B rill [ 6 ]. The frictional pressure drop term o f the Beggs and
Brill two-phase pressure drop correlation is expressed as
(11.59)
where
The two-phase friction factor values were found by solving the total pressure
drop equation for / tp, values of / tp were then divided by the no-slip friction fac
tor. The following relation was reached by regression analysis
(11.60)
f
J ns
where S is defined as
(11.61)
11.36 FLU ID FLO W HANDBOOK
y is defined as:
A
(11.62)
[HLm 2
3 Ql
Q l +Q c
The no-slip frictional factor is obtained from the Moody diagram or, for smooth
pipe, from
Re.
/
J ns
= 2 log (11.63)
4.5223 log Re„s- 3.8215
Re = [PjA ± M I z M Y m R . (11.64)
/ i t A + A/c ( 1 -A )
or simply
GmD
Re... = (11.65)
f i LA + juc (1 -A )
AP _ 2 G j f vm
( 11.66)
L g, D p m
where
(11.67)
F l ^
+ l .281- 0 .4 7 8 / + 0.444y 2 - 0.094y' + 0 .0 0 8 4 3 /4 (I l.69)
y = - ln(A) (l l.70)
= — + £ s i {- l ~ X r ( 11.71)
P,„ Hi. P,„ hg
where X is the input liquid content
The Dukler method has been reviewed by many investigators who have rec
ommended this correlation for general application.
\ 19/8
AP 0.3164 M n2m
t x 114 1+ (1~ A)Pg I (11.72)
L (a i mD iic r 2D p, X £ cP l
where
x Gas-phase quality
£c Parameter defined in Figure 11.13
AF = (n73)
L ScDPl.
where
f
A™ = ,
(11.74)
fu„ Pl M n
ftm Single-phase friction factor using liquid properties and total mass flow;
i.e. Reynolds number is found by
Re^ = % ^ (11.75)
Figure 11.15 shows the / tpm/ / Lm as a function of gas mass ratio, and gas/liq
uid density ration. In a different study Hoogendoorn [17] suggested the following
correlations based on the flow regime.
11.38 FLU ID FLO W HANDBOOK
Equation 11.76 is valid for M g/M m < 0.8. C is a function of pipe diameter
and pipe roughness. Table 11.5 shows C values.
AP _ 2f pCGCs ( I ! ??)
L 8cDPg
where
where
(p Correction factor
Liquid-phase friction factor
Gm Mass flux
Pipe Diameter
in (mm) Air-gas-oil Air-Spindle-oil
E
a
p Gl p Lx 103
_ 64
Re < 1055
L ~ Re
(11.80)
Re
0.86859 In Re > 1055
1.964 ln ( R e ) - 3.8215
\ Pc J Hi.
(11.81)
\P g j W lJ v Hl j
(11.82)
11.40 FLU ID FLO W HANDBOOK
where
PlJ?G
A = ( 1 - jc)2 + * 2 (11.83)
KP g ^ l>
W e= *?~ (11.84)
gcp v
x Liquid mass fraction
EXAMPLE 11.3
Calculate the frictional pressure drop fo r the air-water flo w system in
Examples 11.1 and 11.1.
Solution
_'dP\ _ f pGmVm
vd Z jf 2 grD
where
A 0.023
y (Equation 11.62), y = -----------y = --------- r = 4.2
7 4 [ / / t (0)] [0.074]
S (Equation 11.61),
S = [ln(4.2)]/{-0 .0 5 2 3 + 3.182 ln(4.2) - 0.8725[ln(4.2)]2 + 0.01853[ln(4.2)]4}
= 0.513
f ns (Equation 1 1.63)
( Re.
f
J rts
= 2 log
^ 4.5223 log R ew- 3.8215
-2
1144602.4
2 log = 0.01138
4.5223 log 1144602.4 - 3.8215
Two-phase frictional pressure drop per unit length of pipe, Equation 11.60
' dP_ 0.019x167x110.855
dZ 2 gt D 2x32.174x2/12
Ibf
= 32.8 —^ — = 0.2277 psi I f t = 5 .1 5 kpa / m
ft ft
Dukler correlation
The no-slip frictional factor is found using Equation 11.68 where the Re„, is
the no-slip Reynolds number as defined in Equation 11.65 and calculated
above
7 = - ln(0.023) = 3.77
3.77
F = 1+ = 2.53
1.281 - 0.478x3.77 + 0.444x3.772 - 0.094x3.773 + 0.00843x3.774
Hoogendoorn correlation
The flow regime for this air-water flow is annular as predicted in Example
11.2. Equation 11.77 should be used to find the frictional pressure drop:
AP 2f Gl 2 x 0 .0 0 9 9 6 5 * 8 .1 12 ^ Ibf
= 3.26 — f — = 0.0226 psi / ft
L gcD p c 32.174 jc 2 /1 2 .t0.0749 f r .ft
= 0.512 kpa / m
- f d Ek (11.85)
dZ)„ \d Z ) T
where
Substituting the above equation into Equation 11.45 and then solving for the
total pressure drop term yields
( 11.86)
i- ek
The total pressure drop may easily be estimated provided there a reliable method
to estimate the two-phase acceleration factor EK. Many attempts were suggested
to estimate EK. For exam ple, Beggs and Brill [6] suggested the following expres
sion to estim ate the E K
PoK PLV l
Ek = (I 1.88)
144 gcdP 1- H ,
W ,A V 2 + WgAV^
Ek = (11.89)
288gcQmdP
where
Barua, et al [5J developed the following terms for the estimating the acceler
ation factor.
' c ^
<p
Ek = (11.90)
where
The authors suggested the following formulas to estim ate the critical two-
phase mass flux for separated and homogenous models respectively
a3 (1 -a )3
gL s - 144— —— + (11.91)
k (a ,P ) x -v c ( \ - ) 2x v l
a 1- a
C,;* = 144 — + ------ (11.92)
ks(a ,P ) VVc;
where
a fav O l - a l rd v ,.)
k , ( a %P) = - (11.93)
v G I dP ) < dP J ,_
where
_ (M c (11.94)
ksG = —
vVG dp Kc
(* t )l
(11.95)
dp J , kl
where
Barua et al. [5] have compared their equation with Beggs and Brill for differ
ent mass fluxes and holdup value. It was shown by the authors that the Beggs and
Brill over-estimated the EK factor. Figure 11.16 shows one of the EK factor com
parison charts developed by Barua et al.
0 4 8
F I G U R E 1 1 .1 6 Comparison of the E K values by Beggs & Brill and Barua et al. Source [5] (T =
350 F, P = 500 psig, H L = 0.34 (No-slip)).
H YD RO DYNAM ICS OF VAPOR-LIQUID TWO-PHASE FLOW 11.45
Solution of gas-liquid flow problems is more complex than homogenous flow. For
rigorous estimation of the two-phase pressure drop and other hydrodynamic vari
able such as the liquid holdup and flow regime, the solution should include the
following calculations:
1. Given the fluid composition, temperature, pressure, and mass flow rate at
the pipe inlet, PT flash is performed to determine the gas and liquid flow
rates and compositions. Mixture enthalpy is also required which may be
calculated as
Hm = Z H g + ( \ - Z ) H l (11.96)
where
2. Estimate the liquid and gas properties at the increment inlet temperature,
pressure and composition. For simplicity, the gas and liquid properties may
be assumed to be constant for the entire length of the increment under
study. Therefore, the smaller the increment length, the more accurate the
gas-liquid physical properties are.
3. Calculate the heat transfer rate between the fluid and the surroundings
based on the flow assumption. If adiabatic flow is assumed, Q = 0.0
11.46 F LU ID FLO W HANDBOOK
BTU/hr-ft2; otherwise the net heat transfer rate must be calculated. The
mixture enthalpy at the exit of the length increment is calculated using the
following formula
H M2 = H m + Q A c + Qf (11.97)
where
4. Select a correlation to predict the flow regime, liquid holdup, and total
two-phase pressure drop. The increment outlet pressure is known now.
5. Perform PH (pressure-enthalpy) flash at the increment exit pressure found
from the step above and the increment exit enthalpy found in step 3. The
PH Hash will determine the increment exit temperature and gas- and liquid-
phase compositions.
6. Repeat the above steps 2-5 for the next length increment starting with the
last increment pressure, temperature, and composition.
The accuracy of the above algorithm may improved if the physical properties
are calculated based on the increment averaged inlet and outlet pressure and tem
perature. The reader is referred to Chapter 9 of this book where similar algorithms
have been explained in detail for compressible flows.
REFERENCES
12.1 INTRODUCTION
12.1
12.2 FLU ID FLO W HANDBOOK
72.2 DEFINITIONS
TU= ( 12. 1)
where i j represent the coordinate directions associated with the ij component. Both
the shear stress (r,y) and the shear rate(%y) are tensors with six independent compo
nents and /x is the (Newtonian) viscosity. In most practical applications, we are con
cerned only with one of these components, as illustrated in Figure 12.1. This “simple
shear experim ent” illustrates the physical significance of the shear stress com po
nent Tyx and the shear rate component y y x :
( 12 .2 )
(12.3)
The shear rate (or rate of deformation) is equivalent to the (local) velocity gradi
ent. In most steady uni-directional flows, only one com ponent o f the shear stress
1A y
Y
h y
X
S IM P L E S H E A R
Shear Stress: ryx = p / a
x
and shear rate tensor is important, which we designate as ran d y, respectively. Thus,
for uni-directional Hows, a Newtonian fluid follows Equation 12 .1 in the form:
r = fi y (l 2.4)
where /jl is the fluid viscosity, and is a constant (independent of flow conditions or
stress levels). Fluids with “simple structure” are typically Newtonian. By “simple
structure” is meant simple molecular or microscopic structure, such as relatively
small molecules, or relatively dilute multi-phase fluids such as very dilute suspen
sions or dispersions o f immiscible solids or liquids in a simple carrier fluid. Gases
and liquids such as water, alcohols, acetone, gasoline, diesel fuel, non-polymeric
oils, etc. are typically Newtonian.
Fluids with “complex structure” are usually non-Newtonian. This includes
“multi-phase” fluids such as concentrated solid suspensions, emulsions, and foams,
as well as complex molecules such as high polymer melts and solutions. Exam
ples range from toothpaste to blood, catsup, shave cream, sewage sludge, paint,
coal slurries, fine sediments, dilute solutions of high polymers, etc. For such flu
ids, Equation 12.4 is obviously not valid because the shear stress (r) is a more
complex function of shear rate (y). This function can be expressed as a relation
between r a n d y, or equivalently, as a non-Newtonian viscosity function ( 17):
~T = y ( y ) or 7](t) (12.5)
y
It is evident that the viscosity r](y) in Equation 12.5 is defined the same as /x in
Equation 12.4, except that for a non-Newtonian fluid the viscosity is a function (of
either y or r) instead of being a constant as for a Newtonian fluid. The actual math
ematical form of this function will depend upon the nature (i.e. the “constitution”)
o f the particular material. Some very common examples of non-Newtonian flu
ids are mud, paint, ink, mayonnaise, shaving cream, dough, mustard, toothpaste
and sludge.
Actually, some fluids and solids have both elastic (solid) properties and viscous
(fluid) properties. These are said to be viscoelastic and are most notably m ateri
als composed o f high polymers. The complete description of the rheological
properties of these materials may involve a function relating the stress and strain
as well as derivatives or integrals of these with respect to time. Because the elas
tic properties of these materials (both fluids and solids) impart “m em ory” to the
material (as described previously) which results in a tendency to recover to a pre
ferred state upon the removal of the force (stress), they are often termed “mem
ory materials” and exhibit time dependent properties.
12.4 FLU ID FLO W HANDBOOK
T
T; = T = ( 12.6 )
27tR}L
to
•CH
j
4J &
k
R) 0)
O £ c
0 Ew
U
CD o ®
•H
>
Q) 4-1
Q aIt
(0 -H
3 O
0»
00) QH
•H
♦ >
C Jh
O
1 u CO CO
o Q
u M
w a g J
M o
> Ik CO
U q
Ih
O
Nx
cf c |S
Mh l!
h»I K «!
II
O *8
•rf
%
CD
Classification of Materials
o
x •H
4->
CO
ft
J?
CD
£
G 5
H oH
o
12.1
►J
o
CO
o
w
TABLE
12.5
12.6 FLU ID FLO W HANDBOOK
and the shear rate at the bob is determined from (Darby [ l ]):
2ft
ri (1 - / T " )
where:
n’ = d lo g T (12.8)
d log f t
is the point slope of the log-log plot of T versus ft, at each value of ft, and /3 =
RifR0. If the gap is small (i.e. /3 > 0.98), the following “small gap” equation may
be used for y,:
Another common viscometer is the tube flow or Poiseuille viscometer, in which the
pressure drop (A<& where 4> = P + pgz, which includes any static head change)
and the flow rate (Q) are measured for laminar flow through a tube (diameter D,
length L). The shear stress at the tube wall is related directly to the pressure gradient:
-A <t>D
Tw=— — (12.10)
4L
Here F is the wall shear rate for a Newtonian fluid (n ' = l),
r = J 3 2 £ = 8y (12.12)
ttD D
and ri is the point slope of the log-log plot of-A<I> versus Q, evaluated at each
data point:
This value of ri corresponds to the same parameter determined from the cup and
bob viscometer for a given fluid. As before, if ri = 1 (i.e. A4> - Q), the fluid is
Newtonian. The viscosity is given by 77 = Tw/ y w. Many viscometers utilize flow
from a reservoir through a capillary tube in which the measured pressure is that in
the reservoir. In such cases, there are correction factors that must be applied for the
entrance effects from the reservoir to the tube (e.g. Darby [2]).
The viscosity function 77( 7 ) takes various forms, depending on the nature of the
fluid. Some of these are illustrated in Figure 12.3 (as t v s y) and in Figure 12.4 (as
77 vs y). The classes of behavior illustrated are the Bingham plastic, pseudoplastic,
dilatant, and structural. Both the Bingham plastic and the pseudoplastic are shear
thinning, since the viscosity decreases with increasing y, whereas the dilatant
fluid is shear thickening. The structural fluid exhibits Newtonian behavior at very
low and very high values of y and is shear thinning at intermediate shear rates.
f Bingham Plastic
(0
(0
0)
k.
4-»
0)
k.
(0
o
-C
V)
Dilatant
Shear Rate
F I G U R E 12.3 Classes of non-Newtonian
fluids-shear stress versus shear rate.
12.8 FLU ID FLO W HANDBOOK
Pseudoplastic
CA gingham Plastic
O
O Structural
</>
Dilatant
--------------------------- -— _ ►
Shear Rate
F I G U R E 12.4 Classes of non-Newtonian fluids-
viscosity versus shear rate.
A “plastic” is a material that exhibits a yield stress ( t0), which means that it behaves
as a solid below this stress level and as a fluid above the yield stress. Many concen
trated two-phase suspensions exhibit such properties, such as paint, ink, toothpaste,
shave cream, catsup, various slurries, etc. The simplest model that incorporates
this behavior is the Bingham plastic:
For t < t0 : y = 0
The two parameters that determine the viscosity of this material are the yield stress
( t0) and the limiting viscosity (/x*>). The Bingham plastic approaches Newtonian
behavior with a viscosity of /x<» as the shear stress becom es large relative to t0.
Note that a log-log plot of 77 versus y approaches a slope o f - 1 as y gets very
small, and a constant value equal to as y gets very large. The viscosity can also
be expressed as a function of shear stress instead of shear rate, from Equation 12.14:
He
7J(t) = (12.15)
Another popular and versatile model for non-Newtonian fluid behavior is the power
law model:
n-
•n / •\ I • | 1 \/n \ \—_1
t - my or ri(y) = m \ y \ = m \r\ n (12.16)
NON-NEWTONIAN FLUID S 12.9
where m is the consistency parameter and n is the flow index (m is also the viscos
ity at y = l s~ ) . Data for either t v s y or 77 vs y which are linear on a log-log plot
can be represented by this model. Note that if n = 1, the model reduces to a New
tonian fluid with m = /x. If n < 1, the fluid is shear thinning (as is the Bingham plas
tic, hence the designation pseudoplastic), and if n > 1 the fluid is shear thickening
(or dilatant). It should be noted that for a shear thinning fluid, this model predicts a
continuously decreasing viscosity as y increases, and a continuously increasing vis
cosity as y decreases, both of which are physically unrealistic in the limit. For this
reason it is not recommended that this model be used outside of the range of experi
mental data for which the parameters were determined. (This is not true for the Bing
ham plastic model, however, as it does approach physically realistic behavior at both
the high and low y limit. However, some materials that may appear to have a yield
stress may actually approach Newtonian behavior at sufficiently low values of y
(see below).
The structural fluid shows more complex behavior, and thus requires a more com
plex model. This type of behavior is typical of fluids with a “structure” which forms
in the undeformed state but which breaks down as y increases. The structure then
re-forms as the shear rate is reduced. Many fluids show this characteristic, such as
flocculated sediments, latex suspensions, solutions of entangled (high molecular
weight) polymers, blood, etc., as illustrated in Figure 12.5. Each of these fluids
exhibits a “power law” region at intermediate shear rates, but deviates from this
behavior at both upper and lower limits of shear rate. One model which well rep
resents this type of behavior is the Carreau-Yasuda model [4J:
T?=T?“ + - [ 1 + W r r (1 2 J 7 )
The parameters rfa and 7ja are self-explanatory. The time constant parameter, A, is
the reciprocal of the value of y at which the low shear viscosity intersects with the
power law region of the viscosity curve, which has a slope o f - 2p, and the param e
ter a is a measure of the range over which the transition from low shear Newtonian
to power law behavior occurs. The time constant A represents the characteristic time
12.10 FLU ID FLO W HANDBOOK
Viscosity (Pa s)
Vo A P n*
Solution (Pas x 10) (s) (Pas X 1000)
required for the fluid to respond to a deformation rate (i.e. shear rate). For a - 1,
the model is called the Carreau model [5].
For many fluids, data over 6 to 8 decades of shear rate may be required to
clearly define this complete function and to enable determination of all five para
meters. When this range o f data is not available, sim pler models which are spe
cial cases of this model may represent limited regions of the data, as follows [3]:
a. Low to Intermediate Shear Rate Range—If t]x « ( 77, tj0), the Carreau model
reduces to a three parameter model ( t]0, A, and p ) that is equivalent to a power law'
model with a low shear limiting viscosity, also known as the Ellis model:
NON-NEWTONIAN FLUID S 12.11
F I G U R E 12.5b Viscosity data and Carreau model parameters for structural polyacrylamide solu
tions (Darby [3]).
12.12 F LU ID FLO W HANDBOOK
( A y 2)
W y ) = — -rh — (12.20)
(A2y 2)
where the power law parameters m and n are equivalent to the following com bi
nation o f the Carreau parameters:
m n = \-2 p (12.21)
d. Bingham Plastic Behavior—If the value of p is set equal to 1/2 in the Sisko
model, the result is equivalent to the Bingham plastic model:
where the yield stress r0 is equivalent to r^/A, and r)x is the limiting (high shear) vis
cosity (/Xc*). In practice, extensive data are usually required to identify enough detail
of the viscosity function to justify anything more complex than a two-parameter
model. Also, the two-param eter models represented by the Bingham plastic and
the pow er law are often quite adequate to describe flow behavior in many fluids
in laminar and turbulent pipe flow, particle settling in non-Newtonian media, etc.
Thus we shall concentrate on these two models for describing pipe flows.
A3>r
TyJc = — =~T„- (12.23)
~2T " R
NON-NEWTONIAN FLUIDS 12.13
where <t> = P + pgz. That is, the local shear stress varies linearly as the distance
from the centerline. Also, from the definition of the volumetric flow rate Q :
Q = 77- f vv d ( r 2) = - 7 T I r 2dvr = - t t ( r2 L dr - - i t I r 2 y rx d r
ja ja o dr Jo
( 1 2 .2 4 )
which follows from integration by parts, and relates the flow rate Q to the local shear
rate, y . Thus, for laminar flow, any viscosity model that relates the shear stress (r rx)
to the shear rate ( y ) can be used to eliminate r rx and y rx from Equations 12.23 and
12.24 to obtain a relation between the flow rate and pressure drop.
For example, if the fluid is Newtonian, Trx = f i y rx. Using this to eliminate rrx and
y rx from Equations 12.23 and 12.24 gives:
4M 8fiL 128\fiL
which is known as the Hagen-Poiseuille equation and applies to the steady laminar
flow of an incompressible Newtonian in a uniform tube. It can be written in dimen-
sionless form as follows:
/= — (12.26)
Re
where
R e = D ¥ P = 4QP- (12.28)
/x TrDfi
is the Reynolds number. Note that the Bernoulli equation for pipe flow can be
written -A4>/p = ef , where ef is the “friction loss” (i.e. the energy dissipated per
12.14 F LU ID FLO W HANDBOOK
unit mass o f fluid). The friction loss is characterized by the loss coefficient, Kf,
which for pipe flow is related to the Fanning friction factor,/, by:
Kf = — 2 ~ (12.29)
f Vi D
since both the shear stress and shear rate are negative. Using this to eliminate rrx
and y rx from Equations 12.23 and 12.24 gives:
/= — (12.32)
RCpi
where
= %PnV 2 ~ np = 27 - 3”p Q 2 - ”
pl m[2(3n + 1)/«]" ttnr2 “ "D4 “ 3"[(3n + 1)/«]”
For a Bingham plastic, the situation is somewhat different. Because of the finite
yield stress, there is no relative motion for \rrx\ < t0 , which means that there is a
solid-like “plug flow” region from the pipe centerline (where rrx = 0) to the point
where —r rx = r0> i.e. at r = r0 = - R t0I t w (note that r rx is negative). The result is
that the flow integral, Equation 12.24, becomes:
Q = - i r f * r 2y „ d r (12.34)
NON-NEWTONIAN FLUID S 12.15
Q =
4/A,*
oc 1- 1
3 Vtu.
(12.35)
_16 1 He 1 He
/ = + — (12.36)
Re 6 Re 3 / 3/te
DVp
Re = (12.37)
jU cc
and
D 2pT„
He = (12.38)
f-hoo
Since it is not possible to analyze turbulent flows rigorously from a purely theo
retical perspective, it is necessary to rely on data and suitable generalized dim en
sionless correlations for the required relations. The most useful of these relations
are given below.
12.16 FLU ID FLO W HANDBOOK
The friction factor for a Newtonian fluid in turbulent flow is a function of both the
Reynolds number and the pipe relative roughness, e/D. This relation is commonly
represented graphically on a plot called the Moody diagram, which is based on data
from artificially roughened surfaces using sand grains. (For this reason, the wall
roughness is sometimes referred to as the “sand grain roughness” .) Roughness val
ues for various pipe surfaces can be found in various books and handbooks (e.g.
Darby [3]). The turbulent part of the Moody diagram (for Re > 4000) is accurately
represented by the Colebrook equation:
l e/D 1.255
= - 4 log (12.39)
V f 3.7 + R e V f
Note that if the Reynolds number is large enough, the friction factor depends only
on e/D:
e/D
= - 4 log (12.40)
V f, 3.7
1/12
1
/= 2 3/2 (12.41)
where
16
A= 2.457 In (12.42)
0.276
D
and
16
B= (12.43)
Re I
Note that the Churchill equation is explicit in / , but the Colebrook is implicit.
NON-NEWTONIAN FLU ID S 12.17
For non-Newtonian fluids, the laminar boundary layer is typically much larger than
for Newtonian fluids, which means that pipes that are “rough” for a Newtonian
fluid may be “smooth” for the non-Newtonian fluid. Thus, it is generally concluded
that wall roughness is not significant for such fluids. Consequently for a fluid that
obeys the power law model the friction factor is dependent only upon the power law
Reynolds number (Equation 12.33) and the flow index. An equation was developed
by Darby and Chang [7] based on data from the literature which represents this
relation for the entire range of conditions, for all values of the Reynolds number
and flow index:
a
f= ( \- a ) fL+ (12.44)
where
(12.45)
0.0682n m
fL= (12.46)
,0.414 + 0.747/1
f , r = 1.79 X 10“4 exp (-5.24/7)Re"j (12.47)
a = (12.48)
and
is the value of the Reynolds num ber at which transition from laminar to turbulent
flow occurs.
For a Bingham plastic, the turbulent friction factor is a function only of the Reynolds
number (Equation 12.38) and the Hedstrom num ber (Equation 12.39). Darby and
1 2.18 F LU ID FLO W HANDBOOK
Melson [8] developed a correlation based on data from the literature that applies
for all values of the Reynolds and Hedstrom numbers, i.e.:
f = { f L m + f r m) Vm (12.51)
where f L is the laminar friction factor (Equation 12.36) and f T is the turbulent fric
tion factor, given by:
10"
fr = (1 2 -5 2 )
and
7 +i -----------
m = i1.7 40’000 (1 2 .5 2 )
Re
These equations for both the power law and Bingham plastic models have been com
pared against pipe flow data for a variety of slurries and suspensions by Darby et
al., [9] and found to agree well with available data [9]. In many cases, the extent of
viscosity data is limited to three decades of y or less, and it is possible to fit either
the power law or the Bingham plastic to these data with comparable precision. How
ever, given the choice, the Bingham plastic model is preferred for slurries and sus
pensions because such fluids usually exhibit a yield stress (to varying degrees), and
it is more physically realistic for extrapolation beyond the range of measured data,
as explained above.
The relations given above apply to purely viscous non-Newtonian fluids in steady
laminar or turbulent flow in pipes. However, there are other “non-Newtonian”
fluid properties in addition to the non-Newtonian viscosity function that can also
have a pronounced effect on the flow behavior in pipes. These result mainly from
the viscoelastic properties that are associated with very large molecules (e.g. high
polymers) which entangle and interact extensively and impart elastic properties to
the fluid. These elastic properties do not affect the steady laminar flow behavior,
which is controlled by the viscous properties only. However, turbulent flow is a
NON-NEWTONIAN FLUID S 1 2.19
locally unsteady, fluctuating stochastic process that does interact with the fluid elas
tic properties in some remarkable ways.
For example, the pressure drop in a pipe with a viscoelastic fluid in turbulent flow
can be a small fraction o f that for a purely viscous fluid with the same viscosity at
the same flow rate in the same pipe. This effect can be seen in some very dilute solu
tions (measured in parts per million) of certain very high molecular weight polymers,
and is referred to as “turbulent drag reduction,” or the Toms effect. This effect can be
explained by the fact that viscous properties dissipate energy which must be contin
uously added to the fluid to sustain the flow, whereas elastic properties store energy
in unsteady deformations. Thus, the energy stored by the fluid elasticity remains in
the fluid and sustains the turbulent motion so that less external energy must be added
to maintain the flow. This can be described quantitatively for dilute polymer solu
tions by incorporating an additional dimensionless term in the definition of the
friction factor which accounts for the energy storage effect, i.e.
fs (12.55)
where f [} is the friction factor for the viscoelastic (polymer) solution and f , is the
friction factor for the Newtonian purely viscous solvent at the same flow rate in the
same tube (i.e. the same solvent Reynolds number, Res), and De is the Deborah num
ber, which is a dimensionless time constant for the fluid relative to a characteristic
time of the energy dissipating turbulent eddies. The Deborah number is a complex
function of the fluid rheological properties and the turbulence characteristics, but can
be represented by the following simplified equation for aqueous polymer solutions:
(12.56)
where the parameters k x and k2 depend only on the specific polymer and its concen
tration. A table o f values for k\ and k2 for a variety of polymers obtained from drag
reduction data in the literature for these polymers is given in Table 6.2, i.e. Darby
and Pvisa-Art [ 10].
The friction loss, eft (or irreversible pressure drop, - A Pf = Aej) in valves and fit
tings is characterized by the dimensionless loss coefficient, K/.
(12.57)
0) 00
O C
rH c
e tyi a; _
AJ
>
i~ii * f0 r—,
<u '
*™% n £ in
M C «-<
U ° u ^
4-i m
0) iti H
* i H£ V4
_,_
*4-1 2C — Q >
& •H o
<U
PC S
0.
0.
0.
0.
0.
o o o o
0.
0.
o o o o o o o o
o
ro i/) m oo in m . . o rn
O O O ^ rH O O O rH C
O
Tl CT\ O 00 CT\ O *_7
O ^ O O rH CM to
Parameters for Equation 12.57 for Various Polymer Solutions
o o o o o o o o o
O O O O O O o o o o
r- VO t—4 O m ^0
(M r- 0 CN CTl <T O
»
— < O O a\ o •
a '
8 • 11 » • • ■ ro
o rH fH o Csl
• O' o o o o o o o o o o o o o o o o o
u x; o m o o in o r-H rH O D o o
a ' cm m
o o o
o
O O' <nj m h cj m h in cj no r-
O £
Q «u
1 'Of O
ro
o
ro
o •r* i i T>
C
M e — OD
< Si X!</> S T>0) ro <y>
15 'H r- »-h
<d cm w
H M M c<0 a> a
«o cr l
(0 <0 <0 O ^ S
D <0 td &4 »-i 0)
to Q
3 *71 CJ> a a s§
TABLE 12.2
O a> V
04 w
12.20
«h
fO H-l
VJ Vh
•H c — M^ o
cs H r*- <y 00 M cn XJ
> rH M t-H •H rH w
4-1 (0 — O
c/> o > — <D
< 5
o
CQ
m 00 rH eg CO PO o to
rH VD r- r - tp xH
o Oo o o o o m
o o
o o 1o• o o o o o
• * • « .
o Q o o o o o o o o o o
o
TABLE 12.2 (continued) Parameters for Equation 12.57 for Various Polymer Solutions
m in
CTk rH o m CO
m
o
CD
O
CM
to
H ON
<N <N * (N in *
o V0 i—1 O
IT
)Oo o oo o o o o o O in o
M mo O O O O t£> H CN Ul H O o
i (N m rH C\J i n o oo o
CM
<l> cn
> a
x: a> (\J
r- rH a
cri <i> § > lO id
m > 0 U x:
rH w o ■p
( x 1 4J
o <y c
Darby and Pivsa-Art
u > to
T5 O ’—i X
> o
sc CL
12.21
12.22 F L U ID FLO W HANDBOOK
For Newtonian fluids, this loss coefficient depends on both the Reynolds number
through the fitting and the size or scale of the fitting. The most accurate correla
tion for this dependence is provided by the “3-K Method” of Darby [3]:
<l258)
where Dn in. is the nominal diam eter o f the fitting in inches, and the values of the
three “AT’s” are given in Table 6.3 for a variety of valves and fittings.
There are very little data in the literature for friction loss in fittings with non-New
tonian fluids that are suitable for validating any generalized correlations. However,
in accordance with other similar relations, it is logical to conclude that a function
similar to Equation 12.58, appropriately modified to apply to non-Newtonian fluids,
should be applicable. This can be done if the Newtonian Reynolds number is replaced
by a comparable group for the non-Newtonian fluid. For power law fluids, we
have seen that this group is given by Equation 12.33, i.e.
For a Bingham plastic, two dimensionless groups have been used to describe pipe
flows, i.e. the Reynolds and Hedstrom numbers (Equations 12.37 and 12.38). How
ever, it is possible to combine these into one group by the following reasoning.
The Reynolds number for a Newtonian fluid can be written:
Re = £ ¥ p = J t L
\x fiV /D
=
M l
rw
(i2 .60)
where tw is the wall shear stress, equal to /x(8VZD). For a power law fluid, introduc
ing t w from Equation 12.30 into Equation 12.60 results in Equation 12.59 for the
power law Reynolds number. For a Bingham plastic, the wall stress is given by
t w = t q + /Xoo7w where y w is the shear rate at the wall, which is not a simple func
tion of the flow properties. The wall stress is also related to the flow properties
by Equation 12.35, which however cannot be solved explicitly for rw. However,
if the last term in Equation 12.35 is neglected, it can be solved for rw to give:
•^r
O rH o CM o cr> (? ) CS| o t-if CM <N o o o o O
O • • • « • • • • • • • • • • • •
00 o <r vr ro co •'31 vr vr •^r •tr •^r •^r vr
rH »H VO vo m 00 m »—i CM VO CM
r~~ CT) in VO r~ r~ vp ro m 00 in ro CM
0 .1 0
o O o o o (M o o o o O o CM i— )
• • • • • • • • • • • • • •
o o o o o O o o o o o o o O
o o o o
ro o o o o o o o o o o o o o o
1000
o o o o o o o o o o o o o o
00 00 00 00 00 «H 00 00 in m m m rH »—1
*0
<b
u Q VOO CMT r* O m 00 V£
> m o
c; rH C
Mi—
1«H rH VOrH rH rH vo in
q
o c
H
Q m m
1 .5 )
<
D •..—.-
— .-
•
— .-
— . -—, • U
\ •u rH tH cm vo rH rH Q
QJ
II II II II II II ii o
=
*
in
3-K Constants for Loss Coefficients for Valves and Fittings
+ Q Q Q Q Q 0 0 0 Q Q 0 • T3 Q
*0 N N \ \ \ o m O \ \ in CM ^ \
(r/D
MM M M M cr» ro M M CM jjj M
<TJ ~ CQ ~
Cl XJ
•H M T
> g
03 M
t I
C
XJ
G
<
/
(/) T3
3 C
)
<0
X)
w p
4J
f0 •
rH <V V) V) C rH 0)
Q) 4-J XJ C Q X>T3 TJ (d a) cr:
C
O <Q <
— i •
—i rH ■ P to S
■u O
h 0) a> a; 3 0)
X) 3: s s •H rH w
V) X) Cna > XJ XJ o
•H <U c X) rH CMro <D < 0XJ '—I
T3 o rH XJ 0) (J
aJ 0) (0 u
a> s 0) C T <D
M c •P T3
XJ T) v x : o •H <D
0) T3 T3 H 2 T3 X3
X) a) 0) f0 0)
<0 O' M <U D»
0) a 0) M
M (0 4J XX C (0 .
XX •— i •H
H Pm X
o o o
O m O
Ch ■'3* CO
TABLE 12.3
s w
r-i 5
4J o
■U
H
w
12.23
o o o o O O o o IX) o CTi o a\ m< T
>o 00
• • • • • • • • • • • • • • • • •
sr ■
^r CO TCO cn
CO •< vT CO
'sr rH r* r~ m
r^ 00 T cn «H O m (n o rH 00 co rH a\ VDin
CM rH CM CO o O CM <D r- o rH o o V£> •«r 00
• • • • • • • ■ • • • • • • • • •
o o Oo o o o O i—i o o o o o o o CM
o o o o o O
o o o o o o O o o o o o o o o o o O
o o o o o in o m o mo o o o o o m O
m 00 00 rH CM rH rH o\ rH rH m CO CO co CO rH «H CNJ
10
Q)
*5 Cj O
g o O in o o o oo cd co o o
\ C
M C
M m m (Ti h o o o
t - i CO fO i V
f— O
C
H
3-K Constants for Loss Coefficients for Valves and Fittings
o c
Q ID ,c
<D o>
•u
"O QJ o
M
* e ,G
■P
+ .g
tJ C
Q.C
O.h tr rH i—I
P 0
« a
rP A
*TJ 0 rH II (I
a; a> M HH
* G N Nc£l 3 .G —' CQ. CO.
•H •H •H O H a •—
a<
W V r—i V s i—j
§ XJ A TJ 4-> <0 X3 T3d m o
c
* u u <V A SUM co *r
Q) G G G <0 A O' 1 <ti <ti
\ 3 (0 <d •H U •H <D X) T3 H || u
O M M ►H G G (0 0) G G
■u
i— i CQ JQ 03 M M (0 (0 f i c c
<U X) T3 «H «H 4-> P .G P P S a - a
•H <u T3 G d) X> G (/) CQ CO H W W q > >
U) 73 <1) H a> •H u<
«t3 (0 O' I Cn I Ph V M
a> G jQ <D G A o o > <L> <v <D O
nJ p M H p m O 1—1 > > > e a> u
,G x: •P rcJ rH r—i r—iD> 0)
0 H CO (S) > «T3 <d (0 (0 u .c
c I 1 > > > M o
A <D <v A Cn
M Cn > O' 0) rH CX G 4-»
CQ 3 rH 0 P p rH (0 •H
1 O (X) <H r—i (0 •H * •H
TABLE 12.3 (continued)
A M > Ch cn Q cn
O' ,G
O d>
o rH
u G cn
a 3
h a: 5
G w
*h 0)
■u 10 >
•u a; H
a> «3
£ H > Q
12.24
NON-NEWTONIAN FLU ID S 12.25
Inserting this into Equation 12.60 results in the following expression for the
equivalent Bingham plastic Reynolds number:
ReBP = , R * (12.62)
1 + Bi/6
B i = — = —^ (12.63)
Re VfJLoc
It therefore seems logical that the 3-K expression for friction loss in valves and
fittings should give reasonable results for a Bingham plastic fluid if the Reynolds
number in Equation 12.58 is replaced by Equation 12.62.
Many slurries and suspensions of solid particles in a Newtonian liquid are non-New
tonian if the solids concentration is sufficiently large (e.g. cw > 0.05). If the parti
cles are sufficiently small and or light the suspensions may be stable (non-settling),
and the suspension behaves as a pseudo-homogeneous continuous fluid with non-
Newtonian properties. The flow behavior of such system can be described by the
usual non-Newtonian models, such as the Bingham plastic or power law model. For
larger and/or denser particles, the solids will tend to settle and the suspension is het
erogeneous, in which the particle-fluid interactions must be considered separately
from the fluid-pipe interaction. Also, many slurries contain a broad distribution of
particle sizes, in which the fines form a stable, non-Newtonian continuous medium
through which the larger particles will tend to settle out. The fluid-particle inter
action for such systems is normally characterized by the drag coefficient, which
characterizes the force exerted on or by the fluid and the particle:
F J = C o A p V 2 ( l 2 6 4 )
Equation 12.64, the resulting velocity is the terminal velocity o f the particle fall
ing at steady state in the fluid:
(12.65)
For creeping flow in a Newtonian fluid (i.e. Re < l), to about ±2% , CD is given
by the Stokes equation:
( 12.66 )
where
(12.67)
In Stokes flow, two-thirds of the drag results from viscous drag and one-third from
form drag, i.e. the non-uniform pressure distribution around the particle. For higher
Reynolds numbers, a turbulent wake forms behind the particle in which the pres
sure is low and which results in increasing form drag. For Reynolds numbers above
about 1000, the drag is dominated by form drag and the drag coefficient is approxi
mately constant. If the fluid is Newtonian, the drag coefficient can be determined to
acceptable accuracy over the entire range of Reynolds number (up to about 2 X 10 5)
by the Dallavalle equation [3]:
( 12 . 68 )
where
is the Archimedes number. The diameter of a particle that will fall at a given veloc
ity can also be determined by using the group CD/Re, which is independent of d (from
Equation 12.65), dividing Equation 12.68 by Re and solving for \ ^ R e to give
24
CD = X — — (12.72)
Repl
where
and X is a function of the flow index, n. For the entire range of Reynolds number
and fluid properties, Darby [21] has shown that the drag coefficient can be deter
mined from the following equation which is analogous to the Dallavalle equation:
— ( i 2 -74>
and
l .33 + 0.37/?
x = (12.76)
I + 0.7m3'7
2 / A A \2 - n
P \ / 4gA p\ in + 2
Nd = C £ ~ " 1 Repl (1 2 .7 7 )
Xm
2(2 - n)
1/(2 - n) \ n / 2 ( 2 - n)
ReP l Re P l
N,{ = C, + 4.8 (1 2 .7 8 )
X X
This is implicit in Rept (i.e. the dimensionless V), which must be determined from
Equation 12.78 by iteration. When the value of Repl has been found, the terminal
velocity is given by
1/(2 - /,)
mRe pi
V,= (1 2 .7 9 )
pdn
The diameter of a particle that will fall with a specified velocity can be determined
by first calculating the value of the following group which is independent of d :
^ CD
n X _ / 4 g A p \ 7 Xm \ 2)
(1 2 .8 0 )
Re pi
2n
X X
C, + 4 .8 1 (1 2 .8 1 )
ReP l Re pi
This can be solved for Rept by iteration (or using a nonlinear equation solver), which
is the dimensionless diameter:
1In
mRe Pl
d = (12.82)
pV 2 -/i
NON-NEWTONIAN FLU ID S 12.29
Most highly loaded slurries with a homogeneous component of fines can be best
described by the Bingham plastic model. For a large particle to settle in the Bing
ham plastic medium, the weight of the particle must be large enough to overcome
the yield stress which supports the particle. Since the stress distribution is not uni
form around the particle, the yield criterion cannot be determined exactly, but can
be represented in terms of a dimensionless “gravity yield” parameter:
Yg = ~TT~ (12-83)
gdAp
Equating the vertical component of the yield stress over the surface of the particle
to the weight of the particle for a sphere gives a value of YG = 0.17 for the critical
non-settling condition. However, experimental data appear to fall into two ranges:
one for which YG « 0.2 and the other for which Yc ~ 0.04-0.08. This may be due
to the manner in which the yield stress is measured, since for some materials, the
value determined from “static” measurements differs from the value determined
under “dynam ic” conditions.
The drag coefficient on a sphere in a Bingham plastic should be a function of
the Reynolds number and either the Hedstrom or the Bingham number. However,
we can define a Bingham Reynolds number for the sphere using the concept of the
ratio of momentum flux to the wall stress, as was done for flow through fittings.
For example, in creeping (Stokes) flow, the force per unit area of the sphere is
F_= C o p V ^
A 2
R e = dVp= \2 p V 2 = 12pV2
fi fi(\2 V /d ) t
Introducing the expression for the shear stress for a Bingham plastic, this becomes
Re _ 12p V 2 = Re
( 1 2 .8 6 )
€bp~ n J \ 2 V l d ) r „ ~ \+ B H \2
12.30 FLU ID FLO W HANDBOOK
If the effective shear rate over the sphere is taken to be V/d, the equivalent expres
sion is
ReBP= (12.87)
1 + Bi
Correlations based upon experimental data (e.g. Chhabra and Uhler) can be rep
resented by the following equivalent Bingham plastic Reynolds number [22]:
-
1 + 2.93#/
4 8 V /2
CD = 10.632 + j-* (12.89)
V r e bp
the particle terminal velocity or the diameter of the particle that will fall with a given
velocity can be determined by a procedure similar to that used for the power law.
Because of the implicit nature of the equations, the most straightforward proce
dure is to equate Equation 12.89 to the definition of CD from Equation 12.66 and
solve for either V or d by iteration (or using a nonlinear equation solver).
NOTATION
GREEK
Ap ps - p, [MIL?]
A fluid time constant parameter, [/]
/jl viscosity (constant), [M/Lt]
Bingham limiting viscosity, [M/Lt]
7] viscosity (function), [M/Lt]
p fluid density, [Af/L3]
ps solid density, [MIL?]
t shear stress in uni-directional flow, [FIL2 = M /L t2]
Ty shear stress component, [F/L2 = M /L t2]
t 0 yield stress, [F/L2 - M /Lt2]
Tyx force in x direction on y surface (shear stress, or r), [F/L2 = M /L t2]
4> flow potential = P + pgz, [F/L2 = M /L t2]
(2 angular velocity of cylinder, [ \/t]
co eddy frequency, [/]
SUBSCRIPTS
1 reference point 1
2 reference point 2
i inner
L laminar
0 outer
0 zero shear rate limiting parameter
P polymer (viscoelastic) solution
T turbulent
tr transition
w value at wall
00 high shear rate limiting value
6 coordinate directions
00 high shear limiting parameter
REFERENCES
1. Darby, R. 1985. “Couette Viscometer Data Reduction for Materials with a Yield
Stress,” J. Rheology, 29, 369.
2. Darby, R. 1976. “Viscoelastic Fluids,” Marcel Dekker.
NON-NEWTONIAN FLUID S 12.33
3. Darby. R. 2001. “Fluid M echanics for Chem ical Engineers,” 2nd Edition, M arcel
Dekker.
4. Yasuda, K. et. al. 19 8 1. “Shear flow Properties o f Concentrated Solutions o f Linear
and Star Branched Polystyrenes,” Rheol. A c ta , v. 20, p 163.
5. Carreau, P.J. 1972. “Rheological Equations from M olecular Network theories,”
Trans. Soc. R heol., v. 16, p 99.
6. C hurchill, S.W . Nov. 7, 1977. “ Friction Factor Equation Spans All Flow R egim es,”
Chem. Eng., p. 91.
7. Darby, R. and H.D. Chang. 1984. “A G eneralized Correlation for Friction Loss in
Drag Reducing Polym er Solutions,” A IC h E J ., 30, 274.
8. Darby, R. and J. M elson. Dec. 28, 1981. “ A Friction Factor Equation for Bingham
Plastics, Slurries, and Suspensions,” Chem. Eng., p. 59.
9. Darby, R, R. M un, and D.V. Boger. Septem ber 1992, “Predict Friction Loss in
Slurry Pipes,” Chem. Eng., p. 116.
10. Darby, R. and S Pivsa-A rt. 1992. “An Im proved Correlation for Turbulent Drag
Reduction in D ilute Polym er Solutions,” Canad. J. Chem. Eng., 69, 1395.
11. W ang, C.B. 1972. “C orrelation o f Friction Factor for Turbulent Pipe Flow o f Dilute
Polym er S olutions,” Ind. Eng. Chem. Fund., 11, 566.
12. W hite, A., J. 1966. “T urbulent Drag R eduction with Polym er A dditives,” M ech.
Eng. Sci., 8, 452.
13. W hite, D. Jr. and R.J. Gordon. 1975. “ Influence o f Polym er Conform ation on T ur
bulent D rag R eduction,” A IC hE J ., 21, 1027.
14. Virk, P.S. 1975. “D rag Reduction Fundam entals,” A IC hE J., 21, 625.
15. H offm ann, L. and P. Schum m er. 1978. “Experim ental Investigation o f Turbulent
Boundary L ayer in Pipe Flow o f V iscoelastic fluids,” Rheol. Acta, 17, 98.
16. Astarita, G., G. G reco Jr., and L. N icodem o. 1969. A Phenom enological Interpreta
tion and C orrelation o f Drag R eduction,” A IC hE J., 15, 564.
17. Savins, J.G. 1969. “V iscous Drag R eduction,” C.S. W ells (Ed.) p. 183, Plenum
Press.
18. G oren, Y. and J.F. N orbury. 1967. “T urbulent Flow o f Dilute Aqueous Polym er
Solutions,” A S M E J. Basic Eng., 89, 816.
19. Virk, P.S. and H. Baher. 1970. “The Effect o f Polym er Concentration on Drag
R eduction,” Chem. Eng. Sci., 25, 1183
20. Bew ersdorff, H .W ., and N.S. Berm an. 1988. “The Influence o f Flow Induced Non-
N ew tonian Fluid Properties on T urbulent Drag R eduction,” Rheol. Acta, 27, 130.
21. Darby, R. 1996. “D eterm ining Settling R ates o f Particles in N on-New tonian Flu
ids,” Chem. Eng., v. 103(12), pp 107-112.
22. C hhabra, R.P. and P.H.T. Uhlerr. 1988. Chap. 21, Encyclopedia o f Fluid
M echanics, v. 7, N.P. C herem insinoff (Ed.), G ulf Publishing Co., Houston TX .
CHAPTER 13
TWO-PHASE FLOW:
LIQUID-SOLID AND
GAS-SOLID FLOW
The term “two-phase flow” covers an extremely broad range of situations, and it
is possible to address only a small portion of this spectrum in one book, let alone
one chapter. “Two-phase flow” includes any combination of two of the three
phases: solid, liquid, and gas, i.e. solid-liquid, gas-liquid, solid-gas, or liquid-liq
uid. Also, if both phases are fluids (combinations of liquid and/or gas), either one
of the phases may be continuous and the other distributed (e.g. gas in liquid, or
liquid in gas). Furthermore, the mass ratio of the two phases may be fixed or vari
able throughout the system. Examples of the former are nonvolatile liquids with
solids or non-condensable gases, whereas examples of the latter are flashing liq
uids, soluble solids in liquids, partly miscible liquids in liquids, etc. In addition,
in pipe flows the two phases may be uniformly distributed over the cross section
(i.e. “homogeneous”) or they may be separated, and the conditions under which
these states prevail are different for horizontal than for vertical flow. For uni
formly distributed homogeneous flows, the fluid properties can be described in
terms of averages over the flow cross section. Such flows can be described as
“one dim ensional,” as opposed to separated or heterogeneous flows in which the
phase distribution varies over the cross section.
We will focus on two-phase flows in pipes, which includes the transport of
solids as slurries and suspensions in a continuous liquid phase, paper (pulp) slurry,
and pneumatic transport of solid particles in a continuous gas phase. Although it
may appear that only one additional “variable” is added to the single-phase prob
lems previously considered, the complexity of two-phase flows is greater by orders
of magnitude. It is emphasized that this is only an introduction, and literally thou
sands o f articles abound in the literature on various aspects of two-phase flows. It
should also be realized that if these problems were simple or straightforward, the
number of papers required to describe them would be orders of magnitude smaller.
13.1
13.2 FLU ID FLO W HANDBOOK
13.1 DEFINITIONS
Before proceeding further, it is appropriate to define the various flow rates, veloc
ities and concentrations for two-phase flows. There is a bewildering variety of
notation in the literature relative to two-phase flow, and we will attempt to use a
notation which is consistent with the definitions below for solid-liquid and solid-
gas systems.
The subscripts L, s, and G represent the local two-phase mixture, liquid
phase, solid phase and gas phase, respectively. The definitions below are given in
terms o f solid-liquid (S-L) mixtures, where the solid is the more dense distributed
phase and the liquid the less dense continuous phase. The symbol </> is used for
the volume fraction of the more dense phase and £ is the volume fraction of the
less dense phase (obviously </>= 1 - £). An important distinction is made between
(</>, e) and (</>,„, £m). The former (</>, £) refer to the overall flow-average (equilib
rium) values entering the pipe, for example:
0 = ~ Qs n = 1 - e (13.1)
Qs + Q ,
whereas the latter (</>„„ £m) refer to the local values at a given position in the pipe.
These are different ((f)m =£ </>, £m £) when the local velocities of the two phases
are not the same (i.e. slip is significant), as will be shown below.
Mass Flow Rate (m) and Volume Flow Rate (Q ) :
mm = ms + mi - ps Qs + PL Q L = Pm Q m (13.2)
Cm = — = <5, + Gl = ms + mL (13.3)
V o lu m e F lu x :
Jm = Js + J L = — = + — = Qs Qi = Vm (13.4)
Pm Ps PL A
P h a s e V e lo c ity :
= 'T = 7 ^ ; Vl = ~ = T ^ T (1 3 .5 )
(/) 1 - £ £ 1 - (j)
Vr = VL - Vs \ Sr = — = 1 + — (13.6)
TWO-PHASE FLOW: LIQUID-SOLID AND GAS-SOLID FLOW 13.3
Note that the total volume flux (Jm) of the mixture is the same as the superfi
cial velocity ( Vm), i.e., the volumetric flow divided by the total flow area. However,
the local velocity of each phase (V/) is greater than the volume flux of that phase
(7,), since each phase occupies only a fraction of the total flow area. The volume
flux of each phase is the total volume How rate of that phase divided by the total
flow area.
The relative slip velocity (or slip ratio) is an extremely important variable. This
occurs primarily when the distributed phase density is greater than that of the con
tinuous phase, and the heavier phase tends to lag behind the lighter phase for var
ious reasons (explained below). The resulting relative velocity (slip) between the
phases determines the drag exerted by the continuous (lighter) phase on the distrib
uted (heavier) phase. One consequence of slip (as shown below) is that the con
centration or “holdup" of the more dense phase within the pipe (cf>m) is greater than
that entering or leaving the pipe, since its residence time is longer. Consequently,
the concentrations and local phase velocities within a pipe under slip conditions
depend upon the properties and degree of interaction o f the phases and cannot be
determined solely from a knowledge of the entering and leaving concentrations
and flow rates. Slip can only be determined indirectly by measurement of some
local flow property within the pipe such as the holdup, local phase velocity, local
mixture density, etc.
For example, if </> is the solids volume fraction entering the pipe at velocity
V, and <f>m is the local volume fraction in the pipe where the solid velocity is Vs,
a component balance shows:
Vs <P j Vl 1 - <P
— = — and — = ---------- (13.7)
V <Pm V 1 - <P,„
Substituting these expressions into the definition of the slip velocity (and divid
ing by the entering velocity, V, to make the results dimensionless):
- = v, = v _ i_ j_ v _ s _ _ s _ t_ = <p„, - 0
' V VL + Vs S + 1 0„, ( 1 - <t>J ^ f
_ 1/2
1 - (1 / Vr )+ { i d / Vr) - I ] ' + 4 $ / Vr }
(13.9)
For example, if the entering solids fraction </>is 0.4_the corresponding values of the
local solids fraction <£m for relative slip velocities ( Vr) of 0.01,0.1, and 0.5 is 0.403,
0.424, and 0.525 respectively. There are many “theoretical” expressions for slip, but
practical applications depend on experimental observations and correlations (which
13.4 FLU ID FLO W HANDBOOK
will be presented later). In gas-solid flows, </>„, will vary along the pipe since the
gas expands as the pressure drops and speeds up as it expands, which tends to
increase the slip, which in turn increases the holdup of the denser phase.
The mass fraction (jc) of the less dense phase is x = m Ll(m s + % ) , so that the
mass flow ratio can be written:
/ \
mi _ x _ p L VL A em _ s Pl
ms 1 - x p s Vs A ( 1 - e m ) Ps ( 1 - £m )
(13.10)
This can be re-arranged to give the less dense phase volume fraction in terms of
the mass fraction and slip ratio:
Pm = £rn PL + ( 1 “ £m ) PS (13.12)
which depends on the slip ratio S through equation 13.11. The corresponding expres
sion for the local in-situ holdup of the more dense phase is:
S ( 1 - x ) ( p, / p s )
<!>m = ( 1 - em ) = -------— — Y r — ( , 3-13>
x + S ( 1 - x ) y p L / ps )
Note that both the local mixture density and holdup increase asthe slip ratio
(5) increases. The “no slip” ( 5 = 1 ) density or volume fraction isidentical to the
equilibrium value entering (or leaving) the pipe.
The conveying o f solids by a fluid in a pipe can involve a wide range of flow con
ditions and phase distributions, depending on the density, viscosity and velocity
of the fluid and the density, size, shape, and concentration of the solid particles.
The flow regime can vary from essentially uniformly distributed solids in a “pseudo-
homogeneous” (symmetric) flow regime for sufficiently small and/or light particles
above a minimum concentration, to an almost completely segregated or stratified
(asymmetric) transport of a bed of particles on the pipe wall. The demarcation
between the “homogeneous” and “heterogeneous” flow regimes depends in a com
plex manner on the size and density o f the solids, the fluid density and viscosity,
TWO-PHASE FLOW: LIQUID-SOLID AND GAS-SOLID FLOW 13.5
the velocity of the mixture, and the volume fraction of solids. Figure 13 .1 illustrates
the approximate effect of particle size, density, and solids loading on these regimes.
Either a liquid or a gas can be used as the earner fluid, depending on the size
and properties of the particles, but there are important differences between hydraulic
and pneumatic transport. For example, in liquid (hydraulic) transport the fluid-
particle and particle-particle interactions dominate over the particle-wall interac
tions, whereas in gas (pneumatic) transport the particle-particle and particle-wall
interactions tend to dominate over the fluid-particle interactions. A typical “prac
tical” approach, which gives reasonable results for a wide variety of flow condi
tions in both cases, is to determine the “fluid only” pressure drop and then apply
a correction to account for the effect of the particles from the fluid-particle, par
ticle-particle, and/or particle-wall interactions. A great number of publications have
been devoted to this subject, and summaries of much of this work have been
given by Darby (1986), Govier and Aziz (1972), Klinzing et al. (1977), Molerus
(1993) and Wasp et al. (1977) [4, 12, 13, 19].
Pseudo-Homogeneous Flows
If the solid particles are very small (e.g. typically less than 100 fim) and/or not
tremendously denser than the fluid, and/or the flow is highly turbulent, the m ix
ture may behave as a uniform suspension with essentially continuous properties.
In this case, the mixture can be described as a “pseudo single-phase” uniform
fluid and the effect of the presence of the particles can be accounted for by appro
priate modification of the fluid properties (density and viscosity). For relatively
dilute suspensions (e.g. 5 percent by volume or less) the mixture will behave as
a Newtonian fluid with a viscosity given by the Einstein equation:
H = flL (l+2.5(p) (1 3 .1 4 )
where /jll is the viscosity of the suspending (continuous) Newtonian fluid and (/>
= (1 - e) is the volume fraction of solids. The density of the mixture is given by:
Pm = P l ( 1 - <t>) + P s <t> (1 3 .1 5 )
Figure 13.2 shows how the pressure gradient and flow regimes in a horizontal
pipe depend on velocity for a typical “heterogeneous” suspension. It is seen that
13.6 FLU ID FLO W HANDBOOK
PARTICLE DIAMETER
(Largest 5%)
TYLER
MESH INCHES MICRONS
Log (VJ
F I G U R E 13.2 Pressure gradient and flow regimes for slurry H o w in a horizontal pipe
the pressure gradient exhibits a minimum at the “minimum deposit velocity,” which
is the velocity at which a significant amount of solids begin to settle in the pipe.
A variety of different correlations have been proposed in the literature for the min
imum deposit velocity, one o f the more useful being (Hanks, 1980):
where s = Ps IPl • At velocities below this, the solids settle out in a bed along the bot
tom of the pipe. This bed can build up and plug the pipe if the velocity is too low, or
it can be swept along the pipe wall if the velocity is near the minimum deposit veloc
ity. Above the minimum deposit velocity, the particles are suspended, but are not uni
formly (“symmetrically”) distributed until the turbulent mixing is high enough to
13.8 FLU ID FLO W HANDBOOK
overcome the settling forces. One criterion for a non-settling suspension has been
given by Wasp (l 977):
(13.18)
A Pm = A PL + A Ps (13.19)
where APL is the “fluid only” pressure drop and APs is an additional pressure drop
due to the presence of the solids. For uniform sized particles in a Newtonian liq
uid, APL is determined as for any Newtonian fluid in a pipe. For a broad particle
size distribution, the suspension may behave more like a heterogeneous suspen
sion o f the larger particles in a carrier vehicle composed of a homogeneous sus
pension of the finer particles. In this case, the homogeneous carrier will likely be
non-Newtonian and the methods given in Chapter 12 for such fluids should be
used to determine APL.
The procedure for determining APs which will be presented here is that of
M olerus [13]. The basis of the method is a consideration of the extra energy dissi
pated in the flow as a result o f the fluid-particle interaction. This is characterized
by the particle terminal settling velocity in an infinite fluid in terms of the drag
coefficient, Q :
Crf = <1 3 -2 ° )
j v i
Vr_ (13.21)
7
. V2
N hp = --------- r — (1 3 .2 2 )
( s - I)dg
Vi
N \n = --------( 13. 23 )
( s - 1 ) Dg
TW O-PHASE FLOW: LIQUID-SOLID AND GAS-SOLID FLO W 13.9
(13.24)
where Vro is the dimensionless “single particle” slip velocity as determined from
the diagram, which in turn is used to define the parameter X a:
(13.25)
Using this value o f X0 and the entering solids volume fraction (</>), a value of
X is determined as follows:
(13.26)
Knowing X determines APs, which is added to APL to get the total pressure
drop in the pipe. The above procedure is straightforward if all the particles are the
same diam eter (d). However, if the solid particles cover a broad range of sizes,
the procedure must be applied for each particle size (diameter dh concentration
fa) to determine the corresponding contribution of that article size to the pressure
drop APSi. The total solids contribution to the pressure drop is then 2 A PSi. If the
carrier vehicle exhibits non-Newtonian properties with a yield stress, particles for
which d < r„/(0.2gAp) (approximately) will not fall at all.
r^in co 04
h- lo CO 04
13.10
co 04 T
cooi
FIGURE 133 State diagram for suspension transport [20].
TWO-PHASE FLOW: LIQUID-SOLID AND GAS-SOLID FLOW 13.11
For vertical transport, the major difference is that no “bed” can form on the
pipe wall but, instead, the pressure gradient must overcome the weight of the
solids as well as the fluid/particle drag. Thus the solids holdup and hence the fluid
velocity are significantly higher for vertical transport conditions than for hori
zontal transport. However, vertical flow of slurries and suspensions is generally
avoided where possible due to the much greater possibility of plugging if the
velocity drops.
EXAMPLE 13.1
Determine the pressure gradient (in psi/ft) required to transport a slurry at
300 gpm (68.13 m'Vhrjthrough a 4 inch (101.6 mm) sch 40 pipeline. The slurry
contains 50 percent (by weight) o f solids, having a SG o f 2.5, in water. The
slurry' contains a “bimodal ” particular size distribution, with h alf o f the parti
cles below 100 mm and the other h a lf about 2000 mm. The suspension o f fines
is stable, and constitutes a pseudo-homogeneous non-Newtonian vehicle in
which the larger particles are suspended. The vehicle can be described as a
Bingham plastic, with a limiting viscosity o f 30 cP and a yield stress o f
55 dyn/cm
S olution:
where s = ps/pL• Half of the solids are in the non-Newtonian “vehicle” and half
will be “settling,” with a volume fraction of 0.143. Thus the density of the “vehi
cle” is
We then calculate the contribution to the pressure gradient due to the contin
uous Bingham plastic vehicle, as well as the contribution from the “non-homoge-
neous” solids. For the first part, we use the method presented for Bingham plastics
(see Chapter 12). From the given data, we can calculate N Re BF = 9540 and N He
= 77,600. The friction factor o f/ = 0.0629 is calculated from equations in Chap
ter 12. The corresponding pressure gradient, (APIL)f - 2fpV2!D, is 1.105 psi/ft
(25 kPa/m).
The pressure gradient due to the heterogeneous component is determined by the
Molerus method. This first requires the determination of the terminal velocity of
the settling particles, which is done using the method given in Chapter 12 for the
larger particles settling in a Bingham plastic. This requires determining NRe BP,
NBh and Cd for the particle, all of which depend on Vt. This can be done using an
13.12 F LU ID FLO W HANDBOOK
iterative procedure to find V,, such as the “solve” function on a calculator or spread
sheet. The result is V, = 19.5 cm/s (0.640 ft/s). This is used to calculate the parti
cle and tube Froude numbers, N ^ rp = 25.6 and N^rt = 0.0358. These values are used
with Figure 13.3 to find V J w s = 0.05, which corresponds to a value of X = 0.00279.
From the definition of X, this gives (AP/L)s = 0.0312 psi/ft (0.7057 kPa/m), and
thus a total pressure gradient o f (AP/L), = 1.14 psi/ft (25.8 kPa/m). In this case,
the pressure drop due to the Bingham plastic “vehicle” is much greater than that
due to the heterogeneous particle contribution.
Saltating flow
G as velocity
are a variety of correlations for the saltation velocity, one of the most popular
being that of Rizk [15]:
= — = 10 s N L (13.27)
me
where
8 = 1.44 d + 1.96
* = 1.1 d + 2.5
N Frs = (13.27a)
gd
Conveying Solids
Solids Loading Velocity AP Volume
Conveying M o d e (e.g. kgs/kgg ) ft/s (m/s) psi (kPa) Fraction
Here fis is the “solids loading” (mass of solids/mass of gas), Vgs is the saltation gas
velocity, and d is the particle diam eter in mm. (It should be pointed out that corre
lations such as this are based, of necessity, on a finite range of conditions and have
a relatively broad range of uncertainty, e.g. ± 5 0 -6 0 % is not unusual).
Horizontal Transport
There are two major effects that contribute to the pressure drop in horizontal flow:
acceleration and friction loss. Initially the inertia of the particles must be over
come as they are accelerated up to speed, and then the friction loss in the mixture
must be overcome. If Vs is the solid particle velocity and m s - psVs( 1 - em)A is
the solids mass flow rate, the acceleration component of the pressure drop is:
A P ., + A p „ = V, + S cL ^k =
A 2 2 t tl G V g
(13.28)
The slip ratio VG/VS = S can be estimated, for example, from the IGT correla
tion (e.g. Klinzing, et al. [12]).
1 V4 , 0.68 d 092 p V
s ' t “ - ~ V <l3' 29)
in which d and D are in meters and ps and pG are in kg/m 3. For vertical transport,
the major differences are that no “bed” on the pipe wall is possible but, instead,
the pressure gradient must overcome the weight of the solids as well as the fluid/
particle drag, so that the solids holdup and hence the fluid velocity must be sig
nificantly higher under transport conditions.
The steady flow pressure drop in the pipe can be deduced from a momentum
balance on a differential slice of the fluid-particle mixture in a constant diameter
pipe. For steady uniform flow through area Ax\
Z F X = 0 = d F xp + d F xg + d Fxw
= - A x dP - [ p s ( 1 - £„, ) + p G £,„ ] g As dz
- [ TwS + TwG ] W p dX
(13.30)
where TwS and rwG are the effective wall stresses resulting from energy dissipa
tion due to the particle-particle as well as particle-wall and gas-wall interaction,
TWO-PHASE FLOW : LIQUID-SOLID AND GAS-SOLID FLOW 13.15
and Wr is the wetted perimeter. Dividing by Ax, integrating, and solving for the
pressure drop, - A P = Px - P2:
-A P = [ p s ( 1 - e m ) + p G e m ] g A z + ( Us + Twc ) 4 L / Di,
(13.31)
where Dh = 4AX/Wp is the hydraulic diameter. The void fraction £m is the volume
fraction of the gas in the pipe, i.e.:
ms
= 1 - (13.32)
Ps Vs A x + S ( J - x) p G / p s
A P fS fa .. .,2 A PfG
rws - ~ T P s ^ ' tin ) V s 4 L / Dh
4 L /D ,
(13.33)
Here APfG is the pressure drop due to “gas only” flow (i.e. the gas flowing alone
in the full pipe cross section). Note that if the pressure drop is less than about 30%
of P,, the incompressible flow equations can be used to determine APfG using the
average gas density. Otherwise, the compressibility must be considered and the
appropriate methods used to determine APfG. The pressure drop is related to the
--------------------------------------------------------- 1
Pi
Pi - P 2 = / - (13.34)
/ /
1
^1
fs = 1
“ I (13.35)
8 KPc.) Kd) L vs J
1.15
I - £ N Ret Vc, / £
fs = 0.117 ( 1 - £) ( 13. 36 )
N Rep fzD
13.16 F L U ID FLO W HANDBOOK
-0.869
1 - £ N Rel
f s = 0.0206 ( 1 - e) (13.37)
N Rep
where
- Vs pG
d Vt Pr
N R et = ------------- , N Rep = --------------- (13.38)
Vc Vc
Vertical Transport
The principles governing vertical pneumatic transport are the same as those given
above, and the method for determining the pressure drop is identical (with an appro
priate expression forf s). However, there is one major distinction in vertical transport,
which occurs as the gas velocity is decreased. As the velocity drops, the frictional
pressure drop decreases but the slip increases since the drag force exerted by the gas
entraining the particles also decreases. The result is an increase in the solids holdup,
with a corresponding increase in the static head opposing the flow which in turn
causes an increase in the pressure drop. A point will be reached at which the gas can
no longer entrain all of the solids and a slugging, fluidized bed results with large
pressure fluctuations. This condition is known as choking (not to be confused with
choking that occurs when the gas velocity reaches the speed of sound), and repre
sents the lowest gas velocity at which vertical pneumatic transport can be attained
at a specified solids mass flow rate. The choking velocity, Vc , and the correspond
ing void fraction, £c , are related by the following two equations (Yang [20]):
Vs
^ = i + (13.39)
V, V, ( 1 - Ec )
2.2
2 g D { Ec47 - I )
= 6.81 x 10s — (13.40)
( V c - V, f \ Ps
The Technical Association of the Pulp and Paper Industry, TAPPI, has issued many
technical information sheets to provide general methods to predict frictional pressure
T W O - P H A S E FLOW: LIQUID-SOLID A N D GAS-SOLID F L O W 13.17
loss for many types of pulp suspensions under various flow conditions in pipes.
The following presentation is based on TAPPl Technical Information Sheet num
ber 0 4 10 - 14 (TAPPI 1988). A typical frictional loss versus pulp velocity and
oven-dry consistency, C = 2% -6% , is shown in Figure 13.5.
The graph categorizes the flow-pressure loss conditions into three distinct areas.
At low pulp suspension velocities, the frictional head increases linearly with bulk
velocity. Any further increase in suspension bulk velocity will result in a decrease in
frictional loss. A minimum point in frictional pressure loss is reached after which the
frictional loss starts to increase again in a nonlinear function of the bulk velocity.
The pulp frictional loss-velocity curve is characterized by two reflection points,
unlike the mechanical suspension curve which is characterized by one reflection
point. The suggested TAPPI method divides the pressure curve into three areas
based on the curve reflection points. The first region extends from a low bulk veloc
ity to a bulk velocity after which the frictional loss starts to decrease. This velocity
is designated as Vmax, see Figure 13.5. The second region extends from the max
imum velocity point, Vmax, into the onset of drag reduction point. The onset of
drag reduction point is defined as the bulk velocity at which the pulp suspension
frictional pressure loss intersects with the pure water frictional loss line (see water
(dotted line) in Figure 13.5). The symbol Vw is used to designate the bulk veloc
ity which corresponds to the onset of drag reduction point. The third area extends
beyond the onset of drag reduction point. The TAPPI Technical Information Sheet
recommends a correlation to estimate the frictional pressure loss in the first area,
assumes a constant frictional loss in the second area, and recommends the use the
water frictional loss curve for velocities higher than the onset drag reduction point.
TAPPI recommended frictional loss curve is shown as a dotted line in the second
and third regions of Graph 13.5. The following equations are recommended by
TAPPI technical sheet to estimate the maximum velocity and the onset o f drag
reduction points:
V max = K C a (13.41)
where
C Pulp Consistency (% dry-oven)
K Constant for a given pulp type (see Table 13.2)
<7 Constant for a given pulp type (see Table 13.2)
Vmax Velocity at maximum point, m/s.
AH / L = F K V aC l>D r (13.43)
where
F Correction factor
D Pipe internal diameter, mm.
C Pulp Consistency (% oven-dried)
V Pulp bulk velocity, m/s
A H/L Friction loss (m w ater/100 m pipe length)
a, j8, and y Constants for a given pulp type, see Table 13.3
K Constant for a given pulp type, see Table 13.3
The correction factor may be used to correct for data in Table 13.3 for tem
peratures other than 95°F (35°C). The flow resistance increases/decreases
by 1% for each 1 degree decrease/increase in temperature.
2. Pulp bulk velocity is larger than Vmax but lower than Vw. Assume the
bulk velocity is equal to Vmax and use Equation 13.43 to estimate the
friction loss. The true frictional loss is less than what is estimated by
Equation 13.43, but for a conservative and simple approximation one
may use Equation 13.43.
T W O - P H A S E FLOW: LIQUID-SOLID A N D GAS-SOLID F L O W 13.19
3. Pulp bulk velocity higher than Vw. Friction loss of pure water flowing at
the same bulk velocity may be used to estimate the frictional loss of the
pulp suspension. Any of the equations for frictional pressure drop o f an
incompressible fluid; see Chapter 8, such as Darcy equation or the Blasius
equation may be used.
EXAMPLE 13.2
Estimate the frictional loss o f the following pulp type when the average flow
rate o f the pulp suspension is 25 GPM. Repeat fo r an average flow rate o f
50 GPM.
TABLE 13.3 Pulp-Type Constants for the Calculation of Frictional Pressure Loss
(Table values are valid for pulp temperature of 95°F (35°C))
Pulp type K a 0 Y
Solution
Since the suspension bulk velocity (2.55 ft/s) is less than the maximum veloc
ity (4.623 ft/s), flow is in region 1 and Equation 13.42 must be used to estimate
the friction loss:
T W O - P H A S E FLOW: LIQUID-SOLID A N D GAS-SOLID F L O W 13.21
w h e re
F= I
D = 50.8
C = 3%
V= 0.7787 m/s
a, /3, and y = 0 .3 1, 1.81, and -1.34 respectively (see Table 13.3)
K = 1334 (see Table 13.3)
which is equivalent to 5.03 ft of water per one foot of pipe length, or 2.09 psi per
foot of pipe length assuming water density of 60 lbm/ft (961 kg/m ).
Flow is 50 GPM:
Pulp bulk velocity = 50 GPM / 60 / 7.481 / 0.0218 = 5.109 ft/s (1.557 m/s)
Since this velocity is higher the Vmax, the onset of drag reduction velocity
must be calculated using Equation 13.42:
Pulp bulk velocity (5.109 ft/s) is lower than the Vw velocity, therefore, Vmax is
assumed as the pulp bulk velocity and pressure drop is found using Equation 13.43:
NOTATION
L length, [L]
m mass flow rate, [M/t]
NFr Froude number, [ - ]
N particle Froude number, Eq. (13.22), [ - ]
Frp
SUBSCRIPTS
1, 2 reference points
C choking condition
T W O - P H A S E FLOW: LIQUID-SOLID A N D GAS-SOLID F L O W 13.23
/ friction loss
G gas
L liquid
m mixture
S solid
w wall
REFERENCES
1. Aude, T.C., et al. 1971. “Slurry Piping Systems: Trends, Design M ethods, G uide
lines,” Chem . Eng., p. 74, June 28.
2. Brecht, W . and H. Heller. 1950. TAPPI, 33(9): 14A.
3. Brodkey, R.S. 1967. The Phenom ena o f Fluid M otions, A ddison-W esley, 1967.
4. Darby, R. 1985. “ H ydrodynam ics of Slurries and Suspensions,” Ch. 2 in E ncyclope
dia o f Fluid M echanics, Vol. 5, N.P. C herem isinoff (Ed.), G ulf Publishing Co.
5. Duffy, G. G. and A. L. Titchener. 1974. TA PPI, 57(5): 162.
6 . Duffy, G. G., M oller, K., Lee, P. F. W., and S. W. A. M ilne. 1974. A ppita, 27(5):
327.
7. Fan, L-S. And C. Zhu. 1998. “Principles o f Gas-Solid Flow s,” C am bridge U niver
sity Press.
8 . Ferguson, M .E.G. and P.L. Spedding. 1995. “M easurem ent and Prediction o f Pres
sure Drop in T w o-Phase Flow ,” J. Chem. Tech. Biotechnol., v. 62, pp 262-278.
9. Hanks, R.W . February 3-7 1980. ASM E Paper No. 80-PET-45, A SM E Energy
Sources T echnol. C onf., New Orleans, LA.
10. Hetsroni, G. (Ed.). 1982. Handbook o f M ultiphase System s, H em isphere Publishing
Corp.
11. H olland, F.A ., and D.R. Bragg. 1995. Fluid Flow for Chem ical Engineers, 2nd E di
tion, Edw ard Arnold.
12. K linzing, G .E., R.D. M arcus, F. Rizk and L.S. Leung. 1997. Pneum atic Conveying
o f Solids, 2nd E dition, Chapm an and Hall.
13. M olerus, O. 1973. Principles o f Flow in Disperse System s, Chapm an and Hall.
14. M oller, K., G. G. D uffy, and A. L. Titchener. 1973. Appita, 26(4): 278.
15. Rizk, F., D issertation, U niversity o f Karlsruhe, 1973.
16. Shook, C.A . and M .C. Roco. 1991. “Slurry Flow -Principles and Practice,”
B utterw orth-H einem ann.
17. TA PPI TIS 0410-14. 1988. “G eneralized M ethod for D eterm ining the Pipe Friction
Loss o f Flow ing Pulp Suspensions.” Technical Association o f Pulp and Paper
Industry. A tlanta.
18. W allis, G.B. 1969. O ne-D im ensional Tw o-Phase Flow, M cG raw -H ill, New York.
19. W asp, E.J., J.P. K enny and R.L. Gandhi. 1977. Solid-Liquid Flow in Slurry Pipeline
T ransportation, T rans-T ech Publications, Clausthal, G erm any.
20. Yang, W .C. 1983. Pow der Technology, 35, pp 143-150.
CHAPTER 14
MOLECULAR FLOW
While a gas is generally treated as a continuum fluid without any internal structure,
it is actually comprised of a myriad of discrete molecules. The word “molecules”
is used here as a generic term that includes the single atoms of monatomic gases.
Whether or not the continuum model is valid depends on the value of the Knudsen
number defined by
Kn= —a (14.1)
The mean free path A is the mean distance traveled by the individual molecules
d
between intermolecular collisions and is a typical dimension of the flow. The
mean free path is equal to the mean molecular speed relative to the flow velocity
divided by the collision frequency. The mean molecular speed depends on the state
of the gas, while the collision frequency depends also on the collision cross-section
of the molecules. The values for a gas in a state of equilibrium are almost univer
sally employed and most quoted values of the mean free path are based on the simple
hard sphere molecular model. However, the effective diameter and collision cross-
section of real molecules are a function of the relative speed in collisions and are
therefore dependent on the temperature of the gas. This effect is taken into account
by the inverse power-law molecular model that is widely employed in classical kinetic
14.1
14.2 FLUID F L O W H A N D B O O K
theory, but this has an indefinite cross-section and does not yield a simple expres
sion for the mean free path. This led to the introduction of the variable hard sphere,
or VHS, molecular model that has become the preferred model for engineering stud
ies. This employs the hard sphere collision mechanics, but the molecular diameter
v
has an inverse power-law dependence on the relative speed in collisions. Molecu
lar speeds are proportional to the square root of the temperature, so that
S= re,(T/Trefr\
8 (14.2)
The reference diameter is obtained from the coefficient of viscosity at the refer
v
ence temperature and the power law is obtained from the variation of the coef
ficient of viscosity with temperature. For * jx Tw,
the power law v =co A
- and
the final result for the mean free path in an equilibrium gas is
A= (14.3)
V2 ; r S > ( Trrf/ r r n
co A
The exponent is for a hard sphere gas and the mean free path then becomes
independent of temperature. The simpler hard sphere expression has been commonly
employed for the definition of Knudsen numbers in the non-dimensional presenta
tion of experimental data. This has adversely affected the quality of the information
when significant variations in temperature are involved.
Table 14.1 lists the molecular properties under standard conditions for a range
of common gases. The degree of complexity of the molecules is indicated by the
k
number of degrees of freedom £. The specific heat ratio that is commonly employed
for this purpose in continuum studies is related to£ by
(14.4)
The effective molecular diameters are based on the measured coefficients of vis
cosity and that is why the mean air diameter is greater than that for either oxygen
or nitrogen. The values in Table 14.1 have been used in conjunction with Equation
14.3 to produce the mean free path chart of Figure 14.1.
The molecular or particulate nature of a gas must be taken into account unless
the Knudsen number is extremely small in comparison with unity. High Knudsen
numbers are associated with either large mean free paths, and therefore low gas
densities, or extremely small physical dimensions. Although the Knudsen num
ber is the only parameter that deals directly with the molecular structure of gases,
it is not independent of the more common dimensionless parameters. The mean
>> 3
r-~ sC sO ro Tf in r-~ — — in
» 2 | sO sO — sO r- r- r» r- 00 as o 00 00
£ g -.E d © — o o o d d d d — d o
> oj
Pa and 273 K
U r— oo Os' f_N s
00 in ___ CN
BS. ©
u
.2 0
o
—
in 3
On r-
w
in
Os
o 00
Os cn
^-4
COrn
—
r-
sO f
cn 00
co
•— •
Os SO 00
CN in 00
CN
'—' '—' -—- '-' "
■°~ x
on X CN m f" OS r^- Os r- CN 00 sO •<fr
Os cn Os r» — o — sO O' r-
X co
Standard”Conditions of 10 1325
CN CN in CN Tf in sb rt in
% ^ sO sO
so m oo in ‘O CN CN sO
1 f'“r<N (N m
so
cn
o o os c r—
cm "
• sb — >n
sO
Os >0 o
00
— CN O
CO
- 2 2 r-
3OJ Av .x_- CN m in in —
CN
—-
o c m so m r- on oo
in so — m —
s * — — CN
Representative Gas Properties Under “
1/5 c
« U_ O «n
& o -g ^ cn r- —
;
U OJ u ~ ) c n s d r r> » n > o > n ‘n r n s o s o m r n
o o
X)
E T
ri
X O z
C/5 z U Tf
u-
o
CN
m <u
TC3 72.
’
x -o
ed o
c ’2
TABLE 14.1
O _o
X
U
S
Q.
«/3
c
<u
Ot) c
.2
c
o
B
c
c
<o
eo c
c
(U - c
<u c
o c
O 2 I E 5 o c3 oo § 5 I M- o
E
03
C3 "o>i aOi
o X I <
E J3 .ts w
U Z £
^X ox
o
u: c3
< U
sz £
U
c
<u
X
14.3
14.4 FLUID F L O W H A N D B O O K
ic T T o 3
Density / Standard Density
F I G U R E 14.1 M e a n free path in typical gases.
free path can be expressed in terms of the coefficient of viscosity, rather than the
molecular diameter, and it is readily shown that
__ 2 k ( 5 - 2 ( 0 ) { 1 - 2(o ) M a
Kn = J ------------ (14.5)
K 15 Re
For air, the Knudsen number is equal to 1.21 times the Mach number divided by
the Reynolds number.
The density is defined as the mass per unit volume of the gas and is therefore equal
to the number density times the average molecular mass of the molecules, i.e.
P = n m avg (14.6)
MOLECULAR FLOW 14.5
The molecular mass is, of course, a constant in a single species or simple gas. The
number density at standard temperature and pressure is the same for all gas species
and is equal to 2.6867 X 1025/cubic meter (7.6079 X l0 23/cubic foot). The mean
n
spacing between molecules is equal to 1/3 and is therefore equal to 3.3 X 10~9
m (1.1 X 10” 8 ft). The mean free path in air under standard conditions is 4.9 X 10-8
m (1 .6 X 10-7 ft). Air under standard conditions therefore satisfies the “dilute gas”
condition that the mean free path is large in comparison with the mean molecu
lar spacing which, in turn, is large in comparison with the molecular diameter.
Many of the results in this chapter implicitly assume that the gas is dilute. How
ever, the dilute gas condition is not satisfied by a large margin under standard con
ditions and gases at densities greater than two or three times the density of standard
air must be treated as “dense gases.”
In the continuum model, the flow properties are assumed to apply at a “point”
in the gas. The number density is the number of molecules per unit volume and is
meaningless unless it applies to a finite volume. Flow features scale with the mean
n
free path and the number of molecules in a cubic mean free path \ ? is a useful
reference quantity. For air, this is 3860 (njn)2, and the number becomes so small
in a dense gas that the natural fluctuations in the number and therefore in the local
flow properties are significant. These fluctuations cause physical phenomena such
as Brownian motion, but are rarely taken into account by the mathematical models
for gas flows at either the continuum or the molecular level.
The conventional flow velocity is a mass average velocity and is defined at the
molecular level by
The thermal velocity of a molecule is the velocity relative to the stream velocity,
for example,
c' = c - U (14.8)
and is more easily understood if we note that the flux of jc-directed momentum in
y
the direction is
P,y=n(mu'V)avg
14.6 FLUID F L O W H A N D B O O K
A scalar pressure may be defined as the average of the three normal components
of the pressure tensor, i.e.
P='An(mc'2)avg (14.10)
k,
3 k,
3
T,nt=--2% ~
(£ -3 )* ,
(14.12)
The viscous stress tensor is defined as the negative of the pressure tensor with
the scalar pressure subtracted from the normal components. Using subscript nota
tion to indicate the components,
T ee t =-[n(mc;cj)avg- SijP1
0 (14.15)
The flow variables have been defined as averages over the molecular velocities
and they can be evaluated if the distribution of molecular velocities is known. The
velocity distribution function/(c)is defined such that the fraction of molecules in
velocity space element dc is
A stationary homogeneous volume of dilute gas that is free of any external force
quickly reaches an equilibrium state with the equilibrium or Maxwellian distribu
tion. The equilibrium distribution function for the molecular velocity is best written
where
The distribution function for a single component of velocity is also useful and that
x
for the component in the direction is
The two scalar distributions are plotted on Figure 14.2. Because of the high speed
tails on the distributions, the mean molecular speed c'avg
of 2 / ( V 7T/3) is higher than
14.8 FLUID F L O W H A N D B O O K
0.8
0.6
0.4
0 .?
\
X
\
the most probable speed of 1//3. Note that the distributions depend on the molecular
mass and the equations can be applied independently to every species in a gas mixture.
The distribution function in a continuum gas flow conforms to the equilibrium
distribution as long as the flow can be regarded as isentropic. It can therefore be
applied to gas flows that are not affected by viscous shear stresses, heat conduc
tion or diffusion effects, i.e. to flows that can be described by the Euler equations.
When there is a finite stream velocity, the above equations apply to the thermal
velocity components relative to the stream velocity. The flow speed is generally
normalized by the most probable molecular speed to define the speed ratio
The inward normal momentum flux to the element is essentially the pressure due
to the inward molecules and is
P'=[nm/(2^P2)]
X {sco s0 ex p (-s 2 cos 2 6 )+ ^ ( l/ 2 +s cos2 0)[1 + erf(.vcos0)]
2
(14.25)
as would be expected because the inward moving molecules contribute half the
pressure. The inward parallel momentum flux is
T/ = [nmssinO/ (2^[kP )] 2
and is, of course, zero in a stationary gas. Finally, the inward energy flux to the
element is
nm o s2 h-------
. ^ 1
k- 1
2
(14.28)
q. =[(k+\)/(k-\)]nm/(%^P>). (14.29)
The major achievement of classical kinetic theory in the first half of the twentieth
century was the development of the Chapman-Enskog theory for the coefficients
14.10 FLUID F L O W H A N D B O O K
of viscosity, heat conduction, and diffusion in a dilute gas. This theory confirmed
and clarified the transport terms that, when added to the Euler equations, lead to the
Navier-Stokes equations. The Chapman-Enskog theory assumes a power series per
turbation of the equilibrium distribution function and uses only the first term in
the series. The result for the coefficient of viscosity in a VHS gas at temperature
(14.30)
** 2 (5 - 2ft) )(7 - 2ft) ) r2ef
8
This result has been used in conjunction with the experimentally determined coef
ficients of viscosity to determine the molecular diameters in Table 14.1. The coef
ficient of heat conduction is
so that, for a dilute monatomic gas with specific heat at constant pressure equal to
5/?/2, the Prandtl number is 2/3. The viscous stress tensor and the heat flux vector
become, in the subscript or Cartesian tensor notation,
and
(14.32)
The relative diffusion velocity between two species is (Chapman and Cowling, 1970)
2 / \
The first term in the parentheses on the right hand side of this equation acts to
reduce species concentration gradients and is commonly included in the Navier-
Stokes equations to deal with gradients that are established through the boundary
conditions or through chemical reactions. The second and third terms are for pres
sure and thermal diffusion and these act to produce concentration gradients in the
presence of pressure and temperature gradients. These effects are almost univer
sally neglected in the Navier-Stokes formulation.
For the special case of a flow in the ^-direction with gradients only in the y-
direction, the Chapman-Enskog distribution function for a simple VHS gas may
be written
15V/r/?V
(14.35)
Each of the perturbation terms contains a local Knudsen number based on the
scale length of the appropriate macroscopic flow property. The coefficients of the
local Knudsen numbers may be evaluated on the assumption that terms like j 3c'
are of order unity. It may be concluded that the transport terms in the Navier-
Stokes equation break down unless the local Knudsen numbers are small in com
parison with unity.
A property of the Chapman-Enskog distribution is that, when substituted into
Equation 14.8 for the evaluation of density, velocity and temperature, the result is
the same as for the equilibrium gas. However, the distribution is no longer spheri
cally symmetric and the heat transfer term, when evaluated separately for the upward
and downward moving molecules, yields the following results for the temperature
difference between the upward and downward moving molecules:
r + _ r -
30 dT
A
(14.36)
T (5-2co)(l-2co) Tdy
Similarly, the shear stress term can be evaluated separately for the velocity x
components based on the upward and downward moving molecules. The result is
15 A du (14.37)
u (5 - 2co)(l- 2co) udy
14.12 FLUID F L O W H A N D B O O K
These results provide an indication of the magnitude of the velocity and temper
ature slips that are always present at gas-solid interfaces.
The limiting cases of diffuse and specular reflection were introduced by Maxwell
(l 879) are still the most widely used models. Diffuse reflection assumes that the
incident molecules are completely accommodated to the surface temperature and
are reflected in accordance with the half of the equilibrium distribution of Equa
tion 14 .19 that is directed away from the surface. Specular reflection is perfectly
elastic with the velocity component normal to the surface being reversed and that
parallel to the surface being retained. It is never realized in flows over real sur
faces, but is a useful limiting case for theoretical studies.
Experiments that involve “engineering” surfaces yield results consistent with
diffuse reflection and the model is almost invariably employed in analytical and
computational studies. These surfaces are often microscopically rough and are
generally covered by a layer of adsorbed gas. The reflection is as if the molecule
is momentarily adsorbed by the surface and comes to equilibrium with it before
being re-emitted. However, if a microscopically smooth surface is cleaned by expo
sure to ultra high vacuum at elevated temperatures for a prolonged period, the
reflection departs markedly from the diffuse model. A thermal accommodation coef
ficient may be defined as the ratio of the net heat flux to the heat flux that would
occur with diffuse reflection; i.e.
In the case of a light gas and a heavy metal surface, for example helium or hydro
gen on tungsten, the measured thermal accommodation coefficients may be small
in comparison with unity. In addition, the reflected molecules may retain a signif
icant fraction of their incident parallel momentum. In order to take these effects
into account, analytical and computational studies may employ a combination of
diffuse and specular reflection. While this can reproduce the measured accommo
dation coefficients, the distribution function of the reflected molecules is inconsis
tent with measurements in molecular beam facilities. The measured distributions
may be reproduced by a model based on the theory of Cercignani and Lampis
(1974) that was further developed by Lord (1991). This is called the CLL model
but, because the parameters in the model depend on surface finish and contamina
tion as well as the composition of the gas and surface, the development of the data
base that would be required for the systematic application of the model to practical
problems is an almost impossible task.
For a stationary gas in equilibrium with a diffusely reflecting surface, the
shear stress and heat flux are zero and the incident and reflected molecules con
stitute the two halves of an equilibrium distribution. However, when there is finite
MOLECULAR FLOW 14.13
shear stress on or heat flux at the surface, the distribution function of diffusely
reflected molecules will again be the equilibrium distribution at the surface tem
perature, but that of the incident molecules will be a nonequilibrium distribution.
There is then a Knudsen layer with a thickness of the order of ten mean free paths
in which the bimodal surface distributions gradually change to the Chapman-Enskog
distribution of Equation 14.35. The heat flux and shear stress are essentially con
stant through the Knudsen layer, but there is a temperature and/or velocity discon
tinuity at the surface. These temperature and velocity discontinuities are generally
referred to as velocity and temperature slips and their magnitude is similar to the
differences between the quantities based on the upward and downward moving
molecules in the Chapman-Enskog distribution. As noted earlier, Equations 14.36
and 14.37 indicate that the fractional slips are of the order of the local Knudsen
numbers. The boundary conditions at a surface that are commonly employed with
the Navier-Stokes equations for continuum flow assume zero slip. These boundary
conditions are physically invalid for a gas flow, but the errors become tolerable
at sufficiently low Knudsen numbers.
The Knudsen layer is not amenable to analysis and involves physical effects that
disappear in both the collisionless and continuum limits. The best known of these is
the thermal creep that occurs when there is a temperature gradient along the surface.
Thermal creep should not be confused with the thermal transpiration that occurs
in a collisionless gas. Transpiration produces a flow from a hot to a cold gas as a
consequence of the molecular speeds being greater in a hotter gas. The flow veloc
ities associated with thermal creep are from the cold region to the hot region and
are therefore in the opposite direction to transpiration.
The conventional approach has been to classify flows on the basis of their Knudsen
number and to place them into the continuum, slip, transition, or free molecule
regimes. It is then assumed that the Navier-Stokes or, in the case of inviscid flows,
Euler equations provide an adequate mathematical model in the continuum regime.
The slip regime applies when the continuum boundary conditions must be modified
to take account of the velocity and temperature slips at solid surfaces. The free mol
ecule or collisionless regime is the limiting high Knudsen number case in which
collisions between molecules can be neglected. The transition regime also requires
a molecular description, but intermolecular collisions must be taken into account.
Fixed Knudsen number criteria can be meaningfully defined if they relate to a local
Knudsen number based on the scale length of the appropriate macroscopic flow
property. Overall Knudsen numbers are much less useful and the tendency has been
to specify arbitrary Knudsen number criteria that seriously understate the size of
the transition regime. For example, the Knudsen number that defines the bound
ary between the transition and free molecule regimes has generally been set to 10 ,
whereas 100 is generally more appropriate.
14.14 FLUID F L O W H A N D B O O K
Equation (14.35) shows that the transport terms in the Navier-Stokes equations
are based on the first term of the Chapman-Enskog expansion in powers of the
local Knudsen numbers. The use of the Navier-Stokes equations leads to notice
able errors when the local Knudsen numbers are of the order of 0.1 or higher and
the equations cease to be useful at local Knudsen numbers of the order of 0.2. The
next term in the expansion leads to the Burnett equations that have vastly more
complex transport terms. Caution is required in the progression to higher order
solutions because it has been shown that the Chapman-Enskog expansion is not
uniformly convergent and, at sufficiently high Knudsen numbers, it ceases to
converge. The additional terms do not necessarily lead to an improved result and
some early results from the Burnett equations appeared to worsen the agreement
with experiment. It now appears that these were in error and that the Burnett equa
tions can extend the validity of the continuum model to higher Knudsen numbers.
However, the degree of extension is uncertain and the complexity of the Burnett
equations is such that particle-based methods appear to be preferable for prob
lems in the relevant Knudsen number range.
The transport terms are absent from the Euler equations for inviscid flow and,
since mass momentum and energy must be conserved at the particle as well as at
the continuum level, it might be thought that these equations are applicable at all
Knudsen numbers. However, it has been noted that isentropic flow at the continuum
level corresponds to molecular flow with the equilibrium or Maxwellian velocity
distribution at the molecular level. The equilibrium distribution function is isotropic
in velocity space and, if the density of a flow that is non-isotropic in physical space
is continually reduced, the number of intermolecular collisions eventually becomes
insufficient to maintain the local equilibrium distribution in velocity space. The con
tinuum model then fails because the pressure and temperature cease to be scalar
quantities. Bird (1970) studied gaseous expansion flows and showed that the break
down of equilibrium could be correlated by the breakdown parameter
D (lnp)
(14.39)
Dt
For steady flows, this equation can be written
V?r A dp
P. =----
2
.v—
p dx
(14.40)
and we recover a local Knudsen number similar to those that appear in the Chapman-
Enskog distribution. The breakdown effects become significant when this parameter
exceeds about 0.05. The transition regime extends from local Knudsen numbers
of the order of 0.01 to those of the order of 100. This flow regime poses the great
est challenge and the free molecule or collisionless flow model provides a relatively
straightforward approach for problems with Knudsen numbers above this range.
MOLECULAR FLOW 14.15
(14.41)
This may be compared with the idealized continuum result that applies at the oppo
site limit of mean free path very small in comparison with the dimensions of the
hole. It is assumed that the sonic line lies in the plane of the hole and, using the
standard one-dimensional steady flow equations for the ratios of the sonic to the
stagnation flow properties,
k+1
(14.42)
The ratio of the collisionless to the continuum mass flux is a function of the spe
cific heat ratio and is 0.5494 for a monatomic gas and 0.5826 for a diatomic gas.
The axially symmetric spatial distribution of the effusing molecules is of inter
r
est in many applications. Consider a point at distance from the center of the hole
A r
of area and at angle to the outward normal through the center. For very large
6
in comparison with the linear dimensions of the hole, the ratio of the density to the
stagnation density is readily shown to be
n A cos# (14.43)
n 4nr
0 2
This jet of molecules from a stationary gas through a small hole is sometimes called
a thermal molecular beam. High energy molecular beams are produced from a
14.16 FLUID F L O W H A N D B O O K
supersonic stream by the molecules passing through a small hole at the apex of a
conical skimmer. The flux in this case is given by Equation 14.23.
Because there are no intermolecular collisions, the result of Equation 14.41 may
be applied separately to gases on both sides of the surface containing the hole or
holes. In equilibrium, there must be equal mass fluxes in each direction. Denoting
A
the gas on one side by subscript and that on the other side by subscript Equa
tion 14.41 gives
Pa — Pb
(14.45)
/
nI\> I
(14.46)
n\2
Cascades of gaseous diffusion cells containing uranium hexafluoride have been the
most widely used process for the enrichment of uranium. The inverse mass ratio of
2^5 238
~ UF 6 to UF 6 is 1.00860 so that the theoretical enrichment factor of each stage
is 1.00429. The naturally occurring fraction of uranium 235 is about 0.007 and a
very large number of stages are required in the cascade.
These problems are amenable to exact analysis and provide another opportunity to
compare the limiting case of extreme rarefaction with the well-known continuum
MOLECULAR FLOW 14.17
solutions. First consider the heat transfer between two stationary plates normal to
cl
the v axis and of infinite extent that are separated by a distance that is small in
comparison with the mean free path of the monatomic gas between the plates.
Assume diffuse reflection with complete thermal accommodation to the temper
atureTu at the upper plate and TL at the lower plate. The uniform number density
n of the gas between the plates is equal to the sum of the number density riu
of
the molecules reflected from the upper plate and the number nL
reflected from the
lower plate. The flux of diffusely reflected molecules from a surface is equivalent
to the effusion of molecules from a fictitious gas on the reverse side of the sur
face. The separate density components can then be derived from the effusion Equa
tion 14 .4 1 as
The upward and downward directed heat fluxes may be obtained by substituting
the fictitious gas properties into Equation 14.29. Noting that the fictitious gas den
sity is twice that of the effusing gas and using Equation 14.47, the net collision-
less heat flux is
q = - 2 m p n - m Rm j T X ( ^ - j T L) (14.48)
The heat transfer in the continuum regime is -k(dt/dy) and, for k= CT°where C
is a constant, the corresponding continuum result is
C( T”+'-Tr')
q = ------------------ - 1 (14.49)
(C0+\)d
The collisionless or free molecule heat transfer is proportional to the gas density and
independent of the plate spacing, whereas the continuum heat transfer is inversely
proportional to the plate spacing and independent of the density. This means that the
ratio of the continuum to the collisionless heat transfer is proportional to the
Knudsen number of the flow.
Now consider the case in which the upper plate is moving in the x direction with
speed U. The shear stress may be obtained through Equation 14.27 as
r = -r £ ^ l ^ - (14.50)
and is again proportional to the density and independent of the plate spacing. The
gas between the plates is macroscopically uniform and the velocity distribution func
tions of the upward and downward moving molecules are half-range (in the direcy
tion) equilibrium distributions. These may be integrated over the appropriate ranges
14.18 FLUID F L O W H A N D B O O K
to determine the stream velocity and temperature from Equations 14.7 and 14.11,
to give
u=ujrL/(jrj-jrL) (14.5D
T = [ ^ { T , + ' A U 2 / R ) + J f u T L] / + -ft) (1 4 -5 2 )
The differences between these values and the plate values are the velocity and tem
perature slips at surfaces in the limiting case of collisionless flow.
The pressure and shear stress due to the incident molecules on a surface element at
a given inclination to a stream are given by Equations 14.25 and 14.27. For diffuse
reflection, the shear stress due to the reflected molecules is zero and the pressure due
to these molecules is given by the stationary gas Equation 14.46 with the “fictitious
effusion” number density that is required for molecule conservation. The incident
number flux is given by Equation 14.25 and the fictitious effusion number density
is the density in the stationary gas number flux Equation 14.24 that, at the surface
temperature, matches the incident flux. For the less realistic case of specular reflec
tion, the pressure due to the reflected molecules is equal to that due to the incident
molecules and the reflected shear stress is exactly opposite to the incident shear stress
so that the net shear stress is zero. For a surface element at incidence a in a stream
s T
of speed ratio and temperature U, the pressure ratio is
_P_ 1+
,—
£ssin
. — £ —T
a+ ------
1
e x p (-s 2sin 2a )
P~ Vn 2 V t
£ Tr
where is the fraction of specular reflection and is the temperature of the reflected
molecules. The corresponding result for the shear stress is
T
= ~J=~scos « [ex p (- 52 sin a)+ -[kssin a { l + erf (s sin a)}j
2
P oo
(14.54)
MOLECULAR FLOW 14.19
As long as the shape of the object is such that reflected molecules cannot re-
impinge on its surface, the forces are readily determined by integration over the sur
face elements.
The drag force per unit area of a vertical flat plate with diffuse reflection and
complete thermal accommodation to the surface temperature Tw
is obtained by sub
tracting the pressure at an incidence o f - 7r/2 from that at an incidence of 7r/2 . The
drag coefficient is the ratio of the drag per unit frontal area to ZipU ps
= , so that
CD= - 7 ^ e x p ( - r ) + f2 + - ^ le r f ( s ) + ^ - (14.55)
Vtf .v V s J s Tm
The limiting drag coefficient at high values of the speed ratio is two and this is con
sistent with the corresponding continuum value. The opposite limit at low speed
s s
ratios is less clear, but erffo) —> exp(-.v 2) / V 7Tfor small and the drag coeffi
2
cient tends to (4/V tT + \ttTJt2Is). The pressure and shear stress may be inte
grated over a sphere and the result is
(14.56)
The low speed ratio limit for diffuse reflection and a surface temperature equal to
the stream temperature is (6 / V 7r~ + 2 V7r/3)/v.
The large values of the free molecule drag coefficients for very small speed
ratios are of interest but are not relevant to very small particles in air under stan
dard conditions. The mean free path under standard conditions is approximately
5 x 10-8 m (1.5 x 10 7 ft) and, for free molecule theory to be applicable, the sphere
diameter would have to be less than 5 x 10 9 m (1.5 x 10~ 8 ft). This is less than
twice the mean molecular spacing and a sphere of this size would be more appro
priately treated as a large molecule.
The pressure due to the diffusely reflected molecules is proportional to the square
root of the surface temperature so that there will generally be a net force acting on
an immersed body with a non-uniform surface temperature. The temperature vari
ation is most commonly caused by thermal radiation and the phenomenon is then
called photophoresis. Consider a thin flat plate with temperature Ty
on one side and
T lon the other side immersed in a stationary equilibrium gas at temperature under T
free molecule conditions. The pressure due to the incident molecules is the same
14.20 FLUID F L O W H A N D B O O K
on each side, but that due to the reflected molecules differs. Equations 14.24 and
14.26 show that
The resultant force on the plate causes it to move toward the colder side with speed
such that the drag force balances the photophoretic force. The analysis leading to
Equation 14.55 may be repeated for the case with a temperature difference and the
drag set to zero. The result is
erf(s)
2 s e x p (-r)
2
(14.58)
/
KS
- ( 1 + 2 .r ) e r f ( s ) - =0
Consider the case in which TL=T T(Jis only slightly larger than T.The speed
and
ratio is then small and may be written
Figure 14.3 shows how this limit fits into the more general results from a numeri
cal solution of Equation 14.58. A temperature difference of one or two degrees is
sufficient to produce thermophoretic speeds of the order of a few tenths of a meter
or foot per second. It is tempting to use this result to explain the behavior of the
well-known Crookes radiometer, but the pressure in these devices is such that the
flow is in the transition or continuum regimes.
Thermophoresis relates to a force that acts on a body immersed in a gas with a
temperature gradient even when there is a uniform surface temperature. This force
is a result of the asymmetry of the Chapman-Enskog distribution of Equation (14.35).
This distribution may be averaged through Equation 14.18 to calculate the pressure
and shear stress on a small element. A half-range integration may be made for the
incident molecules on a surface element that is sufficiently small in comparison with
the mean free path for the free molecule assumption to be satisfied. The result is
_J_______ 45 cos
Pj_ A dT
6 ^ r, _ 45 sin 6 AdT
p~ 2 4(5-2co)(l-2co)T~dy ™ ~p” 8 (5 -2 a))(7 -2 c o ) T~dy
( 14 . 60 )
MOLECULAR FLOW 14.21
F I G U R E 14.3 The speed ratio of a thin surface that has a temperature Tv on one surface and the
other surface at the gas temperature T.
where is the angle between the surface normal and the direction of the temperature
6
gradient. These equations apply directly to thin flat plates and may be integrated over
more complex shapes. The pressures and shear stresses due to the reflected mole
cules are unaffected by the temperature gradient. A flat plate at any inclination to
the direction of the temperature gradient experiences a force in this direction. The
ratio of the induced forces per unit area to the pressure is of the order of the local
Knudsen number based in the scale length of the temperature gradient.
This topic is of central importance to the designers of high vacuum systems. The
vacuum literature deals with the “conductance” of the tube and this is expressed in
liters of gas (at the inlet conditions) per second. This can be regarded as an effective
flow velocity at the inlet conditions times the cross-sectional area and is a device
dependent quantity. An orifice can be regarded as a tube in the limit of zero length
14.22 FLUID F L O W H A N D B O O K
to diameter ratio. Equation 14.41 shows that the effective velocity is l/( 2 V 77/30)
and the conductance is the product of this velocity and the cross-sectional area.
Now consider free molecular flow through a uniform tube or channel from x
= b.
0 to jc = The number flux of molecules per unit perimeter length to element
dx x
of the tube at location directly from the upstream gas is denoted by Ny(x) x.6
Let P(x\ x) be the probability that a molecule that is reflected from element dx'
at x ' is then incident on element dx.
The total number flux at location is x N\(x)
and is given by the solution of
h
N(x)=Nt(x)+j P(x,x)N(x)dx (14.61)
0
This is a Fredholm integral equation of the second kind and it generally appears in
the solution of collisionless flows that involve multiple reflection of the molecules.
The evaluation of N\(x) and P(x\ x)
requires a conventional but tedious integra
tion over the relevant distribution function for each location and pair of locations,
respectively. Similar calculations must then be made at each location to calculate
the number flux N{Jthat passes directly through the tube without any surface inter
actions and the probability Pe(x)
that a molecule reflected from location passes x
directly through the tube. The total number flux through the tube is then given by
b
N,=Nd+jN(x)Pe(x)dx (14.62)
0
Despite the formidable task, solutions have been obtained through these equa
tions for problems with simple geometry (Clausing 1932). Clausing’s results have
been summarized by Loeb (1961) and are shown in Tables 14.2 and 14.3. These
results are widely used in the vacuum engineering handbooks and information sites
such as the Semiconductor International site at www.semiconductor.net.
TABLE 14.2 Theoretical Results for the Mass Flux through a Circular Tube
TABLE 14.3 Theoretical Results for the M ass Flux through a Thin Slit
The largely analytical approach becomes quite unworkable for more complex
internal flows, but this type of problem is ideally suited to the test particle Monte
Carlo method (Davis I960). Because there are no intermolecular collisions, the mol
ecular trajectories are independent of one another and the overall flow may be deter
mined from the superimposition of a large number of representative trajectories.
Random numbers are employed for the sampling of the trajectory properties from
the appropriate distribution functions. The entering molecules are uniformly distrib
uted over the entry plane and the cosine distribution of the effusion theory applies
to both the entering and reflected molecules. The molecular speeds are not required
and the problem is essentially one of three-dimensional geometry. The following
theorem provides a straightforward solution for problems that involve quadric sur
faces. The equation of a molecular trajectory from point y„ can be written (x„ zd
x=xi+l\s, y=yi +m\S, i =Zi+ti\S, (14.63)
where sis the distance to an intersection at (x, y, z) with the quadric surface
f(x,y, z) = anx2+a y2+a 33z2+ aiyxz+ aM
2 2 zx 2 2
+ anxy+ 2 a[4
2 x+ 2 a24y+ 2 aM
z+ au= 0
An intersection point is given by setting sin Equation 14.63 to the smaller of any
positive real roots of the quadratic
A{s +2A2s+A}= 0
2 (14.64)
where
^3 = / ( - W mZ,-)
The method has been implemented in the following program for the free molecule
flow through a tapered tube with a circular cross-section. The tube is part of a cone
with apex semi-angle a, so that the non-zero quadric coefficients are
au- tan 2 a , a
22 = - 1, a 33 = - 1, al4 = ta n a , a =1
44
Advantage may be taken of the axial symmetry to set z, = 0 for each trajectory seg
A$
ment. The coefficient is zero for trajectories starting on the surface and there is then
just one non-zero solution. The Fortran 90 program is written in free source form.
It is easily ported to other languages and the comments make it self-explanatory.
PROGRAM fmtat
! calculates the free molecule flux ratio through a tapered tube
IMPLICIT NONE ! all variables must be declared
REAL :: ranf,b,c,ang,sang,cang,tang, rm,al,am,an,xm,cl,c 2 ,c 3 ,S I ,S 2 ,s
INTEGER :: tnt,n t ,np,nd,nr,i
! ranf is a random fraction between 0 and 1
! b is the length of the tube (the inlet is unit radius atx=0)
! c is the exit radius
! ang is the taper angle (the tube centerline is along the x axis)
! sang,cang,tang are the sine,cosine,tangent of ang
! xm,rm x coordinate and radial coordinate of the molecule
! al,am,an are the x,y,z direction cosines of the trajectory
! cl,c2,c3 are working variables
! s the distance to the next intersection with the surface
! sl,s2 solutions of the quadratic for s
! tnt is the total number of trajectories, nt is the progressive number
! np is the number of molecules that pass through the tube
! nd is the number of molecules that pass directly through the tube
! nr is the number of reflections from the surface
! i working integer
WRITE (*,*) 'Input the ratio of the length to entry radius : ' ; READ (*,*) b
WRITE (*,*) 'Input the exit to entry radius ratio : ' ; READ (*,*) c
WRITE (*,*) 'Input the total number of trajectories : ' ; READ (*,*) tnt
np=0 ; nd =0 ; nr=0
ang=atan2(c-1.,b) ; sang=sin(ang) ; cang=cos(ang) ; tang=sang/cang
DO nt=l,tnt ! loop over the independent trajectories
CALL random_number(ranf) ; rm=sqrt(ranf) ! uniformly distributed radius
CALL dcs(al,am,an) ! the effusion is in the x diection
cl=sqrt(rm*al*tang*(rm*al*tang-2.*am)+am**2+(1.-rm**2)*an**2)
c2=al*tang-am*rm ; c3=al**2*tang**2-am**2-an**2
sl= (-c2+cl)/c3 ; s2=(-c2-cl)/c3 ; s=l.e6
IF ( (sl>0.).AND. (sl<s)) s=sl ! s is set to the lowest +ve value, or
MOLECULAR FLOW 14.25
The computation time depends on the average number of surface interactions per
trajectory. For a uniform tube, this number is remarkably close to the length to radius
ratio. For a length to radius ratio of four, the computation rate on a PC is more than
300,000 trajectories per second per gigahertz CPU speed. A sample of ten million
trajectories is sufficient to reduce the statistical scatter to a negligible level and is
generally obtained in less than a minute. Typical results are shown in Figure 14.4.
The results for a uniform tube agree with Clausing’s results for relatively short tubes
but are several percent below his values for the longer tubes.
A similar program has been written for the flow through a uniform tube with
a constriction at its midpoint. This takes the form of a thin surface normal to the
14.26 FLUID F L O W H A N D B O O K
C
o
CO „ _
d 0.5
3=
LU
2 4 6 8 10
Length / Inlet Radius
F I G U R E 14.4 The effective conductance of tapered tubes that have a circular cross-
section.
axis that contains a circular orifice that is centered on the axis. The listing follows
and representative results are shown in Figure 14.5.
PROGRAM fmcot
! calculates the free molecule flux ratio through a circular tube
! that is divided into equal segments by a plane orifice constriction
IMPLICIT NONE ! all variables must be declared
REAL :: ranf,b, c ,a l ,am, an, rm, s t ,s c ,r c ,xc
INTEGER :: tnt,n t ,np,nr,nd,i
! ranf is a random fraction between 0 and 1
! b is the length of the tube which has unit radius
! c is the radius of the orifice in the plane constriction at x=b/2
! rm radius of the location of the molecule
! al,am,an are the x,y,z direction cosines of the trajectory
! st is the distance to the next intersection with the tube
! sc is the distance to the next intersection with the constriction
! rc radius of intersection with constriction
! xc x coordinate of intersection with the surface
! tnt is the total number of trajectories, nt is the progressive number
! np is the number of molecules that pass through the tube
! i surface indicator
WRITE (*,*) 'Input the ratio of the length to the radius : ' ; READ (*,*) b
MOLECULAR FLOW 14.27
WRITE (*,*) 'Input the constriction to tube radius ratio: ' ; READ (*,*) c
WRITE (*,*) 'Input the total number of trajectories : ' ; READ (*,*) tnt
np=0 ; nd=0 ; nr=0
DO nt=l,tnt
CALL random_number(ranf) ; rm=sqrt(ranf) ! uniformly distributed radius
CALL dcs(al,am,an) ! the effusion is in the x direction
st=(-rm*am+sqrt(am**2+an**2*(1.rm**2)))/(am**2+an**2) ! distance to tube
sc=b/(2.*al) ; rc=sqrt((rm+am*sc)**2+(an*sc)**2) ! constriction plane
IF ( (roc) .AND. (sc<st) ) THEN
xc=b/2. ; rm=rc ; i=2 ! the first reflection is from front of constriction
ELSE
xc=al*st ! impact point on cylinder
IF (xc>b) THEN
nd=nd+l ; np=np+l ; i=0 ! molecule passes directly through tube
ELSE
i=l ! the first reflection from point xc on tube
END IF
END IF
DO WHILE (i/ = 0) ! i will be set to 0 when the molecule exits the tube
nr=nr+l
IF (i==l) THEN ! reflection from inside of tube at z=0, rm=-l
CALL dcs (am, a l ,an) ! the effusion is in the x diection
14.28 FLUID F L O W H A N D B O O K
The vacuum handbooks suggest that the conductance of a variable area tube should
be based on the minimum cross-sectional area. This provides reasonable guidance
for contractions, but does not cope with expansions.
A further test particle Monte Carlo program has been written for the flow through
a uniform cylindrical tube that is divided into equal sections by a rectangular elbow.
MOLECULAR FLOW 14.29
This is a three-dimensional (low, and the listing of the Fortran 90 source code is as
follows:
PROGRAM fmelb
! calculates the free molecule flux ratio through a circular tube
! that is divided into equal segments by a rectangular elbow
IMPLICIT NONE ! all variables must be declared
REAL :: ranf ,b ,a l ,am, a n ,xm, ym, zm, rm, xmi ,ymi ,zmi,a ,c ,s i ,s2 ,s
INTEGER :: tnt,n t ,np,nd,nr,i
! ranf is a random fraction between 0 and 1
! b is the length of the tube which has unit radius
! the tube entry is at x=-b/2 and the exit is at y=b/2
! xm,ym,zm x,y,z coordinates of the molecule
! xmi,ymi,zmi x,y,z coordinates of the molecule
! rm radius of the location of the molecule
! al,am,an are the x,y,z direction cosines of thetrajectory
! a,c,i are working variables
! s is the distance to the next intersection with the surface
! sl,s2 are the solutions of the quadratic for s
! tnt is the total number of trajectories, nt is theprogressive number
! np is the number of molecules that pass through the tube
! nd is the number that pass directly through without collision
! nr the number of reflections from the surface
WRITE (*,*) 'Input the centreline length to radius ratio : ' ; READ (*,*) b
IF (b<2.) STOP ! each segment must be at least one radius inlength
WRITE (*,*) 'Input the total number of trajectories : ' ;READ (*,*) tnt
np=0 ; nd=0 ; nr=0 ! the counters have been initialized
IX) nt=l,tnt ! loop over the independent trajectories
CALL random_number(ranf) ;rm=sqrt(ranf) ! uniformly distributed radius
CALL random_number(ranf) ;a = 6 .28318531*ranf
xmi=-b/2. ; ymi=rm*sin(a); zmi=rm*cos(a) ! uniformly distributed position
CALL dcs (al,am, an) ! the effusion is in the x
direction
CALL quad(am**2+an**2.,am*ymi+an*zmi,ymi**2 +zmi**2-1. ,s i ,s 2 )
s=sl ; IF (s<0.) s=s2 ! geometry is such there is a +ve and a -ve root
xm=xmi+al*s ; ym=ymi+am*s ; zm=zmi+an*s ! interaction coordinates
IF (xm <= -ym) THEN
i=-l ! the reflection will be from the entry segment
ELSE
CALL quad(al**2+an**2,al*xmi+an*zmi,xmi**2+zmi**2-l.,si, s2)
s=sl ; IF (s2>s) s=s2 ! the larger of two +ve roots is applicable
xm=xmi+al*s ; ym=ymi+am*s ; zm=zmi+an*s ! impact coordinates
IF (ym<b/2.) THEN
i=l ! the reflection will be from the exit segment
ELSE
i=0 ; np=np+l ; nd=nd+l ! molecule passes directly through the elbow
END IF
END IF
DO WHILE (i/=0) ! i is set to 0 when the molecule exits the tube
nr=nr+l ; xmi=xm ; ymi=ym ; zmi=zm ! reflection from xmi,ymi,zmi
14.30 FLUID F L O W H A N D B O O K
Figure 14.6 compares the flow rate through the elbowed tube with that through
a straight tube of the same length. This would appear to conflict with the handbook
rule from the Semiconductor International site that the conductance is halved by
each rectangular elbow. However, this rule is based on a tube length measurement
along the shortest part of the outside of the tube. The length in Figure 14.5 is the
centerline length and the tube length of one would be zero if measured along the
shortest outside length. This brings this case and other short tube cases into reason
able agreement with the handbook rule. The centerline length is the logical measure
of the tube length and the handbook result is deceptive in that the halving of the con
ductance is a consequence of the measurement convention rather than the physics
of the flow.
The Boltzmann equation is the standard mathematical model of a gas at the mol
ecular level. This has the velocity distribution function for the molecules of class
cas the only dependent variable and, for a simple gas of VHS molecules, can be
written
00 4/r
7i
— [n /(c )]-h c .— [« /(c)] = J J 'i 2[/* (c )/* (c ,) - /( c ) /( c , )]crel 2dQ d c ,.
8
(14.65)
The left hand side contains the convective terms that are similar to those for the
conserved quantities in the equations for continuum flow. The right hand side is the
collision term that deals with the molecules scattered in an out of class c. The colli
sion partner is of class c, and the integration is over all collisions that involve class c
molecules. While there is only one dependent variable, the independent variables
are those of phase space. For a simple monatomic gas, phase space is comprised of
physical and velocity space. A homogeneous gas problem becomes one-dimensional
because the velocity distribution is then spherically symmetric. A one-dimensional
problem in physical space involves an axially symmetric velocity distribution func
tion and the Boltzmann solution becomes a three-dimensional problem. For two
or three-dimensional flows the velocity distribution is three-dimensional and the
problem becomes five or six-dimensional. Time is an addition dimension if the
flow is unsteady and each internal degree of freedom adds yet another dimension to
phase space. Moreover, simultaneous equations are required for gas mixtures. Not
surprisingly, there are very few analytical solutions of the Boltzmann equation for
non-trivial problems. Direct numerical solutions of the Boltzmann equation have
generally been limited to one-dimensional steady flow problems, although Aoki
(2 0 0 1 ) has obtained results that show the effect of cross-section on small-disturbance
pipe flow. The difficulties with the Boltzmann equation are such that most results
of engineering interest for transition regime flows have been obtained through direct
physical simulation of the gas at the molecular level rather than through solutions
of the mathematical model.
\ N(N)avgFn(t crel)maxAfm/ Vc
7 8 2
(n crcl)l(n: crel)mm
8 2 8 2 (*4.66)
The flow properties at the location of a cell are obtained by applying Equa
tions 14.6 to 14.16 to the molecules within the cell. All DSMC calculations are for
time-accurate unsteady flows although, for flows that eventually become steady,
time-averaging may be employed at sufficiently large times.
14.34 FLUID F L O W H A N D B O O K
0.9
x
D
U_
E
g 0.8
C
o
O
<D 0.7
0.6
FIGURE 14.8 Streamlines and contours of stagnation pressure ratio in the flow through a cir
cular orifice at a Knudsen number of 0.005.
10“ ft '). The imposed pressure ratio is 4.54. The results are in good agreement
with DSMC calculations that employed implicit boundary conditions (Fang and
Liou, 2001).
A simple result such as that for the orifice flow may give the impression that,
if the high and low Knudsen number limits are known, the transition regime behav
ior is qualitatively predictable. On the other hand, it was noted in Section 14.1.5 that
there are physical effects in the transition regime that do not appear in either the
free molecule or the continuum limit. In particular, care must be taken to allow for
thermal creep when there are large temperature variations in transition regime flows.
Many proposals have been made for processing in space because of the absence of
thermal convection in the zero-gravity environment. However, the convection cur
rents due to thermal creep can be of the same order as the gravitational convection
currents at sea level. For example, Figure 14.10 shows the vortex that is produced
by thermal creep at the tip of a flat plate in a gas at 300 K when one side of the
plate is at 450 K and the other is at 150 K. The semi-height of the plate is equal
to 25 mean free paths in the undisturbed gas. The maximum velocity at the edge
of the plate is 23 m/s (71 ft/s) but, over most of the vortex, it is of the order of 3
m/s (10 ft/s). The flow at the tip is from the cold to the hot side even though the
pressure on the front of the plate is higher than that at the rear. While this may
appear anomalous, it should be remembered that the pressure on the plate is due
to just one component of a tensor quantity. The net force on the plate is from the
hot to the cold side, as in free molecule flow, but the ratio of the force per unit
area to the undisturbed pressure is only 0.012 compared with 0.259 from Equa
tion 14.57) for free molecule flow.
V A R I A B L E S _____________ _____________________________
SUBSCRIPTS_____________________________________
SUPERSCRIPTS____________________
thermal component
+ based on molecules moving in the positive direction
based on molecules moving in the negative direction
* post-collision value
MOLECULAR FLOW 14.39
REFERENCES
1. Alder, B.J. and W ainw right, T.E. 1957. “Studies in M olecular D ynam ics.” ./. Chem.
Phys. 27, 1208-1209.
2. Aoki, K. 2001. “ D ynam ics o f R arefied Gas Flows: Asym ptotic and Num erical
A nalyses o f the Boltzm ann E quation.” AlAA P aper 2001-0874
3. Bird, G.A. 1963. “ A pproach to Translational Equilibrium in a Rigid-sphere G as.”
Phys. Fluids 6, 1518-1519.
4. Bird, G.A. 1970. B reakdow n o f Translational and Rotational Equilibrium in
G aseous Expansions.” A lA A Journal 8, 1998-2003.
5. Bird, G.A. 1994. M olecular G as D ynam ics a n d the D irect Sim ulation o f G as F low s,
Oxford University Press.
6 . Cercignani, C. and Lam pis, M. 1974. In Rarefied Gas Dynamics (ed. K. Karamcheti)
p. 361, A cadem ic Press, N ew York.
7. Chapm an, S. and C ow ling, T .G . 1952. The M athem atical Theory o f N on-uniform
Gases (2nd edition), C am bridge U niversity Press.
8 . Clausing, P. 1932. A nnin Phys. 12, 961.
9. Davis, D.H. 1961. “ M onte C arlo C alculation o f M olecular Flow Rates through a
Cylindrical Elbow and Pipes o f O ther Shapes.” J. Appl. Phys. 3 1 ,1 1 6 9 -1 1 7 6 .
10. Koura, K. and M atsum oto, H. 1991. “ Variable Soft Sphere M olecular M odel for
Inverse-pow er-law or Lennard-Jones Potential.” Phys. Fluids A 3, 2459-2465.
11. Liepmann, H.W. 1961. “G askinetics and G asdynam ics o f O rifice Flow .” J. Fluid
Mech. 10, 6 5-79.
12. Fang, Y. and Liou, W .W . 2001. “ Predictions o f M EM S Flow and Heat T ransfer
Using DSM C with Im plicit B oundary C onditions.” A lAA Paper 2001-3074.
13. Loeb, L.B. The K inetic Theory o f G ases, 3rd edition. Dover Publications.
14. Lord, R.G. 1991. “Som e E xtensions to the C ercignani-Lam pis Gas Scattering K er
nel.” Phys. Fluids A 3, 7 0 6 -7 1 0 .
15. M axwell, J.C. 1879. Phil. Trans. Roy. Soc. 1, Appendix.
CHAPTER 15
FLOW METERING
15.1
1 5 .2 FLUID FLOW HANDBOOK
Subscript
1 Upstream or Inlet Condition; reference plane 1
2 Downstream condition; reference plane 2
Flow relates to the motion of matter, and a meter is a device that measures or
monitors events. So. a flowmeter is a device that measures or monitors the motion
of matter. In the real world, a flowmeter is a device that measures the flow of mat
ter across a reference plane or a cross-sectional area and flow metering is the
process that accounts for the flow of matter. Flow metering is generally accom
plished by measuring the flow in units of mass or volume for a known time period
and thereby, inferring the average mass or volume flow rate per unit time (mVhr,
kg/sec, gpm, barrels/day, etc.) or by recording the total amount of matter (m3, kg,
lbm, MCF, etc.) that crosses a reference plane from a defined time of reference.
The precision of flow measurement depends on the type of accuracy desired.
If the flow measurement is to control the flow or a process in the plant, the mea
surement accuracy may be relaxed to meet the minimum accuracy limit accepted
for the control. For Fiscal or Custody Transfer measurement, the accuracy
requirements are precise and often defined by the National, State, and Interna
tional Standards.
Matter has three distinct phases or forms: solid, liquid, and gas. Solids have
definite shape, mass, and volume; liquids have definite mass and volume, while
the shape depends on the container or boundary of containment; and gases have
F L O W METERING 15.3
definite mass but the boundary or volume of the container defines volume and
shape. Therefore, the flowmeter output and flow measurement determination
method, for each state of matter, are often different. Precise flow metering of mat
ter in any of the three single phases is often difficult and the flow measurement
becomes more complex when the flow is in more than one phase (multiphase);
e.g., unsaturated steam (liquid and gas), muddy water (solid and liquid), fine dust
or powder (solid and gas), or a real-world field production of crude oil, immisci
ble fluids such as oil, water, and free gas.
Flow measurement of solids with identical shape and size can be accom
plished by simply counting the total number of pieces or measuring the weight of
solids passing through the reference cross-sectional plane. The counting method
of flow measurement is typical in an assembly line; e.g., automobile industry,
bottling plant, packaging/conveyor (egg cartons, medicine strip), etc. Mass mea
surement or the weighing method is prevalent in flow measurement of solids with
irregular shapes or granules like that of coal on a conveyor, sugar production, or
a nail manufacturing plant.
Flow rate measurement of fluids (liquid or gas) is accomplished by many dif
ferent techniques which vary widely, from the simple counting of volume flow of
equal measure in container (bottle, bucket, etc.) to the very complex electro
mechanical device which displays flow rate outputs either in mass or volume units.
For fluids, complexity in flow measurement arises from the fact that the volume
change of fluids due to changes in temperature and pressure is relatively higher
than that of solids. For most real-world applications, liquids may be considered as
incompressible, which makes precise measurement of liquid flows relatively eas
ier than that of gas flows. Since, a change in liquid volume is primarily affected
by the temperature change only, volume correction to the reference or base condi
tion is easily achieved, as the coefficient of thermal expansion of all common liq
uids are known. Compared to liquids, a volume change of gas due to temperature
and pressure is more profound and is more complex when the flowing gas is a mix
ture of several gases. To define volumetric flow rate in an unambiguous unit, the
relationship of volume and mass as a function of fluid temperature and pressure
must be known, and by monitoring the flowing temperature and pressure, the vol
umetric flow rate can be corrected to a reference or base temperature and pressure.
Most gases are non-ideal gases, whose fluid properties (density, viscosity, com
pressibility or expansibility, etc.) are functions of temperature and pressure. To
convert volume from the flowing condition to a reference or base condition vol
ume, the composition of the flowing gas must be known.
The operating principles of flowmeters are based on laws of physics. The two
laws most used are conservation of mass and conservation of energy. That means,
both mass and energy remain unchanged between two different reference planes
15.4 FLUID F L O W H A N D B O O K
where
INLET-PLANE 1
/
FLOW
OUTLET-PLANE 2
4. Direct Mass Meters: Thermal mass meter and Coriolis force mass meter
(using laws of Coriolis force), etc.. Some ultrasonic How meters can infer
mass flow rate by determining fluid density from the speed of sound
through the flowing fluid.
5. Energy Meters: Measuring flow rate in units of energy; i.e., Btu or joules
per unit of time. Most meters determine the fluid flow rate and energy con
tent per unit mass or volume of fluid and thereby, infer energy flow rate
across the reference plane or the meter.
6. Other Types of Meter: The principle of operation of these types of meters
employs other physical laws and is not classified under any of the above
categories. Some of these meters are open channel flow meters, notches
and weirs, dye concentration and dispersion; critical flow nozzles, choked
flows, etc.
7. Multiphase Meters: Most commercially available multiphase meters for
multiphase or immiscible liquid flows use multiple measuring techniques
and different laws of physics to determine the total flow rate and volume or
mass fractions of the different phase components.
The selection and performance of flow meters are largely dependent upon two
important factors: the properties of the flowing fluid and the flowing conditions.
There are many fluid properties; some are of fundamental importance while oth
ers are relevant only for certain types of meters and not for other meter types.
Some of the more important parameters are temperature, pressure, density, spe
cific gravity, viscosity, compressibility, and the fluid phase. The other parameters
that can influence the meter performance are vapor pressure, boiling point, elec
trical conductivity, thermal expansion, speed of sound, specific heat, average
velocity, and flashing and/or cavitation.
Another important factor is the unit of measurement. The fundamental units
are the mass, length, time, and temperature. Corresponding customary units in the
United States are pounds-mass, feet, seconds, and degrees Fahrenheit or degrees-
Rankin, while the SI dimensional Units (Le Systeme International d ’Unites) are
kilograms, meters, seconds, and degrees Celsius or Kelvin. The U.S. customary
unit of length can be in feet or in other derived units like inches, yards, miles, etc.
Details of dimensions or system of units are not discussed in this chapter.
parameter is important for the vortex flowmeter, which utilizes the principle of
flow-induced instability; but this nondimensional parameter has no noticeable
bearing on the performance of a turbine meter. The details of different nondi
mensional parameters and its physical implications on different flowmeters are
beyond the scope of this chapter.
Most flowmeters require certain conditioning of the flow. The flow profile in
the metering section should be free from fluctuations, vibrations, flow profile dis
tortions, secondary flows or swirls, etc. Some flowmeters are more prone to spe
cific flow distortions than others. For example, a turbine meter flow rate
measurement may be biased if swirl is present in the flow at the flow meter, while
the swirl present in the flow has no discernible influence on most displacement
meters. To achieve good flow measurement, the upstream-downstream straight-
pipe lengths, installation requirements, and the flow profile conditioning at the
metering section should conform to the requirements specified by vendors or
accepted standards for that type of flowmeter.
There are two classes of linear velocity meters. Those that employ physical laws
that generate meter response, a linear function of the average flow velocity,
include electro-magnetic (or mag), bladed turbine, vortex, ultrasonic, etc. Other
meters employ physical laws that in principle generate a nonlinear meter response
to the flow velocity, but secondary electronics associated with the primary flow
sensor generate an output that is a linear function of the average velocity. The
meters in the second group include the thermal mass flowmeter, the Doppler-shift
ultrasonic flowmeter, different types of insertion meters, etc.
Magnetic flowmeters, commonly known as mag meters, measure the flow rates
of electrically conductive liquids in a closed conduit. These meters are available
in sizes as small as 1/10th of an inch to about 96 inches in diameter.
Mag meters operate on the physical law of electromagnetic induction defined
by Faraday’s law, which states that when a conductor is passed through a mag
netic field, a voltage is generated and its magnitude is
E = constant x B x L x V (15.2)
where
E magnitude of voltage
B magnetic field density
F L O W METKRING 15.9
L path length
V average flow velocity
The generated voltage is proportional to the magnetic field density and veloc
ity of the conductor. Therefore, fluid temperature, pressure, density, or viscosity
has no influence on the voltage magnitude. The conventional construction of a
mag meter has coils mounted on the outside of a nonmagnetic pipe section. By
passing current through the coils, a magnetic field is generated inside the pipe
section. A schematic of magnetic flowmeter is shown in Figure 15.2. As conduc
tive liquids flow through the pipe, a voltage is generated, which is extracted
through a pair of electrodes. Pipes constructed with conductive material are lined
with nonconductive material to insulate the pipe from the electrodes and gener
ated voltage. The most common use of mag meters is for corrosive liquids, waste
water, sludge, and sewage. Mag meter rangeability may vary, as the maximum
velocity through the meter can be adjusted. Meters with rangeability of better
than 20:1 with accuracy of +/-0.5% are available. As magnetic field strength and
liners may be affected at a high temperature, there is a high temperature limit for
most mag meter designs.
Turbine meters measure How rates of both liquid and gas flows. Meter sizes vary
from a fraction of an inch to over 20 inches in line size. There are numerous
designs of turbine flowmeter, but the basic working principle is the same. A tur
bine meter consists of a rotor with multiple blades that are at an angle to the flow
velocity and the rotor is mounted on a bearing and shaft. As the blades restrict the
flow, momentum is transferred to the blade from the fluid, forcing the rotor to
turn. The rotational speed of the blades is proportional to the velocity of the flow
ing fluid and is monitored by several different methods. The most common
method is by sensing the passage frequency of the blades on the rotor and from
the buttons or slots on the shroud attached to the blades. Some turbine meter
designs (mostly gas flows) monitor the rotational speed of the shaft directly,
rather than via the blades. The signal pickup can be mechanical, electrical, mag
netic, optical, or any combination of these. Insertion turbine meters are also avail
able, to measure point velocity in a large flow area, but usually with a reduced
measurement accuracy.
Generally, liquid turbine meter rangeability is 10:1 but some special designs
can achieve 25:1. For compressible fluids, rangeability varies with the density of
the gas, which is a function of temperature and pressure. A gas turbine meter with
rangeability of 10:1 at atmospheric condition can have a rangeability of better
than 100:1 at 1400 psi pressure.
A turbine meter has one moving part and unless the bearing friction and/or
angle or integrity of the blade changes, turbine meter performance should not
change. At high flow rates, momentum transfer is high and the effect of frictional
force of a failing bearing can still be relatively small to affect the meter perfor
mance. In general, with proper maintenance and care, turbine meters are highly
repeatable and reliable devices. For liquids, turbine meter repeatability can be +/-
0.02% or better with a linearity of better than -»-/-0.25%. For gas flows, meter
repeatability is of the order of +/-0.1% for linearity of +/-1.0% or better. A tur
bine meter output is affected if a swirl present in the flow and the output bias is
positive or negative depending on whether the swirl aids or retards the rotation of
the blade. In compressible flows, presence of a measurable flow pulsation is
likely to result in a positive bias on the meter output. If profile distortion, flow
swirl, or pulsation is expected in the flow, the flow must be conditioned to elim
inate or reduce those flow disturbances to achieve repeatable and precise mea
surement by a turbine flowmeter.
As the name implies, the sonic signal transmitted to determine flow velocity is
beyond the audible range of frequency for humans. Two types of ultrasonic
flowmeters (UFM) in pipe flows are: the transit time, which usually uses pulse
transmission and the Doppler shift, which uses continuous signal transmission. It
is interesting that the first U.S. design patent on UFM was issued in the late 1920s
while the first commercial version of UFM used for fiscal measurement was
available in the 1980s, only after the microprocessor speeds were adequate for
signal processing.
F L O W METERING 15.11
Transit time UFM can be in time domain or frequency domain. Both types
transmit pulses that propagate through the flowing media to a receiver. A pair of
transceivers, each of which is both a transmitter and a receiver, is so installed on
the pipe wall that the transmitted and received signals travel in an upstream direc
tion (against the How) and in a downstream direction (with the flow). The signal
paths are at an angle to the flow. The velocity equation of the transit time UFM
is independent of the velocity of sound through the flowing fluid and is a func
tion of the length of the signal path (between paired transceivers), the difference
in upstream and downstream transit times, and the angle between the flow and
signal path. The speed of sound through the flowing fluid can be calculated from
the average of upstream and downstream transit times.
Doppler shift meters transmit a continuous sonic wave at an angle to the flow.
The beat frequency between the transmitted signal and that scattered from mov
ing particles in the flow provides the flow velocity information. For Doppler
meters to work, the flow must have particles (which may be invisible to the naked
eye) to scatter the transmitted signal and for accurate measurement, the particle
velocity must match that of the flow. For the Doppler shift UFM, calculation of
the flow velocity requires the knowledge of the sonic velocity through the flow
ing media.
Measurement accuracy of UFM depends on the symmetry of the velocity pro
file through the pipe and is improved by installing multiple paths of paired trans
ceivers. The rangeability of multipath UFM can be 20:1 or better with
measurement accuracy of +/-0.5% for gas and better than +/-0.25% for liquids.
The transit time meters are inherently computation intensive and most meters
provide many self-diagnostic features. The relative position of paired-sensors
puts one transceiver upstream and the other downstream of flow, irrespective of
flow direction; hence, The transit time UFM is inherently bi-directional.
There are different designs of sensor mount and precision of the meter often
varies with the meter type and sensor mount. The flow rate calculation of UFM
depends on the flow regime (laminar and turbulent) and the meter performance
normally deteriorates in the transitional flow regime. Application of UPM over
high pressures and temperatures may be limited if the transceiver performance is
affected above a certain operating condition.
When a flowing media strikes an obstruction, it moves around the object and con
tinues to flow downstream. For a non-streamlined object, at the separation point,
a vortex swirl or an eddy is formed under certain flowing conditions. As this
occurs, the formation of a vortex on one side causes a decrease in velocity and
increase in pressure, while the velocity increases with the decrease in pressure on
the other side of the object. As the vortex is shed from the object from one side,
the process is reversed and a vortex is formed on the other side of the object. The
frequency of formation and shedding of alternate vortices is proportional to the
15.12 FLUID F L O W H A N D B O O K
velocity of the fluid past the obstructing body, which is also referred to as the bluff
body or vortex shedder. In 19 12, Theodore von Karman studied the phenomenon
of vortex shed by bluff bodies. The relationship of shedding frequency to the flow
ing velocity is given by a nondimensional parameter, the Strouhal number
St =^ f (15.3)
where
More recent work has demonstrated that use of a non-streamlined object with
sharp and well-defined edges generates more consistent and repeatable vortices.
Many techniques are in use to detect the vortex shedding frequency. The sensi
tivity of detectors determines the rangeability of the meter. Since flow velocities
in all line sizes are of similar ranges, proportional increases in the width of the
vortex shedder of larger line sizes has a decrease in shedding frequency and sen
sor response to the vortices becomes relatively difficult to detect. Advances in
electronics and sensor technology have improved the sensitivity of detection of
the shed vortices, hence improved measurement precision and rangeability, and
extended the size vortex meter size to at least 12 inches in diameter.
Vortex meters are most often used in measuring steam flows, as in most
designs, detection sensors are away from the flowing fluid and can withstand
much higher temperatures than sensors in most other types of flowmeters. The
rangeability of the vortex meter is better than 10:1, while 40:1 may be achieved in
some meter sizes. Generally, the pipe Reynolds number should be above a mini
mum specified by the manufacturer, which is typically between 10,000 and
20.000. Vortex shedding ceases when the Reynolds number is between to 3,000 to
5.000. The typical accuracy of vortex meters in liquid flows is +/-0.75 percent or
better while for gas flows, it is about +/-1.5 percent or better. Vortex meters can
be over-ranged by as much as 150 percent without damaging the meter internals.
The internals of a fluidic meter are so designed, that as the flow enters the meter
body, the flow stream attaches itself to one wall of the flow path, which is due
to the flow phenomenon know n as the Coanda Effect. A small portion of flow is
then diverted through a feed back path, causing the flow stream to separate from
that side of the meter and attach itself to the other side and the process is
repeated. The frequency of the oscillating flow stream is directly proportional to
the flow velocity. From the output of a sensor in the feedback path of the fluid,
the flow rate is determined.
F L O W METERING 15.13
Fluidic meters are available in I inch to 4 inch line sizes. For larger line sizes,
a I inch meter may be used to measure the How rate as a bypass meter, provided
the bypass flow is proportional to the flow rate in the main flow. Rangeability of
30:1 is achievable by fluidic meters with accuracy of +/-1.0 percent or better.
There are no moving parts in fluidic meters and the meter may be over-ranged by
as much as 400 percent without damaging the meter internals.
The thermal mass flowmeter, as the name implies, depends on variations of heat
transfer characteristics at a heated sensor in the flow. The theory is applicable to
all fluids, but commercial versions are limited to measuring gases only. The heat
transfer rate from a heated element in the flow follows the classic King equation
for a hot wire, where flow rate is a function of thermal conductivity of the fluid,
fluid density, and the temperature difference between the heated element and the
flowing fluid. Although sensor response to the heat transfer rate is a nonlinear
function of flow velocity, the electronics in the feedback loop of the sensor gen
erates an output that is a linear function of the flow velocity.
Most commercial versions of thermal mass flowmeters share similar design
features and outputs are in 0-5 VDC or 4-20 mA that are proportional to flow.
Thermal mass flowmeters are generally installed in a bypass flow that is propor
tional to the flow in the main flow stream. Rangeability of thermal flowmeters
varies with application. The meter accuracy is about +/-l.0% of full scale with a
repeatability of +/-0.2%. Most thermal meters can be installed in any orientation
but for some designs, the pressure effect and drift may depend on the sensor posi
tion, so each meter may have a preferred orientation. Gas must be relatively clean
to eliminate coating on the heated element, which is likely to alter the heat trans
fer characteristics and thereby the measurement. Therefore, flows to the meter
may have to be installed with filters.
The operating principle of this type of meter is similar to that of a turbine meter
but the rotor is shaped like the propeller of a ship. This type of meter is predom
inantly used to measure irrigation and sewage flows. The revolution of the rotor
is normally monitored by a magnetic drive coupled to a mechanical display,
thereby isolating the display from the fluid being measured.
For many years, the differential pressure (DP) flowmeter has been the most
widely used flowmeter for fluid flows in pipes, because of its low cost of owner
ship, ease of use and maintenance, and precision of measurement. The most
15.14 FLUID F L O W H A N D B O O K
widely used head-type device is the orifice flowmeter, whose first recorded use
as a flowmeter was in late 1700s but the modern day orifice meters date back to
the very early years of 1900. The operating principle of all head type devices
relates to two simple physical laws; conservation of mass (Equation 15 .1) and
conservation of energy. The conservation of mass or the Continuity Equation is
discussed in Section 15.3. The modified equation of conservation of energy in
fluid flows is the Bernoulli equation, named after the physicist who derived the
equation. Referring to Figure 15 .1, the generalized Bernoulli equation is
defined as
(15.4)
where
For incompressible flows, the change in fluid densities between the two ref
erence planes is negligible and densities can be assumed as constant, while for
compressible flows, density change is normally significant and must be consid
ered in the flow rate equation.
From Equations 15 .1 and 15.4, theoretical flow rate through the system can
be shown to be proportional to the square root of the differential pressure between
the inlet-outlet planes, the density of the fluid, and the area ratio of the inlet-out-
let plane. Due to the viscosity of real fluids, wall roughness, and abrupt changes
in flow path, there will be energy dissipation. The actual pressure difference
between the two reference planes will always be higher than that, which is cal
culated by Equation 15.4 for ideal flows, which assumes no loss of energy or
energy dissipation between the two reference planes. The ratio of the actual flow
rate to the theoretical flow rate calculated from the actual differential pressure
between the two reference planes is the correction factor and referred to as the
Discharge Coefficient, Cd. The generalized equation of volume flow rate, Qv, for
all differential pressure flowmeters is given by
(15.5)
where
Qm=p Qv (15.6)
This DP device is a thin, concentric square-edged hole or a bore in a flat plate, held
between two flanges or installed in a fitting. The hole in the plate is the flow
restriction and the ratio of the bore diameter to the pipe diameter is the beta ratio,
(3. Differential pressure is measured at locations upstream and downstream of the
plate. There are many accepted locations of upstream and downstream pressure
taps. In the U.S., the most preferred location is l inch upstream and downstream
from each face of the plate, and the taps are referred to as the flange taps, as the
taps are normally on the upstream-downstream flanges that hold the orifice plates
in place. The preferred location by the International Standards Organization (ISO)
is at the upstream- downstream corners of the inside pipe wall and the orifice plate
15.16 FLUID F L O W H A N D B O O K
and is known as the corner taps. Orifice meters are available in line sizes of 2
through 48 inches. The orifice meter discharge coefficient is in the range of 0.59
to 0.62 and is defined by a number of empirical equations, and the discharge coef
ficient is a function of beta ratio, line size, and the Reynolds number.
The hole in the plate is similar to that of the concentric plate but is eccentric with
the pipe axis. For liquids with suspended solids or wet gas flows, the hole is
eccentric to the bottom of the pipe, while for liquid flows with gas, the hole is
eccentric to the top. Pressure taps on the pipe wall are upstream and downstream
of the plate and are located diametrically opposite to the eccentric bore or at the
center of the pipe, 90° to the axis of eccentricity. The plate is so installed that the
edge of the hole is aligned with the pipe wall.
This orifice plate is similar to the eccentric orifice with the shape of the hole
being that of a segment of a circle. The circular shape of the plate bore matches
the pipe diameter and aligns with the pipe wall when the plate is installed in the
pipe. The application is similar to the eccentric orifice but the pressure taps are at
the diametrically opposite walls of the segmental bore.
The hole in the plate is concentric but the inlet of the hole is rounded. It is also
referred to as a quarter-circle plate as the rounded section of the bore entrance is
one quarter of a circle. Plates are normally thicker than regular orifice plates to
allow for the inlet curvature of the hole. These plates are used to measure viscous
liquids, especially over the low Reynolds number range in the laminar and tran
sitional flow regime and low Reynolds number turbulent flows.
Used in similar application as the quadrant edge orifice, but the inlet to the hole
is beveled, instead of being rounded.
15.7.6 Nozzles
A nozzle shaped fixture is either installed between the flanges or welded to the
pipe. There are variations in the nozzle design but all nozzles have a contoured
profile that constricts the flow gradually to a straight section of the throat, which
F L O W METERING 15.17
abruptly ends and so the flow has to undergo a sudden expansion downstream of
the throat of the nozzle. The nozzle design depends on the beta ratio range vary
between 0.2 and 0.8. The high-pressure tap is on the pipe wall upstream of the
nozzle while the downstream low-pressure tap is on the pipe wall or at the
straight section of the throat. The operating range of nozzles is limited between
certain Reynolds number ranges.
The upstream section is similar to that of a nozzle design but instead of abruptly
terminating the nozzle after the straight section of the throat, a gradual diverging
cone is attached to the end of the throat. This improves the pressure recovery
downstream of the throat, which results in a lower permanent pressure loss than
that of a nozzle of same size and geometry.
This device is named after the scientist who used it to measure the flow rate. The
device is a simple tube that is bent through 90° and installed into the How in a
manner that the hole of the tube directly faces the How. As the How stagnates at
the hole, the pressure increases to what is called, the total pressure or stagnation
pressure. The pressure difference between the stagnation pressure at the Pitot
tube and the static pressure monitored at the wall, normal to the flow, is the dif
ferential pressure for flow rate Equation 15.5.
The working principle is same as the Pitot tube, but the tube is double walled
where the inner tube monitors the stagnation pressure while the outer tube, form
ing the annular space, has holes on the wall that are perpendicular to the flow
velocity and thereby, monitor the static pressure. The pressure difference between
the inner lube and the outer tube relates to the flow rate.
A tube with multiple holes in the front and in the back is so designed that the
holes in the front are isolated from the holes in the back. When the tube is
installed in the flow stream, the front holes stagnate the flow and the mechani
cally averaged pressure of all the front holes is monitored at the high-pressure tap
while the averaged downstream-hole pressure is monitored at the low-pressure
tap. The differential pressure between the two sets of taps is used in Equation
15.5 to calculate the flow rate. There are many variations in the design of the
15.18 FLUID F L O W H A N D B O O K
inserted flow element and the location and the number of holes in the front or
back. For each design, the manufacturer must provide the values of the discharge
coefficient (Cd) and the “Meter Constant” for the specific meter. These values
may vary with the meter size and flow application. The multiport-averaging
Pitots are often referred to as Annubar type meters, as the first commercial pro-
prietary-design of the Multiport Pitot flowmeter was marketed as “Annubar.”
15.7.11 Venturi
The operating principle of an elbow meter is based on the physical law of cen
trifugal force. A pressure gradient is developed from the inner radius of the elbow
to the outer wall as flow follows the curvature of the elbow. The pressure taps are
installed on the inner and outer walls of the elbow at the diametrical axis. The dif
ferential pressure between the taps relate to the flow velocity. The accuracy and
applicable range, of these types of meters, are limited and mostly used for flow
control where precision is not essential or critical.
This flowmeter consists o f a V-shaped cone element placed at the center o f the
pipe, leaving an annular space for the passage o f the fluid. Downstream o f the
base o f the V-cone tapers in the shape o f another cone but the height o f the down
stream cone much shorter than that o f the upstream cone. The beta ratio o f this
device is calculated from an imaginary bore diameter that equals the smallest
annular open area of the meter. The slope o f the upstream cone is more gradual
than the downstream cone. The effect o f the V-cone on the flow is like that o f a
venturi while the flow separation is well defined at the sharp edge o f the V-cone
at its largest diameter. The high-pressure tap. on the pipe wall, is upstream o f the
V-cone. The low pressure o f the V-cone meter is sensed at a hole located at the
apex o f downstream cone. The low-pressure tap on the pipe wall o f the meter is
connected to the low pressure opening o f the downstream cone through the
mountings o f the V-cone element in the pipe. The discharge coefficient o f the V -
cone is in the range o f 0.85. The design o f V-cone meter makes the meter suit
able for liquid flows w ith solids and gases and for wet gas flows.
In an effort to increase the flow range o f the orifice flowmeter, a few proprietary
designs use a fixed orifice with a spring loaded cone that repositions in response
to the differential pressure across the cone. These meters are rarely used in the
U.S.. As the spring stiffness can d rift with use, the measurement precision in the
long-term use o f these meters is o f a concern.
The inlet o f flu id to a pipe with a contoured inlet is from a very large supply
chamber. For airflows, the supply is often from a large chamber w ith a bell
shaped entry to the pipe. The contoured entry minimizes the loss at the entrance
to the pipe. As the ratio o f cross-sectional area o f the pipe to the area o f the sup
ply chamber is very small, the flow velocity in the supply chamber is negligible
and the contribution to the total inlet energy by the inlet velocity (V ,) in the sup
ply chamber in Equation 15.4 is negligible. Therefore, the flow rate in the pipe
can be inferred from the differential pressure between the static pressures o f the
Chamber and in the pipe. This metering technique is often used in the aerospace
and automotive industries to measure airflows.
Variable area meters are a special type o f differential pressure flowmeter. In most
commonly used differential pressure flowmeters, the area o f the opening is fixed
and the flow rate is measured as a function o f differential pressure across the
opening. In a variable area flowmeter, the differential pressure across the open
ing is constant and the flow rate is calculated as a function o f the area o f the open
ing. The “ float” or the obstruction is free to move w ith the changes in flow rate
to produce the variable open area and its position relates to the flow rate through
the meter. The variable area meter can measure both gas and liquid flows. There
are three basic variations o f design for these meters:
1. Rotameter: The float material is more dense than the fluid that flows
through an upright conical tube, whose smaller area is at the bottom. As
fluid flows through the tube, the float that is contained by the conical tube
is free to move vertically to the position o f equilibrium where upward
forces on the float due to the fluid viscosity, differential pressure across the
float, and buoyancy balance the gravitational force on the float. The verti
cal float location defines the flow rate.
2. Orifice and Tapered Plug Meter: This device has a fixed orifice in a vertical
cylinder and the float is tapered with the smaller end at the bottom. The
working principle is identical to the rotameter with a difference that the taper
is on the float and the position o f the float defines the variable open area.
open to the flow is proportional to the vertical position o f the piston. Flow
through the open ports exits into an outer chamber that discharges flow
into the downstream pipe o f the meter. The position o f the piston defines
the flow rate.
The basic definition o f a positive displacement (PD) meter is stated in Section 15.3.
PD meters have three basic components; the external housing, measuring unit and
the counter drive train but the PD meter designs vary widely. The liquid PD meter
designs are generally different from the designs o f gas PD meters. Almost all PD
meters absorb a very small amount o f energy to overcome the internal frictions to
drive the flowmeter and the measurement display. PD meters are unique in that a
PD meter with a mechanical display for flow rate and totalized flow through the
meter can be installed in remote locations in the absence o f power.
The housing is the pressure vessel that contains the product being measured.
It can be single- or double-cased construction. Single-case units have the housing
and measuring chamber wall as the one integral unit. For the double-case meter,
the external housing is separate from the measuring unit and is the pressure con
taining vessel. Therefore, it absorbs all the piping stresses without affecting the
precision o f the measuring unit. As the pressure on either side o f measuring units
is basically subjected to same line pressure the measuring unit can be thin walled.
The counter mechanism receives the internal motion o f the measuring unit to
convert it to a useable output signal. Most PD Meters use a mechanical gear train
that is driven by a shaft, which requires rotary shaft seals or a packing gland to
prevent the flow ing flu id from leaking. The pick-up mechanisms vary widely and
are electromagnetic, magnetic, mechanical, electromechanical, photoelectric, or a
combination o f these.
The measuring unit is the most important part o f the PD flowm eter and is
made up o f a displacement mechanism and a measuring chamber. The different
designs o f measuring units in use to measure liquid flows are: the piston, the slid-
15.22 FLU ID FLOW HANDBOOK
ing vane, the oval gear, the nutating disk, the tri-rotor, and the bi-rotor. A
schematic diagram o f different measuring units and their working mechanism is
shown in Figure 15.3.
The liquid PD meters are available in sizes ranging from a fraction o f an inch
to 16 inch lines. Most meters operate over a large flow rate range. For liquid PD
flowmeters, the capillary action o f the metered products forms the liquid seal
between the moving and stationary parts. Meters are designed to very close tol
erances, so flow ing fluid must be very clean and most PD meters require filters
i f particles are expected in the flow . Product slippage is the most crucial problem
affecting the accuracy of liquid PD meters, especially over low flow rate ranges.
The viscous properties o f the flow ing fluid seals the clearances through capillary
action, so increasing viscosity extends the low flow rate lim its significantly but
also increases the pressure drop resulting in wear on the parts and more leakage
through the capillary seals.
The working mechanism o f a PD meter for gas flows is sim ilar to that o f liq
uid PD meters, but the sealing requirements and applicable pressure ranges are
different. Meter designs vary significantly. Several different working mecha
nisms o f gas PD meters are available, but the two most commonly used meters
are the Roots type and diaphragm type with variations in design. Most gas PD
meters are lim ited to low-pressure applications, while the double-case Roots
meter design is for high line-pressure application. Generally PD meters are larger
than other meters for similar flow rate application because the measuring units
are large. This insures an adequate sealing surface, to prevent leakage across the
F IG U R E 15.3 Different designs o f positive displacem ent flowmeters for liquid flows.
FLOW METERING 15.23
unit, and achieves lower velocity through the measuring unit to lim it the inertia
of the moving parts. Single home gas meters are predominantly the diaphragm
meters that operate with approximately 7 inches o f water column o f pressure d if
ferential. A diaphragm meter can measure from the very low How volume to a
gas pilot to the maximum flow volume that occurs when all the gas appliances in
the house are operating.
Most PD meters do not respond well to sudden flow rate changes due to the
large inertia o f moving parts. Large step changes in flow rate can damage the
moving mechanism and cease the meter; some designs o f PD meters can actually
prevent further flow through the meter. If flow interruption is unacceptable for
system operation, the selection o f such a PD meter design should be avoided.
Most Liquid PD meters perform with a linearity o f +/-0.25 percent over the
operating flow rate range and at any flow rate, the meter repeatability is o f the
order o f +/-0.02 percent or better. Gas PD meter linearity is usually +/-2 percent
or better, and some designs consistently achieve better than +/-1 percent.
Repeatability o f Gas PD meters is +/-().5 percent or better. At a given flow rate,
the deviation o f differential pressure across the measuring unit from differential
pressure reading at the time o f initial installation, or after repair, is a good indi
cator o f wear on the meter parts and bearings, especially for gas meters.
displacement at the detectors are theoretically identical, but with flow present, the
C oriolis forces act on the oscillating tube(s), causing a small displacement, de
flection, or twist which can be observed as a phase difference between the sens
ing points. The Coriolis forces that distort the tube(s) exist only when both axial
flow and the forced oscillation are present. The presence o f forced oscillation
w ith no flow or flow in absence o f tube oscillation w ill not generate Coriolis
force, so the meter output w ill display no flow condition. From the frequency and
the phase difference between the two tubes, the meter can output the mass flow
rate and the density o f the flow ing fluid at the operating condition. Therefore, the
meter output in volume units is available too.
Direct mass Coriolis force meters are available in a fraction o f an inch to 6
inch size, and many different material options are available for wetted parts.
Meter sizes larger than 6 inch become prohibitively large and heavy. A t the max
imum allowable flow rate through the meter, the velocity in the measuring sec
tion o f the tube can be significantly higher than flow velocities in other types o f
meters. Also, the pressure drop across the bent tube design o f the Coriolis force
meter at maximum flow rate can be significantly higher. As the operating range
o f the meter is in mass units, the flow velocity through the meter can be sign ifi
cantly higher fo r low-density fluids. There are many variations o f mass meter
designs and the rangeability and linearity claims by different vendors may vary
too. Typical rangeability o f a mass meter is 25:1 or better, with a linearity o f +/-
0.25% or better for liquid flows. For flow rates o f less than 4% (25:1) o f fu ll
scale, meter performance generally deteriorates. The physical law o f Coriolis
force is applicable to any mass flow . Therefore, the meter can measure liquid,
gas, slurries, and immiscible liquid flows. Free gases in liquid flows may absorb
FLOW METFRING 15.25
significant energy o f the vibrating-tubes and thereby affect the meter output. Like
other gas meters, the direct mass meter linearity for gas flows has higher error
bounds than that o f the liquid meters. For gas flows to match the mass flow rate
range o f the meter, the flow velocity must be high for low density gas or the gas
density must be high for low flow velocities. For natural gas application, the mass
flow rate range can be achieved at relatively high pressures and/or at significantly
high flow velocities. Gas velocities through the oscillating tubes can be sign ifi
cantly high for the direct mass meters.
The line vibrations from mechanical equipment or flow induced vibrations can
affect the direct mass meter performance. When extraneous disturbance fre
quency, its harmonics, or sub-harmonics are close to the tube excitation frequency,
the meter performance may be affected. In addition, meter performance may be
influenced by changes in line pressures, temperatures, and fluid viscosities. The
meter output can be compensated for the changes in operating conditions by mon
itoring the pressure and temperature changes at the oscillating tube. Most mass
meter designs are inherently self-draining and do not have protrusions or crevices
in the flow path and are suitable for sanitary applications.
Most commercially available energy meters are inferential meters. These meters
measure the flow rate in mass or volume at the operating condition. Mass being
a primary unit, flow rate measurement in mass does not require temperature or
pressure compensation for the base or reference condition; by knowing the heat
15.26 FLA'ID FLOW HANDBOOK
or energy content per unit mass, the energy flow rate is determined. When the
measured flow rate is by volume, the heat content may be determined in several
different ways. I f the energy per unit volume o f the fluid at the operating condi
tion is known, the energy flow rate is the product o f the volume flow rate and the
energy content per unit volume. Energy content may be known from prior tests
o f representative samples drawn from the line or from online live heat content
measurement by direct measurement o f heat content or by an online gas compo
sition analyzer. I f the energy content per unit mass is known, then from the den
sity, at the operating conditions (live or from prior test), the flow rate volume is
converted to mass and the energy flow rate is calculated from the mass flow rate
and the energy content per unit mass.
For gas flows, the energy value is determined by several methods. Basically,
the energy content is determined for unit mass or volume at the operating or base
condition by knowing the gas composition or by directly measuring the heat con
tent o f the flow ing fluid. Gas compositions can be determined by either analyz
ing a grab sample, a composite sample, or by an online gas chromatograph.
Due to the increase in natural gas prices and also that actual gas volume or
mass flow rate may not be the true indicator o f total heat content, reporting flow
rate in units o f energy is becoming the preferred mode o f operation, especially for
fiscal measurements. For example, an increase in nitrogen or carbon dioxide con
tent in natural gas flows w ill indicate a higher measure in mass flow rate. I f the
flow is assumed to be all natural gas, without correction to true heat content, the
fiscal measurement in the energy units w ill be in error. Since most meters are
inferential type, the energy measurement system is a combination o f a flowmeter
and an energy determination system, which can be an online calorimeter, a gas
chromatograph, a sampling system, or from lab analysis o f heat value and/or gas
composition o f a representative sample o f the flow.
Meters listed under this category are not in a class o f any specific group o f meters
listed above. Measurement o f flow is not always in closed conduits and there are
meters for open flows. There are meters that utilize physical laws o f conservation
o f energy and mass, yet the meter output is not by monitoring the differential
pressure. Some o f those varied groups o f meters are listed below:
Unlike a closed conduit running full, the open channel meters have flows with
free surface. There are many methods o f determining the rate of flow in open
channels and some o f the more common include the timed gravity, dilution,
velocity area, hydraulic structures, and slope-hydraulic radius-area methods. The
timed gravimetric and dilution methods are generally not used for routine flow
FLOW MFTHRING 15.27
measurement as they (Jo not provide continuous flow rate reading. These meth
ods are normally used to either verify or calibrate the performance o f other
devices that provide continuous How rate readings.
D ilu tio n M e th o d . The flow rate is determined from the rate o f dilution o f a tracer
solution added to the flow. Commonly used tracers are radioactive and fluor
escent dyes, while brine solutions were used in the past. The two main tracer
dilution methods are slug injection and constant rate injection. In the slug
injection method, a known amount o f tracer is injected into the flow , and the
dilution o f the tracer is accounted for by the complete measurement o f its mass at
a downstream location, for which this method is sometimes referred to as total
recovery method.
In the constant rate injection method as the name implies, the tracer is injected
at a constant rate o f flow . A t a downstream location, the flow reaches a constant
level o f concentration after the equilibrium is achieved, and the flow rate is a
function o f the concentration level.
Whenever a flow goes past an obstruction, a force develops that exerts a push or
drag on the obstruction in the direction o f flow. When the obstacle is constrained
by a support, the restraining force on the support must be equal to and opposite
in direction o f the drag force. The magnitude o f the force relates to the flow rate
and fluid properties. Tw o contributing forces are the friction or the viscous drag
force, that relates to the fluid viscosity and the flow rate, and the pressure drag, a
force resulting from pressures immediate upstream and downstream o f the object.
That pressure drag is a function o f fluid density, flow velocity, and the area and
shape o f the flow blockage. The most commonly used target flowmeter flow
blockage is a circular disk placed at the center of a pipe.
The accuracy and rangeability varies significantly for target flowmeters and
the typical accuracy is o f the order o f +/-0.5 percent to +/-2.0 percent. A target
meter application is generally restricted to a flow control device or as a flow sen
sor in process control.
By definition, insertion flowmeters infer flow rate over the entire cross-section o f
a flow plane by measuring flow rate at a point or over a part o f the total flow cross
section. In practice, an insertion flow meter is defined as a meter that measures
the flow rate at a single point and the measuring element is inserted into the flow
cross section. By definition, Pitot tube and Pitot-static tube are insertion flow m e
ters. Three main components o f an insertion flow meter are the flow sensing ele
ment or primary device, the probe assembly for inserting the primary element to
a desired location on flow cross section, and finally, the secondary processing
unit that provides the output.
Insertion flowmeters are available in most traditional flow measuring tech
nologies like turbine, vortex, magnetic, differential pressure type, thermal sensor,
etc. The range o f flow ing conditions and fluid properties for the Insertion meters
are similar to that o f the inline or fu ll-flo w meters. The rangeability o f insertion
flowmeters is o f the same order as the inline meters, but as the total flo w is
inferred from a point measurement, the overall accuracy o f the insertion flow m e
ter is normally not as reliable as its inline counterpart.
Insertion flowmeter applications are generally restricted to flow control and
check metering, where the primary meter is the custody transfer meter. Most
insertion flowmeters can be removed from the line without having to interrupt or
FLOW METERING 15.29
stop the flow. For high pressure application, the removal and insertion o f the
meter normally requires a special fixture design. Some devices may have a lim it
to the maximum allowable line pressure during insertion or withdrawal o f the
meter and that safe meter removal pressure lim it can be less than the operating
pressure or the designed pressure rating for the meter.
This class o f meters infers flow rate o f compressible fluids by creating a differ
ential pressure across the inlet and outlet o f the flow element, but the principle of
operation is very different from any head-type device. W ith no differential pres
sure between upstream and downstream o f a flow restriction, in a compressible
fluid media, there is no flow across the restriction. When upstream pressure is
held constant and the downstream pressure is gradually reduced, the fluid starts
to flow through the restriction. W ith decreasing downstream pressure, flow rate
across the restriction increases. I f the restriction is designed in a certain nozzle
shape, there is a critical pressure ratio between the upstream and downstream
pressures when further reduction in downstream pressure does not increase the
flow rate. The flow is said to be choked, a condition which exist when the flow
rate at the throat o f the nozzle reaches the speed o f sound. Once the flow veloc
ity at the throat o f the nozzle reaches the sonic velocity, further reduction o f the
downstream pressure does not increase the flow rate.
There are specified designs for critical-flow nozzles and with a well-designed
nozzle, the reduction o f downstream pressure to a value o f about 93 percent of
the upstream pressure can cause the flow to reach the sonic velocity at the noz
zle. The critical pressure ratio is the ratio o f the downstream to upstream pressure
at the nozzle. For the above example, the critical pressure ratio would be 0.93. By
maintaining a downstream pressure o f less than the pressure for the critical pres
sure ratio, it can be ensured that flow is chocked at the throat. For a given
upstream condition, a critical flow nozzle achieves only one flow rate. To achieve
a variable flow rate, either the upstream conditions must be changed or a set o f
different size nozzles, which can be opened or closed in any combination, is to be
installed in the line. A device with multiple sets o f critical flow nozzles is com
mercially available, but the main concern is the ability o f the device to insure that
the flow has no leak-path through the nozzles that are supposed to be closed. The
flow rate o f a critical flow nozzle for a choked condition is precisely defined for
the upstream temperature, pressure, fluid properties, and for the nozzle geometry.
Normally, critical flow nozzles are used in compressible flows as primary ele
ments to calibrate other meters.
Many industries require a “ sanitary condition” for their equipment. Here the word
“ sanitary” does not refer to waste treatment applications; rather it refers to a
15.30 FLUID FLOW HANDBOOK
highly “ clean and hygienic” condition. The design requirements and specifica
tions for sanitary flowmeters originated in the dairy industry, where the handling
and packaging o f perishable products like m ilk required equipment that did not
contribute to the bacterial growth or decay o f product. The most stringent sani
tary requirements are in the pharmaceutical and biotechnology industries. Flow
meters used in industries requiring sanitary designs can be o f any o f the meters
described in the different meter groups, but the meter must meet the requirements
o f the sanitary application. Sanitary flowmeters must be designed to ensure that
the flow ing product or residue does not get trapped or be caught and left to spoil.
The sanitary meters must be designed to eliminate cracks, crevices, or dead ends
where product or residue could collect.
Multiphase flow metering is one o f the most challenging areas o f flow measure
ment. W ith increasing demand o f energy sources, exploration in the deep-sea,
and exorbitant cost o f space on offshore platforms, online measurement tech
niques o f different phase fractions, with any reliability, are o f interest to the
hydrocarbon producers. Complexities in measurement are contributed by the
presence o f immiscible liquids and/or gas in the flows. Accuracy o f online mea
surements o f liquid phases (water-cut) and GOR (Gas O il Ratio) are the two most
important elements o f measurement by a MPM. MPMs can be grouped in two
broad categories; the standard MPM that measures all phases (Water-cut and
GOR) and others that measure wet gases, for gas flows containing hydrocarbon
condensates. Many MPMs were originally developed for a specific field with
known water-cut and GOR ranges, and the meter performance was optimized for
the operating range. As phase fractions o f the field changed with the life o f the
field, meters were adjusted to optimize the performance characteristics and
achieved a wider range o f applicability. So, some M PM designs have a “ sweet
spot” for operating condition and the meter performance may not be as accurate
as when the meter is operating outside the sweet spot.
The text presented in this section is based on the state o f the art o f MPM tech
nology at the time o f w riting this chapter. Efforts in developing MPM are very
high and measurement techniques are likely to improve in the near future.
Wet gas is an interesting application; to date, most wet-gas meters monitor
flow rates by wet-gas Venturis, where liquid loading must be known to determine
the gas rate by the Murdoch method. By monitoring two sets o f differential pres
sures at different locations on the meter, it is possible to obtain the liquid loading
and determine the gas rate. Success with the venturi design led to the evaluation
o f other differential producers with gradual change to flow cross-section for wet-
gas measurement, like the V-cone flowmeter. For standard M PM , all commonly
used techniques seem to fall into two broad categories; those using gamma ray
absorption for composition and the others using dielectric properties and density
FLOW METERING 15.31
Flowmeters must be properly installed and maintained to measure flow rate cor
rectly. The effect o f incorrect installation and/or improper maintenance varies
from reduced measurement accuracy to the meter failing to operate at all. Nor
mally, a flowmeter performance is best when the flow is steady with no external
disturbance or internal distortions to affect the flow profile at the metering sec
tion. In the real world, such a disturbance-free flow is not practical. In order to
minimize the effect o f flo w disturbances, certain installation requirements are
specified by the manufacturers or relevant standards for the type of metering
device. Most flowmeters require:
Any valve or pipe fitting that abruptly changes the flow cross section or gener
ates secondary flows must be installed a sufficient distance upstream of the meter,
so that, at the measuring section o f the meter, any profile distortion and/or flow
swirl generated by the fitting are insignificant to affect the meter performance.
Ideally the flow regime is defined by the Reynolds number and the inside pipe
wall roughness. Any profile distortion caused by the piping configuration is even
tually restored to the fu lly developed or a near ideal profile by the mixing action
o f the fluid as it moves downstream. The restoration process is often augmented
by the installation o f a flow conditioner in the pipe. The use o f a flow conditioner
in most flowmeter installation can significantly reduce the straight length require
ment o f upstream pipe without a flo w conditioner, often referred to as a bare tube.
The use o f a flo w conditioner must be applied with discretion. Because each flow
conditioner generates its own signature on flow profile, the installation o f a flow
conditioner may affect the measurement and result in measurement bias. Any
flowmeter and flow conditioner combination must be flow tested to determine the
optimum location o f the conditioner with respect to the meter and the type o f fit
ting upstream o f the conditioner. Some generic tests protocols are defined for cer
tain flowmeters but not for all types o f flowmeter. The other important items for
proper meter performance are:
For the ease o f maintenance, clearance and accessibility must also be consid
ered. Most flowmeters lose accuracy as the meter gets dirty from the process fluid,
so they require routine cleaning. Performance verification or calibration o f meter
should be performed for new installations and after the meter is cleaned or over
hauled, prior to being installed in the line and used for measurement. Meters with
moving parts should be calibrated periodically or inspected for possible wear or
damage that can influence meter performance. For fiscal measurement, where
measurement accuracy is o f the essence, dedicated provers (online meter calibra
tion devices) are often installed to calibrate the meter when changes in fluid prop
erties and/or operating conditions are known to affect the output o f the meter.
Why measure flows accurately? This is the question that must always be ans
wered to justify the selection and installation o f a flowmeter. The flow is mea
sured accurately because o f the requirements o f a competitive business
environment and the desire to lim it waste. This requirement is also imposed by
the accountability for products, economic viability, and environmental regula
tions, even when fiscal measurement is not warranted.
15.16 ACKNOWLEDGMENTS
It is impossible for me to acknowledge all the people who helped me learn about
flow measurement and flow meters. Many people who I have associated with,
have supported me with my work, and helped me in many different ways to gain
experience. A ll have made an impact. In addition to my own efforts and experi
ence; over the years, my association with colleagues at different standards com
mittees has contributed the most to my understanding and knowledge. For the
development o f this chapter, I must acknowledge and thank Dr. Windsor Letton
for his enlightening discussion on multi-phase meters, Mr. John Roussel o f Tex
aco, M r. Dale Goodson o f Daniel Industries, and my dear daughter Zeena Husain,
for their time and effort in reviewing this chapter. More importantly, I am
indebted and thankful to my wife and children for their understanding, patience,
support, and sacrifice in giving me the opportunity to complete this chapter while
some other pressing tasks could not be accomplished.
CHAPTER 16
FLOW CONTROL
Flow control loops are the most common control loop used in the chemical process
industry (CPI). In addition to the simple flow loops, there are a large number of con
trol loops whose manipulated variable is a set point to a flow rate control loop in a
cascade control arrangement. A schematic o f a flow control loop is shown in Fig
ure 16.1. The pressure drop across the orifice meter is measured by a differential
pressure (DP) cell, Figure 16.2. The pressure drop is correlated to the volume flow
o f the liquid. An I/P converter w ill convert the 4 -2 0 mA current signal from the
DP cell to a 3-15 psig instrument air pressure to drive the control valve. Other types
o f flow meters include rotameters, turbine meters, vortex shedding flow meters,
and magnetic flow meters. The objective o f a flow control loop is to maintain the
flow at the set point in the face o f upstream and downstream pressure changes (dis
turbance rejection) and to follow the set point changes (set point tracking). The
dynamics o f the flow process and the sensor are quite fast compared with the dynam
ics o f the control valve. A flow control loop consists o f a flow sensor, a flow control
ler, and a final control element (actuator and valve). The elements w ill be discussed
in later sections.
16.1
16.2 F L l'ID FLOW HANDBOOK
Control engineers make extensive use o f drawings with standard symbols devel
oped by the Instrument Society o f America (ISA) to represent equipment, transmit
ter, controller, and control valves w ith appropriate arrows to indicate material and
signal flows in a control loop [2]. A ll equipment (piping, vessels, valves, etc) is
drawn in solid lines. Sensors are designated by a circle connected to the point in the
process where they are located [3]. The abbreviations usually consist o f two or three
letters. The first letter in the instrumentation symbol indicates the type o f variable
measured or controlled. Some o f the more common designations are [3, 4]:
A or C Composition analyzer
D Density
F Flow
L Level
P Pressure
PH PH
S Speed
T Temperature
The second and the third letter indicate the type o f function performed, e.g., “ A ”
for alarm, “ C” for a controller, “ H” for high lim it, “ I” for an indicator, “ S” for switch,
“ T ” for transmitter, and “ Y ” for any other calculation. For example, “ FC” is a flow
controller, “ PIC” means pressure indicator and controller, “ TS” uses a temperature
measurement to open/close a switch. A dashed line (or a solid line triply cross-
hatched interm ittently) represents an electrical signal (4 to 20 mA, 0 to 5 V, or
-5 to +5 V); a solid line doubly cross-hatched intermittently is a pneumatic signal
(3 to 15 psig); and a broken line inserted with 0 ’ s is a digital signal (0 to (2n ~ ‘ ),
n is the number o f bits), Figure 16.3. A regular solid line indicates a material flow.
E lectric
or
/ / / / / / / / /
/ / / / / / / / /
// // / /
// // // Pneum atic
FLOW SENSOR
The sensor system is composed o f the sensor, the transmitter, and the sampling sys
tem. The most commonly used flow sensor is an orifice meter, which measures the
pressure drop across a restricted flow area and derives the flow rate according to
the well-known relation [1,6]:
(16.1)
F «■ (A P ) U2 (16.2)
which means the flow rate is proportional to the square root o f the measured pres
sure drop. A straight run o f pipe preceding the orifice meter is required to avoid
an excessive measurement error. A typical precision is ±0.3 to ±1% [1, 7].
Other types o f flow meters include rotameter, vortex shedding flow meters
and magnetic flow meters [7, 8]:
The vortex shedding and magnetic flow meters are more costly to install but
cheaper to operate because there are no pressure tap maintenance problems. Flow
measurements are typically installed upstream o f the control valve to provide a
low-noise measurement. The calibration o f a differential pressure sensor is quite
sensitive to the condition o f the sensing lines (e.g., condensate buildup in a sup
posedly dry line) [1 ,4 ].
FLOW CONTROL 16.5
• Square root compensation not applied properly e.g., applied for non-differen
tial pressure type indicator, not applied for differential pressure type indica
tor, or applied twice (once in transmitter and once in DCS)
• O rifice plate installed backward or damaged
• Plugged line to differential pressure sensor
• Flashing o f liquids as they flow through an orifice meter.
The actuator system (final control element) consists o f the I/P converter, the instru
ment air system, the valve actuator, and the control valve. The valve actuator pro
vides the force necessary to overcome pressure forces, flow forces, friction from
the valve packing, and friction from the guide surfaces to move the valve stem posi
tion. When a change in the control signal is made to the I/P converter, the instru
ment air pressure to the valve changes. This causes the diaphragm in the control
valve to expand or contract which in turn moves the valve stem position to increase
or decrease flow. The UP converter is an electromechanical transmitter which con
verts 4 -2 0 m A signal from the controller to a 3 - 15 psig instrument air pressure
to the valve actuator. More details on the most common final control elementothe
throttling control valve are given below.
Most throttling control valves consist o f a plug on the end o f a stem that opens
or closes the orifice opening as the stem is raised or lowered. As sketched in Fig
ure 16.4, the stem is attached to a diaphragm that is driven by the air pressure
above (or below) the diaphragm. The force o f the air pressure is opposed by a spring.
The dynamics o f the valve stem position to changes in the instrument air pressure
can be very slow. The dynamics o f the control valve can be represented by a lin
ear first order lag f 1],
d F _ (F spcc- F )
(16.3)
dt Tv
where
--------------------- In le t fo r In s tr u m e n t A ir
r
The dynamics o f a flow control loop or a control valve with a positioner can be
described by the same equation with a time constant r v = 0.5 -2 seconds.
Valve Characteristics
By changing the shape o f the valve plug and valve seat, a control valve can have
different inherent valve characteristics, i.e., different flow area versus valve stem
position relationships. The flow rate for non-flashing liquids for a given pressure
drop across the control valve is [ l , 3, 6, 8]:
(16.1)
f ( x ) = jc Linear
= R x ~ 1 Equal Percentage (= % )
where
Equal percentage valves are used in almost 90% o f the control valve applica
tions while linear valves are used in the remaining cases where the pressure drop
across the control valve is relatively constant. A third kind o f valve plug, even
though seldom used, is “ quick opening” where flo w increases rapidly at the initial
stages o f the stem travel, Figure 16.5 [7, 8]. An example o f a constant pressure
drop across a control valve is to control steam flow from a constant-pressure supply
header through the shell side o f a heat exchanger. The shell side pressure drop is
small compared to that across the control valve. When the overall pressure across
the process equipment (e.g., heat exchanger) plus the control valve is constant as
in most applications, the = % valves actually exhibit a more linear behavior in addi
tion to a wider rangeability [6]. This near-linearity is the so-called “ installed char
acteristics” for an =% valve. The wide rangeability or a large turndown can also
be seen from the “ inherent characteristics” ,/ ( * ) , o f an equal percentage valve. If
the theoretical rangeability R = 50, for every valve travel Ax = 0 .1 7 7 2 ,/U ) w ill
double. For example, x = 0.5316, f ( x ) = 0.16; x = 0 .7 0 8 8 ,/(*) = 0.32, etc. The
aforementioned rangeability (R ) and turndown ( T ) are important valve properties
in addition to the valve characteristics [5, 7, 8]:
% of Baled Flow
0 20 40 60 80 100
% of Rated Travel
F IG U R E 16.5 Three types o f inherent valve characteristics.
FLOW CONTROL 16.9
Usually valves are sized so that T = 0.7/?, where R lies between 20-50.
Figure 16.6 shows an example o f sizing the centrifugal pump and control valve
in a classical engineering trade-off. The flow rate o f at the design condition is 50 gpm,
the minimum and maximum flow rates are 25 and 150 gpm, respectively, the pres
sure in the feed tank is atmospheric, the pressure drop through the heat exchanger is
10 psi at 50 gpm. The f ( x ) ranges from 0.1 to 1 from pinch down to wide-open for
the minimum and maximum flow rates. The liquid specific gravity is 1. The pres
sure drops o f the control valve, the heat exchanger, and the system are A Pv , A PH,
and AP T. We have:
1/2
F V
1/2
150 = Cv (1.0)| A P r - 1o f
' \ ^ 1/2
25 = Cv (0 .1 )|A /V ~ 101 25
.50,
The solution is C v = 21.3 and A P r = 139.2 psi. I f the water pressure at the exit
is 2 psig, the pump head w ill be A PP = 139.2 + 2 = 141.2 psi.
Heat
Exchanger
Pump
F IG U R E 16.6 A pump, heat exchanger, and control valve assembly.
16.10 FLUID FLOW HANDBOOK
More detailed equations in control valve sizing are available from manufac
turers [9] that handles flows o f gases, flashing liquids, and critical flows.
Actuators are also designed to be fail-open or fail-closed based on the safety con
siderations when the instrument air pressure is lost or unit is shut down. For exam
ple, a cooling water control valve prefers to operate in a fail-open mode while a
steam, fuel, or chemical reagent feed valve needs to operate in a fail-closed mode.
Here, fail-open means air-to-close (AC) and fail-closed indicates air-to-open (AO).
The pressure acts on the diaphragm/spring system from the top to close the valve
as the air pressure increases in the air-to-close actuator while the air enters below
the diaphragm (spring at the top in this case) to open the valve for the air-to-open
actuator. The valve can also be made air-to-open by reversing the action o f the plug
to close the opening in the up position and wide-open in the down position f 1,4, 6].
The diaphragm is constructed o f an air impermeable, flexible material on which the
force generated by the instrument air pressure is balanced by the compressed actu
ator spring force. A control valve with an air-to-close valve actuator is also known
as a reverse-acting final control element. By the same token, a control valve with
an air-to-open valve actuator is a direct-acting final control element.
Booster relays are installed to provide extra air flow capacity for large valves
to reduce the dynamic response time. Booster relays use the pneumatic signal as
input to adjust the pressure o f a high-flow-rate instrument air system to the dia
phragm o f the valve actuator.
Note that the pressure at the vena contracta in a throttling control valve is the lo w
est compare to both the upstream and downstream pressures. A phenomenon called
“ cavitation” can take place i f the pressure drop across a control valve is too big. That
is, the pressure at the vena contracta drops below the vapor pressure o f the liquid.
As a result, vaporization w ill occur. Due to the surface tension, the vapor phase w ill
first appear as bubbles. As these bubbles travel downstream, the pressure recovers
to above the liquid vapor pressure and bubbles collapse. The bubble-collapse process
causes noise, vibration, material removal, and reduced flow. The violent, asymmet
rical implosion forms a high-speed micro-jet and induces a pressure wave [8, 10]. I f
implosion occurs next to a solid surface, minute solid pieces can be removed which
w ill lead to a rough surface overtime. Ways to reduce or eliminate the effect o f cavi
tation include selecting harder materials, placing the valve at a higher-pressure loca
tion, or using a lower recovery valve.
I f downstream pressure is lower than the liquid vapor pressure, the vaporization
process w ill continue until all liquid phase is gone. This phenomenon is called “ flash
ing.” The huge volume expansion o f the phase change leads to a high speed stream
FLOW CONTROL 16.11
and impingement o f entrained liquid droplets in the vapor w ill cause material dam
ages. Flashing damages can be mitigated by choosing hard materials and directing
the 2-phase je t away from solid surfaces [7, 8, 9].
• O n/off valves usually have tight shutoff and low pressure drops when they
are wide open. Applications o f on/off valves include feeding catalyst and
reactant to a batch reactor and pressure isolation fo r maintenance/installation
needs. Gate valve is one o f the most commonly used on/off valves. Solenoid
valves with a switch are also used in certain on/off applications 15, 8, l l ] .
• The pressure relieve valve is an automatic valve that opens when the pressure
exceeds a pre-determined lim it and automatically shuts o ff when the normal
pressure restores, Figure 16.7. W ithin this class there are relieve valve, pilot-
operated pressure relieve valves, and safety valves. The pilot pressure
relieved valve has an auxiliary pressure relieve valve (pilot) to improve the
sensitivity in over-pressure conditions. The safety valve is fu lly open when
the pressure exceeds the rated lim it.
• Check valves allow a relatively unimpeded flow in the desired direction and,
in the mean time, prevent flow in the reverse direction. In the forward direc
tion, flow forces overcome the weight o f the closure member or a spring to
open. In the reverse direction, flow forces push the closure member into the
valve seat to close.
• An adjustable speed pump can be used as a final control element in low flow
applications such as catalyst addition or base injection. Adjustable speed pumps
offer energy savings, no need for instrument air, and dynamic performance
advantages over throttling control valves. The disadvantages are high capital
cost, no fail-closed/fail-open feature. Pump speed can be adjusted by using
turbines, motors with magnetic or hydraulic couplings, or electric motors.
• Regulator is a compact device combining sensor, controller, and final control
element into one unit. Regulators are designed to maintain at a specified level
in the presence o f disturbances. As a result, they are not suitable for applica
tions with constant set point changes.
'///A
FLOW CONTROLLER
A proportional plus integral (PI) controller, which can eliminate the offset, is usually
the proper choice for How control. The PI controller w ill also absorb any unmea
sured disturbances through the feedback loop. The derivative action, the D part in
a tunable PID controller, is not recommended in a flow control loop due to the fluc
tuations in a turbulent flow as in most industrial applications. A wide proportional
band (PB = 150) or low gain is used due to the noisy flow measurement or the neces
sity to filter, which introduces a time lag. A low value o f integral or reset time (Tj
= 0 .1 min per repeat) is used to get fast, snappy set point tracking [1,6, 12]. The con
troller action (direct acting or reverse acting) depends on the process requirements
and the actuator selection. Usually an actuator (direct acting or reverse acting) type
is determined first based on safety and process needs, then a proper controller action
can be chosen accordingly.
Deadband is a common phenomenon in flow control valves. That is, the actual
valve stem position does not instantaneously change with the instrument air pressure
change. The air pressure delivered to the valve has to increase or decrease by a
significant amount (> deadband) to realize a move in the valve stem position, Fig
ure 16.8. The “ stickiness” is a result o f friction between the valve stem and valve
Time
F IG U R E 16.8 Valve stem position versus air pressure to the valve actuator for a “sticky” valve.
16.14 FLUID FLOW HANDBOOK
packing and a minimum force is required to move the valve. When it does move,
significant valve travel may be experienced. Excessive friction may be caused by
packing damages, excessive tightening to avoid fugitive emission, and so on. I f a
valve positioner is available, the high frequency P-only feedback control provided
by the positioner w ill check the valve stem and make sure the desired valve posi
tion is attained thus the effects o f the deadband are eliminated. A typical deadband
for industrial control valve is 10 to 25%. W ith a positioner, the actual average flow
can be controlled to within 0.5% or even 0.1% [1, 10].
Commonly encountered flow controller/DCS problems include:
The flow control described above is the so-called “ basic” P1D control. Often, other
advanced flow control schemes are needed to satisfy process requirements as de
scribed in the follow ing sections.
Ratio control involves keeping constant the ratio o f two or more flow rates. The
flow rate o f the “ w ild ” or uncontrolled stream is measured and the flow rate o f the
manipulated stream is changed to maintain a constant or prescribed ratio. Common
examples include [4, 6, 12, 13]:
• Holding a constant reflux ratio between the reflux stream and the distillate
stream on a distillation column
• Keeping a little greater than the stoichiometric ratio between air and fuel to
guarantee complete combustion
• Purging o ff a fixed percentage o f the feed stream to prevent the accumulation
o f inert gases
• Maintaining a composition by using a ratio controller in lieu o f a composition
feedback controller when a suitable composition analyzer is not available
Two ratio schemes are commonly employed: (1) a ratio multiplier scheme, (2) a
ratio divider scheme. Figure 16.9 (a) shows a ratio m ultiplier (ratio station) scheme
where the wild flow is measured and the signal is multiplied by a predetermined or
prescribed ratio (from another controller) [6, 12]. The output o f the m ultiplier is the
FLOW CONTROL 1 6 .1 5
(b)
FIGURE 16.9 Ratio control (a) multiplier scheme (ratio station) (b) divider scheme
(ratio controller).
16.16 FLU ID FLOW HANDBOOK
setpoint o f a remote-set flow controller for the manipulated flow. An example o f the
multiplier ratio control scheme is given as follows. An acid-water-ratio control sys
tem for continuous pickling automatically control the addition o f fresh acid in cor
rect proportion to the water used [13]. The accurate flow control results in a stricter
adherence to the pickling specifications and a minimum consumption o f acid. An in
stallation is schematically shown, Figure 16.10. The ratio factor is set by a ratio relay
or m ultiplying unit. This unit would be located between the w ild flow (water or
Flow A ) transmitter and the flow controller set point. This output becomes the set
point o f the flow controller that regulates the manipulated stream (acid or Flow B)
in a preset ratio. The ratio factor has a factor adjustment range (e.g., from 0.3 to 3)
or can be manipulated by another controller. The ratio relay may be incorporated
in the ratio flow controller, and a ratio dial may replace the normal set point dial.
This m ultiplier ratio control can be viewed as a simple form o f feedforward con
trol to maintain flows at desired proportion [3].
Another scheme is a ratio divider, Figure 16.9 (b). The two flow rates are
measured and their ratio is computed. The ratio signal is then fed to a controller to
compare with a ratio set point and the output is directed to the control valve. The
divider ratio control can be viewed as a feedback control with a calculated con
trolled variable [3],
Care should be taken in selecting a transmitter range in order to allow system
flexibility. Try to choose a range so that the ratio factor is normally in the middle.
However, the transmitter range should also cover all process conditions [13]. Flow
meters used in a ratio control system are often orifice meters or other constriction
meters with differential-pressure transmitters. Thus, the transmitter output is pro
portional to the square o f flow . For example, when the ratio relay provides 0.3:1
to 3.0:1 ratios between differential pressure transmitter outputs, the actual flow
ratios w ill be 0.6: l to l .7 :1. However, transmitter output in magnetic flowmeters,
turbine flowmeters, rotameters, and so on is proportional or linear to the flow.
@ -----■©*--------
FIGURE 16.11 Multi-cascade configuration for the bottom composition control of a distillation
column.
16.18 FLU ID FLOW HANDBOOK
faster than the temperature loop which, in turn, is much faster than the composi
tion loop. The high frequency feedback action o f the slave loop eliminates the
effects o f steam pressure changes before they can affect the master loop [1, 3, 6].
Mass flow control can be achieved by computing using several measurable vari
ables. In Figure 16.12, a mass flow rate can be inferred by utilizing temperature,
pressure, and pressure drop measurements and the calculated mass flow rate is fed
to a feedback PI controller for mass flow control.
When a distillation column encounters an extreme weather change, such as a
thunderstorm, the excessive sub-cooling o f the reflux causes the condensation o f
the upward vapor. Thus, the internal reflux, which determines the composition, is
substantially greater than the external reflux.
Internal reflux control in a distillation column can be achieved by combining
information from external reflux flow and temperature [14]:
* (Ta - Tl )
Lt = L 1+ C (16.4)
AH
Mass
flow
rate
com puter
where
The internal flux rate thus can be inferred and the appropriate set point can be sent
to the external reflux controller.
Flue gas
heater. The measured How rate disturbance is feed-forwarded lo adjust the temper
ature feedback controller output before being sent to a flow controller as the set
point in a cascade manner [3].
Mass flow controllers (MFCs) are used for accurate measurement and control o f gas
flow . MFCs w ill not operate with liquids. In general, MFCs consist o f two main
components, Figure 16.14:
The mass flow controller divides the flow between a heated sensing tube and a
flow bypass. In the heated tube the flow is actually measured and the overall flow
Operation Parameters
The mass flow meter is designed to provide fully developed laminar flow within the
sensor and the bypass. By adding heat to a gas and monitoring the change in tem
perature in the sensing tube, the mass flow rate can be determined:
(16.5)
where
Given Cn and A T , the rate o f heat loss clQ /dt is converted to an electrical signal
that represents the mass flow rate dm/dt. Zero and span adjustments are usually
accessible [15, 16]. Flow rate ranges from
The sensor is a long, thin stainless steel (e.g., type 316) tube, often called a ‘cap
illary tube” . The large ratio o f length to inside diameter ratio ensures laminar gass
flow within the sensor tube. A t the midpoint o f the capillary tube, two wire coils,,
wrapped side by side, serve as both heaters and temperature sensors. Heat is trans
ferred through the thin wall o f the sensor tube to the gas flo w inside. Since the:
resistance o f coils varies with temperature, the coils function as resistance tem
perature detectors (RTDs) to measure the gas temperature. As gas flows, heat iss
picked from the upstream heater and is carried towards the downstream heater..
This movement o f heat shifts the temperature gradient so that a temperature d if
ference develops between these two heaters. This temperature difference can be:
linearly correlated to the mass flow rate, Figure 16.15. Since the coils are the two)
resistive legs in a Wheatstone bridge circuit, the temperature difference betweem
the two coils presents a difference in resistance and is am plified by the bridge cir
cuitry into a 0 -5 VDC output.
The bypass is composed o f a variable number o f capillary tubes. These flow/
splitters, also known as restrictor flow elements, serve to smooth the main strearm
o f gas into laminar flow and split a small stream into the sensor tube. The num
ber o f capillary tubes adjusts the flow splitting ratio to allow a wide range of masss
flow rate measurements.
Tub* Length
FIGURE 16.15 Temperature profile of a thermal sensor, which is symmetric
without flow but shifted downstream as flow increases.
FLOW CONTROL 16.23
Controller
In an MFC, the difference between the output voltage from the How meter and the
set point voltage is used to adjust the drive voltage to the control valve. With a high
gain proportional controller, the valve w ill proportionally open or close to approach
the set point. Computer-tuned PID controllers with a minimum thermal time lag are
also available for MFCs [15, 16. 17]. A pre-charged capacitor may be used in a PID
controller to speed up the integral action. Because the sensor and matched circuit
board are temperature-compensated, the MFC is insensitive to changes in room
temperature. Over a wide range o f flows and operating conditions, a good MFC
provides zero overshoot and stable control regardless o f temperature, pressure, or
vibration. The controller set point can be 0 -5 V d c or 4 -2 0 m A analog signals. A
typical settling time is from 1-30 second. Some fast settling time can be as short
as 4 milliseconds.
Control Valve
The dynamic behavior o f a heat exchanger that involves the flow rate o f a heat
transfer fluid and the transfer o f heat through a heat exchanger metal tubes. The
dynamics are affected by the heating/cooling o f the metal and the transportation
delay o f the heat transfer fluid. The follow ing equation lumps a flow and a heat
transfer process [1]:
(16.6)
where
xH includes l - 6 seconds time constant for changing the temperature o f the heat
transfer tubes and 5-30 seconds time constant for the flow through tubes.
Valve Positioner
Valve positioner is a hardware box mounted on the side o f the valve actuator and
implements the valve position control, Figure 16 .16. The valve positioner is espe
cially needed for large control valves that are often “ sticky” , i.e., have a significant
deadband from packing friction, friction clue to viscous fluid, or process pressure
changes and can not achieve the exact percentage stem position demanded by the
controller output. The positioner itself is a high frequency, high gain P-only feed
back controller that looks at the measured valve stem position, compares it with the
signal sent to the valve, and make the necessary instrument air pressure adjustment.
Therefore, the whole arrangement is a cascade control configuration with the valve
position loop being the secondary one. A control valve with a 25% deadband can
meter How rate to a precision o f 0.5% with a positioner [1 ,4 , 6, 8, 18).
REFERENCES
http://www.unit.com
http://matec.0rg/Catal0g/M 0duleDesc/M I I4.htm
http://www.accessperrys.com
http://www.mhhe.com
http://www.isa.org
http://che.ttu.edu/pcoc/software
CHAPTER 17
FLUID MACHINES
INTRODUCTION
Stationary and mobile power generation in power plants, airplanes and cars would
be impossible without steam turbines, gas turbines, water turbines and reciprocat
ing piston engines. It is estimated that pumps consume about one third o f all elec
tric power generated in the United States. Domestic, commercial and industrial air
conditioning, refrigeration and compressed air service all require compressors. Fluid
mechanics analysis plays an indispensable role in the design o f these machines.
Fluid machines— pumps, turbines, compressors— serve to increase or decrease
the energy o f a flu id stream. The energy o f a fluid element is usually expressed
as a head, as an equivalent height o f fluid column in feet or meters. This practice
makes the calculation independent o f the density or specific weight o f the fluid. The
Bernoulli equation expresses the principle o f conservation o f energy, applied to a
fluid stream. The total energy, the stagnation pressure, stagnation enthalpy or stag
nation head, in the preferred form, is equal to the contributions o f the static head
corresponding to the static pressure, velocity head corresponding to the kinetic
energy and the elevation corresponding to the work against gravity. W ithin fluid
machines, elevation changes are usually negligible. Consequently the static pres
sure and velocity fu lly define the state o f the fluid, at any one location. Since, to
define the velocity, a vector quantity, three orthogonal components are required, a
total o f four variables are needed to define the state o f the fluid, one o f these being
the pressure or head. Typically the force-balance equations in three directions, which
are the three Navier-Stokes equations, and the equation o f conservation o f mass,
which is the equation o f continuity, provide the four equations needed to solve for
the four unknowns. Often the assumption can be made that the flow follows the
17.1
17.2 FLUID FLOW HANDBOOK
flow passage walls, in which case the direction o f the velocity is given and only
its magnitude is unknown. The number o f variables is then reduced to two, and
the energy equation, B ernoulli’ s equation is used instead o f the complex Navier-
Stokes partial differential equations. As w ill be seen, in rotating machinery the
conservation o f angular momentum plays an important role and its mathematical
expression substitutes for the Bernoulli equation.
In fluid machines, moving flu id boundaries increases the energy o f the fluid
stream. Efficient transfer o f energy demands a cyclical, unsteady process that is evi
dent in reciprocating or in positive displacement pumps, for example, but is also
present in dynamic fluid machinery using bladed rotors 11 J. The necessary pressure
difference across the blades results in a pressure fluctuation in the stationary frame
o f reference at the rotor exit. Energy can also be transmitted in a steady process by
viscous friction, and some fluid machines do exist that use such a process. Ejectors
and disk pumps exem plify such machines. Viscous friction being inherently dis
sipative, such devices only reach lower efficiencies, and are used under unusual
conditions [2, 3J.
Positive displacement machines, prim arily piston pumps or compressors, fill
and seal o ff a cavity and then shrink the cavity to increase the static pressure [4, 5].
Simpler devices, especially for incompressible fluids, rely on a back flow from the
high-pressure exit to raise the pressure in the cavity, and do work against exit pres
sure by expelling the fluid from the cavity. Since positive displacement devices offer
high efficiencies at high static pressures and relatively low flow rates, the kinetic
energy o f the flow can be neglected.
Axial pumps, turbines, and compressors increase, or decrease, the energy o f the
fluid by changing the kinetic energy o f the fluid. In centrifugal pumps, turbines, and
compressors about ha lf or more o f the energy increase, or decrease, is due to a
change in kinetic energy. I f the energy increase in pumps or compressors is desired
in the form o f static pressure, then a diffuser must follow the impeller in which
velocity head is transformed into pressure head. Diffusers are usually inefficient
and account for most losses in fluid machines. Dynamic machines are preferred in
low head, high flow rate applications. High flow rate demands large flow passage
cross-sections to minimize wall friction losses. Large flow passages result in large
machines and higher cost. Therefore size is reduced and particular attention is paid
to the flow pattern through the machine to minimize losses. Typically a single com
ponent, a shaft mounted bladed impeller imparts energy to the fluid. The angular
velocity and the torque o f the shaft define rotary input power which must be bal
anced by the increase o f angular momentum and flow o f fluid mass through the
impeller. This equation, expressing a balance o f moments, usually referred to as
Euler’ s equation, fu lly defines the energy increase, or decrease, achieved by the
fluid machine.
An interesting device, the hydraulic ram, combines a cyclic operation resem
bling a positive displacement device and a dynamic effect, using the kinetic energy
o f the fluid. [6].
The dimensionless Specific Speed provides a useful guide for classifying fluid
machinery according to optimal performance. The performance o f a fluid machine
FLUID MACHINES 17.3
* = (17.D
g (H )y4
In commercial practice in the United States, the Specific Speed is not dimension
less and is expressed directly in RPM. GPM, and feet, gravitational acceleration
being ignored. In international trade, the Specific Speed is calculated using RPM,
cubic meters per second, and meter. In commercial practice, the Specific Speed of
compressors is expressed in a sim ilar form but using the units RPM, icfm and ft.
It was found that the efficiency o f fluid machines generally correlates with Spe
cific Speed. Positive displacement machines have relatively high efficiencies at low
Specific Speeds, and dynamic machines have relatively high efficiencies at high
Specific Speeds. The classification is obviously only approximate, but it is not arbi
trary. In positive displacement devices, designers devote particular attention to leak
age losses. High pressures enhance leakage flow which subtracts from the useful
flow rate and has a greater undesirable effect at low flow rates. In dynamic pumps
or turbines designers pay particular attention to the flow pattern in the device to
minimize flow losses which increase rapidly with higher through flow velocities.
Some internal leakage can be accepted.
The classification according to Specific Speed even applies w ithin the class of
dynamic machines. For example: Pelton impulse turbines, Francis centrifugal, or
Kaplan axial water turbines are chosen depending on increasing Specific Speed [9].
Dimensional analysis applied to dynamic flu id machines predicts that for sim
ilar devices the flow rate w ill vary in direct proportion to the rotational speed and
to the cube o f the nominal diameter, while the head w ill vary with the square of
the rotational speed and the square o f the diameter, provided that the gravitational
acceleration is also introduced in the nominator to balance the units:
Elim inating the diameter from these expressions leads to the Specific Speed.
Another dimensionless number, the Reynold's number, provides a measure of
the relative importance o f inertia versus viscous forces. It consists o f the product o f
a typical velocity and typical diameter divided by the kinematic viscosity of the fluid.
Re = — (17.3)
17.4 FLUID FLOW HANDBOOK
A majority o f pumps and liquid turbines handle water, which has a relatively
low viscosity. Consequently viscous effects are not very important and are mostly
restricted to the neighborhood o f the flow passage walls. Compressors and gas tur
bines often handle gases o f higher viscosity. Viscous effects near passage walls
result in boundary layers having reduced velocities w ith respect to the walls. Such
boundary layers affect machine performance in two ways: wall friction can produce
an energy loss, and the presence o f the low velocity layer can restrict the flow pas
sage cross-section and produce a blockage. An idealized analysis o f flow pattern in
dynamic fluid machines usually assumes inviscid flow as a first approximation, and
applies some correction to account for the above two viscous effects. Modern com
puter programs can handle a solution o f the full Navier-Stokes equations, but the cal
culation for a typical pump, for example, may require up to a m illion mesh points
and needs to run overnight on a work station. Such calculations can only be ju s ti
fied for expensive machines.
Since the trailing edge condition, the Kutta condition, has a crucial effect on
the calculation o f the flo w past the airfoil, much research has been devoted to
clarify its role. The practical first order assumption, first proposed by Kutta and
Joukowski, and the inviscid flow assumption provide a surprisingly good esti
mate o f the lift. The Kutta condition requires that the strength o f the distributed
vortices, and therefore the circulation around the airfoil, be adjusted until the rear
stagnation point coincides with the trailing edge o f the airfoil. The circulation
then provides an estimate for the lift. Various modifications have been proposed
to account for the presence o f boundary layers, and especially for blunt trailing
edges. The exact nature o f the Kutta condition is still debated. Unsteady flow
experiments and calculations, and also such phenomena as the starting vortex,
suggest a stability criterion. Indeed, at the onset o f the airfoil motion, or at a
change o f its velocity, a vortex is shed into the flow , the strength and sense of
rotation o f which is equal and opposite to the change in circulation around the
airfoil. Conditions gradually reach steady state as the starting vortex is swept
downstream to infinity.
W hile an inviscid analysis can provide a good estimate for the lift, inviscid
flow calculation predicts no drag (D ’ Alambert’ s paradox.) Since designers delib
erately minimize drag on airfoils, its principal remaining component on slender
airfoils is skin friction drag, which depends on viscosity. Consequently bound
ary layer research, including turbulence, became an extremely important issue
and has recently almost entirely dominated fluid mechanic research. Boundary
layers on the a irfoil shaped blades o f gas turbines, for example, are very impor
tant. In industrial machines, they do affect the maximum efficiency that can be
obtained at the design condition; however, design is often compromised for bet
ter off-design operation since conditions often vary. Boundary layers in a cas
cade also produce a blockage and affect the amount o f turning o f the velocity
vector, which is less than the theoretical based on the blade exit geometry. A cor
rection is applied in the form o f the “ deviation angle” which subtracts from the
turning angle.
The practical pressure ratio through a single compressor stage can only be ot
the order o f l .2 to 1.5. Consequently axial compressors usually consist o f several
stages, up to 30 in high performance gas turbines. Compressibility effects are small
w ithin one stage and constant pressure can often be assumed.
Knowing the inlet and exit absolute tangential velocities o f the stage, the head
rise can be calculated from the change in angular momentum, Euler’ s equation:
h2 - h ]= (17.4)
g
In order to convert the velocity head at the exit o f the stage to pressure head, the
rotating stage must be followed by a row o f stator vanes to turn the velocity vector
back and to recover the pressure. These stator vanes define the direction o f the flow
17.6 FLUID FLOW HANDBOOK
entering the next stage. It can be assumed that the flow enters the first stage w ith
out any swirl and consequently the tangential component o f the inlet velocity is zero.
In the following stages, this may not be the case and the flow may enter with a cer
tain tangential velocity.
In steam or hot gas turbines, a set o f stationary vanes precedes the first stage.
A t the expense o f a pressure drop, they produce a swirl in the flow which impacts
tangentially the moving blades that extract the kinetic energy. Some o f the pressure
drop might also take place in the moving blade row. I f all the pressure drop occurs
in the stator vanes, and the flow through the moving vanes is at constant pressure,
designers speak o f a pure impulse turbine. The stationary vanes act as nozzles direct
ing a free stream at the rotor blades. I f a pressure drop is also present in the rotating
blade row they refer to a reaction turbine.
It the blades are very long, the blade root and tip circumferential velocities can
be very different and demand different blade profile shapes. A need for high e ffi
ciency suggests that the same energy be imparted on each o f the streamlines passing
through the blade row at different radii. The feasible blade shapes and their stack
ing lim it the length o f the blade. Various radial velocity profiles have been conceived
and are in use to solve this problem. I f the lift o f the airfoils, the energy imparted
to the fluid stream, is not the same radially along the blade, then a shear flow exists
downstream and trailing vorticity is spilled into the flow. In blades designed for uni
form energy addition radially, more energy w ill be imparted at the blade tips than at
the root at reduced flow . The stream lines turn in a radial direction within the blade
row. In the extreme, the flow may reverse and flow back at the inlet near the tip, and
flow back from the exit into the blade row near the root at the exit. Such flow rever
sal also occurs in mixed flow pumps and w ill be discussed below. Flow reversal
has an unfavorable effect on performance and in compressors can affect surge.
The expression for the ideal isentropic work offers a first order estimate o f the
work done on a compressible fluid. It relates the energy increase o f a unit mass of
fluid to the pressure ratio. The corresponding head is obtained by dividing with the
specific gravity o f the fluid.
(1 7 .5 )
In centrifugal pumps, compressors and radial inflow turbines a shah mounted impel
ler imparts or extracts energy from the fluid. [12, 13, 14]. In pumps and compressors,
the flow typically enters the impeller through the impeller eye in an axial direction,
turns radial in the wheel and exits radially at the periphery. A disc, perpendicular
to the shaft, in the rear supports the blades which are usually curved in the direction
opposite to the sense o f rotation. A front cover, or shroud, may or may not exist.
Pumps usually have three to seven blades, but one bladed sewage pumps also exist.
Pump designers usually prefer an exit blade angle o f about 70 degrees from radial.
Industrial practice measures pump exit angles from the tangential direction, typically
20 degrees. Compressors can have 20 to 30 blades, usually backwards curved about
30 degrees from radial for better range and efficiency. Material strength limits cen
trifugal compressor speed. It depends on the maximum allowable tip speed U 2 which
can be o f the order o f several hundred to 1000 ft/sec. Consequently, small compres
sors may turn at 50,000 RPM or more.
A simple, first order flow pattern analysis usually assumes that the flow enters
w ithout swirl, all streamlines having the same energy. At design conditions, the
velocity o f the entering flow is aligned w ith the leading edge of the blades. The
17.8 FLUID FLOW HANDBOOK
relative velocity W, in the rotating frame o f the impeller becomes the vectorial
sum o f the absolute inlet velocity V\ and the circumferential velocity (orx\
The expression o f the energy o f the inlet flow without sw irl, in the form o f a
stagnation head h, becomes in the stationary and rotating frames respectively:
i/2 ,i/2 2 2
h , = h„ + = h„ + (17,7)
2# 2g 2g
From the impeller exit velocity diagram, assuming backwards curved blades,
one can derive the relationships, since the radial component o f the absolute and
relative velocity are the same:
V 2 = w ? - a f r 2 + 2 0)r2Vt2 (17.9)
u u y2 L Wi a)2r 2 car2Vt2
h 2 = h s2 + = hs 2 + — ------- + —
2g 2g g
As pointed out before, the last term corresponds to the head increase pre
dicted from the conservation o f angular momentum, Equation 17.4. Combining
the expressionsfor the total head at the exit h2, Equation 17.7, and at the entrance
h,, Equation 17.10, and subtracting the expression for thehead rise, Equation
17.4, it becomes apparent that the three terms, usually designated as Rothalpy,
remain constant throughout the impeller.
u /2 fi)2r 2
Rothalpy = / = h s + — — - ------ (17.11)
2g
The Rothalpy takes the place o f the Bernoulli equation w ithin the impeller. It
does not depend on the direction o f the relative velocity, only on its magnitude. It
can be used to predict the local static head h s or the pressure, provided that the rel
ative velocity W can be estimated, from the flow rate and passage cross-section, for
example. While presented here for incompressible flow , the Rothalpy also applies
to compressors, in which case the enthalpy replaces the head.
From the expression o f the Rothalpy, it becomes apparent that in centrifugal
compressors and pumps an appreciable portion o f the static pressure increase comes
from the centrifugal acceleration, unlike in axial machines. The energy addition is
FLUID MACHINES 17.9
in the form o f pressure, not as kinetic energy. Centrifugal machines can provide a
much greater pressure rise or pressure ratio in a single stage than axial machines.
Compressor pressure ratios o f 3 or 4 are not unusual. Sonic velocity at the inlet and
excessive temperatures at the exit lim it the maximum attainable compression ratio.
The exit velocity diagram also leads to simple analytical expressions tor the head
and How rate characteristics of pumps. Naturally it assumes uniform or averaged
flow velocities in the flow passages. The relationship between the ideal or theoret
ical head rise H th and the flow rate Q becomes:
In this expression, B 2 designates the width o f the impeller at the exit. (i2 the
exit blade angle measured backwards from the radial, D 2 the impeller diameter
and U2 the tip speed.
The symbol <7stands for a correction coefficient, the slip which accounts tor the
fluid stream leaving the impeller at an exit angle greater than the geometrical. It
results from the rotation o f the fluid relative to the impeller, in a sense opposite to
the shaft rotation, because o f its inertia. Busemann calculated the slip in a radial
bladed impeller assuming potential flow and using conformal transformations. Sev
eral simple, practical expressions have been proposed since by designers, in partic
ular by Wiesner, who reviewed Busemann’ s results and incorporated experimental
data from some three dozen pumps 115]. The slip coefficient amounts to about 0.8
in typical pumps.
Centrifugal pump or compressor blades are designed by applying the conserva
tion o f angular momentum or Euler’ s equation in incremental, radial steps to deter
mine the local blade angle. By choosing different blade angle distributions from
inlet to exit, blade loading, the rate o f energy addition, can be shifted towards the
leading or trailing portions o f the blades. Gradual relative velocity changes are pre
ferred to avoid the possibility o f flow separation. If warranted, different blade shapes
can be used on different streamlines, such as the hub, mean and shroud streamlines.
However, the ideal impeller should impart the same energy to the fluid on all stream
lines. Practical considerations, such as avoiding a badly twisted blade shape or ease
o f manufacturing, often compromise the blade shapes.
The above averaged impeller flow model has proven adequate for general ser
vice pumps and compressors, but ignores some important flow phenomena and sec
ondary flow effects which must be accounted for in large, expensive custom designed
pumps. Unlike in axial flow machines, the transverse Coriolis acceleration strongly
influences the flow in radial flow impellers. A strong circumferential pressure gra
dient exists between two consecutive blades which produces important secondary
flo w effects. Since the Coriolis acceleration, and related pressure gradient, depend
17.10 FLUID FLOW HANDBOOK
on the flow velocity, it does not exist in the side wall boundary layers where the
velocity vanishes. Consequently, a strong convective flow is induced in these bound
ary layers from the pressure side, leading face o f the blades to the suction side, where
low velocity fluid accumulates and often flow separation appears. Also the trans
verse, Coriolis acceleration tends to stabilize the shear layer between the separated
region and the tree stream [ 16, 17]. Consequently, a distinct je t and a wake region
can exist in the impeller. This flow pattern is more likely to appear in highly loaded
impellers, those with substantially radial blades. Such conditions often prevail in cen
trifugal compressor but not in pumps. Designers refer to such a flow pattern as a jet-
wake flow. It produces a strongly fluctuating flow at the impeller exit in the absolute,
stationary frame o f reference.
Large, expensive custom designed pumps and compressors warrant an extensive
analysis employing the latest Computational Fluid Dynamic computer programs
[18, 19, 20]. Only an appropriate modeling o f turbulence and boundary layer flow ,
suitable for rotating impellers, make such calculations meaningful. The simple k - £
turbulence model, for example, does not give correct results in the presence o f strong
transverse accelerations which are known to be present in impellers. Mathematical
turbulence models are artificial by necessity since turbulence is an unsteady process
while practical calculations demand steady, averaged numbers. Turbulence models
contain empirical coefficients which unfortunately are not universal. Turbulence lev
els in real pumps often depend on upstream conditions which remain unknown. Solu
tions o f the full Navier-Stokes equations have been successfully executed using up
to a m illion mesh points. However, valid results require previous calibration o f the
calculation methods on actual test data. Care must also be devoted to the formula
tion of the inlet and exit flow assumptions. The assumption o f a uniform, steady inlet
flow automatically rules out any unsteady effects, non-axisymmetric inlet flo w or
flow re-circulation, often encountered in fluid machines, especially at o ff design con
ditions. The modeling o f the unsteady interaction between impeller exit and vaned
diffuser inlet flow presents especially difficult problems. The unsteady m ixing pro
cess in the vaneless space at the impeller exit is not well understood. In addition, in
compressors sonic flow may appear at the impeller exit adding further difficulties.
Convenience suggests that the performance plot o f the pump, head versus flow
rate, be expressed in nondimensional form by using the quantities o f head and flow
coefficient.
_ g H lh
ui
(17.13)
Flow coefficient = u/= —— Q
(J2 (2/rD2£ 2(/2)
Some other definitions o f the head and flow coefficient are encountered in the
technical literature. Some authors define the head coefficient using the actual, net
head delivered by the pump instead o f the theoretical head, others introduce a factor
o f two. Some define the flow coefficient by omitting the impeller width B 2 and squar
ing the diameter D 2. However, the definition offered here ties the impeller exit veloc
ity components, radially and tangentially, directly to the performance. Note that the
head coefficient only depends on the energy input while the flow coefficient only
depends o f the flow rate.
The above relationship for the theoretical head corresponds to a straight line, de
clining with flow rate, on a performance plot. The inclination increases with increas
ing blade angle, measured from the radial, and with a narrowing ot the impeller exit
width. The actual head-flow curve is obtained by subtracting the various head losses
from the theoretical head. The primary losses consist o f the skin friction and d iffu
sion losses which typically increase with the square o f the flow rate, shock losses as
the flow enters the diffuser at the impeller exit which typically increase with the
square o f the absolute exit velocity and therefore the head, and incidence losses which
increase on either side o f the design flow rate. The resulting curve resembles an
inverted parabola that may or may not have a maximum.
The mechanical losses o f the drive system and disk friction, fluid friction of
the front and back faces o f the impeller, detract directly from the shaft power sup
plied to the pump. They reduce the efficiency but leave the head unaffected.
X
d
aj
45
U ♦V*
•H
P
3 >
s U
> G
<u
H
H O
113 H
*-> 4-.
O <H
H GJ
2 c
01
w
u
o
O.
H-4
*1.0-1
QJ
J] &
£ o
w a
0 J
C c tp a t; i c y , Q
The analysis o f disk friction offers an interesting case of viscous How in a radial
clearance space between a rotating disc and stationary housing wall. The alterna
tives o f a net outward or inward flow can complicate the solution [21].
A sim ilar parasitic loss results from inlet recirculation. A t reduced flow rate, a
point can be reached where the flow at the inlet reverses at the periphery and flows
back from the impeller into the inlet passage. The corresponding energy represents
a power loss which, to a large extent, is independent o f the pump design except for
its inlet configuration. The lost power typically increases rapidly, approximately in
proportion to the 5/3 power o f the flow reduction below the design flow . Its exact
value also depends on the upstream configuration o f the piping, and not only on the
im peller design.
In centrifugal pumps, an appreciable portion o f the energy increase is in the form
o f a velocity head. Therefore a diffusing device follows the impeller. In refined,
high efficiency pumps a radial vaned diffuser recovers pressure. It consists o f vanes
which turn the flow gradually in the radial direction and slow the velocity down. D if
fuser design requires particular care. The area increase must be very gradual and,
i f possible, upstream disturbances must be avoided. Good design practice usually
recommends an included angle o f six degrees for straight diffusers. More refined
guidelines are available in the literature [22, 23]. Designers rely on the pressure
recovery rating o f different designs to optimize performance. Pressure recovery,
a dimensionless figure o f merit, expresses the fraction o f the velocity head at the
diffuser entry that the diffuser recovers in the form o f static pressure.
Perfect, loss free flow conditions offer an upper lim it for the pressure recov
ery coefficient. Use o f Bernoulli’ s equations and continuity lead to the analytical
expression, A designating the diffuser cross-sectional area:
(17.16)
A lower lim it becomes available from the application o f the continuity equa
tion and o f the momentum equation to a je t exhausting into a cylindrical mixing
tube, and assuming that a uniform velocity profile exists at the diffuser exit. This
assumption implies that the device is long enough to completely mix the flow, even
i f flow separation takes place. Calculation o f this “ sudden expansion” process leads
to the expression
F LUID MACHINES 17.13
At an area ratio o f two, for example, the diffuser ideally recovers a maximum
o f 75% o f the inlet velocity head: however, a sudden expansion process can still
recover 50%. Sudden expansion pressure recovery reaches a maximum at an area
ratio o f two. Designers rarely use diffusers with area ratios in excess ot about two.
In industrial pumps for general use a housing o f spiral shape, a volute or scroll,
surrounds the impeller, that collects the flow from the impeller circumference and
guides it to a common exit. Designers size the volute to match the velocity in the
volute exit, the throat, which is known from its cross-section and the flow rate, to
the velocity coming o ff the impeller. A mismatch, a sudden change in velocity, results
in a loss. Such a loss is unavoidable at reduced flow rates.
Conditions at the impeller exit, however, are far from ideal. A stationary observ
er would see an unsteady, oscillating flow leaving the impeller. It has been suggested
that the mixing in the space between impeller and diffuser inlet does not take place
by viscous shear, but by pressure exchange, by alternating high and low energy
slugs o f fluid follow ing each other 1241. Pressure pulsations at the impeller exit can
induce pipe resonance downstream at the blade passing trequency, the rotational
frequency times the number o f blades, which accounts for the dominant mode ot
excitation o f fluid induced vibrations.
Vertical pumps, consisting o f impellers mounted on a vertical shatt and low
ered into the ground, often have several stages. For example, the bore hole size
lim its the impeller diameter o f submersible oil well pumps; consequently a litt ot
several thousand feet often requires one or two hundred stages. Large boiler teed
pumps and pipeline pumps have several stages. High pressure ratio compressors
for the chemical industry also have several stages. A return flow passage is then
required which guides the fluid from one stage to the next. These return passages
also diffuse the flow using appropriate guide vanes to assure a swirl free entrance
flow to the next stage.
Disc pumps belong in the category of centrifugal pumps. They consist ot several
tightly spaced discs mounted on a shaft. They have no blades but rely on the viscous
friction o f the fluid on the discs to entrain the flow and impart energy. This fluid
mechanic problem o f viscous flow between closely spaced, rotating discs admits of
an interesting analytical solution [25]. Because o f viscous losses, their efficiency
usually falls below that o f bladed centrifugal pumps. They are prim arily suited for
viscous fluids, or fluids containing contaminants. They can also be used as turbines.
These pumps belong into the class o f dynamic pumps, but deliver a performance
intermediate between centrifugal pumps and positive displacement pumps [26].
They consist o f a shaft-mounted rotor in a concentric housing. Inlet and exit ports
are arranged on the circumference sim ilarly as in vane or internal gear pumps. On
the periphery o f the rotor, symmetrically on either side, circular cuts are made to
produce numerous small blades. The housing forms a recess on either side ot the
rotor tip to allow two toroidal flow patterns, on either side, circulating repeatedly
17.14 FLUID FLOW HANDBOOK
in and out o f the rotor. The repeated impulses from the small blades gradually in
crease the pressure around the circumference from the inlet to the exit. Such side
channel pumps can produce higher pressures than centrifugal pumps o f the same
size. The flow rate is typically less and the efficiency is also inferior. Still, the sim
p licity o f the design and the higher pressure capability makes these pumps some
time preferable to others.
Ejectors
Ejectors have the great advantage o f no moving parts. In ejectors a high power,
primary jet, discharged through a central nozzle entrains and mixes with a parallel
stream of low energy, secondary fluid. M ixing is completed in a mixing tube o f con
stant cross-section. A diffuser follow ing the m ixing tube reduces the high velocity
at the exit o f the m ixing tube and recovers the pressure. Since the flow velocities
retain an axial direction and the geometry is cylindrical, a simple analytical model
o f the m ixing tube, based on the momentum equation, becomes possible which
predicts the performance o f an ejector well [27].
S crew Machines
Screw machines belong in the class o f positive displacement devices, suitable for
relatively higher pressure and lower flow rates than dynamic machines. They con
sist o f one, two or even three rotors having spiral shaped teeth which mate with each
other, or with the housing in the case o f a single screw [28, 29]. The special profile
shapes trap a certain volume o f fluid between the rotors, and move it gradually from
an axial inlet to an axial outlet. Internal leakage and back flow limits the efficiency
since the mating rotors must mesh with a sufficient clearance to accommodate man
ufacturing tolerance variations, thermal expansion and unavoidable deflection. Mod
em numerically controlled manufacturing machine tools produce more precise rotors
and have improved the competitive position o f screw pumps and compressors. The
pressure ratio o f screw compressors does not exceed 2 or 3, unless oil is injected
which improves sealing between the rotors and can also cool the gas during com
pression, approximating a more efficient isothermal compression process. However,
oil injection penalizes the compressor assembly by requiring oil cooling and filte r
ing equipment. Screw pumps are favored for high viscosity fluids.
Gear pumps and motors, a set o f gears in a tight fitting housing, constitute perhaps
the simplest pump types. Since they are fabricated with fixed clearances they are best
suited for viscous fluids and have found general use in oil lubrication systems and
FLUID MACHINES 17.15
servo actuators. They typically can produce one or two hundred psi pressure and
deliver up to ten GPM (35 liter) How rate. The fluid is trapped between the gear
teeth and the housing and is transported to the exit port. Pressure in the cavity rises
because o f back flow from the high-pressure exit port. The meshing o f the gears
expels the flu id from the cavities between the gears.
Gear pump designs can take a variety o f shapes, internal gears, external gears,
or the trochoid profile shaped. Wankel fam ily o f machines [30). In a combustion
engine configuration, they were the only different engine type that could success
fu lly challenge the piston engine in automotive applications, and were manufac
tured by the millions. In these positive displacement machines special apex seals
solved the leakage problem.
Since oil pressure power actuation in o ff the road equipment, for example, re
quires pressures up to 5000 psi or more, designers reduced external gear pump leak
age by movable side plates that adjust the running clearance automatically using
the principle o f hydrostatic bearings. In these applications o f hydrostatic bearings
pump exit pressure is fed into the cavity behind the movable side plate. A small
orifice, properly sized, meters the flow to the bearing clearance. A change in clear
ance automatically reestablishes the balance o f fluid pressure forces on the side plate
and readjusts the clearance. Hovercraft also use the principle o f hydrostatic bear
ings. Bearing loads resulting from the transverse pressure force on the gears lim
its the capacity o f very high pressure gear pumps.
Vane pumps and compressor solve the problem o f sealing by vanes in radial slots
in shaft mounted rotors which slide on the inner, cam surface o f an eccentric hous
ing. The space between consecutive vanes traps and compresses the fluid. Variable
delivery becomes possible with a circular cam shape the eccentricity of which can be
changed. An oval cam shape results in two strokes per shaft revolution and also bal
ances the pressure force on the rotor to minimize the bearing load. Such vane pumps
find application in oil powered actuation systems. Vane compressors are sometimes
used as refrigeration compressors and as moderate pressure vacuum pumps.
The earliest pumps and compressors used the reciprocating piston configuration.
Circular bores and pistons can be machined very accurately and can be honed to
very high accuracy. Also circular clearances are the easiest to seal reliably. Pis
ton ring and O-ring seals have been developed to a very high degree o f perfec
tion. Not surprisingly, reciprocating piston engines, for example, have dominated
engine technology. The analysis o f the unsteady flo w in engine cylinders almost
became a separate discipline and has generated a vast literature [3 1 1.
17.16 FLUID FLOW HANDBOOK
High pressure piston pumps for o il servo actuation, capable o f up to 6000 psi,
illustrate the highly refined state o f piston pump technology [32, 33J. In the preferred
configuration, several solid pistons, honed to very high accuracy, are arranged axi
ally on the periphery o f a rotating barrel. On one end, a separate valve plate, honed
flat to w ithin a few light bands and in a precise axial hydrostatic balance to m ini
mize leakage, incorporates the kidney shaped inlet and exit ports. On the other end
o f the assembly the spherical ends o f the pistons rest in slippers, also hydrostatically
balanced, which slide on a wobble plate, or swash plate. It is inclined with respect to
the perpendicular to the axial direction and imparts the reciprocating motion to the
pistons. In a variable capacity configuration o f the pump, the wobble plate inclina
tion is adjustable, increasing or decreasing the stroke o f the pistons. High pressure
oil operated motors also use this axial piston configuration. This refined design, in
which very high static pressure forces are carefully balanced, results in power actu
ation components o f exceptionally high power to weight ratios, making them the
preferred systems for aircraft servo actuation, for example, where weight is at a
premium, and more generally in o ff the road equipment and machine tools.
High pressure piston pumps in the chemical industry and in oil well service
(triplex pumps) use more conventional crankshaft and connecting rod drive mecha
nisms. Some designs use a cross head at the end o f the connecting rod and a straight,
reciprocating rod attached to the piston. Such an arrangement has the advantage
o f making a double acting operation possible. The seal towards the exterior, relo
cated from the piston circumference to the straight, reciprocating rod, becomes sim
pler and functions under more favorable conditions. The pressure drop and proper
functioning o f the pressure actuated valves o f such compressors remains a key
issue. A very great number o f piston compressors find application in refrigeration
and compressed air service.
The process o f selection and application o f fluid machinery starts with the estab
lishment o f the proper specifications [34]. Fluid machines are power conversion
devices, transforming shaft power, a product o f torque and rotational speed, into
fluid power, a product o f pressure and flow rate. Consequently, as a minimum, the
pressure or head //, the flow rate <Q, the rotational speed /V, and the power must be
specified. As pointed out above, these quantities enter into the expression o f the
Specific Speed, which generally correlates with efficiency and suggests the best
flu id machine configuration for the desired service. A xial flow machines are best
suited for high Specific Speed, followed by centrifugal machines, positive displace
ment screw, gear and piston machines with decreasing Specific Speed. Centrifugal
pumps reach their highest efficiency at a specific speed o f about 2300, when ex
pressed in feet, GPM and RPM.
Establishment o f the needed head for pumps usually presents some difficulties.
Specifications define the head rise o f the pump as the total, stagnation head at the
pump exit minus the total, stagnation head at the inlet. In rare instances an ambi
in ID MACHINES 17.17
guity remains whether the exit velocity head should be considered as useful or as
a loss. The exit head usually consists o f an elevation, a lift to a higher level, a fric
tion loss due to the piping and throttling devices, and the velocity head.
The inlet o f the pump might be submerged, in which case the inlet head w ill
exceed the ambient pressure, usually the reference pressure from which all pres
sures are measured. If the pump aspirates fluid from a lower level then suction
pressure, negative pressure below ambient, w ill prevail in the inlet. Its magnitude
w ill consist o f the level separation between the free fluid surface ahead o f the pump
and the velocity head o f the inlet flow. The vapor pressure o f the fluid, the pres
sure at which the fluid starts to boil, w ill set a lim it to the suction head of the pump.
Theoretically at this pressure vapor bubbles w ill form in the pump inlet, cavita
tion w ill set in and w ill impair operation. In cavitating flow small vapor bubbles
form locally and, swept downstream, collapse. Such bubble collapse generates
locally very high pressure pulses which may destroy the material o f the impeller.
W ith a further reduction o f the pressure large vapor filled cavities form which
change the flow pattern in the impeller and reduce the head. In practical operation
cavitations may set in at negative inlet pressures that are higher than the vapor
pressure o f the fluid because o f high local velocities on the blade surfaces, espe
cially at o ff design operation, at incidence. Pump manufacturers usually show the
required suction head on the published performance curves of pumps. Laboratory
pump tests detect cavitations lim its by the reduction o f the head delivered by the
pump, usually a 3% drop. However, cavitation damage can occur sooner. Conse
quently the estimate o f the permissible suction head must be conservative and
usually remains somewhat uncertain. Applications in aviation can present a partic
ularly d ifficu lt problem when pumping boiling liquids and volatile hydrocarbons,
such as those handled by aircraft or rocket fuel pumps for example. The vapor pres
sure o f cold water approaches absolute vacuum.
>1 160—n
u
£
01
•H
u
•H 1 2 0 —
U 80 —
01
&
o
ft
0)
Irt
u 40 —
o
£
od 0 —
£
Capacity, gpro
FIGURE 17.2 Pump performance curves at varying submergence.
17.18 FLUID F L O W H A N D B O O K
C avitation can be prevented by boosting the pressure ahead o f the im peller inlet.
Instead o f a separate booster pump, often an inducer, attached to the pump inlet, is
used to increase the inlet pressure. Inducers have spiral, screw shaped blades extend
ing for several revolutions around the shaft w hich induce an axial flow into the
eye o f the im peller and produce a very sm all pressure rise, ju st enough to prevent
cavitations [35, 36].
A proper m atch o f a pum p to the load usually dem ands that a broader view be
taken o f the perform ance, on either side o f the nom inal operating condition. For this
purpose the head, efficiency and pow er dem and curves o f the pum p are plotted
against the flow rate. A curve, corresponding to the load characteristics and plot
ted on the sam e graph, intersects the head-flow curve o f the pum p at the operating
point. A typical load curve consists o f tw o elem ents: a head rise corresponding to
an elevation and a pipe friction loss. The first rem ains constant with flow rate while
the second varies with the square o f the flow rate. The resulting load curve appears
as a parabola on the perform ance plot with its origin on the vertical axis at a value
that corresponds to the elevation.
G enerally speaking, centrifugal pum ps have a characteristic that resem bles an
inverted parabola in the perform ance plot. For backw ards curved blades the curve
usually retains everyw here a negative slope, or perhaps reaches a m axim um at zero
flow rate. For steep exit angles, approaching the radial, or for very low radial exit
velocities, due to large impeller exit widths, the characteristics can have a m axim um
at som e point below the nominal flow rate. In such a case an approxim ately horizon
tal load curve corresponding to an elevation only, will have tw o intersections with
the pum p curve, on either side o f the m axim um . O nly the intersection at the larger
flow rate, w here the slope o f the pum p curve is negative, will correspond to a sta
ble operating point. A reduction o f the flow rate below the value corresponding to
the m axim um will lead to unsteady surging, an uncontrolled fluctuation o f the flow
rate, and to associated surges in torque and pow er that can reach destructive am pli
tudes. T herefore, in pum ps intended for general use a m axim um in the character
istics is to be avoided.
In axial and in mixed flow pumps, having flow passages inclined from the radial
towards the axial direction, exit velocities are usually low. Also the blades at the exit
extend to a larger radius near the shroud than near the hub. As a consequence a max
imum appears in the pump characteristics, and an unstable operating range can exist.
H ow ever, w ith further reduction o f the flow rate the pum p head often starts to rise
again. This head increase is due to im peller exit re-circulation w hen the exit flow
reverses near the hub, the sm aller radius o f the outlet passage, and flow s back into
the impeller. At progressively decreasing flow rates the rem aining stream lines grad
ually crow d tow ards the shroud at a larger radius and acquire m ore energy and a
higher head. The flow reversal results from uneven exit energy on the stream lines
leaving the impeller. If a mixed flow im peller is designed with blade angles that give
uniform energy distribution on all stream lines at the exit, as would be desirable at
design conditions for high efficiency, then at reduced flow rate less energy is imparted
on the stream lines at the sm aller radii as on those at the larger radii and a rotating
FLUID M A C H I N E S 17.19
shear fe w appears at the im peller exit. The shear flow m anifests itself in the form
o f a swirling flow which can reverse near the core .The flow rate corresponding to
the onset o f exit flow re-circulation can be calculated from the head increase on the
stream lines at different radii at the exit and from the pressure rise due to centrifu
gal acceleration in the exit flow. An uneven energy distribution at the exit can result
from several causes. The ideal im peller blade configuration is often com prom ised
because o f practical reasons such as ease o f m anufacturing for exam ple. Since only
the impeller exit geom etry controls the onset o f exit re-circulation, it can be adjusted
to chose the flow rate o f the onset. Som etim es, when a maxim um o f the character
istics copears, the onset o f re-circulation can be m oved to larger flow rates to sup
press the instability range below the m axim um .
Apolication engineers often match existing standard pumps by selecting a differ
ent speed. Standard electric m otor speeds restrict practical pump speeds to an even
fraction o f the 3600 RPM o f 60 Hz m otors, or o f the 3000 RPM o f 50 Hz m otors.
Frequency converter driven variable speed m otors allow greater flexibility but im
pose a cost penalty. A utom atic flow control in chem ical processes can often ju s
tify the expense. Som e applications, such as for exam ple high speed boiler feed
pum ps in pow er plants use variable speed steam turbine drives.
A less precise, inexpensive m atch can som etim es be accom plished by changing
the im peller diam eter in an existing pum p housing. Pum p catalogs often show the
head curves for different impeller diameters and the corresponding efficiency islands.
However, a sm aller im peller in an existing housing necessarily leads to a mismatch
o f the impeller and diffuser or volute, and to an efficiency penalty, as would be appar
ent from the application o f the perform ance calculations presented above.
Centrifugal pum p and radial inflow turbine principles have been applied to
the design o f torque converters, present in practically every car on the road [37].
Torque converters achieve a variable speed transm ission o f shaft torque that allow s
the smooth start up o f vehicles. They consist o f a concentric arrangem ent o f a cen
trifugal pum p and inflow turbine on separate shafts, and a set o f stationary guide
vanes. The flow describes a toroidal path as it repeatedly cycles through these
machine elem ents. The blade shapes are designed for perfect incidence at the design
conditions. Torque converters achieve exceptionally high efficiencies at design con
ditions, since the through flow velocity hardly changes and diffusion never takes
place. Input and output shafts being aligned, a net difference between exit and inlet
torque requires that a reaction elem ent, stationary vanes be present. Fluid couplings
do exist, that have no stator vanes, but they can only produce a slip, a difference in
shaft velocity, the torque rem aining the sam e.
Transient operation o f centrifugal pum ps, start up and shut dow n, m ust also be
considered when planning an installation since they can lead to special problem s.
In large pum ps, delivering into long conduits, as for exam ple pow er plant cooling
tower pum ps, the possibility o f colum n separation, vapor filled voids, can appear.
When these collapse, dangerous pressure waves, w ater hammer waves are produced.
Refined analytical techniques have been developed to trace the propagation o f such
water ham m er w aves and to estim ate the local pressures which can reach destructive
17.20 FLUID F L O W H A N D B O O K
levels [38]. Theoretically vapor cavities would be expected to appear when the abso
lute pressure decreases to the vapor pressure o f the fluid. In reality, water always con
tains dissolved gasses, such as air, which evolve at absolute pressures higher than the
vapor pressure. The presence o f such air bubbles affects the bulk com pressibility o f
the fluid and the pressure wave propagation velocity. W ater ham m er analysis fre
quently requires that pump characteristics be know n not only for normal operation
but also for reversed flow through the pum p and reversed shaft rotation. Designers
refer to such perform ance plots as four quadrant pum p characteristics. The flow and
head variables are norm alized by dividing with the rotational speed and its square
respectively, according to similarity rules. The axes o f the plot, normalized head and
flow , are extended to negative values.
Pressure oscillations and wave propagation can also create problem s in positive
displacem ent pum ps. Their cyclic mode o f operation is inherently unsteady. Instal
lation designers make a special allow ance for the instantaneous acceleration o f the
flow aspirated at the inlet and the corresponding negative pressure which can cre
ate cavitations and also may prevent a com plete filling o f the cylinder.
Perform ance plots o f centrifugal com pressors resem ble those o f pum ps. C hok
ing at high flow rates and surging at low flow rates limit the useful range o f opera
tion. Sonic flow at the im peller inlet o r at the vaned diffuser inlet result in choked
flow. Surge in centrifugal com pressors can be attributed to a system instability. The
head-flow characteristics o f the com pressor reaches a m axim um at the surge point.
A dynam ic analysis will show that, when the com pressor delivers into a volum e
capacity, operation below the flow rate corresponding to the surge point excites pres
sure oscillations in the system. Surge can take different form s; it can be a harm onic
oscillation, or can resemble a limit cycle oscillation with a sudden collapse and grad
ual recovery. O peration below this flow rate becom es possible and perform ance
data can be taken if a throttling device im m ediately follow s the com pressor. Such
an arrangem ent m ight be encountered in industrial installations. Surging in indus
trial applications, such as in natural gas pipeline com pressors for exam ple, are not
necessarily destructive im m ediately but cannot be tolerated for any length o f time.
A n autom atic surge preventer, bypassing som e flow from the exit to the inlet, can
usually protect the com pressor.
The perform ance characteristics o f centrifugal com pressors can be altered using
variable speed o r inlet guide vanes. The m axim um electric m otor speed is usually
insufficient for centrifugal air or refrigerant com pressors, and a gear set increases
the shaft speed. High speed, variable frequency drives offer an alternate possibility
w ith the added advantage o f variable speed as a m eans o f capacity control. Inlet
guide vanes are also used. They im part a pre-rotation to the inlet flow. The equa
tion o f conservation o f m om entum , E uler’s equation show s that a pre-rotation of
the inlet flow in the direction o f rotation will reduce the energy im parted to the
fluid by the im peller and therefore reduce the head rise through the com pressor.
Hot gas turbine driven air com pressors, such as turbochargers, have inherently
variable speed. Som e diesel engines use turbochargers to com press the inlet air and
raise the inlet pressure in the cylinder. A higher m ass o f intake air increases the
pow er capacity of an existing engine size. A higher pow er to engine w eight ratio
FLUID M A C H I N E S 17.21
reduces the cost per pow er o f the engine. The hot engine exhaust gases drive the tur
bine that is directly coupled to the com pressor, recovering som e o f the heat, avail
able and norm ally lost, in the exhaust. A waste gate, a bypass valve, provides
som e control over the operation o f the turbocharger.
Positive displacem ent pum ps deliver practically constant tlow rate regardless
o f pressure. The corresponding perform ance plot consists of a vertical line at the
nominal flow rate. At high pressures and considerable leakage the curve deviates
from the vertical w ith increasing pressure tow ard reduced How rates. Such a ch ar
acteristics assures stable operation for all load types. Because of the constant How
delivery o f these pum ps, the variable o f the “displacem ent” , volum e per shaft rev
olution, specifies the capacity o f positive displacem ent pumps.
Because o f their cyclic operation, positive displacem ent com pressors produce
strong pressure fluctuations at the exit. These pressure oscillations can induce reso
nance in the dow nstream piping, can even react back on the com pressor and inter
fere with the proper operation o f the pressure actuated valves. It the com pressor
exhausts into the atm osphere unacceptable noise levels can be produced. Interest
in these phenom ena m otivated designers to develop extensive analytical m ethods
which can trace the stationary and propagating pressure waves. At relatively small
am plitudes and intensities, the linear sim plifications o f acoustic theory can be
applied w hich, how ever, rem ain inadequate at high am plitudes. These analytical
m ethods are used to design m anifolds and mufflers for positive displacem ent com
pressors and engines.
REFERENCES
34. Florjancic, D. 1989. Sulzer Centrifugal Pump H andbook, Elsevier Science Publish
ing. New York.
35. Brennen, C. E. 1994. Hydrodynamics of Pumps, Oxford University Press. New York.
36. Rohatgi, U. S. 1995. “ Sizing o f Aircraft Fuel Pumps,” A SM E Journal o f Fluids
Engineering. I 17, 298-302.
37. Brun, K... Flack. R. D. 1996. “ The Flow Field Inside an Automotive Torque Con
verter,” SA E Technical Paper 960721.
38. W ylie, E. B., Streeter, V. L. 1993. Fluid Transients in Systems, Prentice Hall,
Englewood C liffs, New Jersey.
_________ CHAPTER 18_________
FLUID FLOW NETWORKS
Flow o f fluids in netw orks o f pipes is very com m on in many areas such as gas and
oil gathering netw orks, fire-w ater netw orks in cities, plants and refineries, w ater
cooling netw orks, steam and condensate networks, ventilation systems, and munic
ipal w ater utilities. Unlike single pipe solution equations, the design and operation
o f fluid networks is much more complex since it requires a system of energy, momen
tum , and m ass balance equations to solve for all o f the network param eters. When
fluid is com pressible, the com plexity increases by the necessity o f adding equation
o f state to the above mentioned equation system to predict the pressure-temperature-
density change along the fluid flow path. W hen m ulti-phase flow formation is en
countered, flash calculation (tem perature-pressure and pressure-enthalpy flash
m odels) must be added to the system equations to predict the gas and liquid mass
ratios, gas- and liquid-phase com position and thus physical properties o f each phase.
Fluid flow netw orks m ay consist o f large num ber o f pipes and equipm ents result
ing in a large system o f equations to be solved. This chapter will focus on develop
ing the netw ork system equations for various fluid flow networks including simple
w ater networks, com pressible, and 2-phase flow. M athematical solution techniques
to solve the system equations will be briefed next.
18.1
18.2 FLUID F L O W H A N D B O O K
T o start, Figure 18 .1a shows a schem atic o f a sim ple flow netw ork o f incom
pressible fluid, the network is a w ater-cooling netw ork with m any heat exchanger
loops. The first step in solving fluid flow netw orks problem is to create a node
branch schem atic diagram to represent the actual cooling pipe netw ork. Figure
18.1b show s the branch-node schem atic representation o f the netw ork show n in
Figure 18.1a. A schem atic netw ork consists, at least, o f nodes and branches.
18.1.1 Nodes
Figure 18.2 shows different schem atics o f nodes. N odes are im aginary points
added at locations in the netw ork where:
O
£
o
c3
£
ou
c
O
o
c
c
c.
'-I
2
co
JC
-C
CJ
c
r—3
i
CQ
<u
■g
Z
x
w
as
o
18.4
FLUID F L O W N E T W O R K S 18.5
o f nodes will increase the solution com plexity since each node will add an
equation and an unknown.
18.1.2 Branches
Branches represent pipes; one or more than one pipe may be included in a branch.
Figure 18.3 shows different schem atics o f branches with different pipe orienta
tions such as inclined branch or a branch with diam eter change. A branch has to
have beginning and ending nodes and a defined resistance. A pipe length, diam
eter, and internal roughness factor will define the pipe resistance factor K , as was
discussed in C hapters 8, 9, and 10. W hen flowing fluid is incom pressible, the pipe
diam eter is also required in order to predict the change in the fluid velocity and
thus the acceleration pressure drop value. A heat transfer m odel must be assumed
for the branch such as isotherm al, adiabatic, or non-adiabatic.
A network will include many other pieces o f flow related equipm ent such as
control valves, pumps, flow meters and pipes fittings such as elbow s, reducers etc.
All o f the netw ork flow -related equipm ent must be included in the netw ork m ath
em atical model to account for pressure drop. A lthough fittings will cause minor
losses in pressure drop, the m ass flow distribution in the netw ork is a sensitive
function o f pressure drop and inclusion o f all pressure-drop term s will increase the
solution accuracy o f the m ass-pressure calculation outputs.
♦— O — >---------0
Junction
Nodes
O
Demand
Mode
i
FIGURE 18.2 Nodes schematics for network analysis. The small arrow without a source or exit
node—no inlet or outlet circle—represents an external net flow rate, not a branch.
18.6 FLUID FLOW HANDBOOK
H r iz e n ta l
B ra n c h
In c lin e d
B ra n c h
B ra n c h w i t h Q — —— Q O- Q
change in
d ia m e te r c
T he follow ing variables and conventions will be used to represent the netw ork
variables in order to set up the netw ork m athem atical model:
Nodes. Each node has tw o variables, the node pressure and the node net external
m ass or volum e How rate. The net external flow rate should not be confused with
the node internal flow. The node net external flow rate is the flow rate that enters
or leaves the netw ork through the node. W hen the node net external flow rate is
equal to zero, there is still an internal flow in the node which is equal to the sum o f
all flow rates com ing from branches linked to the designated node. A negative sign
will be used to indicate a net outlet flow rate w hile a positive sign indicates a net
external inlet m ass flow rate. The volum etric or m ass flow rates may be used; how
ever, since pressure and tem perature are not constant and thus fluid density will
change in different netw ork sections, especially for com pressible fluids, it is rec
om m ended to use a m ass flew rate o r volum etric flew rates at standard conditions.
The node-pressure variable will be designated as: Pj, w here j is a node reference
index. The node mass flow rate is given the follow ing variable, Qj. A gain, j is the
node reference index. For non-isotherm al heat transfer m odel, a node tem perature
variable is also required. V ariable Tj w ill be used to represent the m ixed node
tem perature. V ariable 7}, w ill be used to represent the node net external flow rate
tem perature w hich may be different the actual node tem perature due to m ixing
w ith other inlet branches feed.
Branches. The branch inlet and outlet points will have the sam e pressure as the
branch source and outlet n odes' pressure. T herefore, no new variables have to be
FLUID F L O W N E T W O R K S 18.7
introduced. The branch inlet tem perature is the sam e as the source node tem pera
ture; how ever, w hen tw o branches mix at a node, the branch exit tem perature may
be different than the branch exit node tem perature. The pipe m ass flow rate will
be designated as cjj w here i is a pipe num ber index. The sym bol Tu and Tio will
be given respectively for the branch inlet and outlet tem peratures.
It is very im portant prior to setting up the netw ork system equations to give
each node and pipe in the netw ork a unique num ber. N odes as one group each
w ith a unique num ber and branches as another group with a unique branch num
ber for each branch in the branches category. Subscripts l and 2 will designate
the branch inlet and outlet respectively.
T he follow ing equations have to developed for each pipe and node in the system .
(18.1)
w here
T h e above equation should be developed for each node in the netw ork. The sign
o f Q j depends on the flow direction, positive sign for net inlet external node flow
rate and negative sign for net outlet external flow rate. T able 18.1 show s the above
equation for different nodes orientation.
A node tem perature is calculated based on the assum ption o f ideal m ass m ixing o f
all inlets including any external flow rate. A node tem perature equation is needed
w hen the heat transfer m odel is o th er than isotherm al; fo r isotherm al assum ption
this equation may be ignored. The follow ing equation m ay be used to calculate the
node tem perature
T. = (18.2)
18.8 FLUID F L O W H A N D B O O K
w here
TA B LE 18.1 Mass Balance Equation for Different N o d e and Pipe Flow Orientations
O
0
II
1
M
o -— >— o — *—
’l q2
<?i + -<?2 = ° 0
. X .
r ti2 ^
" I .
3
*3
c1)
Q\ + ~ 42 ~ 100 = 0.0
Q - -1 0 0
o 1 \ > 2n
r+ * r°
3
93
c>
II
1
©
0
o
O
K»
100
*3
\ ^ g Q
*2
FLUID F L O W N E T W O R K S 18.9
Table 18.2 show s node tem perature equation developed for different nodes-
branch orientations.
TABLE 18.2 Temperature Equation for Different N o d e and Pipe Flow Orientations
T v T io +
’ <?1 + <?3
V i °
nl . A . 2_^
3
c3
Q = 1 0 0 ;T=75
T + < liT u + 2,^75
r «2 ^ <7l + + <2|
93 3
c3
j- 9l^lo + ‘fo'Go
0 = -1 0 0
?1 + 4j
.1 "i .
t , =r„
where T u = 75 F
Q = 100 ; T -7 5 F
M — o
18.10 FLUID F L O W H A N D B O O K
The Bernoulli energy equation is developed for each branch. A com prehensive
form o f the branch Bernoulli equation is
P V2 P V2
-— + Z \ + - T - + lh > - hf - h r ----- L - Z, - = 0.0 ( 18.3)
PS pg 2g
where
Z Elevation
P A bsolute pressure
V Velocity
hp Pump head
hf Frictional head
hc Equipm ent (such as control valve, meter, etc) head
Subscripts I and 2 refer to the branch inlet and outlet points respectively. All
term s in the above equation m ust be consistent in units, i.e. all term s must be in
pressure or head units. Branch m inor losses, K h are included in the pipe frictional
pressure drop, which may be w ritten as
Equations for other flow -related equipm ent such as How m eters and control
valves should replace the he term. Since the flow -related equipm ent pressure drop
equations may be functions o f other unknow n netw ork param eters such as the
branch flow rate or the branch inlet pressure, careful review o f the overall num
ber o f equations and unknow ns is required prior to solving the netw ork system of
equations. Table 18.3 shows the resultant Bernoulli equation developed for selec
ted different types o f branches. O ne branch variable for each branch m om entum
equation must chosen. For analysis problem s, the branch diam eter is known and
in m ost cases, one will solve for the velocity, i.e. flow rate. For design problem s,
a specific flow rate per branch is required and one will solve for the pipe diam e
ter required to maintain the required distribution o f flow rates.
Thermal energy balances for fluid flow in pipes is discussed in detail in C hapter 27.
The branch inlet tem perature is alw ays set equal to the source node m ixed tem per
ature. The branch exit tem perature is calculated based on the heat model assum p
tion as discussed in C hapter 27.
FLUID F L O W N E T W O R K S 18.11
TABLE 18.3 T em perature Equation to r Different Node and Pipe Flow Orientations
H + Y L - (J L + y K ) ^ . A _ ^ = 0 .o
pg 2g D, 2g pg 2g
y K ) v l - p> =o.o
pg D, D2 ^ 1 2g pg
0 -------- — o Horizontal constant-diameter branch,
1 Li L2 2 compressible fluid.
1 Dj D2
---- o
4 2 f L + z i + hp - ( ^ Z ^ I) f - A - Z ! = 0.0
o -g p - Di pg ‘ 2g pg
1
(For compressible fluids, it is recommended to
add to a node at the pump discharge to isolate
the pump and then treat the remaining section
of the branch as case 1 of this table.
The fluid physical properties o f each branch will also be required; this m ainly
includes the density and viscosity. For incom pressible flow, the fluid density and
viscosity m ay be calculated using the branch average inlet and outlet tem perature
and pressure. Any equation o f state o f other correlations may be used to find the
fluid density and viscosity, see C hapter 2. For com pressible fluids, density m ust
be calculated at the inlet and outlet conditions separately in order to find the fluid
com pressibility factor and thus the true fluid velocity which will be used to cal
culate the acceleration pressure drop.
18.12 FLUID F L O W H A N D B O O K
W hen fluid is com pressible, i.e. gas or 2-phase flow , critical flow equations des
cribed in C hapters 9 and 11 m ust be added into the solution algorithm to set the
boundary for the m axim um flow rate allow ed for each branch.
Tw o-phase gas-liquid pressure drop correlations are dependant on the gas and liquid
m ass flow rates, therefore, the gas or liquid to total mass ratios have to be estim ated
for each branch. A conservative approach is to find the gas and liquid ratios at the
branch average tem perature and pressure w hich is the average o f inlet and outlet
branch conditions. O r a more robust approach is to divide the branch into sm aller
sections and perform T P flash at each section to have better representation o f the
change in gas-liquid ratios.
The m ass and energy equations for a fluid flow netw ork are nonlinear in nature.
The non-linearity o f the equations is clear in the frictional pressure drop term in
E quations 18.3 and 18.4. Solution o f non-linear equations m ay be accom plished
by m any m ethods and the topic is w ell docum ented by m any num erical analysis
books. The solution m ust yield a unique solution vector for a given set o f input
data. N ew ton-R aphson iterative procedure is one o f the m ost com m on m ethods to
solve a system o f non-linear equations. It is im portant that the num ber o f equations
be equal to the num ber o f netw ork unknow n param eters in order to be able to solve
the system equations. An under- or over-specified problem will result when the num
ber o f equations is not equal to num ber o f unknow n variables. Furtherm ore, the
solution may fail when contradiction between input param eters exists, such as when
the user sets a pipe flow to a certain am ount w hile the exit node net external flow
rate is set to 0.0. Logically that is an im possible case and the m athem atical solu
tion will yield a sparse m atrix. N ew ton-R aphson m ethod requires the setup o f a
Jacobian matrix which is constructed by perform ing a partial derivative o f all equa
tions with respect to each variable. Since the equations are non-linear, a num erical
derivation is the best technique to create the Jacobian m atrix. A com plete descrip
tion o f the N ew ton-R aphson m ethod and the num erical differentiation technique
is found in A ppendix A. N etw ork variable specifications will be discussed next
since it is usually a source o f confusion. Figure 18.4 show s a schem atic of a simple
netw ork that consists o f four nodes and three branches. T o sim plify, an isother
mal heat m odel will be assum ed, therefore, the branch energy and node tem pera
ture equations are ignored for this exam ple. N odes and branches are labeled as
/V,, yv2, W3, and Pj, P2, P3 respectively. Each node has tw o variables: pressure and
FLUID F L O W N E T W O R K S 18.13
Node 2
^2 Branch 2
Branch 1
02 q2,D2,L2
o
ql, Dl, LI
o ------------------ »“ o
Node 1 Node 3
PI P3
Q3
Q1
Branch 3
q3, D3,
L3
0
Node 4
P4
Q4
FIGURE 18.4 Common variables (nodes and branches) of a network.
flow rate, each branch has one variable: flow or diam eter. The system o f equations
for the netw ork show n in Figure 18.4 w ill consist o f a total o f seven equations:
four node m ass balance equations, ju st like Equation 18.1 and three branch
m om entum equations, Equation 18.3.
N odes m ass-balance equations (see Figure 18.3)
0 - 9 , = 0 .0 Node 1 (18.5)
where qt is the m ass flow rate (lbm /sec) o f pipe i. N otice that the m ass flow rates
for all o f the nodes have positive signs since the flow direction has not been set
yet. P ip es’ m om entum equations are developed fo r each branch as
18.14 FLUID FLOW HANDBOOK
Since the fluid is incom pressible, all nodes are at the sam e elevation level,
and no pum ps or other flow -related equipm ents exist in the netw ork, the above
equations are sim plified into (ignoring the velocity head to pressure change at the
diverging tee for sim plification)
p2 V^ P,
f '/L = 0.0 Branch 2
PH ( d 2-8 2 J PH
_ ' fL
- ] = 0.0 Branch 3
PH { D 2g, 3 PX
w here
For the system o f mass and momentum equations shown above, only seven vari
ables must be chosen to be the unknown param eters and all other network variables
m ust be specified. For each node, one variable m ust be specified, and for each
branch one variable must be specified. Depending on the nature o f the problem , rat
ing or designing, one may choose the follow ing variable specifications
No d e Pressure Flow
l Known Unknown
2 unknown k n o w n = 0 (or any inlet or outlet flow rate)
3 unknown k n o w n = - (outlet flow)
4 unknown k n o w n = - (outlet flow)
I unknown known
2 unknown known
3 unknown known
l Known Unknown
2 unknown k n o w n = 0 (or any inlet or outlet flow rate)
3 k nown u nknown
4 k nown u nknown
FLUID F L O W N E T W O R K S 18.15
1 unknown known
2 unknown known
3 unknown known
In the above specification, the boundary pressure o f all supply and dem and
nodes are set and is required to find the flow distribution in the branches in which
all are o f know n diam eters. O ther com binations o f variables may be used to solve
sim ilar netw ork rating problem s. A t least one node with a known pressure m ust
be defined in the netw ork specification. In netw ork design problem s, the supply
node pressure and the dem and n o d e’s flow rate are usually specified and one
requires the pipe diam eters to be found for a given branch velocity (flow rate).
Known Unknown
unknown k n o w n = 0 (or any inlet or outlet flow rate)
unknown k n o w n = - (outlet flow)
unknown k n o w n = - (outlet flow)
known unknown
known unknown
known unknown
T he solution algorithm for the netw ork m ass and energy system o f equations
is shown in Figure 18.4; the above netw ork will be used as an exam ple o f the
solution steps.
1 100(689.48) unknown
2 unknown 0.0
3 unknown -4.1667 (-1.8919)
4 unknown -4.1667 (-1.8919)
(18.8)
/ r\ ir\ , i a , x? \
100/ 2 .1 5 8 4 e - 4 0.019x300 { q j A, / p)~ p
= 0.0
62.4x32.174 3 /1 2 2x32.174 /l 62.4x32.174
Branch
■>\ p
Pi 0.018x200 ( q J A J p )
F = 3 = 0.0
62.4x32.174 2 /1 2 2x32.174 / 2 62.4x32.174
Branch
P, ( 0.018x400 ((q,
a , / AA,, //op))‘M P,
F, ------------- ±----------- = 0.0
62.4x32.174 2 /1 2 2x32.17 / 3 62.4x32.174
Branch
where
f4 = G , - * , = 0.0 Node 1
[y ^ [ A X ,] = -[/• ] (18.10)
where
J Jacobian matrix
X C orrection vector o f unknow n variables (q u q2, q 3, F 2, P 3> P4» anc* Q i)
F V ector o f energy and m ass functions (F, ... F7)
AP, =— (18.11)
9<y,
AP,
*5 . *6
dqt
AG, A
K
dqt
9F7 3F7 SF1 dF-, dF-j
dq, dq2 dq, 3P, 3P, 3P, 9(2,
1 8 .1 8 FLUID F L O W H A N D B O O K
3/* _ Fi (X + A X ) - F i (X )
dX. AX
where
X U nknow n variable
F N onlinear equation.
AX An increm ental change in V ariable X
A X = \% X
[AX,] = - f t ] [ / „ f (18-13)
where
O nce the correction vector, AX, is found, the unknow n variables are updated as
A converged solution o f the netw ork occurs w hen the sum o f squares o f the
mass and energy equation is less than a chosen tolerance value such as
If the above criterion is not satisfied, a new iteration shall resum e w ith the lat
est corrected set o f variables. T he above problem converges in a few iterations
The m ass and energy equations (E quations 18.9 and 18.9a) yield the follow
ing values w hen evaluated using the above values
.3044 0 0 -4 .9 1 6 7 E -0 4 0 0 0
0 -1.3458 0 4.9803E - 04 -4 .9 7 8 7 E -0 4 0 0
0 0 -3.646 4 .9 8 0 3 E -0 4 0 -4 .9 8 6 E -0 4 0
.999 0 0 0 0 0 .999
.999 -.9999 -1 0 0 0 0
0 1.000 0 0 0 0 0
0 0 1.00 0 0 0 0
The follow ing matrix equation is solved by G auss-Seidel for the unknow n cor
rection values
-.3044 0 0 -4 .9 1 6 7 E -0 4 0 0 0
0 -1.3458 0 4 .9 8 0 3 E -0 4 -4 .9 7 8 7 E -0 4 0 0
0 0 -3.646 4 .9 8 0 3 E -0 4 0 -4 .9 8 6 E -0 4 0
-.999 0 0 0 0 0 .999
.999 -.9 9 9 9 -1 0 0 0 0
0 1.000 0 0 0 0 0
0 0 1.00 0 0 0 0
The solution o f the above equation yields the follow ing values
A<7, 4.333
Aq2 0.6667
Aq, -0 .8 3 3 3
AP2 — 465350
A /5, 458991.7
ap4 452709.7
a <2,_ 2.3334
FLUID F L O W N E T W O R K S 18.21
<7. '8.333
4.1667
4.1667
- 465440
P,
P, 459076
P-i 452779.7
8.3334
Q i.
Equations 18.9 and 18.9a are reevaluated w ith the latest unknow n variables
values to find the sum o f squares w hich will be checked for convergence versus
a given tolerance, say 10'5
K -3 .6 5 1 1 3
-0 .1 5 2 2 8
6 .6 8 3 2 ^ - 3
— -1 .9 0 7 * - 6
Fs 2.0504^ - 5
-4 .8 8 2 ^ - 6
Fi_ 2 .4 9 7 8 ^ - 6
A sm all absolute value, say less than 0.001, o f the equation is an indication
o f convergence o f that particular equation, not the system o f equations. The sum
o f squares o f all equations has to be less than the criterion to assum e problem has
converged, the sum o f the equations squares is
which is not less than the convergence criterion o f 1.0 X 1O'5, therefore, a new
solution iteration m ust be resum ed.
Iteration 2. F ollow ing the steps outlined in iteration 1, the follow ing values are
obtained for the unknow n variables:
18.22 FLUID F L O W H A N D B O O K
"8.333
<?2 4.1667
4.1667
P2 - 458108
P, 451439
P* 445461.5
g ._ 8.3334
R eevaluating the m ass and energy equations w ith the above set o f variables
yields a satisfactory sum o f squares, 1.19e'6, and it can be assum ed that the problem
has converged.
T he units for the pressure variable in the above solution set are in lbm /ft-sec2,
w hich m ay be converted to psia and kpaa as
EXAMPLE 18.2
F o r the network structure described above in Example 18.1, let us assume that
the fo llo w in g changes in branches and nodes specifications are made
T he follow ing boundary (nodes) specification w ill be used to solve a netw ork
rating problem
A tank is placed at each o f nodes 1, 3, and 4. The pressure at each tank will be
atm ospheric. The flow driving force will be the elevation change between nodes l
and 3 and nodes l and 4. The Bernoulli equation for each is developed as
P R (18.16)
—- + Z, —(h f ) ] ----- —- Z 2 = 0 .0 Branch 1
pg Pg
— + Z 2 - (/ i, ) 2 - — - Z 3 = 0.0 Branch 2
Pg Pg
Substituting know n param eters into the above branch equations yields
2 \
F _ 1 4 .7 /2 .1 5 8 4 ^ - 4 | ( 0 .0 1 9 * 3 0 0 0 (q x / A, / p )
(18.17)
62.4*32.174 ^ 3 /1 2 2*32.174
Pi
- 5 = 0.0 Branch 1
62.4*32.174
P2 f 0.0 1 8 * 2 0 0 0 (q2 / A 2 / p ) 2 \ _ 1 4 .7 /2 .1 5 8 ^ - 4
Fl ~~ 62.4.v32.174 +' [ 2/12 2 x32.174 J 2 62.4x32.174
The follow ing is the list o f unknow n variables; q h q2, <73, Pi, Q \, Q3, and 04.
18.24 FLUID F L O W H A N D B O O K
Q2, which is the net external flow at node 2 is 0.0 since no flow leaves or enters
the network at this node. A sum m ary o f the iterative solution is shown below assum
ing the following initial values for the unknow n variables
<2, 12
a . 15 _
Iteration l
Aqt '-6.29128 ' "8.70872 "
Aq2 -3.53416 <?2 4.46584
Aqy -3.75711 4.24289
AP2 = 85770.86 y ^2 = 85870.86 ; 2 =3149.8
AG, -6.29128 a 8.70872
A ft -16.46599 a -4.4659
AQ4 -19.24294 -4.2429
a
Iteration 2
A?, -3.00895 '5.69977 '
A<?2 -1.35455 Qi 3.11129
A<?3 -1.65443 2.58846
a p2 = 564.5428 » Pi = 86435.4 I / - 2 = 115.5
AQ, -3.00895 Q\ 5.69977
aq3 1.354684 a -3.11129
Aq 4 1.654489 a -2.58846
Iteration 3
Aqx ' -0.84095 <7, '4.8588 '
A^2 -0.30199 <h 2.80929
A^3 -0.53894 % 2.04951
AP2 = 24.23266 * 86459.4 X / - 2 =1.21
P2
AG, -0.84095 Qs 4.85881
ag3 0.302006 a -2.8093
ag4 0.538939 a -2.0495
FLUID F L O W N E T W O R K S 18.25
Iteration 4
Ac/, - 9 .0 8 8 ^ - 2 <7i "4.76793"
Aq2 -\J 3 2 e -2 <7: 2.79197
A<7, -1355e-2 ft 1.97596
A ft 10.07846 ft — 86469.7 ; £ / - 2 = 6 .3 2 e - 4
AG, - 9 .0 8 8 ^ - 2 ft 4.76793
A<2, 1 .7 3 2 6 ^ -2 ft -2 .7 9 1 9
A& 7 .3 5 6 0 ^ - 2 -1 .9 7 5 9
.ft.
Iteration 5
Aqt - 1 .8 5 0 ^ - 3 4.76608
A q2 -1 .2 1 4 ^ - 4 2.79185
A q3 - 1 .7 2 8 ^ - 3 1.97423
A P2 = 0.3849515 •> ft = 86470.1
a q, - 1 .8 5 0 ^ - 3 ft 4.76608
Af t 1 .2 1 4 2 ^ -4 G, -2 .7 9 1 8
1 .7 2 8 6 6 ^ -3 -1 .9 7 4 2
JP
>
.ft.
i
Since the sum o f squares is less than 1.0e-5, the problem is assum ed to be
converged w ithin the specified accuracy tolerance, the solution vector in appro
priate units becom es
EXAMPLE 18.3
To examine the effect o f m inor losses on flo w distribution, a flo w resistance ele
ment w ill be added to branch 2. A 0.5 inch flo w meter (square-edge orifice) is
placed at the entrance o f branch number 2. Resolve Example 18.2 with a 0.5
inch flo w meter placed on branch 2.
Solution
The flow m eter pressure drop is a function o f the m eter diam eter and fluid flow
rate as
18.26 FLUID F L O W H A N D B O O K
M° (0 .0 4 3 8 d a2 C f p j (l8 2 0 )
w here
( 1 8 -2 1 )
T he energy and m ass equations and unknow n netw ork variables w ill be the
sam e as in Exam ple 18.2 except for the branch 2 m om entum equation. The ori
fice head loss m ust be inserted into the equation as
(18.22)
p p
— + Z, - (hf ) , ----- —- Z2 = 0.0 Branch 1
Pg Pg
P P <?2
— + Z2 - (hf )2 ---- —- Z3 - - 0.0 Branch 2
PS pg 0.0438 d 2
oCfp j
P, „ ... PA
— + Z2 - (hf )3 ----- —- Z 4 = 0.0 Branch 3
Pg pg
Since Cd and d0 are know n, the flow coefficient is calculated using Equation 18.
0 595
Cf = - , — = = 1.0577 (18.23)
V (l- 0 .5 /2 ) 4
Substituting the flow coefficient and other know n param eter into the branch
equations and then sim plifying the equations yields
(18.24)
= 43.923 - 0.037899 q \ - 0.000498P2 = 0.0 Branch 1
The nodes' mass balances are the same as those in the previous example. Solv
ing the m ass and energy equations by N-R w ith the initial guess provided in Exam
ple 18.2 yields the follow ing iterative solution
Iteration I
Aq{ -6 .5 1 1 8 6 “ 8.48814
Aq2 -3.75899 4.24100
Aq} -3.75285 <h 4.24714
AP2 = 86279.12 * p2 = 86379.1 = 24.6*+ 6
AQ, -6.51186 Qi 8.48814
AQ, -16.2411 Gj -4.2411
-19.2472 g4 -4.2472
AC4
Iteration 2
Aq, -3.30446 9, "5.18367"
Aq2 -1.65667 <?2 2.58433
Aq3 -1.64783 <?3 2.59932
AP, = 627.1425 ^2 -
87006.2 ^ / r 2 = 212.56
Iteration 3
Aqx -1.07502 " 9, 4.10865"
Aq2 -0.54188 <72 2.04245
Aq3 -0.53311 ?3 2.06620
AP2 = 3 .4 8 8 * -2 * P2 = 87006.3 I / - 2 =2-51
AC?, -1.07502 G. 4.10865
A ft 0.54188 ft -2.0424
Iteration 4
Aqx -0.146699 <7. "3.96195"
Aq2 - 7 .4 9 3 * - 2 <?2 1.96752
Aq3 - 7 .1 7 6 * - 2 93 1.99444
AP2 = 1 .5 9 4 * -2 y A = 87006.3 = 1.39«?-3
A(2, -0.146699 G, 3.96195
AG, 7 .4 9 3 * -2 g3 -1.9675
7 .1 7 6 * -2 -1.9944
a g 4. 04_
18.28 • I.I ID F L O W H A N D B O O K
Iteration 5
Aq, -3 .4 6 1 9 * -3 "3.95849
Aq2 -1 8 0 7 1 5 * -3 <72 1.96571
Aq} -1 .6 5 4 7 * -3 <?3 1.99278
AP2 - 1.70353*-3 P2 = 87006.3
A Q, -3 .4 6 1 9 * -3 Q\ 3.95849
AG, 1.80715*-3 a -1.9657
A Q4 1.65466*-3 -1.9928
By com paring the solution vectors o f exam ples 18.2 and 18.3 one may notice
that the total flow rate in Exam ple 18.3 is about 16 percent less than the total flow
in Exam ple 18.2,3.958 lbm/sec (1.797 kg/sec) versus 4.766 lbm/sec (2.355 kg/sec).
T his due to the addition o f m ore resistance (O rifice) w hile the flow driving force
(elevation differences in both exam ples) is constant.
Branch 2: 2.5 lbm /sec (1.135 kg/sec) (E xam ple 18.2: 2.7918 lbm /sec,
1.267 kg/sec)
B ranch 3: 1.75 lbm /sec (0.7946 kg/sec) (E xam ple 18.2: 1.974 lbm /sec,
0.896 kg/sec)
Find the orifice diam eters (on braches 2 and 3) required to control the flow
at the desired set points.
Solution
In this exam ple the flow rates in branches 2 and 3 are set w hile the orifice diam
eter, do2 and dQ3, are unknow n. The vector o f unknow n variables will become: Q {,
4»2> Q\i 035 Q a '
FLUID F L O W N E T W O R K S 18.29
Flow rates in branches 2 and 3 are gives as 2.5 and 1.75 lbm/sec respectively.
The branches’ Bernoulli equation becom es
Branch
V
P , 7 , _ A _ 7 = 0.0 Branch 2
pg 2 f pg 3 1.0.0438 4 C /P
= 0.0 Branch 3
~pg+ Z 2 ~ { h f ) 3 ~ M ~ Z * " (0 .0 4 3 8 4 C /p J
(18.25)
The flow coefficients, Cf, in the Bernoulli equation for branches 2 and 3 are
function o f the orifice diam eter, but since the orifice diam eters are unknown vari
ables, the flow coefficients m ust be recalculated in each iteration using the latest
predicted orifice diam eters. Rearranging and simplifying the above equations gives
2.5
F2 = 0.000498/*2 - 4 0 .2 6 - = 0.0 Branch 2
0.0438 d ;2Cf2(62.4)
1.75 V
F3 = 0 .000498 P2 - 4 0 .0 3 4 - 1 = 0.0 Branch 3
0.0438 4 C / 3(62.4)
(18.26)
F4 = Q \ - q x = 0.0 Node 1
F5 = q x- 2.5 - 1.75 + 0 = 0.0 Node 2
(18.27)
F6 = 2.5 + a = 0 .0 Node 3
Fj = 1.75 + 0 4 = 0 . 0 Node 4
The above seven equations F\ ... F-j contain 9 unknow ns: q\, do2, da3, P 2, Q\,
Q i, Qa» C / 2, and Cy3. T herefore, tw o m ore equations m ust be added to be able to
solve the system o f nonlinear equations. Equation 18.21 w hich relates the Cf to
d(> is developed for each orifice and w ill be used to update the flow coefficients
as a function o f the orifice diam eter in each o f the solution iterations.
18.30 FLUID F L O W H A N D B O O K
Solving the above system o f equations by N ew ton-Raphson and using the initial
unknowns variable values provided in Exam ple 18.2 (use do2 = 0.4, do3 = 0.45) yields
Iteration 1
1 .9 1 7 4 0 E -0 2 doi .4 6 9 1 7 4
A P2 = 96044.5 » Pi — 9 6 1 4 4 .5 = 115 .6 1 9
a e, -10.7500 Q\ 4 .2 4 9 9 7
A Qy -14.5000 G3 - 2 .5 0 0 0
A G4 -16.7500 - 1 .7 5 0 0
Q *.
Iteration 2
Iteration 3
A<7, 0.0 <7, "4.25
A^„2 6 .2 4 9 4 E -0 2 4,2 .604863
A<,3 2 .8 5 1 3 E -0 2 4,3 .554024
AP2 = -7 .7 1 8 4 E -0 3 » ^2 = 86824.2 Y F , 2 = 1.469
AG, 0.0 G, 4.25
AG3 0.0 G, -2.5
AG4 _ 0.0 -1.75
&
Iteration 4
00
t/5
—a.
E
c3
X
UJ
M
o
£
e
c3
*
OX)
c
oc
o
E
<u
M
0oD
VO
00
2
o
18.31
18.32 FLUID F L O W H A N D B O O K
Iteration 5
A ft 0.0 q3 -2.5
aq4 0.0 g4. -1.75
In this exam ple, different netw ork setting will be used to illustrate the net
w ork concepts. The follow ing assum ption will be used throughout the entire net
w ork analysis.
• Pipe inside wall roughness is 0.0018 (0.045 m m ) inches and applies to all
pipes.
• Fluid tem perature is constant at 70°F (21.1°C). T his assum ption is ideal and
is used to sim plify the solution process. Liquid viscosity is alos assum ed con
stant and equals to 1.0 cP. The therm al balances o f each pipe and node will
be ignored in this exam ple.
• Since a constant density is assum ed, the pressure drop due to acceleration
effects (velocity change) drops from all o f the branches’ B ernoulli equations.
A lso, the pressure gain due to velocity drop at each o f the m anifold tees is
ignored since it is m inor relative to the frictional pressure drop.
FLUID F L O W N E T W O R K S 18.33
The follow ing data describe the pipes and nodes specifications:
Pump Data (Flow-Head Curve). Both pum ps have the follow ing flow-head curve:
0 (0.0) 50(15.24)
25 (5.6777) 47(14.325)
50(11.355) 44(13.411)
75 (17.033) 41 (12.497)
100(22.711) 38 (11.582)
125 (28.388) 32 (9.754)
In o rd er to be able to include the above flow -head curve into the m athem ati
cal m odel, a polynom ial fit o f the above data yields
where
hp Pum p head in ft
Q Flow in GPM
T he follow ing equation will be used to predict the pressure drop associated
with the control valve
(18.29)
where
Since the netw ork fluid is w ater at the standard conditions, the term p/6 2 .4
m ay be dropped from Equation 18.28. A sw ing check valve w ith resistance co ef
ficient o f 0.9 is added at the exit o f each pum p.
H eat Exchangers. The following heat exchanger resistance coefficients are assumed
M athem atical Model. The B ernoulli equation is developed for each pipe in the
netw ork, the general form at o f the B ernoulli equation, ignoring velocity head is
-£ - + Z, + hP - h f - h e - ^ - - Z 2 = 0 .0 (18.30)
P8 P8
w here
The pum p head, hp, is described in E quation 18 and w ill be included only for
pipes 3 and 8. The he term represents the head loss o f flow -related equipm ent
such as control valves or flow m eters. T he he term w ill appear in one equation
only, the B ernoulli equation o f pipe num ber 16. The h f is frictional head loss due
18.36 FLUID F L O W H A N D B O O K
pipe and other m iscellaneous fittings. W e will include the heat exchangers’ K fac
tors in the follow ing general-form D arcy equation as
hf = ^ + ' L K i + K E * c )^ ; (18-31)
where
The C olebrook-W hite equation will be used to estim ate the friction factor. A
com plete docum entation o f the equation is found in C hapter 9. The pipes’ (pipe 1-
pipe 31) energy equations are listed below (units for pressure variables are in
lbm /ft-sec2, and units for pipe and node flow rates are in lbm /sec)
FLUID F L O W N E T W O R K S
5 l + 5.0 J J k ) W A L r i L_ ^ - 5 . 0 = 0 .0
PS ' voj,, 2g pg
^ 1 + 5 () - f — 1 ^ I A I P& P" - 5 . 0 = 0 .0
pg I d J 12 2g pg
fu + 5 0 J £ l) i q l A I P)“ - 5 . 0 = 0 .0
pg \ D ) l4 2g pg
^ i + 5.o M A L r ik - 5 . 0 = 0 .0
pg ' v D ),5 2g pg
I k + 5 .0 - A P J £ ) m i A - 5 l _ 0 .0 = 0 .0
pg \ d j i6 2g Pg
APcv units m ust be in feet.
fL\ (,q / A I p ) 2
n
+ 0.0 - [ - 0.0 = 0.0
pg D J l7 2g pg
5 l + 0.0 - ( / L ’j {g ! A l p ) ] ,
- 0 . 0 = 0 .0
ps D JIS 2g pg
,/M i q / A I p) ] v Px
+ 0.0 - ( - - 0 . 0 = 0 .0
PS D 2g pg
p /Z A ( q / A I p) ] Q £
20 + 0.0 - i - - 0 .0 = 0 .0
PS D / 20 2g py»
fL\ ( q / A I p )2
lt P»
+ 0.0 - I - - 0 . 0 = 0 .0
Pg D J 2i 2g pg'I
7M (q I A l p )\3 ^
! k + 0.0 - ( L - 5 . 0 = 0 .0
pg D J 23 2g pf.
fL \ (q /A I p ) j P,
Z k + 5.0 - ( 5- - 10.0 = 0 .0
pg . D J 24 2g pt7
p* ( f l A (q / A I p ) 25 1
+ 10.0 - J - - 1 0 .0 = 0 .0
pg V D J 25 2g £
18.38 FLUID F L O W H A N D B O O K
— + 5.0
Pg
exch'26 ( q / A2g/ p ) i P22
Pg
- 0.0 = 0.0
p
— + 5.0
Pg
— +K
,D
exch27 ( q l A2g/ p ) 27 P2I
Pg
Pg
+ 5.0
D
+K exch'28 ( q / A2*l p ) '8 Pi,
Pg
P
— + 5.0
Pg .D
exch'29 (-q / A2 g/ p ) „2 P»
Pg
A lp)]o
— + 5.0 - I
Pg
Ik + K
D
exch'30 Pi*
Pg
— + 5.0 - ■ £ \ ( ! / A l p ) l L _ A _ 0 .0 = 0 .0
Pg pg
T he follow ing list o f mass balance equations are developed for each node in
the netw ork:
N ode I Q l+ q 2 5 -q 1=0.0
N ode 2 ql-q 2 -q29= 0.0
N ode 3 q2-q3=0.0
N ode 4 q3-q4=0.0
N ode 5 q4-q5=0.0
N ode 6 q29-q6=0.0
N ode 7 q6-q7=0.0
N ode 8 q7-q8=0.0
N ode 9: q8-q9=0.0
N ode 10 q5+q9-q 10=0.0
N ode 11 q lO -q l l-q26= 0.0
N ode 12 q l l-q l2 -q 2 7 = 0 .0
N ode 13 q l2 -q l3 -q 2 8 = 0 .0
N ode 14 q l3 -q l4 -q 3 0 = 0 .0
N ode 15 q l4 -q l5 -q 3 0 = 0 .0
N ode 16 q l5 -q 16=0.0
N ode 17 q l6 -q 17=0.0
N ode 18 q l7 + q 3 1 -q 18=0.0
FLUID F L O W N E T W O R K S 1 8 .3 9
N ode 19 q l8 + q 3 0 -q 19=0.0
N ode 20 q 19+ q28-q20= 0.0
N ode 21 q 2 0 + q 2 7 -q 2 1=0.0
N ode 22 q 2 1+ q26-q22= 0.0
N ode 23 q22-q23= 0.0
N ode 24 q23-q24= 0.0
N ode 25 q24-q25=0.0
Solution
The above system o f equations is solved by N ew ton-R aphson with the follow ing
initial guess o f unknow n variables.
Iteration l
q l = 41.81747; q2= 19.78049; q3= 19.78049; q4= 19.78049; q5= 22.03703;
q6= 22.03703; q7= 22.03703; q8= 22.03703; q9= 19.78049; q l0 =
41.81747; q l 1= 35.05716; q l2 = 28.38701; q l3 = 21.77134; q l4 = 15.27678;
q l5 = 8.856343; q l6 = 8.856345; q l7 = 8.856345; q l8 = 15.27678; q l9 =
21.77135; q20= 28.38701; q21= 35.05716; q22= 41.81748; q23= 41.81748;
q24= 41.81748; q25= 41.81747; q26= 6.760357; q27= 6.670188; q28=
6.615705; q29= 22.03703; q30= 6.494607; q31= 6.420472
Q l = 1.955032E-05
P2= 18.60294; P3=16.35285; P4=36.89974; P5=34.14939; P6=18.0568;
P 7= 16.69144; P8=16.00876; P9=36.39816; P10=33.55726; PI 1=33.35252;
P I 2= 32.85115; P 1 3 = 3 2 .4 6 1 14; P 1 4 = 3 2 .18157; P15=32.01042;
P I 6=31.94646; P 17= 27.78898; P18=27.72502; P19=27.55387;
P 20= 27.27431; P 2 1=26.88429; P22=26.38293; P23=23.72129;
P24= 20.53936; P25= 16.94795
Sum o f the square o f errors =71.7555119
Iteration 2
q l= 34.01856; q2= 16.31734; q3= 16.31734; q4= 16.31734; q5= 17.70121;
q6= 17.70121; q7= 17.70121; q8= 17.70121; q9= 16.31734; q l0 =
18.40 FLUID F L O W H A N D B O O K
Iteration 3
q I = 32.91012; q2= 15.84913; q3 = 15.84913; q4= 15.84913; q5= 17.06098;
q6= 17.06098; q7= 17.06098; q8= 17.06098; q9= 15.84913; q l0 =
32.91012; q l 1= 27.49971; q 12= 22.33837; q l3 = 17.31535; q 14= 12.49582;
q 15= 7.790068; q 16= 7.790068; q 17= 7.790068; q 18= 12.49582; q l9 =
17.31535; q20= 22.33837; q21= 27.49971; q22= 32.91012; q23= 32.91012;
q24= 32.91012; q25= 32.91012; q26= 5.410407; q27= 5.161348; q28=
5.023011; q29= 17.06098; q30= 4.819538; q31= 4.705747
Q l =6.908087 E-06
P2= 18.48596; P3= 16.65831; P4=37.37854; P5=34.73935; P6= 18.04285;
P7= 16.93508; P8=16.3812; P9=36.9705; P10= 34.2584; PI 1=33.99515;
P I 2=33.43621; P13=33.0607; P14= 32.82923; P15=32.7039; P I 6=32.65172;
P17= 29.83278; P 18=29.78059; P I 9= 29.65526; P20=29.4238;
P 2 1=29.04828; P22=28.48935; P23=25.06713; P24=21.59267;
P25= 17.59172
Sum o f the square o f errors = 9 .0 5 1 3 7 IE -04
Iteration 4
q l = 32.8569; q2= 15.82692; q3= 15.82692; q4= 15.82692; q5= 17.02998;
q6= 17.02998; q7= 17.02998; q8= 17.02998; q9= 15.82692; q l0 = 32.8569;
q l 1= 27.45169; q 12= 22.29775; q l3 = 17.28394; q l4 = 12.47548; q l5 =
7.781941; q l6 = 7.781941; q l7 = 7.781941; q 18= 12.47548; q l9 = 17.28394;
q20= 22.29775; q21= 27.45169; q22= 32.8569; q23= 32.8569; q24=
32.8569; q25= 32.8569; q26= 5.405214; q27= 5.153936; q28= 5.013814;
q 29= 17.02998; q30= 4.808456; q31= 4.693544
Q 1= -7.993685E -08
P2= 18.48562; P3=16.65719; P4=37.37887; P5=34.73948; P6=18.04223;
FLUID F L O W N E T W O R K S 18.41
EXAMPLE 18.6
Resolve the cooling water network described in Example 18.5 f o r the mass and
pressure distribution when pipe number 26 is blocked due to heat exchanger
maintenance.
Solution
A pipe m ay be blocked by m eans o f increasing the flow resistance; this may be
accom plished by setting the heat exchanger resistance to a very high value such
as 106, or m aking the pipe diam eter very sm all, 0.1 in. The system equations are
the same as described in Exam ple 18.5. A nother m ethod is to delete pipe num ber
26 and therefore delete the B ernoulli equation for pipe 26 and solve the rem ain
ing equations. Both m ethods should yield very close results. To solve this exam
ple, the K value for heat exchanger on pipe 26 will be changed to 106.
The netw ork solution converges in 10 iterations. The converged solution o f
the flow and pressure distribution is show n in Figure 18.6b and listed below for
the last iteration only.
Iteration 10
q l= 30.5893, q2= 14.73407, q3= 14.73407, q4= 14.73407, q5= 15.85523,
q6= 15.85523, q7= 15.85523, q8 = 15.85523, q9= 14.73407, q l0 = 30.5893,
q l 1= 30.53492, q l2 = 24.80668, q 13= 19.2315, q l4 = 13.8825, q l5 =
8.659776, q l6 = 8.659776, q l7 = 8.659776, q l8 = 13.8825, q l9 = 19.2315,
q20= 24.80668, q21= 30.53492, q22= 30.5893, q23= 30.5893, q24=
30.5893, q25= 30.5893, q26= 5.437426E -02, q27= 5.728241, q28=
5.575181, q29= 15.85523, q30= 5.349005, q31= 5.22272
Q 1-1.041907E -08
P 2= l 8.55329; P3=16.9601; P 4= 37.79268; P5=35.21519; P6=18.16698;
P7=17.20122; P8=16.71834; P9=37.43703; P10=34.79593; P I 1=34.56635;
P 12=33.87996; P13=33.41909; P 14=33.13531; P15=32.98196;
P 16=32.91836; P17=28.9253; P I 8=28.8617; P19=28.70835; P20=28.42457;
P 2 1=27.9637; P22=27.27731; P 23= 24.29278; P24=20.98665;
P25= 17.22137
Sum o f the square o f errors = 3 .8 4 2 4 1 8 12E-06
OJ
O.
'a.
2c/5
O.
<u
.5
cID
£
O-
K
3
5 /3
C /5
6O.
C /5
<U
Ic
T)
oo
JU
'S.
E
cxa
UJ
c
.2
s
X>
•S
C /5
‘■a
£
0
G
TCJ
CTj
232
C /5
C/5
1
CL,
S
flows in GPM).
VO
00
o
u*
18.42
<u
Q.
'a.
eg
’ s /5
Q.
C
<o
a
£
r3
O.
1i—)
3/J
is.
O
uJ_
a.
!/)
(U
-a
o
sO
00
JD
a.
E
r3
X
UJ
c
o
X)
3
•a
*
o
c
-a
a
a
£>
3
s /5
C/3
g
CL
22 £
w
a
o £o
u. c
18.43
18.44 FLUID F L O W H A N D B O O K
The flow rate in pipe 26 has been reduced from 5.873349 to 5.4387E-02
lbm /sec due to the pipe blockage. Further decrease in flow rate may be obtained
by increasing the K value to 107. The change in the netw ork flow and pressure
distribution due to blockage may be com pared in Figures 18.6a and 18.6b.
Solution
The B ernoulli equation for pipes 3 and 8 in Exam ple 18.5 are m odified into the
follow ing equations.
(18.33)
All o f other m ass and energy equations are the sam e as described in Exam ple
18.5. The netw ork solution converges in iteration, the result o f the last iteration is
Iteration 7
EXAMPLE 18.8
The network schematic which is shown in Figure 18.4 w ill be used to illustrate the
solution step o f a compressible gas (air) network. The follow ing data w ill be used:
Elevation Pressure Flow
Node ft (m) psia (kPaa) lbm/sec (kg/sec)
Diameter Length
Pipe in (mm) ft (m) Friction factor
A constant friction factor for each pipe is assum ed. Also, isotherm al condition
is assum ed. Any equation o f state m ay be used to estim ate the gas com pressibility
factor and thus the gas density. The Papay’s equation is used in this exam ple to r its
sim plicity, see C hapter 2 for a full docum entation o f the Papay’s equation. D en
sity will change as the fluid pressure changes, therefore, density calculation for each
pipe and node m ust be repeated fo r each iteration. A sim plified approach may be
considered by assum ing a constant gas density for each pipe which is estim ated at
the average pipe inlet and outlet conditions. M ore rigorous approach is to integrate
the pipe frictional head using an increm ental m ethod such as those developed in
C hapter 9. The sim plified m ethod is considered here, the pipes’ energy equations,
in lbm /ft-sec2 units, are
VIin M , ) 2 I p img P2 V2
F = _ _1 ou± Xgp2 = 0 .0
2g D 2g
Branch 1
Branch 2
Branch 3
(18.34)
18.46 FLUID F L O W H A N D B O O K
where
Q 1 —q 1 = 0.0 N ode 1
q l - q2 - q3 = 0.0 N ode 2
(18.35)
q2 — 1 5 = 0.0 N ode 3
q3 - 12 = 0.0 N ode 4
The solution o f the above system o f equations with the follow ing variable ini
tial values
q l = 25; q2 = 15; q3 = 18; Q l = 3 0
P2 = 405000; P3 = 392000; P4 = 400000
Iteration No 1
Correction V ector = [2,-8.448E-07,-5.999,- 3.000, 47647.54, 14766.94,36528]
C orrected V ariables: q l= 27; q2=15; q3=12.0; Q l= 2 6 .9 9 9 9 9 ; P2=452647.5;
P3=406766.9; P4=436528
Sum o f Square errors = 117357764.82
Iteration N o 2
C orrection V ector = [5.5144E-08, 9.6615E -07, -9.8914E -06, 7.679E-06,
112.195, 236.67 5 ,-1 1 5 7 9 .0 3 ]
C orrected V ariables: q l= 2 7 ; q2=15; q3=12; Q l= 2 7 ; P2=452759.7;
P3=407003.6; P4=424949
Sum o f S quare errors = 651.575444
Iteration N o 3
C orrection V ector = 12.774E-10, 2 .8 4 4 E -1 1, -1 .8 1 0 1 E -1 1, 2.7842E-10,
6.227E-02, .262216, -27.5974]
C orrected V ariables = q l= 2 7 ; q2=15;q3=12; Q l = 27; P2=452759.8;
P3=407003.9; P4=424921.4
Sum o f Square errors = 5.3717E-04
FLUID F L O W N E T W O R K S 18.47
REFERENCES
19.1
19.2 FLUID F L O W H A N D B O O K
Subscripts
19.1 INTRODUCTION
In this chapter we give the basic relationships for open-channel flow and illustrate
their application in sam ple problem s. A discussion o f the physics o f open-channel
flow and derivations o f the relationships presented can be found in some textbooks
on open-channel flow , e.g. by V.T. C how (1959), H enderson (1966) and C haudhry
(1993), o r in textbooks on Fluid M echanics, e.g. by R. O lson (1966), V ennard
and Street (1982), and Streeter, W ylie and B edford (1998). Som e softw are for
open-channel flow com putations is also recom m ended. The choice o f relationships
F L O W IN O P E N C H A N N E L S 19.3
Param eters to define channel geom etry include channel shape, cross-sectional
area (A), w etted perim eter (P) and hydraulic radius (R), top width (B), and hydraulic
depth (D). They are given in T able 1 9 .1 for com m on geom etric channel shapes.
Side slope o f a channel (z, z h z2 in T able 19 .1) is given as the horizontal increase
w ith a unit vertical increase. E xcept for rectangular channels, hydraulic depth is
not equal to the w ater depth y (T able 19.1). H ydraulic depth or m ean depth (A/B)
is therefore used and com puted by dividing the cross-sectional area (A) by the
w idth o f the free surface (i.e. w idth B). W etted perim eter (P) is the total length of
the channel boundary at a section w etted by the flow ing liquid, and hydraulic
radius (R) is determ ined by dividing channel cross-sectional area (A) by w etted
perim eter (P). For very w ide natural rivers (B » D ), hydraulic radius is approx
im ately equal to mean depth D.
EXAMPLE 19.1
Solution
1. W hen the w ater depth y = 5 ft (Figure 19.2), there is w ater flow ing in the
main channel (indicated by subscript m) only. The top w idth o f the trape
zoidal channel with equal side slopes (4 vertical versus 1 horizontal) is =
102.5 ft (100 + 2 X 5/4). The hydraulic radius is com puted as follow s
using equations in Table 19.1:
F IG U R E 19.2 Geom etry for a com posite channel used for Exam ples 19.1 and 19.2.
F L O W IN O P E N C H A N N E L S 19.5
2. W hen the w ater depth y = 15 ft (4.57 m), there is w ater flow ing in both the
main channel and the left overbank area (subscript lb). The left overbank
area is a rectangular channel (5 by 200).
= 125.77 f t (38.34 m)
R = u1553
j j . u125
j = 1 I 3 5 f i (3 J e m )
125.77
2553 125
/? = — = 7.718 f t (2.35 m)
330.77
Note that the hydraulic radius fo r the w hole channel is not equal to the sum
o f hydraulic radii o f the m ain channel and the left overbank area.
The flow in an open channel is uniform if the depth o f flow does not vary along the
length o f the channel. The average velocity in a cross-section remains constant. The
volum etric flow rate (discharge in ft3/sec or m 3/sec) in a uniform open-channel
tV
■5 +
D.
r^7 + +
a CN
tv"
side slopes of horizontal versus vertical; dQis diam eter; r is radius for rounded comers o f
k.
& »n c CN
CN
& CN o + I/3 I +
+ + + -CJ >» I c CN
-Cl >o <£> c->
o c3
CD tV
+
.c
•3 ?*s
tV I
£ + <3^ +
u
o
Cn [3 .
n7 tV o (2 k.
I
‘t: N +
3
00 CN
+ t-7 ■i 7 ?
+ K /.
X3
N
CN -CJ Kf co cn (N
+ r}- CN
CN
T3 +
CQ
C*S
+
+ *^3° •<5
z' " -\ (S + +
rectangle and triangle. Angle 0 = 2 c o s-1 [( d 0 - 2 y ) I d Q\ , which ranges from 0 to 360 degrees.
?*s
c 00 CN
t CN
& + ’53
N
+
X £ + 1 CN (N
ts | CN
<5 CO
Geometric Properties of Different Channel Sections
tv
CN
<5 +
-Cl
I
z, and z2are
+
&
■8
v*
rs IV
CN
£ + N| 03
+
CN 00
5 +
rH
width; z,
+ «o
CN o 03 03 I N?
?s
r__
Symbol definition: y is depth; b is bottom
k.
CN
tv + i
+ •Q O
u
tv" tv +
rN X 1
?s + <& *N
I w-> i-7 c
*-*
>*
d <N CN rV.| N,
N
+ I >> |
1
+ m 03
m ^ | <N
e ft d v_ s CQ rf
TABLE 19.1
19.6
F L O W IN O P E N C H A N N E L S 19.7
flow can be determined from the continuity (conservation of mass) equation (Equa
tion 19.1) and M anning's equation (Equation 19.2):
Q=VA (19.1)
Hydraulic
Slope, S radius, R Pivot Velocity, v
m/m or ft/ft m ft line m/s ft/s
-i 0.3 -0.2
0.07- 14-T
- 0.2 0.08- 12-“-40
0.09- -0.3
■ 0.1 - 10 -
9-"
8- r 0.01
- 0.1 -0.4
- 0.09 - 7-
- 0.08 -0.5
- 0.07 6 - "- 2 0
- 0.06 -0.6 5-
- 0.05 0.2 -J
r 0.7
- 0.04 -%0.8 4-
- 0.02
- 0.03 . ^
U.3 - -1.0 % 3-’
- 0.02 -0.03
0.4-
s 3>
\
0.5- \
\ + - -0.04
\
I 0.6 - - 2.0 * ✓' 2- _
- 0.&08 0.7- \ -0.05
- 0.007 n ft_ ~ %
\ + -
- 0.006 r -0.06
- 0.005 : - 3.0 \ V 1- -0.07
*
- 0.004 * 0-9 - -0.08
- 4.0 -0.09
- 0.003 * ^ 0.7 -
-0.1
- 5*. a '' vO ,6-“
- 0.002 2 Q \
-'6.0 0.5-
- 7.0
- 8.0 0.4-
- 9.0
- 0 :88093.0 - - 10.0 0.3-- 1.0 -0 .2
8 888
: : ?
“ 0.00064 Q _ -
- 0.0005 * _
- 0.00045.0 - 0.2- -0.3
FIGURE 19.3 Nomograph for the solution of the Manning’s equation (U S D T Federal Highway
Administration, 1973).
F L O W IN O P E N C H A N N E L S 19.9
Bdc'r Fotwatd Slop Refre*h Home | Search Favwiei Hitfwy | Mai Print Ed* Oncut*
A d d tW ^ t) WtofcalnhVPioiecl^houMangwebMangwrti Nfnl
—
T h e o p e n c h a n n e l flo w c a lc u la to r #
..... - .................................... ... .......... .......... ...... . .... . c A . ^ 1
H— T---- 1 ^ ----TH-----
I----sz---- 1 V ----32----- / _
■^L. \ 0
Sele:t Channel Type |Trepeioid 1 1 If
” S i* ^
Rectangle
K ^ - 2'
Tiapezotd
'
Triangle
!> J l
oCi.cie
- u ------------ m
| Select parameter tor solving jSclect unit system.! Feettft)
..... : . . ------- 2 * 4 ----- -- S g s i
|Chamel slope: | (Ml [Water depth(y): I fa [Bottom width(b) 1 t
.2/3
f V /V D 1/2
l i - p>"■ (19.3)
yLui=i
,v pi
For a composite channel, one should apply M anning’s equation to compute dis
charges for each sub channel, then summ arize them to obtain the discharge in the
whole channel. M anning’s equation can be solved by using an algebraic method,
nomograph (e.g., Figure 19.3), or various com puter software packages, for exam
ple, Flow Master by Haestad M ethods, Inc. (1997), or WwwChannel as a part of
the handbook. W wwChannel is an Internet-based computer program (Open-
Channel Flow Calculator, Figure 19.4) obtainable through any web browser and
does not require other special software.
WwwChannel (http://www.ceserver.lamar.edu/fang/handbook/Channels.html)
was developed by Dr. Xing Fang at Lam ar University to perform common open-
channel computations. All three methods are illustrated in Example 19.2.
EXAMPLE 19.2
The compound cross sections in Figure 19.2 have M anning’s n values o f 0.025
fo r the main channel and 0.060 fo r the overbank areas. The uniform channel
slope is 0.004. Determine the flo w rates when water depths are 5ft ( 1.52m) and
15ft (4.57m) above the channel bottom , respectively.
1 9.10 FLUID F L O W H A N D B O O K
S olution
Computation of the cross sectional parameters (e.g. area and hydraulic radius) for
this com pound channel (Figure 19.2) has been illustrated in Example 19 .1. These
parameters will be used directly to determine the flow rate by M anning’s equation.
1. W hen the water depth is y = 5 ft (1.52 m), the channel area is 506.25 ft 2
(47.03m 2) and hydraulic radius is 4.59 ft (1.40 m). Therefore the flow rate
Q is
If the chart in Figure 19.3 is used, one draws a line to connect S = 0.004
and n = 0.025, which intersects the pivot line. To find velocity V, a line
from hydraulic radius R = 4.59 ft is drawn to the intersection point on the
pivot line, and extended to the vertical line for velocity, V, which gives
V = 11 ft/sec.
o
Therefore-j the flow
_
rate will be 11 ft/sec X 506.25 ft 2
= 5569 ft /sec (157.7 m /sec). The algebraic method provides more accu
rate results.
If W wwChannel (Figure 19.4) is accessed through a web browser (e.g.
Internet Explorer, Netscape 6.0), one can select channel type first (trape-
zoid, rectangle, triangle, circle), then select a unit o f length (feet or meter)
and param eter to be calculated (discharge & velocity, channel slope from
velocity, M anning’s n from velocity or discharge, depth from discharge,
side slopes from discharge). One can enter channel slope of 0.004, water
depth o f 5 ft, bottom width of 10 ft, side slope of 0.25 horizontal versus
1 vertical, M anning’s n o f 0.025, then click on “Calculate” button. Dis
charge is displayed as 5256 ft3/sec with m = 1.486 used in M anning’s
equation in W wwChannel. W wwChannel also provides other related
results, e.g. channel area, wetted perimeter, top width, flow velocity, veloc
ity head, specific energy, critical depth and slope, Froude number, and flow
status (subcritical or supercritical flow).
2. W hen the water depth is y — 15 ft, the total flow rate should be the sum of
flow rates in the main channel and the left overbank area.
1 49
Qm = — — x 1553.125 x 1 2.35 2/3 x 0.0041/2 = 31280 f t 3/sec
1 49
Qh = — x 1000 x 4.8782/3 x 0.0041/2 = 4517 f t 3/sec
F L O W IN O P E N C H A N N E L S 19.11
The mean flow velocity in a uniform open-channel flow can also be deter
mined from the equation.
(19.4)
In an open-channel flow the pipe diam eter D in the original Darcy-W eisbach
equation is replaced by the hydraulic diam eter Dh = AR = 4 X hydraulic radius to
VAR
calculate the flow Reynolds number Re = ------ and the relative roughness (ks /4R),
v
then th e /v a lu e is found from the Moody diagram (Figure 8.1). The absolute rough
ness ks is a function of channel bed material. For natural riv e rs,/can be estimated
from the size d 84 of bed material (84% of the materials in the bed sample are
smaller than the d 84 size) (Limerinos, 1970):
EXAMPLE 19.3
Estimate the discharge in a gravel bed stream (d84 = 2.8 cm) that has an aver
age depth o f 1.0 m (3.28 ft), a slope o f 0.0035, and a width o f 50 m (164.04 ft).
Solution
Method 1: Assume absolute roughness ks = d 84 to estimate friction factor
from Moody Diagram.
R = ^ = 1.0 w (3 .2 8 /r)
1.0
/ = 1.2 + 2.03 log 1.2 + 2.03 log = 0.053
V
2.8/100
19.12 FLUID F L O W H A N D B O O K
When the flow is uniform, the depth in an open-channel is called normal depth. In
channel design, normal depth is usually determined when discharge and channel
geometry are given. It is computed from M anning’s equation using the algebraic
method, Figure 19.5, or by the software WwwChannel, as illustrated in Example 19.4.
S olution
Q = - A R 2nS 112
n
, p 2/3 _ Q n ARm _ Qn _ 3 0 x 0 .0 1 2
S '12 ’ bm S ],2b m 0.002i/2 x 5 8/3
A R lri _ Qn _ 3 0 x 0 .0 1 2
bm S V2b m 0.002 1/2 x 4 8/3 '
z = l / 2 = 0.50
A R in Qn 3 0 x 0 .0 1 2
= 0.067
d0W S ',2d08/3 0.0021/2 x 68/3
in O id O O
—'ii' coii cmii rOii Vn
■■■■■■■■I
FIGURE
o c> c i c>
° p /A p u o q /fi jo s a n |0 A
19.13
19.14 FLUID FLOW HANDBOOK
Channels can be classified as non-erodible channels and erodible channels for design
purposes. Most non-erodible channels are constructed and lined (concrete, rip-rap,
interlocking blocks, geo-textiles, vegetation) channels which can withstand erosion
satisfactorily under all operational velocities. Unlined channels formed in natural
materials (e.g. sand, gravel, sandy loam, firm soil, stiff clay) are generally erodible,
especially under high velocity as encountered, e.g. during floods. In order to prevent
the overtopping of a channel by surface waves and surges, a freeboard is recom
mended. Freeboard is defined as the vertical distance from the top of the chan
nel to the water surface at the design discharge. An average freeboard from 1.4 ft
to 3.6 ft for discharges from 10 to 3000 ft3/sec respectively is recommended by
the U.S. Bureau of Reclamation (1952). In open- channel design the dimensions of
a channel are com puted for uniform flow such that hydraulic efficiency, channel
stability, construction and economy are optimized (Streeter, 1945).
For lined channel design, one may use the best hydraulic section (Table 19.3), where
the channel section has the least wetted perimeter for a given cross-sectional area.
Mean
Wetted Hydraulic Surface hydraulic
Channel Area perimeter radius width depth
definition (A) (P) (R) (B) (D)
Trapezoid V J y2 1 3
2 \Z ly \V iy —y
(half of 2 3 4
hexagon)
Rectangle 1
2y2 4y 2y y
(half of 2
a square)
Triangle y2 2V I > I V I , 2y
1
(half of 4 2
a square)
Semicircle 1
V ity 2y
2*
Parabola f v l / 1 2
| V i, 2 V 2 y
2" 3y
« = ly /iy
The best hydraulic section may not give the most economical channel because of
many other factors, e.g. cost for land acquisition. The semicircle has the least perime
ter among all sections with the same area, but it may not be the most practical to
construct with conventional materials.
S olution
Side slope = z = V 3 / 3 = l / V3
Bottom width = b = 2 V3 / 3 y
Area = A = V3 y 2
Q ^ — A R ^S"1
n
b = = ^ - = 10.8 f t
3
A = *j3y2 = 153 f t 2
After we consider the 1 ft freeboard, the final channel cross section para
meters are:
y[b + 2 y j \ + z 2) = 2 by + 2 z y 2'
:.b = 2y^J\ + z 2 - z)
Q = 6 0 0 = ^ ( 3 -3 2 4 ? 2 ) ( f J ( 0 . 0 0 0 2 ) 172
b. Since the velocity for the best hydraulic channel design is greater than
the maximum permissible velocity, we will use the maximum permissi
ble velocity method to redesign the channel cross section.
1 49 2 i j 49 2 j_
V = 1.75 = —1— R * S 2 = — R> (0.0002 )2
n 0.02
R. = 2 .1 4 /r
A = b y + 3y2 = Q / V = 600 /1.75 = 342.8 f t 2 (a)
Using a trial-and-error method to solve equations (a) and (b) we can get
y = 2.25 ft (0.69 m), b = 145.9 ft (44.5 m). The final channel cross sec
tion with 1 ft freeboard gives
P = b + 2 y ‘J\ + z 2 = ^ + 2>jV5
A = by + zy2 = by + 2 y 2 = 2 A l l y 2
R - y - A - b y * 2y2
2 P b +2 y S
F L O W IN O P E N C H A N N E L S 19.19
by + 2 v2V5 = 2by + A y2
b = >’( 2 V 5 - 4 ) = 0.472y
Q=J
600 = J 8 ^ 2 ( 2 . 4 7 2 / ) ^ V 0 0 0 0 2
/ =F = 4 .3 7 8 x l( T 7 y
1511.3
= 1.2+ 2.03 log
The design velocity V is less than the permissible velocity. The final cross
section parameters with 1 ft freeboard are:
The one-dimensional energy equation for open channel flow (see Figure 19.7) is
(19.6)
19.20 FI I II) FLOW HANDBOOK
where y is the depth of flow, S() is the slope of the channel bottom, hL is the head
loss between cross sections 1 and 2, and a is the kinetic energy correction factor.
The factor is introduced to account for the true kinetic energy in a cross section
due to the spatially variable velocity distribution. The factor a is defined as
\\u \lA
(X= t / v - (19.7)
V A
Here u is the local velocity dependent on the depth (Figure 19.7), whereas V
is the average velocity of flow at a cross section. For a composite channel (e.g.
Figure 19.1), the factor a can be determined in finite increment form:
I
V i=l /
The subscript / indicates the subarea of which and A,- are the velocity and
cross-sectional area, respectively, and N indicates the number o f subareas that the
section is divided into. I'and A are the total area and average velocity for the whole
section, respectively. Kt is the conveyance for the subarea /':
1.49
Kj = —— A R,2n (English units) (19.9)
ni
F L O W IN O P E N C H A N N E L S 19.21
Typically, one assumes that the friction slopes Sf = h\/Ax in the subareas of
one cross section are the same, while the roughness coefficients /i, can be differ
ent, especially in the main channel and overbank areas of a floodplain.
The kinetic energy correction factor (Equations 19.6 and 19.7) is often assumed
to be 1.0 for simple prismatic channels (with a constant cross-sectional shape and
bottom slope), but can be significantly higher than unity in composite channels and
floodplains (Table 19.6). The energy head loss between cross sections (hL) can be
determined from either M anning’s equation or the Darcy-Weisbach equation.
The specific energy E in open channels is the sum of the depth of flow y and the
velocity head a V 2/2g.
(19.10)
For a given discharge and channel geometry, the specific energy is a function
o f the flow depth (Figure 19.8). Flow depths y, and y 2 (Figure 19.8) are the alter-
ate depths, the two possible depths at which a given discharge may flow with the
same specific energy. The critical depth yc is the depth o f flow where the specific
energy reaches the minimum value at a given discharge. When the depth of flow
is greater than ;yc (e.g. y i > yc), the flow is called subcritical flo w (high potential
energy and low kinetic energy). A flow with a depth less than critical depth (e.g.
y 2 < yc) is called supercritical flo w (high kinetic energy).
For a rectangular channel, the critical depth at a given discharge is
1/3 1/3
v 8 > I s b 2)
Value of a
where q (m 2/s or ft 2/s) is the discharge per unit width of channel. For other types
of channels, the critical depth can be determined from Equation 19 .12 or from
Figure 19.9 or the program WwwChannel.
(19.12)
where subscript c stands for the critical flow. In Figure 19.9, the capital Z stands
for Q /V g ", and the small z is the side slope of the channel.
For critical flow, the dimensionless Froude number Fr (Equation 19.13) is
equal to 1 .
V Q IA
F = (19.13)
r J g A lB
where D is the hydraulic depth, and B is the width of flow at the free surface.
S olution
1/3 1/3
( V/3 ( ™2 V/3
f i i f = fI SQB22 1) f1^9.81302* 5 2lJ = 1.54 m
, * )
yc =
30 2 x 5
9.81 x (5 y r)3 “
/. yc = 1.54 m (5.05 ft)
Alternatively using Figure 19.9, the horizontal axis (at the top) has a value
Z / b 25 = Q I -yfg I b25 = 0.17, then y/b = 0.3 on the vertical axis; therefore
the critical depth yc = 0.3 X 5 =1.5m (4.92 ft).
Volues of Z / b 2-8 for trooezoidol sections
C> d
O
O
o
19.24
p/Apuo q/xp S9n|0A
FIGURE 19.9 Curves for determining the critical depth (Chow, 1959).
F L O W IN O P E N C H A N N E L S 19.25
4 + (4 + 4 y ) 8 + 4yt. .
A = --------------- - x y = ------- x _y = (4 + 2 y x) x y c
2 2
7 = 4 + 4y(
302 4+4
x ~---------- ——- j = l
9 8 1 {(4 + 2 , , ) , , }
4 + 4 ,, 9.81
= 0.011
{(4 + 2 , , ) , , } ' 30'
3. For a circular channel (see Table 19.1 for circular channel geometry)
D 2
= —-1 .0 3 = 1.97 w (6 .4 6 /0
2
Alternatively using Figure 19.9, the horizontal axis at the bottom has a value
Z /d2 5 = 0.1, then (y/d) = 0.315 on the vertical axis; therefore the critical depth is:
;y£. = 0.315 X 6 = 1.89 m (6.20 ft).
The hydraulic jum p is a rapidly varied flow phenomenon that occurs when the
flow changes from supercritical flow upstream to subcritical flow downstream.
19.26 FLUID F L O W H A N D B O O K
The depths upstream and downstream of the hydraulic jum p are called conjugate
depth and are typically determined from the momentum equation. For a horizon
tal channel
where p , (upstream) and p 2 (downstream ) are the pressure at the centroids of the
respective areas and A 2. For a rectangular channel with horizontal slope, the
solution of Equation 19.14 yields
3'2 = y ( V l + 8 / V - O o r j , = y ( V l + 8 / rr22 - 1) (1 9 .1 5 )
where the Froude number F r } = Vx/ \ ^ 2 g y [ > 1.0 and Fr2 = V2/ V 2 gy2 < 1.0. The
relationship between conjugate water depths y l and y 2 and Froude numbers F rx for
hydraulic jum ps in rectangular channels with various bottom slopes S0 is shown
in Figure 19.10.
0 1 2 3 4 5 6 7 8 9 10 II 12 13 14 15 16 17 18 19 20
f , = v xn / g d i
velocity for the follow ing horizontal channels: (1) a trapezoidal channel with a
bottom width o f 5 f t and side slopes o f 1 o f I, (2) a rectangular channel with a
bottom width o f 5 ft, (3) a triangular channel with equal side slope o f 5 hori
zontal and I vertical.
Solution
1. For a trapezoidal channel, one can apply Equation 19.14 directly. Given
conditions are Q = 150 ft3/sec (4.25 rrrVsec), water depth upstream of the
hydraulic jum p = 1.0 ft, and top width = 7 ft (2.13 m) since side slope is
1 to 1 and bottom width is 5 ft (1.52 m).
5 +7, 2
2 x 1= 6 f t
Ai.= ------
Q_ = 150 = 25 / sec
A, 6
A2 = by 2 +
- = - r = 2 i Z l ± ^ V l x 6 2.4 (b)
5 + y2
V2 = e = I 5 0 (c)
a2 a2
Using Equations (a) through (d), and applying a trial and error method, one
can find y 2 = 5.03 ft (1.53 m)
Ax= 5 x 1 = 5 f t 2
v G. = ]5 0 =30 f i / s e c
1 A, 5
30
f r. = —= = = = 5.28 (supercritical flow)
V 3 2 .2 x l
1 9.2 8 FLUID F L O W H A N D B O O K
3. For a triangular channel, one has to use Equation 19 .14 again. Surface
width is 10 ft since side slope is 5 horizontal and I vertical.
Ai = -2 x l 0 x l = 5 f t 2
>’i = - f t
•1 3
Q 150
= 30 f t / sec
A. 5
A = 5 v ,2 (e)
v2 = - . v 3 (0
Using Equations (e) through (g), and applying a trial and error method, one
can get y 2 = 4.32 ft (1.32 m)
In most cases, except hydraulic jum ps, non-uniform open-channel flow is gradu
ally varied flow characterized by gradual changes in water depth and water surface
slope, which occur when channel geometry and bottom slope vary in flow direc
tion. The “water surface profile” gives the elevation of the free surface with dis
tance, which provides essential information on flow depths and associated velocities.
W ater surface profiles are classified based on the slope of the channel (mild = A/,
steep = S, horizontal = H , critical = C, and adverse = A) and the flow depth, rel
ative to normal depth and critical depth. Zone l has the flow depth greater than
both the normal and the critical depths, zone 2 has the flow depth located between
the normal and the critical depths, while zone 3 has the flow depth smaller than
both normal and critical depths. Given sufficient distance, all flows tend towards
a normal depth which depends on channel shape, and roughness, bed slope and flow
rate. Therefore the flow depth increases in flow direction for water surface profiles
in zone l and zone 3 (dy/dx > 0), and decreases for water surface profiles in zone 2
(dy/dx < 0). When the normal depth (y n) is larger than the critical depth (yc), the
channel is called a mild slope channel; a steep slope channel is a channel where
F L O W IN O P E N C H A N N E L S 19.29
yn < yc. There are also horizontal channels, channels where yn = yc (critical slope)
and channels with adverse slope. In the classification ot gradually varied water
surface profiles of open-channel flows (Figure 19.11) a letter for the water sur
face profile type designates the bottom slope (M, S, H, C, A), and a number des
ignates the zone (1, 2, and 3).
After applying the specific energy concept, the energy equation (Equation 19.6)
can be rewritten as
YL
i jy2 + a 2
£, - E2 2g
Ax = (19.16)
S / —S{] Sf S0
where A x is the distance between cross sections 1 and 2, Sf = hjcxx is the average
energy (friction) slope, and S0 is the bottom slope. In practice, most surface pro
files are computed by numerical stepwise solution of Equation 19.6 or 19.16, that
is, by dividing the channel or river into short reaches and carrying the com puta
tion reach by reach from one end of the channel to the other. The average friction
slope is determ ined from either M anning’s or the Darcy-W eisbach equation as
n2V n 2 V~
S f = -------z tttt (English units) or S f = _ 4/1 (SI units) (19.17)
j 2.22 R R
or
f v2
S f = - —=- (D arcy - Weisbach equation) (19.18)
7 8g R
where V = (V\ + V2)/2 is the average velocity and R = (R\ + R 2) ^ is the average
hydraulic radius in a reach. There are two common approaches for surface pro
file computation: the direct step method and the standard method.
The direct step method is suitable for prismatic channel with simple geometry.
In the direct step method, the depth, velocity, and channel geometry are known
at a given section of the channel, and one arbitrarily chooses the depth at the other
end of the reach based on whether surface profile decreases (zone 2 ) or increases
(zone 1 and 3), then the length o f the reach is determined from Equation 19.16.
If the flow is subcritical, the com putation starts from the downstream end o f the
channel and moves upwards (negative Ax)', in supercritical flow the computation
starts from the upstream end o f the channel and moves downwards (positive Ax).
19.30 FLUID F L O W H A N D B O O K
| None
I
y* ■« " y»
1 \
-
\
a*
§A f
\
! * ..t f
\
\
i
! * * ■ ----------- ------ = 5 -
t*
* **
S p . -
^ ....
C1 V
t *
1 * y" j yc
i l ij ■i,. .
2_
X ' 1 ............ ..
>« 1
! ?
I* ? { j.— . - < c r ~ '
Non#
.
.....................
1 N
V
y. n
j;
EXAMPLE 19.8
Given a discharge o f 25 f t 3/sec (0.708 m3/sec) and a 2 f t depth at the downstream
end (elevation is 800 f t above mean sea level), compute the water surface profile
through a 5 f t diameter, 2 0 0 ft long, corrugated metal (M anning's n = 0.022)
circular culvert with a slope o f 0 .0 0 1.
Solution
From Equation 19.12 or Figure 19.9 it is found that critical depth yc = 1.39 ft
for given flow conditions, and normal depth y n = 2.54 ft from Equation 19.2 or
WwwChannel. W ater surface profile is therefore of type M2, because the
initial depth of 2 ft is between critical and normal depth. One can use Equa
tions 19.15 and 19.16 and the direct step method to find the water surface
profile as follows:
— 0.00186 0.00178
O
i
The average friction slope is typically determined by one of the following four
methods in HEC-2 (HEC. 19 9 1) and HEC-RAS:
(19.19.1)
water surface profiles in com plex river settings, e.g. having bridges with low flow
or pressurized flow or w eir flow, highway culverts, ineffective flow areas, and
flow junctions and separations.
Hydraulic structures are used typically for three purposes: ( l) flow measurement
(e.g. weirs), (2) flow regulation (e.g. gates), and (3) flow conveyance (e.g. culverts).
Typically there is a fixed relationship between the flow rate and the upstream and/or
downstream water-surface elevation at hydraulic structures.
19.34 FLUID F L O W H A N D B O O K
19.6.1.1 Weirs. Weirs are elevated structures in open channels that are used to
measure flow and/or control outflow elevations from channels. Sharp-crested weirs
consist of a thin vertically plate across the channel; most common are rectangular
and triangular weirs constructed in canals and flumes. Rectangular weirs can be
uncontracted with the opening spanning the entire channel width, or contracted with
opening only in a portion of the channel. Discharge equations are given in Table 19.7.
The broad-crested weir is an elevated structure with a long horizontal crest in the
direction of flow. Critical depth occurs on all weirs, unless the tailwater down
stream from the weir is substantially above the weir crest and causes a submerged
flow. W eirs typically create a considerable head loss in an open-channel flow.
19.6.1.2 Parshall Flume. The Parshall flume or Venturi flume was developed and
calibrated by Ralph L. Parshall (1926), and is widely used in wastewater treat
ment plants and irrigation channels to measure flow rate. Standard Parshall flume
geometry is shown in Figure 19.14. Dimensions can be found in many hydraulic text
books, e.g. Chaudhry et al. (1997) and U.S. Bureau of Reclamation (1978). Critical
flow is produced by reducing the channel width and increasing the slope of the
Rectangular
Q = K ^2 g L H }12 and K = 0.40 + 0.05// / P, h/p £ 10
sharp-crested (Kindsvater and Carter, 1959) . P is the height of the
crest of the weir above the bottom of the channel, H
(uncontracted) is the head from water surface to the crest, and L is
the weir opening (width).
Rectangular
Q = K ^2g (L - 0.2 H) H 3/2, both sides have contraction
sharp-crested
as L/H > 3 (Roberson, Cassidy and Chaudhry, 1997).
(contracted)
Trapezoidal
Q = K ^ 2 g B H y2 , b is the bottom width, a-has the
sharp-crested
same value as for a rectangular weir. The Cipolletti
(Cipolletti Weir)
weir has side slopes 1:4 (H:V).
Triangular g __ /
Q = — K^j2g tan — H 5'2 , K varies from 0.6 to 0.57 as
sharp-crested 15 v 2y
Broad-Crested
Q = 0 . 3 8 5 0 / ^ 2 # H 312 , C has values of 0.86, 0.89,
bottom. The discharge (<2, ft 3/s) through the flume for a throat width (W) from 1 to
8 ft and free flow conditions < 0.7) is
where Hu (ft) is the head above the floor level of the flume measured in the con
verging section (Figure 19.14). Parshall (1926) and Chow (1959) provide more
information on small size flumes (width W < 1 ft) and tor submerged flow condi
tions in the flume.
e = C J i g L H V1 (19.21)
where L is the crest length perpendicular to flow, and H is the total head on the
crest, which includes the approach velocity head. C is a dimensionless discharge
19.36 FLU ID FLOW HANDBOOK
coefficient, and C/Cp is given as a function of H/Ht), where Cp is the discharge coef
ficient under the design head Hn (Figure 19.15). In some textbooks and design
manuals the expression c V 2g is called the discharge coefficient. For spillways
with a radial (tainter) gate above the crest, the discharge is
Q = 2/ 3 ^ C L ( h ?'2 - t f f ) (19.22)
where heads H , and H2 are defined in and the discharge coefficient C is given in
Figure 19.16.
19.6.2.2 Gates.There are many types o f common gates for flow control: the two
most common ones are vertical (sluice) and radial (tainter)gates (Figure 19.17). The
discharge is determined from the relationship
Q - K W y^lgH (19.23)
P
"d
[a. Uiscnargc coefficient C u versus n t i L tor design head
ni
FIGURE 19.16 Coefficient of discharge for gated spillway (U.S. Bureau o f Reclamation, 1978).
where W is the width o f the gate, y is the vertical opening of the gate, and H is
the head above the gate from the channel bottom (Figure 19.17). The flow pass
ing through the gate can be free flow or submerged flow, depending on the down
stream water depth.
The discharge coefficient K for a sluice gate is given in Figure 19.18. For free
flow, K varies from 0.5 to 0.6 as a function of H/y only. For submerged flow condi
tions, K is a function o f H /y, y j y and Froude number Fra (= V / v g y ), where yd
is the water depth downstream o f the gate. For tainter gates, Figure 19.19 gives the
discharge coefficient K as a function of the ratio a/r (Figure 19.16) for a/r = 0.1 (left),
0.5 (middle) and 0.9 (right), respectively. For free flow through a tainter gate, K is
a function of H /r and y /r , and for submerged flow, K is a function of y j r also.
H
y
FIGURE 19.18 Discharge coefficient for vertical sluice
gate (Henry, 1950).
FIGURE 19.19 Discharge coefficient for Tainter gate (U.S. Bureau o f Reclamation, 1978).
A culvert is a conduit placed under a road to convey water from one side to another.
Culvert design involves selection of the type o f culvert and size to pass the design
discharge without overtopping of the road and without erosion on either end of the
culvert. The flow through a culvert can take different forms and is a function of sev
eral variables, such as cross-sectional size and shape (circular, rectangular, or arch),
bottom slope, length, conduit roughness, entrance and exit design, and water depths
at the upstream (headwater) and the downstream (tailwater) of the culvert. The
entrance flow can be broadly classified as submerged or free. There are two types
of flow controls in culverts: inlet control and outlet control. For inlet control, the
hydraulic control section is located at or near the culvert entrance, and the discharge
through the culvert is dependent only on inlet geometry and headwater depth because
F L O W IN O P E N C H A N N E L S 19.39
the conduit can convey more discharge than the inlet will allow. For outlet con
trolf, the control section is at or near the culvert outlet, and the discharge through
the culvert is dependent on all o f the hydraulic factors upstream from the outlet.
Equations, tables and charts for culvert design can be found in textbooks (e.g.
Franzini and Finnemore 1997, Roberson et al. 1997) or design manuals (ACPA
1987, AISI, 1995). Com puter software, e.g. “CulvertM aster’' developed by Haes-
tad Methods, Inc. (1997), is also available and recommended.
The Culvert Calculator in CulvertMaster can solve for size, discharge, and head
water elevation for a culvert. The Calculator organizes information for a culvert
into seven areas: solve for, culvert, section, inlet, inverts, headwater elevation, and
exit results. Figure 19.20 shows an example for the sizing of a single, circular, con
crete culvert (roughness n = 0.13) with square edge entrance and design discharge
of 1.4 m3/sec (49 ft 3/sec) and 11 m (36 ft) long. The result for the culvert size is
750 mm (30 inches); flow has inlet control since the required headwater elevation
for inlet control is greater than the one for outlet control.
W ater Environm ent Federation (formerly the W ater Pollution Control Federation,
W PCF, 1982). Sewers up to 375 mm (15 inches) in diameter are designed to flow
half full at the design flowrate; larger sewers are designed to flow three-fourth full.
The design flowrate is a function of local precipitation characteristics and the design
return period (e.g., 50 years, ASCE, 1982) and the wastewater quantities based on
the present and possible future population. In order to prevent solid deposition, sewer
system needs to maintain self-cleansing velocities under the design flowrate: not less
than 0.60 m/sec (2 ft/sec) or not greater than 3.5 m/sec (10 ft/sec) (ASCE, 1982).
The hydraulic design of sewers includes the selection of pipe sizes and the check
ing o f velocity under the design discharge under uniform flow condition. Channel
geometry for a circular-pipe is given in Table 19.1 including cross-sectional area
A, wetted perimeter P, and hydraulic radius R as functions of angle 6 and diam
eter D. The M anning’s equation with constant roughness and friction slope gives
where the subscript V ’ stands for the pipe flowing full, Hydraulic Radius Ra = D/4.
The velocity (V0) for full pipe flow can be determined from Figure 19.3. Figure 19.21
graphically shows Equation 19.24 to assist in sewer design (Chow, 1959). An illus
tration is given in Example 19.9.
V a lu e o f n/n<>
Subscript "o" indicotes th e full flow condition
1.2 14
1.1 12 13
Volues o f Q / Q 0 , V / V 0 , A R ^ / A o R o 273, o n d R 2/ 3/ R ^ 3
EXAMPLE 19.9
Design flo w rate f o r a circular concrete sewer is 10 ft*/sec. The slope o f the
sewer is 1% and M anning's n = 0.015. Find the pipe diameter if the pipe flow s
three-fourth fu ll under design flo w rate. Determine the discharge and flo w
velocity as the water depth is 1 ft.
Solution
From Figure 19.21, we have Q /Q o = 0.93 as yID = 3/4, therefore the full
pipe flowrate should be QG = 10/0.93 = 10.75 ft3/sec (0.3m 3/sec).
We can get D = 1.6 ft (0.49 m), so we select the standard pipe size of 21"
with full flowrate of 13.73 ft3/sec and velocity of 5.71 ft/sec. If y = 1 ft
(yID = 0.57), we have QIQ() = 0.62 and V/VQ = 1.05 from Figure 19.21,
then Q = 8.51 ft3/sec (0.24m 3/sec) and V = 6.0 ft/sec (0.56 m/sec), which
falls between the maximum and minimum self-cleansing velocities.
For complex pipe network design and analysis of sewer or stormwater
systems, utilization of com puter software is highly recommended. For
example, Visual Hydro (Visual SW M M —Storm Water M anagement Model,
originally produced by the U.S. Environmental Protection Agency) by
CAiCE Software Corporation (1999), StormCAD by Haestad Method Inc.
(1997), Loops II (Eaglin, 1996) and others are available for selection. Visual
Flydro can be used for com bined sewer modeling, sanitary sewer modeling,
storm sewer modeling, master drainage plans, and detention basin design.
REFERENCES
2 0 .1 1NTRODUCTORY CONCEPTS
In general, flows past immersed objects, or external flows, are flows of fluids sur
rounding objects with solid walls, while, ideally, subject to no outer boundary,
except at practically infinite distance from the object. Examples are flows around
ground structures or land, marine and air vehicles. In contrast, internal flows are
those in which the flowing fluid is surrounded by solid walls, such as flows through
pipes, ducts and channels or within fluid machinery. An object immersed in a flow
ing fluid will generally experience a force, whose magnitude and direction depend
on several parameters. The source of this force is the friction between the fluid
and the solid wall. An entirely immersed solid object, having a constant relative
velocity with respect to an unbounded, incompressible, inviscid (i.e. frictionless)
fluid, will experience a zero force (d’A lem bert’s paradox), although not neces
sarily a zero moment [1-3].
In general, the flow pattern around an object depends on the plane of obser
vation, while the various flow parameters, such as velocity and pressure, vary
along any direction. This is the case of three-dimensional (3-D) flow. However,
if the object is assumed to extend to infinity on two opposite sides and has a uni
form cross section normal to a certain axis and if the fluid motion is such that the
flow pattern and all flow parameters are independent of position along that axis,
the flow is called two-dimensional (2-D) or planar (Figure 20.1). Obviously, the
total force on 2-D immersed objects would be infinite and so it becomes neces
sary to refer to a force per unit span.
20.1
20.2 FLUID F L O W H A N D B O O K
A fluid having ^uniform velocity vector Vx far away from an immersed object
will exert a force F on the object (Figure 20.2). This force may be decomposed
into two components: a component FD parallel to V*,, called drag, and a compo-
nent FL normal to V called lift [4, 5]. When a 3-D object (for example an air
plane) has a plane of symmetry and is not parallel to that plane, it is customary
to consider the force component normal to the symmetry plane separately, and
identify it as the side force.
Every elementary surface area dA of an immersed object is subject to an ele
mentary surface force dF, which may be decomposed to a tangential (shearing)
component dFs and a normal component dFp. When divided by the surface area,
these force components become, respectively, the wall shear stress t w = dFJdA
and the wall normal stress a w = dFp/dA. The latter is related to the surface pres
sure p as p —- a w\ the negative sign is necessary because, by convention, pressure
is positive when compressing the surface, while the normal stress is considered pos
itive when in the direction of the outward unit normal vector n. The total surface
force F may be found by integrating the elementary force dF over the surface of
the object, taking into account the change of orientation of the latter. For sim
plicity, let us consider a 2-D flow (Figure 20.3). Then, the force dF on the ele
mentary area dA = bds may be decomposed to two co-planar pairs of components
and one may compute the drag and lift per unit span from the wall shear stress
and the surface pressure as
b
= J
•'contour
P cos Bds + f
•'contour
rw sin
t)
= JPnnfniir p sin Ods + •'pnntmir
f tu. cos
Thus, the drag and lift may be considered as consisting of two distinct contribu
tions: one due to the pressure distribution around the object and another due to
wall shear stresses. In particular, the contribution of pressure variation to the drag
is called the pressure drag, or form drag, while the contribution of wall shear
F L O W PAST I M M E R S E D OBJECTS 20.3
stresses is called the skin friction drag, or simply skin friction. Both contributions
are caused by friction (refer to d ’Alembert’s paradox above). Objects for which
the skin friction is overwhelmingly larger than the form drag are called stream
lined, while those for which the opposite happens are called bluff.
In order to be presented in dimensionless forms, the drag and lift are normal
ized by the parameter ^ p V 2A (p is the fluid density and A is a characteristic area),
which is a measure of the inertia “force” on the fluid. This results into the fol
lowing coefficients
drag coefficient: CD =
lift coefficient: CL
The area A is a projected area and not the surface area of the object. Caution
must be taken to distinguish between the planform area, namely the projected area
of the object on a suitable plane parallel to V*, and the frontal area, namely the
projected area of the object on a plane perpendicular to Voo. For streamlined objects,
the area used for the drag coefficient is the planform area, while, for bluff objects,
it is the frontal area. Reflecting the fact that lifting objects are usually streamlined,
it is the plantorm area that is used, almost exclusively, for the lift coefficient.
Under certain specific conditions, portions of the drag may be attributed to
specific phenomena and are usually estimated separately from the more conven
tional portion. These include the following:
• Drag due to lift, or induced drag (Section 20.3.4): this occurs when a lifting
solid surface is terminated in the fluid.
• Acceleration drag [3]: when an object accelerates through a fluid, the pres
sure distribution on its surface is adjusted such as to generate a force capable
o f accelerating the fluid in the vicinity o f the object. This generates a larger
drag force, compared to that on the same object in steady motion through the
fluid. Idealized analysis expresses this additional force in terms of the mass
o f a certain volume of fluid (added mass) that has to be added to the mass of
the object for an unsteady momentum balance.
• Cavitation drag (Figure 20.4a): if the static pressure in the near wake of an
object immersed in a liquid drops to a critical pressure, called the vapor pres
sure, a vapor-filled cavity forms (supercavitation); the drag coefficient in
cavitating flows is lower than that in single-phase, liquid flows, because the
wake pressure cannot drop below the vapor pressure level [6 ].
• Ventilation drag (Figure 20.4b): when objects are crossing the interface
between a liquid and a gas, gas may be channeled to a cavity that forms in
the low-pressure wake of the object; like cavitation, ventilation reduces the
form drag as it poses a lower limit on the wake pressure [6 J.
bow wave
Vm , p .
liquid
(a)
(b)
interface
(c) (d)
FIGURE 20.4 Illustration of configurations generating a) cavitation drag, b) wave and ventila-
tion drag, c) interfacial wave drag, and d) internal wave drag.
F L O W PAST I M M E R S E D OBJECTS 20.5
• W ave drag (Figure 20.4b): this is the additional drag on an object that pene
trates the free surface of a liquid, due to hydrostatic pressure differences cor
responding to the elevation of the liquid near the stagnation point | 6 |; wave
drag also applies to objects that are fully immersed in one of the phases of a
stably stratified, two-phase system (interfacial waves; Figure 20.4c) or in a
density-stratified, single-phase fluid (internal waves; Figure 20.4d) [7].
• Compressibility drag: for a given Reynolds number, the drag usually increases
as the Mach number increases [8]; this is the additional drag which arises due
to compression of the gas in front of the object and its subsequent expansion
in the wake.
du dv dw „
& + * + e T °
du du du du 1 dp 1' a 2« ~c)2U d 2u
— + u — + v — + w — = ------ — + gx + +
dt dx oy az p ox c)x2 dy2 + dz2
dv dv dv dv 1 dp 1' d 2v a 2v d 2v
— + u — + v— + vv — - — + gy + +
at ax dv az p ay v\ d x 2 dy2 + dz2
dw dw dw dw 1 dp I' ? 2W d 2w d 2w
— + U — + V— + VV— = ----------
+— gz + +
dt ojc dy dz p az v\Kd x 2 dy2 + dz2
where
u, v and vv are the com ponents of the flow velocity V along the Cartesian
axes x, v and z
p is the fluid density
v is the kinematic viscosity of the fluid
gx, gy and gz are the Cartesian components of the gravitational acceleration
vector.
An important dynamic param eter for fluid motion is the Reynolds number Re =
pVcD//j, (Vc is a characteristic velocity, D is a characteristic length and p = vp),
which is a measure of the relative importance of the inertia “force” on the object
compared to the viscous (frictional) force.
The Navier-Stokes equations are quite difficult to solve analytically, and even
numerically at large Re, mainly due to the presence of non-linear terms. For this
20.6 FLUID F L O W H A N D B O O K
du 1 dp ( d 2u d 2u d 2u
* = - ^ T x + g' + v \ ^ +^ +
du _, du 1 dp
Iat
T + = -------
dx p dx 2
I T + & + v \ \ dx
where Voo is the far field velocity, away from any immersed object, and assumed
to be uniform.
Two types of boundary conditions have to be satisfied by the solution o f vis
cous flow equations: a) the no penetration condition, which requires that the nor
mal component of the relative velocity of the fluid must vanish on the surface of
a solid object, and b) the no slip condition, which requires that the tangential com
ponent o f the relative velocity must vanish on the surface as well. (Exceptions to
the latter condition must be allowed, as for example in the case o f very low den
sity gas flows). A significant simplification of the mathematical flow problem is
achieved when the effects of friction are neglected, in which case the Navier-Stokes
equations are simplified into a set of first-order differential equations, called the
Euler equations, as [ 1, 3, 4]
du du du du 1 dp
~dt + U ~dx+ V~dy + W~dz = ~~p Ihc + 8* ^
The effects of friction on the flow around an immersed object should be expected
to diminish with increasing distance from the object. For sufficiently large Reynolds
F L O W PAST I M M E R S E D OBJECTS 20.7
numbers, the direct effects of friction are confined in relatively thin layers adja
cent to the solid walls, called the boundary layers, and in the wake of the object,
which is the flow region downstream of the object that contains the fluid that
originated in the boundary layers around the object (Figure 20.5a). The flow out
side of the boundary layers and the wake is called the free-stream, and, for many
practical purposes, may be considered as inviscid. While the streamwise flow veloc
ity component u increases rapidly across the boundary layer, starting from zero at
the wall, it presents a much slower variation across the free-stream. As a result,
and for the purposes of boundary layer analysis, one may consider the free stream
velocity \Je as independent of distance from the wall and, at most, a function of the
streamwise coordinate x (Figure 20.5b). Within the boundary layer, a different kind
of simplification is possible, known as the boundary layer approximation, which
facilitates the computation of the boundary layer thickness, the wall shear stress
and other parameters [2-5, 8-10, 13].
For simplicity, consider steady, two-dimensional, large Reynolds number flow
past an immersed solid surface and neglect body forces, such as gravity. The sur
face is assumed to be nearly parallel to the x-axis, with the y-axis normal to it.
Then, the momentum equation in the x-direction is simplified as
1 dp d 2u
----- — +
p dx d y2
A result of the boundary layer approximation is that the pressure is nearly con
stant across the boundary layer and so the surface pressure may be taken as equal to
the free-stream pressure at the same streamwise location. The pressure variation
must be provided and it is usually calculated by performing a computation of invis
cid flow past the object by analytical or numerical means, for example using the
vortex panel method [14, 15]. The surface velocity computed by this method is
also considered to be the same as the free-stream flow velocity Ue(x) ju st outside
the boundary layer. Ue(x) is related to the surface pressurep(x) by Bernoulli’s equa
tion, which leads to
L = u
p dx e dx
C) 77 77 7 7 7 7 , ?eP*r*tion
... hbbic
T >0
•0 '''//> > °»
free-stream
d)
on the surface (separation point), and then it may spread into an extended region
near the object (separation bubble). The boundary layer approximation breaks down
when separation occurs and one may not use boundary-layer-type analyses beyond
the separation point, even for rough purposes. If the pressure gradient becomes favor
able again following separation, flow reattachm ent may occur, in which case one
may reconsider the use of the boundary layer approximation.
Like other types of flow, a boundary layer may exist in either a laminar or a turbu
lent state. Depending on the flow conditions, a boundary layer may start in either
F L O W PAST I M M E R S E D OBJECTS 20.9
state, or it may start as laminar and change to turbulent, through the process
known as transition (Figure 20.5d). Under special circumstances, a turbulent
boundary layer may turn laminar, through the process called relaminarization.
For simplicity, the discussion below will be restricted to 2-D boundary layers;
discussion of 3-D boundary layers may be found in References [8, 10, 13]. The
discussion will also be confined to incompressible flows, although the reader
must be warned that compressibility has significant, and even dramatic, effects on
the developm ent and stability of boundary layers |8, 10, 13, 14]. Only general-
purpose analytical and empirical results will be reviewed, while discussion of
numerical methods will be confined to a few relatively simple procedures.
Complete analytical and numerical solutions should provide the detailed veloc
ity variation within the boundary layer. For many practical purposes, however, this
information is not necessary or essential, as one may only be interested in overall
boundary layer effects, such as wall shear stress and free-stream streamline diver
gence. Approximate methods that average out velocity variation across the bound
ary layer are widely used, under the name of integral methods. Such methods are
based on solutions of an averaged differential equation (momentum integral equation).
| ^
• Energy thickness: 8e = f 1 - (-— ] dy. The param eter -zpUe 8e
Jo Ue \U e )
20.10 FLUID F L O W H A N D B O O K
represents the deficit o f mechanical energy flux per unit span, compared to
inviscid flow with the same mass flow rate as that in the boundary layer.
Additional shape factors may also be defined as the ratios of <5* or 6 and 4 [8].
The different boundary layer lengths may be used to define the following
Reynolds numbers, each with a distinct significance:
du
r M. = /x
dy
y = 0
Obviously,
Integration of the 2-D momentum equation in the direction normal to the wall
(i.e. for 0 < y < °o) results in the momentum integral equation
Analytical solutions of the boundary layer equations are available for uniform flow
past a thin flat plate (Blasius solution) and past corners of different angles (Falkner-
Skan solution) [3-6, 8, 10, 13-16]. Such solutions are self-similar, namely result
in universal dimensionless velocity variations when expressed in terms of a dimen-
sionless distance from the wall.
Laminar boundary layer in parallel flow over a flat plate (Blasius problem.
Figure 20.6a):
• dimensionless coordinate: 77 = — \ / R ^ x
x
( dp
• constant pressure I — = 0 1
11 . /3FI 2m
• wall angle: - — = --------
2 m + 1
• free stream velocity variation: Ue{x) - (J\x"\ U\ = const.
K>P-
- J
------ =1
b) c) i
FIGURE 20.6 Sketches of the flow configurations for the Blasius (a) and Falkner-Skan
(b, c, d) problems.
20.12 FLUID F L O W H A N D B O O K
1.0
0.8
0.6
0)
3
(3 = 2m/(m+1)
0 2 4 6 w'w0 1 2 3 4
n = (y /x )(R e x)1/2 5= (y/x)[(m +1 )Rej2]'*
FIGURE 20.7 Velocity profiles for laminar boundary layers over a) a flat plate (Blasius solution)
and b) wedges of different angles (Falkner-Skan solutions).
• dimensionless coordinate: £ = — m +^
x v 2
• for expressions and values for different thicknesses and cy see References [10, 13]
• wedge flow (Figure 20.6b): 0 < /3 < 1 (0 < m < 1), — < 0, favorable pres
sure gradient ^
• flow around an expansion (Figure 20.6d): - 2 < /3 < 0 ( - — < m < 0), — > 0.
adverse pressure gradient ^
A num ber of approaches have been developed for the computation of laminar
boundary layers with arbitrary pressure variation. A popular method is Thw aites’
method, which is based on the solution o f the momentum integral equation. This
method provides the following results [10, 13 - 16].
• skin friction coefficient, for -0 .0 9 < A < 0.25 [10]: C r- -“ -—(A + 0.09)°62
Re0
At sufficiently low Reynolds numbers, all boundary layers are laminar and, if dis
turbed, they would return to their original state (hydrodynamically stable flow). As
Rex increases, however, to a critical value Rexcn disturbances do not decay with time
(hydrodynamically unstable flow) and eventually lead to turbulence, through the pro
cess called transition. Transition depends on many factors and remains a problem
that is hard to describe and predict. A summary of the transition process for an incom
pressible, 2-D boundary layer over a smooth flat plate and in the absence o f free-
stream turbulence, vibrations and acoustic disturbances is as follows [10, 17].
Under the above conditions, a nominal value o f Rex tr for transition may be
taken as 5 X lO6. However, when disturbances are present, even at a small amount,
R ev ,r decreases substantially. Furthermore, Rex cr is more than an order of mag
nitude sm aller than ReA/r, which means that the length o f the transition region in
low-disturbance boundary layers may be much longer (e.g. thirty times) than the
length o f the initial laminar boundary layer. Then, it becomes obvious that the pre
diction of transition location would be subject to large uncertainty.
When significant disturbances are present a different kind of transition is
observed. This is called bypass transition [ 18] and corresponds to a direct break
down o f the laminar boundary layer, without the appearance of intermediate
phases. A realistic value for transition in boundary layers over flat plates with
significant disturbances (e.g. high-turbulence free stream or large surface rough
ness) is Rextr % 5 X 105.
A simple method for predicting transition is Michel’s criterion [10, 13-15] which
states that transition will occur when Ree(computed using Thw aites’ method) ex
ceeds a certain threshold, whose value has been correlated empirically with Rex.
An improved version of M ichel’s criterion is [ 10]
For a flat-plate boundary layer, this criterion provides Rex t r « 2.5 X 106,
which was the experimental value observed in early transition experiments.
O ther transition criteria as well as the effects o f free-stream turbulence, pressure
gradients, surface roughness, unsteadiness, three-dimensionality, density stratifi
cation, com pressibility and non-Newtonian behavior have been discussed by
several authors [8, 10, 13, 19, 20].
Because of the difficulty in predicting and controlling transition, it is sometimes
desirable to induce it at a known location. Early transition may also be desirable for
other purposes, for example to delay or eliminate boundary layer separation. Tran
sition may be induced by the addition of surface roughness or a “trip wire” in the
laminar region. For effective transition of boundary layers over a smooth wall and
in a low-turbulence free-stream, the diameter D o f a trip wire in contact with the
wall must be selected such that ReD = UeD!v> 850 [10].
F L O W PAST I M M E R S E D OBJECTS 2 0.1 5
A fully turbulent boundary layer is separated from the free-stream by a highly con
torted surface, called the turbulent-non-turbulent interface. The disturbance thick
ness 8 may be taken as the time-average local distance of this interface from the wall.
A turbulent boundary layer may be roughly divided into two distinct regions, the
inner region, in which the flow is dominated by molecular (viscous) stresses, and
the outer region, in which the flow is dominated by turbulent stresses; these regions
are connected by an intermediate, “overlap region.” In turbulent flows, all flow vari
ables are random processes and one must consider statistical averages, such as time
averages (to be denoted by overbars, see Chapter 31). In the inner region over
smooth walls, the dim ensionless average velocity u+ = u/uT is only a function of
the dimensionless distance y + = u rylv. Subdivisions of the inner region and expres
sions for u+ in the inner and overlap regions may be found in Chapter 31, section
31.6.1 [8, 10, 13, 21]. In the outer layer, u+ is a function of y/8 and the pressure
gradient parameter (Clauser parameter)
u - Uc
over smooth and rough walls (Figure 20.8) collapse as long as the value of j8c. remains
constant [8, 10, 13]. An approximate expression for the outer and overlap regions,
valid for boundary layers both in equilibrium (i.e. self-similar) and in non-equi
librium, is as follows [10]:
where k ~ 0.41 (von Karman constant), B ~ 5.0 and the Coles wake parameter II
has been related em pirically to the Clauser parameter as
p c = - 0 .4 + 0.76FI + 0.42FI2
A relatively simple method [10] to compute the parameters 0, H and cf for bound
ary layers with arbitrary free-stream pressure variations is to integrate numerically
2 0.16 FLl'ID F L O W H A N D B O O K
the momentum integral equation (Section 20.2.1), supplemented by the above rela
tionship between /3(. and II, rewritten in the form
- 0 .4 + 0.76ri + 0.42n2 = - - H —
Cf Ue dx
2 _ 2 + 3 .17 9 0 + 1.5II2 H
cf k( I + FI) H - 1
Once the above solution is achieved, the disturbance thickness may be computed as
For 2-D flow over a smooth flat plate, the pressure gradient vanishes (&. = 0,
n ~ 0.45) and one may use the following practical expressions [4], approximately
valid in the range 5 X 105 < Rex < 107, provided that the boundary layer is turbu
lent from the start (x = 0):
1/7
J
(not valid very close to the wall, e.g. for y /S <0 . 1 )
A great deal of research has been dedicated to the study of the structure of tur
bulence in the boundary layers, motivated by the need to control the forces between
the fluid and the immersed object. In particular, the bulk of the wall shear stress has
been associated w'ith randomly recurring flow patterns, known as coherent struc
tures [22, 23].
F L O W PAST I M M E R S E D OBJECTS 20.17
FIGURE 20.8 Velocity defect profiles for self-similar turbulent boundary layers with constant val
ues of the Clauser pressure gradient parameter /3(,
_ P - Pm _ Ul_
= l -
pm~ ~pUen u2
Vem
1/2 I/IO
dc pm I Rex
-pm [x - (x„, - X*)] = a\-
dx \ 10c
For the prediction of separation o f boundary layers that undergo transition see Ref
erence 14. The value o f the recovery pressure coefficient at separation is much
larger for turbulent boundary layers than for lam inar ones, reflecting the higher
resistance to separation posed by turbulence.
F L O W PAST I M M E R S E D OBJECTS 2 0 .1 9
The following analytical results are available for Stokes flow past a sphere (Fig
ure 20.9a) [9 -12]:
• drag coefficient: CD - —
Re
(1 + ---R 3T -------
3R \
2r 2r j
( ^3 ^
• azimuthal velocity: u () = sin 0 -1 + — r -------
\ 4r 4r
3fiR V ^
• pressure: p = ------- 2 ■cos 6
T ro -u V y, s'inO 5R 3 3R \
shear s tre s s:-------------------- 1 +
2r 3 4r )
r <_ Fo
L./) —
fiV ^D
For Stokes flow past a sphere, C'D= 37r. It turns out that C'D is not very sensi
tive to the shape of the immersed object, as long as its dimensions in all directions
are comparable. The values of C'Dfor Stokes flow past a few other types o f objects
are given below [5].
, 4 IT
• cylinder normal to flow (Figure 20.9d): Cp = — — — ——
In (2 LID) + 0.5
(viscous force based on L)
20.20 FLUID F L O W H A N D B O O K
2 77
cylinder parallel to flow (Figure 20.9e): C'n =
ln(2L/D) - 0.72
(viscous force based on Lj
For cases in which Re is not negligible com pared to 1, but still less than 1, the
Oseen approximation provides corrections to the Stokes relationships. For the drag
on a sphere and other sim ilar objects, the Oseen relationships are not substantially
closer to available experimental results than the Stokes relationships, which may
therefore be used in the range 0 < Re < 1. On the other hand, the Oseen approxi
mation provides the following useful expressions, for which the Stokes analysis
provides no results [10]; both expressions are valid in the range 0 < Re < 1.
• drag coefficient on one side of a thin flat plate parallel to flow (Rec = Kociv,
where c is the length of the plate):
_ 87T
W) -
Rec[0.423 + In (16//?^.)]
drag coefficient on a 2-D circular cylinder with its axis normal to flow (Re„ =
V^D /v, where D is the diam eter o f the cylinder):
•3
z , P.
k l
6 /a
<— >
D = 2R (b ) (C)
(a)
(d) (e)
FIGURE 20.9 Sketches of low Reynolds number flows past immersed objects: a) flow past a sphere,
b) flow normal to a circular disc, c) flow parallel to a circular disc, d) flow normal to a truncated cylin
der and e) flow parallel to a truncated cylinder.
F L O W PAST I M M E R S E D OBJECTS 20.21
The drag on a very thin flat plate parallel to an unbounded, uniform flow consists
entirely o f skin friction. For incompressible flows, the drag coefficient generally
depends on the Reynolds num ber Rec = Vx c / p ( c is the length of the plate) and on
the surface roughness. If it may be assumed that the boundary layer is laminar
throughout the plate (for smooth plates in most practical environments, this will
happen when Rec < 5 X 105; for rough plates, transition may occur at a lower Rec),
the drag coefficient on one side of the plate can be computed from the Blasius solu
tion, with a correction term for the low-Re range, as [ 10]
1.328
CD = , for 100 < Rec < 5 X 1O'
\/R e c
For turbulent flows, the effect o f surface roughness may be inferred from the
average height of the roughness element e |6]. When e < 100p/V ^, the plate is called
hydraulically smooth and CD is independent of roughness. W hen e > IOOOiVV*,,
the plate is called fully rough and CD is independent of the Reynolds number. In the
interm ediate case, 100v/V* < e < 1000zVVU one must take into account both the
Reynolds number and the roughness effects. If the boundary layer may be assumed
to be turbulent from the start (e.g. due to roughness or a boundary layer trip near
the leading edge), the drag coefficient may be computed as [4, 6, 8, 10]
If the boundary layer starts as lam inar and undergoes transition at some loca
tion on the plate, at which the Reynolds num ber is Rex tn the drag coefficient may
be estimated as [4]
Rex tr
C d = C oiurb (C pturb ~ ^'Dlam)
Re<
where CDturh and CD/am are, respectively, the drag coefficients for fully turbulent and
fully laminar flow at Rex tr. For a smooth plate with Rex tr = 5 X 105, the result is
0.455 1610
Cn —
D (log Rec)2 5* Rec
20.22 FLUID F L O W H A N D B O O K
The one-side drag coefficient for unbounded flow parallel to a smooth plate,
for a wide range o f Re and assuming transition to occur at Retr = 5 X 105, is shown
in Figure 20.10.
To compute the skin friction drag on flat plates with a pressure variation or on
stream lined 2-D or 3-D objects, one must integrate the wall shear stress along the
surface o f the object, taking into account the relative orientation of the surface ele
ment with respect to the free stream. Approximate analytical and numerical meth
ods have been developed towards that purpose (see sections 20.2.2 and 20.2.4).
Bluff objects are mainly subject to form drag, or pressure drag, generated by the fact
that the static pressure in the part o f the object facing upstream is generally higher
than the pressure in the part facing downstream, which is usually in contact with sep
arated flow. The values of drag coefficients presented in this section [3-6, 10] include
both form drag and skin friction. It is understood, however, that the contribution of
skin friction is small or negligible, except at very low Reynolds numbers.
For Re « 1, Q> on bluff objects is proportional to R e~ \ with the proportional
ity coefficient depending on the shape of the object (Section 20.3.1). Such relation-
Re
F IG U R E 20.10 Drag coefficient for one side of a thin, smooth, flat plate.
F L O W PAST I M M E R S E D OBJECTS 2 0.2 3
ships may be used, with increasing uncertainty, for Re up to about 1, beyond which
the decrease rate of Cn with increasing Re progressively diminishes. In the approx
imate range 103 < Re < 105, Cn is essentially constant. What happens as Re increases
further depends on the shape o f the object, particularly on whether the edges of
the object facing the flow are sharp or rounded. Sharp edges (e.g. those of a tri
angular or rectangular cylinder) fix the separation points and the width of the
low-pressure wake, so that bluff objects with sharp edges have a nearly constant
CD for Re > 103. Rounded edges, on the other hand, permit the separation points
to move along the object’s contour, thus affecting the width of the wake and the
drag; a narrower wake would generally produce a lower drag, compared to a wider
wake at the same Re. The CD vs. Re curves of objects with rounded edges present
a characteristic sharp decrease (“dip”) at a critical Reynolds number Recn which,
for smooth surfaces, occurs in the range between 105 and 106. The reason for this
dip is the shifting of the separation points towards downstream, and the subsequent
narrowing of the wake width. Separation delay is the result of transition to turbu
lence, which increases the momentum of the near-wall fluid by mixing it with high-
momentum free-stream fluid. Addition of roughness on the surface of rounded
objects reduces the value of R ex tr for transition to turbulence in the boundary
layer and, therefore, also the value o f Recr.
In general, 2-D objects have a substantially larger CD than 3-D objects with
the same cross sectional shape; for example, the CD for a circular cylinder in the
precritical Re-range is nearly twice as large as that for a sphere. For objects with
one elongated dimension, called the span (e.g. a truncated cylinder), one may define
the aspect ratio as the ratio b/h between the span b and the frontal height h. CD
increases with increasing aspect ratio and approaches the 2-D value only at val
ues of this ratio as high as a few hundred. For objects with non-circular cross sec
tions, CD also depends on the orientation o f the flow with respect to the object’s
contour. As a rule o f thumb, one may compare the estimated widths of the corre
sponding wakes and infer that Cn will be higher for the case with the wider wake.
For example, a hem isphere oriented with its convex side facing upstream will
have a higher CD than the same hem isphere rotated by 180°.
The drag coefficients for 2-D circular cylinders and spheres have been plotted
vs. Re in Figures 20.11 and 20.12, respectively. For convenience, one may use the
following fitted expressions, valid in the corresponding precritical regimes [101,
The drag coefficients for several bluff objects with sharp edges have been listed
in the following tables. These values are generally valid for Re > 103 to 104 and
subject to a typical uncertainty of ±10% .
2 0 .2 4 FLUID F L O W H A N D B O O K
FIGURE 20.11 Drag coefficient for 2-D circular cylinders; the solid line corresponds to a
smooth surface and the dashed line corresponds to a fully rough surface [3-6|.
Re
FIGURE 20.12 Drag coefficient for spheres; the solid line corresponds to a smooth surface
and the dashed line corresponds to a fully rough surface [3-6].
F L O W PAST I M M E R S E D OBJECTS 20.25
k
D
F IG U R E 20.13 Cross-sectional shapes of 2 -D bluff objects [6|.
20.26 FLUID FLOW HANDBOOK
As examples of the aspect ratio effect, the values o f CD for a finite circular cylin
der and a thin flat plate, both perpendicular to the flow at about Re ~ 105 are given
below [6 ].
The reader is advised to read Section 20.4.3 before the present one. The term drag
due to lift is also com m only referred to as induced drag and occasionally as vor
tex drag. It is associated with the vortices that roll around the two tips of a wing,
due to the positive pressure difference between the lower and upper wing surfaces,
and are convected downstream by the flow. These vortices induce a velocity field
which immerses the wing in a normal velocity field (downwash), in addition to the
free-stream velocity field. As a result, the wing is effectively immersed in flow at
an angle of attack that is sm aller than the actual one. The effective lift direction
also changes, such that it has a com ponent in the direction of the flow, which is
an additional contribution to the drag [4, 5]. The total drag coefficient for incom
pressible flow past a wing may be expressed as
CD = CDoc + CDi
where CDx is the drag coefficient o f an infinite wing with the same profile, some
times called parasitic drag coefficient, and CDi is the induced drag coefficient. CDx
is usually evaluated from wind-tunnel tests using a wing with a uniform cross-
section and spanning the entire wind-tunnel width. The minimum value of CDo0
occurs when the angle of attack a is equal to the angle of zero lift a G\ for well
designed, smooth airfoils, the minimum CD«» may be as low as 0.005 to 0.006,
comparable to or even lower than CD for a two-sided thin flat plate parallel to the
flow. In any case, CDoo is rather insensitive to the angle of attack, as long as it
remains well below the stall angle (see Section 20.4.2).
On the other hand, Q*, which vanishes when a = a (), increases dramatically
with increasing a. For typical cruising conditions of commercial aircraft, the induced
drag is about 40% of the total drag on the wing, but it assumes a much larger sig
nificance during take-off, when the lift coefficient is increased. A simple relation
ship for the computation o f CDi is [14-16, 25-27].
c D i ~- <2
ir(AR)e
where CL is the actual lift coefficient (see Section 20.4.2), (AR) is the aspect ratio
of the wing and e is a correction factor, called the M unk or span efficiency factor,
which should ideally be unity, but for practical wings it is closer to 0.9. Twisted
wings may have a non-zero induced drag even under conditions o f zero total lift.
The total drag coefficient of wings is weakly sensitive to the Reynolds number
and the free stream turbulence level. It generally decreases with increasing Reynolds
number but very slightly for Re > 106 (Re based on the chord length). For Mach
numbers greater than about 0.5 to 0.6, an additional compressibility correction
must be applied, which becomes quite significant in the transonic regime [27, 28].
20.28 FLUID FLOW HANDBOOK
Drag impedes motion and, with a few exceptions (e.g. parachutes and drag sails),
it is an unwanted force. Drag reduction usually results in increased speed, lower
energy losses, higher efficiency and a more economical operation of a system. A num
ber of strategies have been developed over the years for drag reduction, and a great
deal of effort is being spent for the development of new approaches and the reso
lution of controversies that remain over existing ones [4-6, 29-34]. For fully
immersed objects, drag reduction strategies are mainly targeted towards reducing
skin friction, form drag or induced drag, but one should keep in mind that a mod
ification of one type of drag may have an effect on another. Methods of drag
reduction may be either active, requiring external power, or passive. In the analy
sis o f the benefits of a drag reduction technique, one must consider the additional
energy losses associated with the devices used, as well as the cost of the imple
mentation of the technique compared to the savings from drag reduction.
Form Drag Reduction. Form drag is the dominant component of drag for bluff
objects and it is usually related to separation in the aft region of the object. Large form
drag also occurs on wings and other streamlined objects at large inclination to the
flow, in which cases it is also due to extensive flow separation. Therefore, the effort
tor drag reduction is focused upon reducing or eliminating flow separation, both
steady (separation bubble) and unsteady (vortex shedding). An obvious and effective
approach is to streamline a bluff object by the addition of an afterbody that ends in a
thin edge. The shape and length of the afterbody may be optimized for a drag
reduction by as much as 95 percent. Rounding or sharpening of the nose of elongated
objects may also result in reduced form drag. Placing an object closely downstream
of another usually results in a total drag that is lower than the sum of the drag forces
on the two objects positioned far apart. Delay or reduction of separation may be
effected by increasing the momentum of the near-wall fluid, such as it may resist the
adverse pressure gradient. The addition of surface roughness or “trips” in a laminar
boundary layer induces early transition to turbulence, which results in enhanced mix
ing of low-momentum, near-wall fluid with high-momentum, free-stream fluid.
Examples include the use of dimples on golf balls and boundary layer trips for low-
speed aircraft wings. Injection of high-momentum fluid through the wall, either tan-
gentially through slots or perpendicularly through a porous surface has a similar
effect. Vortex generators, short slanted airfoils normally mounted to the wall near the
leading edge of wings, transport free stream fluid towards the wall by vortical actions.
The removal of the boundary layer or even the entire separated region by wall
suction results in significant drag reduction, as well as delay of wing stall, although
suction also consumes considerable power. Finally, wall motion, in the form of a
moving belt or a rotating cylinder, has also been used to reduce wall shear stress and
flow separation, while also producing lift via the M agnus effect (Section 20.4.4).
Skin Friction Reduction. Because laminar boundary layers have much lower wall
shear stress than turbulent ones, skin friction on streamlined objects may be reduced
FLOW PAST IMMERSED OBJECTS 20.29
Induced Drag Reduction. The induced drag on lifting surfaces decreases with
increasing aspect ratio, so an obvious strategy is to increase the aspect ratio, while
maintaining an approximately constant surface area. This approach may be readily
observed on many long-distance flying birds (e.g. albatross) and has been applied
successfully on gliders and special types of aircraft. Flowever, the increase of wing
span is limited by structural requirements and, for this reason, most aircraft have
relatively short wings. For a given aspect ratio and wing shape, induced drag may
not be easily eliminated without also eliminating lift. Several approaches involving
the use of multiple surfaces (e.g. biplanes) and different wing configurations, includ
ing upsweep and twist, have been explored, but the most practical ones involve
the use of wing tip devices (e.g. vertical winglets, tip fences and other patented
devices) to reduce the average downwash by blocking or diverting the tip vortices
and even to generate a small amount of thrust.
Besides the above, special methods have been developed against other types
of drag. Examples include the use of bulb-shaped bows and sterns of ships to reduce
wave drag and sweep in wing design to reduce compressibility drag.
The term flow -induced vibrations refers to the oscillatory motion of immersed
objects as a result of aero- and hydro-elastic interactions between the object and
20.30 FLUID FLOW HANDBOOK
V.
Re
FIGURE 20.14 Strouhal number for a smooth circular cylinder normal to the flow; vertical line
indicates the minimum Reynolds number for vortex shedding [47].
FLOW PAST IMMERSED OBJECTS 20.31
frequency of the object, large-am plitude vibrations may occur. This phenomenon
is far more severe for 2-D and elongated objects than for objects with comparable
dimensions in all directions, because the object loading due to long vortices is
stronger. Tables and graphs containing the Strouhal number for objects with com
mon shapes are available [6 , 35, 36]. O f particular interest is vortex shedding
from circular cylinders, whose Strouhal number has been plotted in Figure 20.14
[37]. For Re < 47, no vortex shedding occurs, while for Re > 400, S ~ 0.20 ±
0.02. The uncertainty of S in the critical and supercritical regimes (Re > 200,000)
is significant. As a rule of thumb, the Strouhal number varies inversely to the
width of the wake (and, thus, the value of the drag coefficient). Thus, the high-
Re Strouhal number for a square cylinder is lower than that for a circular cylin
der. Vortex shedding from objects with free separation points is sensitive to
surface roughness and free stream turbulence. W hen the object is free to oscillate,
its motion interacts with that o f the fluid, resulting in com plicated patterns, such
as phase-locking [38].
Flow-induced vibration may also be the result of time-dependent loading of an
object immersed in oscillatory or highly turbulent flow. Finally, it could be due to
hydro-elastic instability, as for example in the case of two cylinders parallel to each
other and normal to a flow stream, in which case the vibration is excited by the
pressure force generated in the low pressure region between the cylinders [5, 351.
20.4 LIFT
A lift force on an im m ersed object is generated any time there is flow asymm e
try about a plane parallel to the flow direction [7]. Examples of lift generation
mechanisms include the following configurations:
• a symmetric object with its axis inclined with respect to the flow direction
(Figure 2 0 .15a)
• any asymmetric object at an arbitrary orientation (Figure 2 0 .15b)
• any object in shear flow (Figure 2 0 .15c)
• any object in density stratified flow (Figure 2 0 .15d)
• a rotating object (Figure 2 0 .15e)
Lift is usually generated with the use of specially designed, streamlined objects,
called wings or blades, depending on the application. A wing is characterized by
its cross-sectional shape, called an airfoil, and its projected surface, called the
planform. An understanding o f lift characteristics may be first based on the study
of 2-D wings, referred to as airfoil theory (Section 20.4.2). The important effects
of finite wing span (Section 20.4.3), taper, twist and sweep may then be exam
ined separately.
20.32 FLUID FLOW HANDBOOK
a) b)
e)
FIGURE 20.15 Illustration of different lift generation mechanisms.
The lift may be viewed as the result of surface pressure variation, while the
contribution of wall shear stress is usually negligible (see Section 20.1.1). For many
practical applications, the computation o f lift based on inviscid flow theory is of
sufficient accuracy (it is reminded that such approaches are entirely inadequate for
the computation o f drag). A particularly suitable method for aerodynamic appli
cations is the vortex panel method [14, 15J.
The flow asymmetry about an object is characterized by the circulation, defined as
FLOW PAST IMMERSED OBJECTS 20.33
where the integration is carried around a closed curve C surrounding the object. [1,5,
14-16]. The lift per unit span b on a 2-D object is related to the circulation by the
Kutta-Joukowski theorem, which states that
v Jr
FL = - p V
—
b
20.4.2 Airfoils
A typical, low-speed (i.e. not intended for supersonic flows), aeronautical airfoil
is sketched in Figure 20.16. The centers of all inscribed circles form a line called
the camber line, or midline. The end o f the airfoil facing the flow is rounded and
called the leading edge, while the rear end, forming a small angle 6, is called the
trailing edge. The straight line connecting the leading edge and the trailing edge
is called the chord line and its length c is the chord. The distance between the
camber line and the chord is called the camber. Airfoils that have a zero camber
everywhere are called symmetric, while those with non-zero camber are called cam
bered. The distance between the two surfaces of the airfoil, in the direction normal
to the chord, is called the thickness. The angle a between the flow direction and
the chord line is called the angle o f attack, or incidence, and it is considered pos
itive if the flow points towards the concave side of the camber line. Then one may
distinguish between the upper (or suction) surface and the lower (or pressure) sur
face of the airfoil.
A large variety of airfoils have been designed, tested and optimized for par
ticular applications, including aircraft wings, propellers, turbine and compressor
blades, hydrofoils and struts [28, 39]. Detailed characteristics for different series
o f airfoils, notably several series o f NACA profiles, are available, so the need for
designing new airfoils only arises under exceptional circumstances.
Typical variations of the lift coefficients of symmetric and cambered airfoils
plotted vs. the angle of attack are shown in Figure 20.17. For a = 0°, the sym
metric airfoil has cL - 0, while the cam bered airfoil has cL > 0. The value cL = 0
is attained by the cambered airfoil at an angle a () < 0 , called the angle of zero lift.
As a increases, cL also increases, essentially linearly, up to a certain value o f a,
typically around 10-15°. In that range of a , one may approximate cL as
C [ CO ----- k r x j ^ C X f t y )
where k^ is the slope of the c*L-line and the subscript indicates that this expres
sion applies to two-dimensional airfoils. As a increases further, cL increases slower
than linearly up to a maximum value cLmax,, occurring around a ~ 15-20°. For fur
ther increases o f a, cL drops dramatically, while the drag coefficient also increases
dramatically in this range. This phenomenon is called stall and it is caused by mas
sive flow separation on the upper surface of the airfoil [4-6].
A fairly accurate prediction for the value o f fc* is provided by the thin airfoil
theory [14, 15], which is based on the assumptions of thin airfoil, small camber and
FIGURE 20.17 Typical variations of the lift coefficients of symmetric and cambered airfoils.
FLOW PAST IMMERSED OBJECTS 20.35
small angle of attack, all of which are satisfied in usual aeronautical applications.
The result is
ky, = 2 tt
where it is assumed that the angles a and a n are expressed in radians. Empirical
linear fits to experim ental results at moderate Reynolds numbers give [25]
hx « 5.7
with the slope increasing slowly towards the ideal value 277as the Reynolds num
ber increases. A slightly more accurate expression for the variation of cL that takes
into account the weak non-linearity of the lift line is [25]
The above expressions are valid for angles of attack well below the stall angle.
In all these expressions, the angle of zero lift (in radians) may be roughly com
puted as
^hmax
a (t = ------------
c
where hmax is the maximum cam ber o f the airfoil and c is its chord length [5]. A
typical range for cam bered aeronautical airfoils is -5 ° < a„ < -2 °. More accurate
methods for com puting a () are also available [16].
A numerical method for the com putation of lift for airfoils of arbitrary thick
ness and camber is the vortex panel method [ 14,15]. This is an inviscid flow method,
by which the surface of the airfoil is approximated by a set of plane elements,
called panels. Each panel is assum ed to be a vortex sheet, consisting of an infi
nite number of distributed, parallel vortex filaments, each with infinitesimal
strength. To obtain a solution, one must satisfy the no-penetration condition at
selected points on the surface as well as the Kutta condition, which essentially
ensures that the trailing edge is a stagnation point. The vortex panel method and
similar approaches may be actually used for wings with multiple surfaces (e.g. a
main wing with a flap) and for arbitrary, three-dimensional, lifting surfaces.
20.4.3 Wings
A few typical planforms of aircraft wings are sketched in Figure 20.17. One may
identify the following geom etrical characteristics [40, 41].
• the leading edge, the trailing edge and the quarter-chord line
• the root (i.e. the cross section on the plane of symmetry) and the two tips (i.e.
the cross sections at the two ends of the span)
• the local chord, c, the root chord, cn the tip chord, c„ and the average chord,
- ,4
c =—
b
• the sweep angle, which is the angle between the quarter-chord line and the
normal to the root plane; depending on whether this angle is zero, positive
flow
U nsw ept wings Sw ept W ings
o
o
I ' 1 CL,
leading edge!
quarter-chord iline
" f
elliptical
backswept
tapered
(trapezoidal)
span
In addition to the main wing, aircraft may utilize other lifting devices, includ
ing the following:
The expressions for cLoc given in Section 20.4.2 were either derived under the
assumption of two-dimensional flow (namely of an infinite wing with a constant
cross-section) or measured using trapezoidal wings spanning the entire width of the
wind-tunnel. When a wing operates with its tips free in the flow, however, the high-
pressure fluid from the pressure side tends to flow around the tip towards the suc
tion side, thus creating a vortical motion. As the aircraft moves through the fluid,
it leaves behind it two counter-rotating vortices, originating at the wing tips and
with axes nearly parallel to the direction of motion. These vortices are called wing-
tip vortices. An alternative explanation for these vortices is with the use of the sec
ond Helmholtz theorem [3, 14] for inviscid flow, that states that a vortex may not
terminate in a fluid. As a lifting wing may be approximated by a vortex (bound vor
tex), it follows that this vortex must bend at the two tips and form a closed loop;
the fourth segment of this loop is the starting vortex, which has an axis parallel to
the wing and is left behind by the aircraft when it first begins its motion.
The wing-tip vortices induce a velocity field (downwash) that approaches the
wing in a direction normal to the flow. This effectively reduces the angle of attack
and creates a lift that is inclined with respect to the normal to the actual flow
direction. The result of this is a reduction of the conventional lift coefficient and
the generation of an additional drag component, called the induced drag (Section
20.3.4) [4, 5, 14-16, 40-43]. The Lanchester-Prandtl theory for elliptical wings
provides an expression for the lift coefficient of a finite wing with an aspect ratio
(AR) as
cL = k(a - a 0)
20.38 FLUID FLOW HANDBOOK
where
k = ------—
1 + k c
tt(AR)
2 i r ( a - a ())
CL —
. + 2
(AR)
k=
1 + ~~ (1 + t)
tt( A R)
where the correction r is typically between 0.05 and 0.25, depending on the plan-
form shape of the wing. When compressibility becomes significant, the above lift
coefficient expressions need additional corrections. A simple correction method is
the Prandtl-Glauert correction, valid in the range of subsonic Mach numbers 0.3 <
M < 0.8, which gives [43]
_ k(a - a 0)
L - /--------
V l - M2
Lift is produced on rotating immersed objects due to the asymmetry of the flow
around them. This phenom enon, generally known as the Magnus effect, has in the
past been explored for propulsion (Flettner rotor) and power generation [5] and
is widely utilized in sports (e.g. curve-ball in baseball [44]). Figure 20.19 illus
trates typical variations of the lift and drag coefficients for smooth spheres [4, 45]
and cylinders [4,45,46] rotating with an angular velocity, a>, plotted vs. the spin ratio
(i.e. the ratio of the surface velocity and the far-field velocity) (oD/2Vx . All coef
ficients seem to approach asymptotes as the spin ratio increases to values near 5.
The results shown in Figure 20.19a for spheres correspond to precritical basic
flows. For spin ratios lower than a certain value (e.g. about 0.4 for Re ~ 107,000)
FLOW PAST IMMERSED OBJECTS 20.39
0.7 a) b)
9
0.5 7
0.4 6
0.3 / ° L 5
c7
4
0.2
3
0.1
/
2
0.0
1
0.1
0
0 1 2 3 4 5 0 1 2 3 4 5
coD/2Vx coD /2
F IG U R E 20.19 Lift and drag coefficients for a) rotating spheres |4 , 44] and b) rotating cylinders
[4, 44, 45].
a negative lift, with reference to the positive direction shown in Figure 2 0 .15e, is
observed. This is caused by a delay o f separation on the advancing side, due to
early transition to turbulence, while the retreating side remains precritical. For
higher values of the spin ratio, or when the basic flow is already postcritical, a
positive lift is produced due to a delay of separation on the retreating side [44].
NOMENCLATURE
A area, ft2 (m 2)
(AR) aspect ratio
Cl coefficient in the turbulent Stratford criterion
B constant for the logarithmic law o f the wall
b span, ft (m)
c chord length, length o f a plate, ft (m)
c average chord length, ft (m)
CD drag coefficient
c Di induced drag coefficient
20.40 FLUID FLOW HANDBOOK
GREEK SYMBOLS
REFERENCES
1. Karamcheti, K. 1966. Principles o f Ideal Flow Aerodynamics. John W iley & Sons,
New York.
2. Batchelor, G.K. 1967. An Introduction to Fluid Dynamics. Cambridge University
Press, Cambridge.
3. Panton, R.L. 1984. Incom pressible Flow. John Wiley & Sons, New York.
4. Fox, R.W. and McDonald, A.T. 1998. Introduction to Fluid M echanics. 5th Edition
John W iley & Sons, New York.
5. Gerhart, P.M, Gross, R.J. and Hochstein, J.I. 1992. Fundam entals o f Fluid M echan
ics. 2nd Edition Addison Wesley, Reading (Massachusetts).
6. Blevins, R.D. 1984. A pplied Fluid D ynam ics Handbook. Van Nostrand Reinhold
Company, New York.
7. Corrsin, S. 1974. Notes on Fluid M echanics. Department o f Materials Science and
Mechanics, The Johns Hopkins University.
8. Schlichting, H. 1979. B oundary-Layer Theory. 7th ed. M cG raw-Hill, New York.
9. Rosenhead, L. (Ed.) 1963. Lam inar Boundary Layers. Oxford University Press,
Oxford.
10. White, F.M. 1991. Viscous F luid Flow. 2nd ed. M cGraw-Hill, New York.
11. Happel, J. and Brenner, H. 1965. Low Reynolds Num ber Hydrodynamics. Prentice
Hall, Englewood Cliffs.
12. Langlois, W.E. 1964. Slow Viscous Flow. Macmillan, New York.
13. Schetz, J.A. 1993. Boundary Layer Analysis. Prentice Hall, Englewood Cliffs.
FLOW PAST IMMERSED OBJECTS 20.43
14. Kuethe, A.M . and Chow, C.-Y. 1998. Foundations o f Aerodynam ics. 5th ed. John
W iley & Sons, New York.
15. Moran, J. 1984. An Introduction to Theoretical and Com putational Aerodynamics.
John W iley & Sons, New York.
16. Thwaites, B. (Ed.). 1960. Incom pressible Aerodynamics. Oxford University Press.
Oxford.
17. Kachanov, Y. 1994. “ Physical Mechanisms o f Laminar-Boundary-Layer Transition"
Ann. Rev. F luid Mech. 26:411-482.
18. Morkovin, M. 1984. “ Bypass Transition to Turbulence and Research Desiderata.” In
Transition in Turbines. NASA Conf. Publ. 2386, pp. 161-204.
19. Reshotko, E. 1976. “ Boundary-Layer Stability and Transition.” Ann. Rev. Fluid
Mech. 8:311-349.
20. Reed, H.L. and Saric, W.S. 1989. “ Stability o f Three-Dimensional Boundary Lay
ers.” Ann. Rev. Fluid Mech. 21:235-284.
21. Pope, S.B. 2000. Turbulent Flows. Cambridge University Press, Cambridge (UK).
22. Robinson, S.K. 1991. “ Coherent Motions in the Turbulent Boundary Layer." Ann.
Rev. Fluid Mech. 23:601-639.
23. Van Dyke, M. 1982. An Album o f Fluid Motion. Parabolic Press, Stanford.
24. Stratford, B.S. 1959. “ The Prediction o f Separation o f the Turbulent Boundary
Layer.” J. Fluid Mech. 5:1-16.
25. Von Mises, R. 1959. Theory o f Flight. Dover Publications Inc., New York.
26. Kroo, I. 2000. “ Drag Due to Lift: Concepts for Prediction and Reduction.” Ann.
Rev. Fluid Mech. 33:587-617.
27. Glauert, H. 1926. The Elem ents o f A erofoil and Airscrew Theory. Cambridge Uni
versity Press, Cambridge.
28. Abbott, I.H. and von Doenhoff, A.E. 1959. Theory’ o f Wing Sections, Including a
Sum m ary o f A irfoil Data. Dover Publications, Inc., New York.
29. Gowen, F.E. and Perkins, E.W. 1952. “ Drag o f Circular Cylinders for a Wide
Range o f Reynolds Numbers and Mach Numbers.” NACA RM A52C20.
30. Hoerner, S.F. 1965. Fluid Dynam ic Drag. Hoerner Fluid Dynamics, Bricktown, NJ.
31. Bushnell, D.M. and Moore, K.J. 1991. “ Drag Reduction in Nature.” Ann. Rev. Fluid
Mech. 23:65-79.
32. Bushnell, D.M. 1983. “ Turbulent Drag Reduction for External Flows.” Proc. A IA A
21st Aerospace Sciences Meeting, Reno, Nevada, January 1983, AIAA-83-0277.
33. Gad-el-Hak, M. 1989. “ Flow Control.” Appl. Mech. Rev. 42:261-293.
34. Chang, P.K. 1976. Control o f Flow Separation. Hemisphere Publishing Co., Wash
ington DC.
35. Blevins, R.D. 1977. Flow Induced Vibration. Van Nostrand Reinhold Company,
New York.
36. Roshko, A. 1954. “ On the Drag and Shedding Frequency o f Two-Dimensional B lu ff
Bodies.” NACA TN 3169.
37. Fey, U., Konig, M. and Eckelmann, H. 1998. “ A new Strouhal-Reynolds-number
relationship for the circular cylinder in the range 47<R e<2xl0'\” Phys. Fluids
10:1547-1549.
38. Bearman, P.W. 1984. “ Vortex Shedding from Oscillating B lu ff Bodies.” Ann. Rev.
Fluid Mech. 16:195-222.
39. Eppler, R. 1990. Airfoil Design a n d Data. Springer-Verlag, Berlin.
40. Bertin, J.L. and Smith, M .L. 1998. Aerodynam ics f o r Engineers. 3rd ed. Prentice
Hall, Upper Saddle River, NJ.
20.44 FLUID FLOW HANDBOOK
41. Anderson, J.D., Jr. 2001. Fundam entals o f A erodynam ics. 3rd ed. M cG raw-Hill,
New York.
42. Glauert, H. 1926. The Elements o f A erofoil a n d A irscrew Theory. Cambridge U ni
versity Press, London.
43. Shevel, R.S. 1989. Fundam entals o f Flight. 2nd ed. Prentice Hall, Englewood
C liffs.
44. Mehta, R.D. 1985. “ Aerodynamics o f Sports Balls.” Ann. Rev. F luid Mech.
17:151-189.
45. Goldstein, S. (Editor). 1965. M odern D evelopm ents in F luid D ynam ics. Dover Pub
lications Inc., New York.
46. Hoemer, S.F. and Borst, H.V. 1975. F luid-D ynam ic Lift. Hoemer Fluid Dynamics,
Bricktown, NJ.
__________CHAPTER 21__________
TRANSPORT PHENOMENA
IN POROUS MEDIA
21.1 INTRODUCTION
Fluid flows through porous materials occur in a wide range of applications. Each
application may have its own body of knowledge and mathematical expressions tor
design and prediction o f its processes. Often the modeling equations do not easily
transfer between applications. It is not possible to cover all of the topics in great
depth and do them justice. Hence the approach used in this chapter is to provide a
generalized model of flows through porous media that in principle can be applied
to most of the applications.
This chapter begins with a brief physical description what is meant by porous
media. This is followed by a shell balance derivation of the general multiphase
equations. The general multiphase equations have more unknowns than equations;
hence constitutive relations must be introduced to obtain mathematical closure. The
remainder of the chapter takes macroscale correlations and applies them to small dif
ferential volume elements to formulate plausible constitutive functions for momen
tum, heat and mass transfer coefficients.
In recent years literature has described the derivation of the general multiphase
continuum equations using a method of volume averaging. The volume averag
ing approach is rigorous and produces the desired set of equations. However, the
volume-averaged derivation is lengthy and is available in literature. Instead, a shell
balance approach is applied to derive the desired equations. The reader may com
pare these equations with those derived using the volume averaged theory and find
that they are analogous. The shell balance approach may have one advantage over
the volume averaged approach: it allows one to better intuitively understand the
meanings of the terms in the equations. This can aid in understanding how to apply
the equations.
21.1
21.2 FLUID FLOW HANDBOOK
The shell balance approach here explicitly applies the area fraction available
for flow to derive the general equations. These equations simplify to those obtained
by volume averaging only after the assumption is made that the area fraction is equal
to the volume fraction. This result suggests that the volume-averaged approach may
be more restrictive than originally thought. Furthermore, this approach is appeal
ing if one considers a media that has properties dependent upon direction (oriented
media). As an example, suppose a porous media has the shape of a cube with (x,y,z)
axes parallel to edges of the cube. The cube has holes drilled through it parallel to
the *-axis and parallel to the v-axis such that the holes in the two directions do not
intersect. In effect this is a multidirectional capillary tube model. The tubes in the
two directions need not be equal in number or in size. The superficial velocities in
both directions may be the same, however the intrinsic pore averaged velocities in
the two directions will differ because of different areas available for flow. With
out taking the area effect into account, the volume averaged approach would give
us the same intrinsic pore averaged velocities because the porosity is a volume aver
aged quantity that does not have a directional sense.
Also in this chapter, macroscopic correlations are used to deduce plausible con
stitutive functions for the model. Literature abounds with data and correlations de
rived from experim ents at the macroscale. Few data are available that truly apply
to the volume averaged continuum scale. To bridge this gap in available informa
tion, the macroscopic correlations are scaled down to the continuum scale. In many
cases the results are reasonable and this approach provides us with mathematical clo
sure on the model equations. Certainly, when available, relations rigorously derived
from theory should be preferred over the scaled down macroscopic correlations.
To understand how to model flows through porous media, one needs to have an
intuitive description o f what is or is not included in the mathematical models. A
porous material is one that is made up o f at least tv/o distinct and separate phases
that are interdispersed. For flow to occur, at least one of the phases must be con
tinuously connected. The other phase may be continuously connected or may be
discrete particles. An example o f the latter is air flowing through a Fixed bed of
Lucite™ spheres (as shown in Figure 21.1). An example o f the former is that of
water flowing though a sponge; the water is the continuously connected fluid phase
and the skeletal matrix o f the sponge is the continuously connected solid phase.
Water flowing through a fluidized bed o f sand is another example of water being
the continuous fluid phase and the sand being the discontinuous solid phase. How
ever, in the fluidized bed the particles o f the solid phase are free to move relative
to each other unlike the sponge or fixed bed.
TRANSPORT PHENOMENA IN POROUS MEDIA 21.3
F IG U R E 21.1 Photograph o f a layer o f a Lucite™ particle bed. The spheres have a range o f sizes
from 2 0 to 150 microns. The photo show s the gaps and pores between particles through which fluid
may How.
Particles in the solid phase may contain smaller particles and have an internal
pore structure. The smaller scale structures o f the particles affect the material behav
ior, such as internal coefficient of friction or stickiness between particles. Feda [11
gives an excellent comparison of such a structural approach to that of a phenomeno
logical approach. In the phenomenological approach the solid phase is treated as a
continuum. The continuum equations do not model the lower scale structural prop
erties of the solid phase, but the lower scale properties are accounted for in the
values of the coefficients in the constitutive relations.
The phenomenological approach is pragmatic in that in applications such as soil
mechanics, the solid particles are a mixture of size, shape, and material content.
Under such conditions it is very difficult to characterize the material structurally
and to predict solids’ behavior starting from the structure description. The phenom
enological approach is to directly measure the continuum defined properties and
to use those properties in the continuum equations. The continuum equations are
typically derived from lower scale continuum balances through the method ot vol
ume averaging [2- 6 ] or by shell balance as is done here.
The diagram in Figure 21.2 indicates where the model equations fit into the
broader context o f mechanics in general. The scope of this chapter is focused on
21.4 FLU ID FLOW HANDBOOK
MECHANICS
CONTINUUM DISPERSION
MECHANICS MECHANICS
FLUID-SOLID FLUID-FLUID
MODELS MODELS
/
STRATIFIED DISPERSIONS
FLOW AND
MODELS EMULSIONS
fluid-solid two phase systems including rigid and compressive solid matrices. Fluid-
solid systems are a subset of two-phase systems that also include fluid-fluid sys
tems such as droplet dispersions in im miscible fluids [7]. Two-phase systems are
a subset o f the more generalized multiphase volume-averaged continuum theory.
The shell balance approach could be easily extended to three (or more) phase sys-
TRANSPORT PHENOMENA IN POROUS MEDIA 21.5
terns such as oil, water, and gas. The three phase systems have three-phase effects
that result in capillary pressure due to surface tension phenom ena | 8 |. To model
three phase systems requires constitutive relations not included here that account
for the three phase effects [9]. The multiphase continuum equations can be derived
from principles of continuum mechanics or from dispersion mechanics, the latter
here meaning models tracking motions of individual particles in particle clouds.
Shell balances provide a simple but effective way to derive equations for flows through
porous media. First, let’s derive a simplified shell balance for one-dimensional flow
through a packed bed section. This is followed by a more generalized derivation.
Consider a shell balance over a differential elem ent for simple steady one-dim en
sional fluid flow through a packed bed as shown in Figure 2 1.2. The flow is in
the Z direction from left to right.
A shell balance only considers the relevant terms. A momentum balance is set
up over the differential section o f thickness Az. The flow is at steady state; hence
the accumulation term is zero. The diam eter o f the packed bed section is large
D IF F E R E N T IA L SEC TIO N
Az
I
D IR
CROSS-
^....... ...............
SECTIONAL
SURFACEA AREA, PACKED BED O F LENGTH L
A*
FIG U R E 21.3 Fluid flow through a packed bed. Indicated in the bed is a differential section of
length Az and diam eter 2R. The fluid How direction is through the circular surface at the end o f
th e bed. A,.
21.6 FLUID FLOW HANDBOOK
compared to the particles, hence the drag o f the fluid along the wall is negligible.
The com ponents of the momentum balance are
The velocity of the fluid phase in the z direction, v{, is the average velocity
within the pore spaces in the interstices between the solid particles. P is the fluid
pressure, p f is the density of the fluid phase within the pores, A{ is the area avail
able for fluid flow through a plane normal to the z direction, and A z is the total
cross section area of the bed in a plane normal to the z direction equal to k R 2. F (
is the drag force component in the z direction representing the momentum lost by
the fluid to the solid phase. We add up the contributions to the momentum balance:
( 2 l . 6)
a{_ = (21.7)
a £ H - ' ,U ) - A ^ / = 0 (21.8)
We now divide Equation 21.8 by Az and take the limit as Az—>0; this gives
dP
aL — = - F ' (21.10)
dz
The drag force is typically modeled as being proportional to the velocity through
the bed as
F / = R f v{ (21.11)
where /J is the fluid viscosity and RJ is a resistance function related to the perme
ability, k, by k = ( a L ) 2fj/R J. Hence, com bining Equations 21.10 and 21.11, we
get the familiar D arcy’s law at the local scale,
k dP
- — = -a lv ! (21. 12)
IJ dz
Integration of Equation 21.12 over the volume of a bed of length L gives us the
macroscale D arcy’s law
Q = n R 1- ~ (21.13)
/i L
This shows that for relatively simple systems the ad hoc approach can be used
to develop the governing equations. For more complex systems a more general
ized shell balance approach is preferred.
Rate of
accumulation of
A jcA vA z (21.15)
(p in the volume dt
A jcA vA z
Rate of
convection in A y & z a l p a<pavax x+AxAzcc“yp a<pav°
across surfaces (21.16)
of volume + A xA ya “p a<pava\_
element
Rate of
convection out A y A z a ° p ap v ° x+Ax + A xA z a “yp a<l>“v° v+Av
Rate of
homogeneous
AxAyAzeap n( f + g a) (2 1 .20)
generation
of 0
Rate of
heterogeneous
-A xA yA z[E a + l a + S a + C " ) ( 21 .21)
generation
o f (p
where
d(eap a<pa )
+ £ ap a ( r + g a )
dt dx dy dz
- ( £ “ + / “ + Sa + G a)
( 2 1 .2 2 )
(21.23)
The area tensor, a « , is a diagonal tensor such that the off-diagonal compo
nents are zero. Using index notation this is written as
< 0 (21.24)
0 (XZ
Substitution of terms from Table 21.1 provides expressions for specific property
balances of mass, momentum, energy, and species. In most applications constitu
tive relations are needed for the flux and generation terms to obtain mathematical
closure on the system o f equations.
In Table 21.1 the term _r is the total stress tensor, sum of the deviatoric stress and
pressure terms t_ = T + S P , q is the thermal heat flux vector, h is the external
energy supply (such as electrical heating), is the species mass flux vector, coA is
the mass fraction of species A, and rA is the rate of mass generation of species A
due to homogeneous chemical reaction.
21.10 FLUID FLOW HANDBOOK
TABLE 21.1 Term s to Substitute into General Balance to Derive Balances o f Mass
C ontinuity, M om entum , Energy, and Species
0 1 V 0 + 1 v2 (0A
2~
(Property/m ass)
i_ 0 I q +1 * v JA
(Flux)
(f + g ) 0 g g - V+ h rA
P
(B ody sources)
In many applications, such as flows through granular materials, the void area avail
able for flow is independent of direction, such that
ea = a a (21.26)
That this must be true is deduced by starting with a volume element of a finite thick
ness. As the thickness of the volume element shrinks to a small dimension, the vol
ume occupied by a fluid phase becomes equal to the area available for flow times
the differential thickness. The volume o f the volume element is the cross-sectional
area times the differential thickness. Thus the ratio of the volumes equals the ratio
o f the areas. Hence, the general property balance of Equation 21.23 becomes
^ - ( e a p a<pa) + V ■( e np a<t>a v“ ) + V ■e a i a
d t' ' (21.27)
- e ap a ( f " + g “ ) + ( E a + Ia + S a + G a ) = 0
TRANSPORT PHENOMENA IN POROUS MEDIA 21.11
In a two phase system one can see from Equation 2 1.27 that as ea approaches
zero, the first four terms o f Equation 2 1.27 become negligibly small, hence the het
erogeneous generation terms, (Eu + l a + S a + G a ), are similarly negligible, as ex
pected due to small interfacial area of contact in a dilute multiphase mixture. When
£a approaches unity, again the interfacial area is small and the heterogeneous terms
are neglected. However, in this latter case. Equation 21.27 reduces to the single
phase balance equation.
Furthermore, this form of the balance is appealing if we consider what we call
a “null” phase. A null phase is one that has no effect on surrounding phases other
than it occupies space. A null phase does not transfer mass, momentum, or energy
between phases. A null phase is a hypothetical condition such that the system
behaves as though the volume space occupied by the null phase is simply empty. For
example, consider a two-phase system such as water flowing through a packed bed
of Teflon coated solid particles, the solid being a null phase. The Teflon allows the
water to slip past the particles with little or no drag force on the particle surfaces.
In this example, all of the heterogeneous terms in Equation 21.27 are zero (because
there are no property transfers between the phases), and as a first approximation, the
single phase constitutive relations can be applied for the constitutive functions.
Another feature o f the multiphase continuum equations is that they must apply
to mixtures as a whole. When the phase averaged continuum equations are summed,
the resulting equations must be analogous to the single phase equations. These region
equations are applied when one wants to treat the mixture as a pseudo-single-phase.
Summing Equation 21.27 over all of the phases, one gets balances that have the
same form as single phase equations:
where
and
^ [ E a + I a + S (X+ G “] = 0 (21.35)
a
Equation 21.28 assumes that the sum of the convection terms may be approx
imated by
(21.36)
a
The approximation in Equation 21.36 is reasonable for dilute systems or for systems
in which the velocities of all of the phases are the same. Next, the equations for the
specific properties are derived.
(21.37)
in which the first term is the mass accumulation in the a phase, the second term
is the mass convection, the third term represents the mass leaving the a phase to
the adjoining phases due to phase change (i.e., melting, solidification, or vaporiza
tion), and the fourth term represents the rate of mass leaving the a phase due to
heterogeneous reaction at the interface with the adjoining phases. Using the defi
nitions in Equations 21.29 through 21.36, the region mass continuity balance is
(21.38)
w h e r e Z ( ^ + ^ ) = 0 (21.39)
a
TRANSPORT PHENOMENA IN POROUS MEDIA 21.13
For the conservation of momentum set (pa = va, = _ra, and f a = g. The slip
term, S m , accounts for capillary force contributions in unsaturated flows through
porous materials, and 1% accounts for the drag forces between the phases. The
combination of capillary forces and drag forces between the phases produces a
net drag force between the phases, less a form drag effect due to variations in the
local porosity. This is given by
S aM +LaM = E a - P V £ a (21.40)
where F a is the drag force of the a phase acting on theadjoining phases, and P
is the fluid pressure. The second term on the right side of Equation 21.40 accounts
for the form drag due to the changing porosity.
The stress tensor is separated into the fluid normal pressure and the shear or
deviatoric components, given by
t_a = P8 + z a (21.41)
— ( £ “ p a Va ) + V - ( £ ap a Va Va ) + £ a V P + V - f a r a + F a + ^ M + ^ M - ^ P ^ = 0
dt v 7 v 7
(21.42)
Here, the first term is the unsteady inertial term, the second term is the steady-state
inertial term, the third term accounts for the pressure forces, the fourth term accounts
for the forces due to shear stresses between the a phase and the system boundaries,
the fifth term represents the drag force by the a phase acting on the surrounding
phases, the sixth and seventh terms represent the momentum leaving the phase due
to phase change and heterogeneous reaction, and the last term is the gravity compo
nent. Equation 21.42 is rearranged with help from the mass balance, Equation 21.37
to obtain:
- e ap ag - v ° ( E m
a + G :) = 0
(21.43)
where the last term comes from the mass balance and represents the change in the
a. phase momentum due to the change in mass. For momentum transfer across the
phase interface, the convection and heterogeneous terms, E aM + are insignifi
cant for most applications and may be neglected.
21.14 FLUID FLOW HANDBOOK
When Equation 2 1.42 is summed over all phases, applying the definitions in
Equations 21.29 through 21.36, the region momentum balance becomes
Note that the form o f equation (21.44) is identical to that of a single phase sys
tem, yet this equation applies to a multiphase system. For example, in flows
through a packed bed we know that there are significant drag forces between the
fluid and the solid particles. This drag force appears in the a phase balance but does
not appear in the region balance. This is because the drag force is transmitted to the
solid phase, which in turn converts the force to solid phase movement or transmits
the force to the system boundaries via the stress tensor. Hence, the drag force is
intrinsically accounted for, it does not appear explicitly in the equation.
The total energy balance for the a phase is obtained in a similar manner as above,
(21.46)
The total energy balance is a mathematical statement o f the first law of thermo
dynamics, that energy is conserved. It accounts for internal energy, kinetic energy,
and potential energy. From left to right, the terms in Equation 21.43 account for the
accumulation of energy in the a phase, convection of energy, heat conduction, rate
o f work done on the a phase by stress forces, work by gravity forces, external sup
ply of energy, and energy transferred to the adjoining phases via phase change,
mechanical transfer, and chem ical reactions.
TRANSPORT PHENOMENA IN POROUS MEDIA 21.15
Introducing the definition in Equation 21.41 and applying the mass continu
ity balance, Equation 21.37, the total energy balance is rewritten as
+ V •e a q a + V - f " P f + Vea • r “ • v“
(21.47)
The mechanical energy balance is the scalar product of the a phase velocity,
va, times the momentum balance, Equation 19-42. The mechanical energy balance
accounts for the energies associated with bulk movement of the a phase and is essen
tially a balance on kinetic energy. With some mathematical manipulations, neglect
ing the + Gm ) terms, and with help from the a phase mass continuity balance,
Equation 21.37, the mechanical energy balance becomes
£ P ^ ( + v “ 2) + v “ • v ( | v “ 2) j + v “ e “ V P
(21.48)
The first two terms are the accumulation and convection of kinetic energy of
the a phase, the third term is the work done by the a phase by pressure force, the
fourth term is the mechanical energy conversion to thermal energy by stress forces,
the fifth term is the irreversible degradation of mechanical to thermal energy due
to the viscous forces, the sixth term is the work done by gravity forces, the seventh
term is the work done by momentum forces transfer from the a phase to the adjoin
ing phase, and the last term is the change in kinetic energy of the a phase due to
mass transferring or reacting with the adjoining phases.
When we subtract the mechanical energy balance, Equation 21.48, from the total
energy balance, Equation 21.47, what remains is the thermal energy balance. The
thermal energy balance is
+V e V + PV e a vn + £ wi a :Vyw- e ap ah(t
+ ( E“ + / “ + 5 “ + G “ ) - va • F a + 4 va l ( E l + G m
a)= 0
(21.49)
21.16 FLUID FLOW HANDBOOK
To make practical use of this equation, one must introduce constitutive functions
for the phase transfer terms. Hassanizadeh and Gray [3] show that the mechanical
transfer o f energy is given by
(/,“ + S “) = Q“ + v a F a ( 2 1.50)
where Qa represents the energy transfer due to a temperature difference. This energy
transfer could be correlated to a local heat transfer coefficient and the temperature
difference between the phases. As an approximation for a two phase system (a and
/J), this is given by
One model for the interphase energy exchange due to phase change, that sat
isfies the constraint in Equation 21.35, is
E ae = U aE m
a + e a( 0 p - { / “ )£,: (21.52)
where the first term on the right side of Equation 21.52 represents the pre-phase-
change energy of the a phase material, and the second term is the additional energy
required by the material changing phase necessary for it to join the /? phase. The
bracketed term, ( 0 ^ - Ua ), is equal to the latent heat of the phase change for mate
rials that do not have a significant change of density (for example, when ice melts
to water). When there is a significant density change, such as boiling a liquid, then
the work against the gas pressure must be included as required by thermodynam
ics, AU = AH - APV. A sim ilar approximation for the energy transfer due to het
erogeneous reaction is
G f = U aG am + e a( 0 p - U a )G" (21.53)
Summing Equation 21.49 over all phases, using the definitions in Equations
21.29 through 21.36, the region thermal energy balance becomes
-p "h ° + ■E a = 0
(21.54)
TRANSPORT PHENOMENA IN POROUS MEDIA 21.17
Further manipulations of the energy equations are possible but are beyond the scope
of this chapter.
The chemical species balance is derived in a manner similar to the previous bal
ances. The vector form of the equation is
(21.55)
where the terms from left to right represent accumulation of species A , convection,
diffusive flux, homogeneous generation (chemical reaction), and the three mecha
nisms of transfer across the phases: transfer due to phase change, diffusion, and het
erogeneous generation (chemical reaction). The three phase slip effect is neglected.
Combining Equations 21.55 and 21.37, we get the species balance in the desired form
Since the sum of the species balances for all species present in a given phase must
yield the mass continuity balance, Equation 21.37, then
Furthermore, the excess terms for species A represent the mass rate at which species
A changes phase. If the rate of mass transfer of A across the phases is proportional
to the mass fraction of A in the bulk phase then the excess terms would cancel each
other out of Equation 21.56, though this is not true in general.
Finally, if Equation 21.55 is summed over all phases we get the region species
balance
| ( p x ) + y ■( p x ; v " ) + y ■ / ; - < + = 0 ( 2 1 .5 8 )
Equations 21.55 through 21.58 are in terms of mass density and species mass frac
tion. These equations could be converted to molar concentration and molar frac
tions through use o f the species molecular weight.
21.18 FLUID FLOW HANDBOOK
The a phase mass, momentum, thermal energy, and species balances given in Equa
tions 2 1.3 7 ,2 1.43, 2 1.49, and 2 1.56 are written in vector-tensor notation. Most appli
cations are associated with a particular geometry coordinate system. These equations
are summarized here in Tables 2 1.2 and 2 1.3 for rectangular and cylindrical coordi
nates respectively.
The set of working equations derived here may be used to model dispersed mul
tiphase systems in general. These equations represent balances on the mass continu
ity, momentum, thermal energy, and species for the a phase. The biggest challenge
in using these balances is finding appropriate constitutive relations for the consti
tutive functions. Several relationships have been introduced for the constitutive func
tions, but for most of the constitutive functions these relationships have not been
introduced. In many of the modules that follow constitutive relations are devel
oped and applied to specific problems.
TABLE 2 1.2 a Phase Mass. M om entum, Therm al Energy, and Species Balances in Rectangular
C oordinates
(A)
I (e“ p " ) ■
+i ; )■
+ h (e“ p “ v>
“ )■
+i (£" pDv- ) + + c : ; = 0
at ox dy dz J ox
(B)
d e aT° + -------+
-------iL d e a T“ ----------*L
d e a T?
- f; + £ “p^ + v?( e : + g: )
dx dy dz
y-com ponent
dv° dv“ dp
e np a
" a T +Vz ~dT dy
(C)
d e ax% | d e ar l > d e aT ° )
- F>a +EaP agy + v ; ( E : + G : )
dx dy dz
c-com ponent
TRANSPORT PHENOMENA IN POROUS MEDIA 21.19
TABLE 21.2 (continued) a Phase Mass. M om entum , Therm al Energy, and Species Balances in
R ectangular Coordinates
dt x dx ' dy dz dx dx dz
-P
(denva de*v° deava)
— + ---------+ -------
chr
dx dy dz dx 0v dz (E)
TABLE 21.3 a Phase M ass, M om entum , Therm al Energy, and Species Balances in Cylindrical
Coordinates
l(e'‘p‘
dtv l)7 +-^(r£"p'\C)
r drv ' +--^(£"p"'’
r dOy en)' +
dT(E
z “PaV")+E:,+G:=0 (A)
TABLE 21.3 (continued) a Phase Mass, Momentum, Therm al Energy, and Species Balances
in C ylindrical Coordinates
^-component
F* L W + r a L dva
+r
{r dr r 30 dz J \ " dr mr de u dz
(E)
Isothermal flows through porous media are commonly found in filtration, ground
water migration, oil recovery from underground reservoirs, and many other appli
cations. The common theme in all of these applications is to relate the flow rate of
the fluid to the driving pressure force.
The derived continuum theory for dispersed multiphase systems provides the
starting equations for a continuum analysis. These equations are general and must be
sim plified for the application considered here. For evaluation the permeability of
clean filter media or packed beds we only need to consider the mass and momen
tum balances for the fluid phase (assuming the solid phase is rigid and stationary).
Permeabilities can be measured directly or they can be estimated from correla
tions. Ergun’s Equation for Newtonian fluid flow in a packed bed is discussed here.
TRANSPORT PHENOMENA IN POROUS MEDIA 21.21
The friction factor for flow through a packed bed, derived from Ergun’s Equation,
is related to permeability for use in the continuum equations. Furthermore, the
friction factor for Power Law fluids and Yield Stress fluids are similarly derived.
The isothermal mass and momentum continuum balances reduce consider
ably for an incompressible fluid-steady flow through a rigid porous medium with
no chemical reactions and no mass transfer, to obtain [5]
Mass V (£ V Y ) = 0 (21.59)
Hassanizadeh and Gray [3] introduce the constitutive relation for the Drag Force
function in the momentum balance as
F* = R \ y ! - v v) (21.61)
In stationary media the solid phase velocity, vs, is zero. R is a second order
tensor resistance function that is characteristic of the material properties and the
flow conditions. Combining Equations 21.60) and 21.61) gives the expression for
the local volume averaged velocity as
vf = - R ~ ]£ f • ? P (21.62)
in which /? ' is the inverse o f the resistance function tensor. The resistance func
tion approach is commonly used in commercial computational fluid dynamic soft
ware. However, most of literature uses Darcy permeability to relate the pressure to
the flow rate. D arcy’s Law (from Section 21.3) may be written in vector form tor
non-oriented media as
vf = - = - = - (21.63)
£J H
By inspection we see that the resistance function and the local permeability are
related by
The local permeability depends upon several factors including the local
porosity and the effective wetted surface area [10]. A number of references pro-
vidie correlations relating local permeability or resistance functions to surface
area, porosity, particle size, and velocity [8, 11]. The significance of the tensorial
21.22 FLUID FLOW HANDBOOK
A material is isotropic when its properties are independent of the direction of the
flux. In this case, the permeability is isotropic at a local point if the same value is
obtained independent of the direction o f the flow rate. This is not to say that the
isotropic permeability does not change with position. When the permeability at all
points in the medium is the same, the material is homogeneous. If the permeabil
ity varies depending on position, it is referred as inhomogeneous [ 12]. Isotropic per
meability is the easiest to model because the permeability tensor reduces to a scalar,
k. The relation between the scalar permeability and the scalar resistance function,
Equation 21.64, simplifies to
k = R~lEf2ii (21.65)
In this work only isotropic permeability is considered. The reader may refer
to works o f W hitaker [13], Guin et al. [ 14], and Dullien [8 ] for more information
on anisotropic permeabilities.
The magnitude o f the perm eability coefficient depends upon the porosity and sur
face area of contact between the fluid and the particles. It is frequently used to
give an indication of the ease with which a fluid will flow through a porous m ate
rial. Permeabilities for some com m on materials are listed in Table 2 1.4.
TABLE 2 1 .4 Typical V alues for the Perm eability Coefficient for Non-
R egular M aterials
(2 1 .6 6 )
The derivation of this equation is reviewed here and the form is modified to
allow direct com parison with the model developed for the Yield Stress fluid. The
friction facto r,/, is defined by the rate expression 115]
(21.67)
A force balance on the fluid over the length of the tube gives drag force to be
Fk = n R 2{P0 - P L) ( 2 1 .6 8 )
Combining Equations 21.67 and 21.68 to eliminate the drag force gives
r { p0 - p ,.)
(21.69)
L p (v )2
For flow in the packed bed, we consider the bed to be a bundle ot capillary
tubes of equal diameters and equal flow rates through each tube. The total flow
rate through all N tubes is
(21.70)
Also, the total flow rate is related to the bed average velocity by
Q = K D l^ V (21.71)
21.24 FLUID FLOW HANDBOOK
Eliminating the flow rates between Equations 21.70 and 21.71 gives
/ V V
<v> = - (21.72)
where the porosity, £, equals the ratio of the total area of the capillary tubes divided
by the area o f the bed.
Knowing that the bed actually consists of spherical particles of diameter dp, and
not capillary tubes, we introduce the hydraulic radius. The radius of the tube in Equa
tion 21.70 is related to the hydraulic radius by
R = 2 Rh (21.73)
However, the hydraulic radius is also related to the porosity and the particle diam
eter by
The bed-specific surface area, a , is related to the specific surface area per vol
ume o f particles by
a = a s( \ - e ) (21.75)
For spherical particles of diameter dp, the specific surface area is related by
dp = — (21.76)
a
Finally, combining Equations 21.69, 21.72, 21.73, and 21.77, we get the general
expression for the friction factor as
e\l„ (P.. - P. )
/ = ------- (21.78)
3(1- f )L pV-
Equation 21.78 applies for all flow regimes, large or small Reynolds num
bers. The Reynolds num ber for flow in the capillary.
R = P^ 2R (21.79)
is related to the packed bed by combining Equations 21.79 with 21.72, 21.73, and
21.77 to obtain
pVcl 4
R = F - - (21.80)
V 6( > - £ )
= P V^ i ’ (21.81)
/i(l-e )
hence
(21-82)
For laminar flow we introduce the laminar How friction factor correlation for the
flow in a capillary.
/.= £ (21.83)
Ke
w here/o is the friction factor value at low Reynolds number. Combining Equa
tions 21.82 and 21.83 we get
24
fo=f~ (21-84)
Rep
21.26 FLUID FLOW HANDBOOK
When we eliminate the friction factor and Reynolds numbers between Equa
tions 2 1.84, 2 1.81, and 2 1.78 we obtain
L d; e
In laminar flow, the assumption of hydraulic radius frequently gives too large a flow
rate for a given pressure gradient. Hence, the number 7 2 is expected to be too small.
Analysis of experimental data led to improvement of the formula by replacing the
72 in the denom inator of Equation 21.85 with 150, and yields the Blake-Kozeny
equation [15]
K - ^ ) _ 1 5 0 ^ ( l- £ ) 2
L ~ d; (2L86)
For large Reynolds numbers, a similar analysis with experimental data pro
duced what is known as the Burke-Plummer equation, in the form
( 5 l Z ^ = 1.75 1 V2 l ^ £
L dp e
Com bining Equations 21.8 and 21.78 yields the friction factor for large Reynolds
numbers to be
L = 0.5833 (21.88)
Ergun 116] found that by adding Equations 21.86 and 21.87 we obtain a cor
relation for the full range of flows, as given in Equation 21.66. This is equivalent
to summing the two asymptotic solutions to obtain the friction factor for the full
range of Reynolds Numbers as
/ = /o + /~ (21.89)
MacDonald et al., [17] extended Ergun’s results to a wider range of materials and
found that the correlation is improved by replacing the 150 in Equation 21.66 with
180, and by replacing the 1.75 with 1.8 for smooth particles or by 4.0 for rough par
ticles. In this work we only consider the smooth particles, hence Equation 21.66
is revised and the Ergun Equation for Newtonian fluids becomes
TRANSPORT PHENOMENA IN POROUS MEDIA 21.27
/ = — (21.91)
!u KP
and
L = 0.60 (21.92)
AP
<V> = (21.93)
\2fp(v)y L
k_AP
(v) = (21.94)
H L
by inspection gives the desired relation between the friction factor and permeabil
ity to be
d„li
k= (21.95)
K2 / p ( v >
'dU l- e )
k= (21.96)
2 . RT /
/ • = 6 0 + 0.6 (21.97)
K
The expression in Equation 21.96 is significant in two ways. First, it can be applied
macroscopically over the whole bed, or if properties vary locally, it can be applied
21.28 FLUID FLOW HANDBOOK
locally by replacing \v ) with v{ for use in the continuum equations. Second, the
expression is not limited to just Newtonian fluids. If the friction factor is corre
lated for non-Newtonian fluids, that correlation can similarly be substituted into
Equation 21.96 to estimate the permeability (see section 21.6.2.2).
For porous media made of materials such as fibers, that are significantly dif
ferent from spheres in their geometry, other correlations have been developed.
Dullien [8] provides a summary of such correlations. One such correlation, by Kyan
et al. 118], proposes a friction factor based on the pressure drop across a fibrous
bed to be the sum of the pressure drops due to viscous losses, form drag, and
deflection of the fibers.
21.6.2.2 Non-Newtonian Fluid Flow Through Porous Media. This topic is only
briefly mentioned here. Christopher and Middleman [19] introduce a modified
Darcy model to represent non-Newtonian flows through porous media in which a
modified permeability accounts for the non-Newtonian behavior. Park et al. [20]
evaluated several approaches to modeling non-Newtonian flows through packed
beds and concluded that the capillary tube approach was the best choice when com
bined with the a particular rheological expression. Marshall and Metzner [21] dis
cuss the effects o f viscoelastic properties on flows through packed beds and the
importance of the Deborah number for such fluids. Hayes et al[22], model the
flow of power law fluids through packed beds from a volume averaged approach
and account for wall effects. Shirato et al [23], demonstrate the effects of non-
Newtonian behavior on cake filtration.
Dachavijit and Chase [24] derived an expression for the friction factor for a
yield stress fluid. The yield stress fluid behavior is defined by
dv, | |
r n = ~li0 —*• ± t a ot r t > r„ (21.98)
dr
and
Following the approach described above tor Newtonian flow, the friction factor
for a yield stress fluid is found to have the form of
TRANSPORT PHENOMENA IN POROUS MEDIA 21.29
where the value of constant C3 is tentatively set at a value of 3.5 based upon lim
ited experimental data. The Hedstrom and Reynolds numbers are defined here as
r pci2 £2
^ = - ^ 7 7 -----7T (21.101)
I-1'. 0-e)
pVd,
Re,,= , (21.102)
iU„0-£)
In the limit as r(, —> 0 Equation (h) reduces to the Newtonian fluid correlation.
Flow occurs in the porous media when zr > T(,, or equivalently, when
c,w ,,
<1 (21.103)
■K
( ^ ^ ( i z f ) |04)
L dp £
The above analysis for flow through a packed bed neglected the inertial and stress
terms. They may be neglected by dimensional analysis arguments, but it is infor
mative to consider under what conditions they are not negligible. In this section
we consider the importance of the shear stress term relative to the dimensions of
the system and the size o f particles in a granular porous medium.
A slit between two flat plates separated by a distance of 2B is filled with a gran
ular material such that the permeability is uniform and isotropic (Figure 21.4). We
assume constant properties, uniform porosity, steady one-dimensional flow, and
t *
F IG U R E 21.4 A N ew tonian fluid flows through a porous medium sandwiched between two
flat plates.
21.30 FLUID FLOW HANDBOOK
dP f i £ f v { d 2v!
0 = ---------- ------ + n — f - (21.105)
dz k dx
This equation says the pressure drop for flow through the porous medium is
controlled by the drag resistance (Darcy’s law) for flow through the pores of the
medium, and due to the momentum transfer by the flow to the walls of the slit (by
the viscous stress term).
The solution to this equation is straight-forward and is
cosh
(21.106)
HL cosh(fi/ 4 k )
Integrating Equation 21.106) over the cross-sectional area gives the flow rate.
With some algebraic manipulation this is rewritten as
where
B
A =
V k
The quantity
2 WBk(P0 - PL)
fiL
is the vaiue we would get if the only mechanism resisting flow is friction in the
porous media. The quantity
tanh(A)
A
TRANSPORT PHENOMENA IN POROUS MEDIA 21.31
is a correction factor for the wall effect on the porous media flow. A plot of the
correction factor is shown in Figure 21.5. This figure shows that when
- 4 = < 100
Vk
the wall effect becomes important. If we assume a low flow rate the friction fac
tor is given by
from Equation 21.97. Substitution of this value into Equation 21.96 gives
(21.108)
v 120
For a typical value of 0.4 for the porosity, this tells us that when B is smaller
than about seven particle diameters that the wall effects become important. This is
consistent with the common rule of thumb for packed beds in cylindrical vessels that
the vessel diameter be 10 to 20 particle diameters for the wall effect to be neglected.
This also tells us that for most industrial applications the wall effects are proba
bly negligible.
10
A=B/sqrt(k)
FIGURE 21.5 Correction factor for wall effect in permeable flows.
21.32 FLUID FLOW HANDBOOK
Among the most challenging problems facing engineers are those involving heat
and mass transfer in porous media. Understanding the mechanisms controlling heat
conduction and diffusion in porous media is essential if one is to predict perfor
mance and design such processes.
The analogous behavior between heat and mass transfer allows us to use corre
lations in one to predict correlations in the other (as long as the mass transfer rate
is small). Heat and mass transfer depends strongly upon surface area through which
the transfer can occur. Porous media have large surface areas and hence heat and
mass transfer can be very effective in porous media.
Because multiphase systems are so complex, we must first consider simplified
process conditions to deduce appropriate correlations for the constitutive functions.
We only consider heat conduction and diffusion through a porous material. Heat
and mass transfer between phases of a two phase system are not considered here.
Compared to single phase systems, multiphase systems are much more complex.
If we consider the single-phase movement of a chemical species flowing through
a conduit there are five mechanisms that can occur: convection due to bulk flow,
diffusion of species A relative to the bulk flow ( v A ^ v), diffusion to or from the
conduit wall (a boundary condition), homogeneous chemical reaction (a body vol
umetric generation or depletion of species A ), and heterogeneous chemical reac
tion at the wall of the conduit (a boundary condition). These mechanisms are
shown in the sketch in Figure 2 1.6.
(21.109)
where the terms from left to right are the accumulation of species A (pA is the mass
concentration of species A , mass of A per unit volume), convection of species /\,
diffusion of species A (motion relative to the bulk flow), and homogeneous reac
tion generation of species A
We define the mass fraction as coA = pA/p. Substitution of this definition and
use of the mass continuity balance. Equation 21.109) can be transformed to
J_A
^ d f + -v V- f f Al , | + V -J - r ,A = 0 (2 1 . 110)
where 0 is the internal energy per unit mass, q is the heat flux vector (heat con-
duction term) and the remaining terms are all part of the heat body source. Com
parison of Equations 21.110) and 21.111) shows analogy between the differential
equations between heat and mass transfer. The pressure term, PV * v, represents
the reversible rate of internal energy increase per volume by compression. The shear
stress term r : Vv is the irreversible rate of internal energy increase per volume by
viscous dissipation, and the S, term represents all other sources of internal energy
generation per unit volume such as from electrical, chemical, or nuclear sources.
The analogy also extends to boundary conditions. A heat flux at a wall surface
is analogous to diffusion at a surface. A heterogeneous chemical reaction at a wall
may generate thermal energy at a boundary condition.
To extend this discussion to flows through porous media, the multiphase ana
log of Figure 2 1 . 6 is shown in Figure 2 1 . 7 . For two phases, (fluid,/, and solid, s)
there are 13 mechanisms, namely: convection due to bulk fluid flow, diffusion of
species A within the fluid phase relative to the bulk fluid flow ( v \ * v f ), diffu
sion to or from the conduit wall (a boundary condition) to the fluid phase, homoge
neous chemical reaction (a body volumetric generation or depletion of species A) in
the fluid phase, heterogeneous chemical reaction at the wall of the conduit (a bound
ary condition) with the fluid phase, convection due to bulk solid phase flow, dif
fusion of species A within the solid phase relative to the bulk solid phase flow
(vA s * vs), diffusion to or from the conduit wall (a boundary condition) to the solid
phase, homogeneous chemical reaction (a body volumetric generation or depletion
21.34 FLUID FLOW HANDBOOK
SOUD PHASE
FLUID PHASE HOMOGENEOUS
BULK FLOW REACTION
SOLID PHASE AND SPECIES
BULK FLOW
AND SPECIES FLOW VELOCITIES
FLOW VELOCITIES
FLUID PHASE V
o f species A) in the solid phase, heterogeneous chemical reaction at the wall of the
conduit (a boundary condition) with the solid phase, diffusion between the solid and
fluid phases, phase change between the solid and fluid phases that carries with it
species A between the phases, and heterogeneous reaction at the interface between
the solid and fluid phases.
This demonstrates the added complexity that occurs in multiphase systems. The
more phases that are present, the more complex the interactions and the modeling.
By analogy one could describe a similar picture for heat transfer.
The corresponding multiphase equations are derived in section 21.3. The species
balance is
c “p ° « ) + v" • )j - < (E l + G“ )
(21.56)
+Y . e a l ° - e a r Z + ( E ° + I aA + G aA ) = 0
+V e V + PV e “ v“ + e ttT°:Vv“ - e ap aha
(21.49)
TRANSPORT PHENOMENA IN POROUS MEDIA 21.35
Let’s consider the situation of diffusion or conduction in porous media but no chem
ical reactions (G " = G " = 0), no phase changes ( E “ = E% = 0), no mass transfer
between phases ( / " = 0), no heat transfer between phases or equal phase tem per
atures (Qa = 0), and no generation terms (/*" = S, = 0). The species and thermal
energy balances reduce to the multiphase conduction and diffusion equations
e ap a
+ V - e * q a + P V - e a va + e a Ta:Vva - e ap aha
= 0
(21.113)
The thermal energy balance is further reduced by assuming that there are no
external energy supplies (such as by radiation, hence ha = 0) and that viscous dis
sipation and compression energy sources are negligible. The energy balance now
takes a form analogous to Equation 21.112.
£ ap a \ ^ [ U a ) + v a -V [ U a +v - £ V = o (21.114)
Analogy between heat and mass transfer exits when the partial differential equa
tions are the same and when the boundary conditions are the same. Equations 21.112
and 21.114 are analogous partial differential equations. We can apply correlations for
heat transfer to mass transfer for the conditions assumed in deriving Equations 21.112
and 21.114 as long as the boundary conditions are also similar.
It turns out that the diffusion and conduction coefficients are affected by flow rate
due to a phenomena called dispersion. Dispersion is discussed later, first consider
stagnant systems where v " = 0. This will simplify the development of correlations.
21.7.2.1 Molecular Diffusion in Stagnant Porous Media. In keeping with the anal
ogy with single phase systems, the simplest constitutive relation for the species
flux in porous media is the Fick’s law model, written as
i : = - p a D aA v w aA (21.115)
where Dif is the effective diffusivity (in flows through porous media this may also
be called “dispersivity”) for the chemical species A diffusion through the a-phase
21.36 FLUID FLOW HANDBOOK
within a porous media. For a fluid phase ( a = f ) reduces to the molecular dif-
fusivity in the idealized case of the capillary tube model in which the porous
medium only affects the diffusion by reducing the available cross-sectional area.
However, diffusion experiments (with no convection, no interphase surface diffu
sion o f species, diffusion only through the fluid phase, and no heterogeneous reac
tions) show that in real porous media the effective diffusivity is not equal to the
intrinsic molecular diffusi vity, but it is equal to an effective molecular diffusivity
o f species A in flu id /, that is symbolized here as v = o,
where the subscript “ V = 0 ” implies that the velocity is zero. In general, 2)“ v =0
is not equal to the intrinsic fluid phase molecular diffusivity of species A ,
The single phase molecular diffusivities are typically found in handbooks and the
effective diffusivities are estimated from correlations.
The flux expression in Equation 21.115 can be generalized for the flux, i_, of
any property through a single phase due to a potential gradient, V0, in the form:
/ = -(TV0 (21.117)
The flux through a porous medium that would be measured at a surface is the
region flux, a combination of the flux through both the solid and fluid phases. The
region flux is related to the applied region potential by
[° = -o°Y<P (21.118)
i° = e f if + e sis (21.119)
i.f = - c r / V0 (21.120)
Maxwell [25], one of the earlier researchers in transport in porous media, con
ducted experim ents with electrical current in porous media. He showed that for a
TRANSPORT PHHNOMENA IN POROUS MEDIA 21.37
non-conducting solid phase, the effective region conductivity is related to the intrin
sic conductivity of the fluid phase, 07, by the relation:
where (Jy=o *s the effective conductivity for zero flow conditions. Bruggeman 126 1
proposed a slightly different correlation for non-conducting spherical particles
3/2
(21.123)
These two equations are frequently cited in literature and are used to compare
experimental data. Much o f the data seems to favor the Bruggeman equation, but
overall the majority o f the data falls between these two correlations.
By analogy, the relationship between effective molecular diffusivity and the
effective fluid phase average diffusivity over a wide range of porosities can be
approximated by M axw ell’s correlation as:
(21.124)
21.7.2.2 Heat Conduction in Stagnant Porous Media. By analogy, under the same
conditions as discussed for diffusion, the effective thermal conductivity is analo
gous to Equation 21.124 as
21.38 FLUID FLOW HANDBOOK
The quantity k° is the effective thermal conductivity for the multiphase region,
treating the multiphase system as a pseudo-single phase, with essentially the same
temperatures between the phases where
=-k°VT (21.125)
and q ° accounts for the combined heat flux through both phases.
When temperatures are large, thermal radiation can be an important mechanism
in the heat transfer even though the particles are in close proximity. Wakao and
Kato [30] present correlations to account for the effects of radiation between parti
cles in a packed bed. One such correlation for orthorhombic packing of spheres
(£ /= 0.395) and for cubic lattice packing (ef = 0.476) for k j k f = 20 ~ 1000 and
Nur = 0 ~ 0.3 is reported to have an 8 percent error
kKv=o
° k° o.%
+ 0.707/V. (21.126)
N= 0 v kf
where N ur = h rdp/ks is the Nusselt number for radiation, k f and ks are the phase
intrinsic thermal conductivities, dp is the diam eter of the spherical particles, and
/?,. is the radiation heat transfer coefficient. W akao and Kato [30] give a correla
tion for the radiation heat transfer coefficient as
0.1952 ( 273+JTV
K = 2------- — I “ 7 ^ I ( 2 1.12 7 )
--0 .2 6 4
I 100
e
21.7.2.3 Example o f Diffusion Through a Packed Bed. Consider the diffusion sys
tem shown in Figure 21.8. Liquid A at the bottom of a porous bed evaporates into
gas B, and we imagine that there is some device which maintains the liquid level
at z = 0. We assume that capillary effects that tend to draw the liquid upward are
negligible. At the gas-liquid interface (at z = 0), the gas phase mass fraction of
species A is wA0. This is taken to be the gas phase concentration of A correspond
ing to equilibrium with the liquid at the interface and is equal to the vapor pressure
of A divided by the total pressure. We further assum e that the solubility of B in
liquid ‘A ’ is negligible, that there is no mass transfer or adsorption between the
TRANSPORT PHENOMENA IN POROUS MEDIA 21.39
Ga S
^ s tre a m o f
A a n d B.
F lu x o f A
th ro u g h
° o o o th e bed
o o
L iq u id A le v e l
c o n s t a n t a t z=0.
F IG U R E 21.8 Diffusion o f chem ical species A through gas B through a packed bed
in a steady state with B not in motion.
gas and solid phase, that diffusion of A in the solid phase is negligible, and that
there are no chemical reactions.
At the top of the bed (at z = L), a stream of gas mixture A-B of concentration
(oAL flows past slowly; thereby the mass fraction of A in the gas phase at the top
of the bed is constant. The entire system is presumed to be held at constant tem
perature and pressure and is in steady state. Gases A and B are assumed to be ideal.
At steady state there is a net motion of species A away from the evaporating surface
and the gas species B is stationary. The rate of diffusion is small so that the velocity
in the fluid phase, v/, is negligible, and the mass concentration of species A is small.
In this system the steady state species balance in Equation 21.112 reduces to
V - £ aj a = 0 (21.128)
— -A
(21.129)
If we assume that the porosity is uniform, then the effective diffusivity in the
fluid phase is a constant, as related by Equations 21.116 and 21.124. Since the
21.40 I I I II) FLOW HANDBOOK
concentration of species A is small then the fluid phase mass density is constant
and Equation 2 1. 129 reduces to
(21.130)
CPJM - 0) A
J (z) = z
(21.131)
« I(, - < / , L
Many of the processes of interest to engineers involve flowing systems. Early exper
imental work showed that in some cases the apparent diffusivity in porous media
with fluid flow is greater than the stagnant effective diffusivity as given by corre
lations such as Equation 2 1. 124 in flow systems. In other cases the effective diffu
sivity is smaller. The former has resulted in deriving correlations for the dispersivity
coefficient, in Equation 2 1. 115, whereas the latter has branched into two approaches:
retardation factors and packed bed mass transfer coefficients. We consider the dis
persivity correlations in this section.
Dispersion is a phenomenon that occurs in all flows in porous media. Q uali
tatively, dispersion can be explained by comparing to one-dimensional laminar
flow o f a fluid through a region of space in the presence of, and in the absence of,
a porous medium [3 1]. In the absence o f a porous medium, the paths of the m ol
ecules under diffusion are straight parallel lines. However, when diffusion occurs
in a porous medium the molecules must follow tortuous paths which slows down
the rate. In a flowing system, the tortuous paths cause a mixing of microscale flow
streamlines that enhances the diffusion rate.
In general, the trajectories of the molecular motions are random which results
in both the axial and the transverse migrations. The transverse migrations ‘disperse’
the molecules away from the straight parallel lines that would have been followed
in the absence o f a porous medium.
It is obvious from this description that the species flux, y'“ , must be a func
tion of the flow rate through the porous medium. In Equation 2 1 . 1 15 only the dis
persivity, D “ , can functionally account for the velocity dependence.
Carbonell and Whitaker [32] apply theoretical arguments to show that the effec
tive diffusivity is a second order tensor with nine components. For isotropic porous
TRANSPOR T PHENOMENA IN POROUS MEDIA 21.41
media the effective diffusivity tensor may be split into two primary parts: a longitu
dinal (axial) component in the direction of flow, and a transverse component normal
to the direction of flow. In the analogous heat transfer problem Howie and Georgiadis
[33] show that the heat flux in the porous medium is given in the form:
where the ijth component of the medium thermal conductivity tensor is given by:
where D{ and D r are the dispersion parameters that account for the property disper
sion in the longitudinal and transverse directions , respectively. These parameters
are correlated by
D , = ^ (21.134)
' £' L
and
d t = ^ (21.135)
T e' L
Howie and Georgiadis [33] report CL = 0.18 and the ratio of the coefficients
falls in the range of
In Equation 21.133 5,, is the Kronecker delta. In terms of the chemical species
diffusion the analogous ij th component of the dispersivity for the fluid phase is:
fVf
D°Aij - D"V=0 8 , + DTEf Y \ 8 ij+{ DL - D T) ef ^
U (21.137)
V
—I /
For a non-diffusing solid phase, the effective dispersion coefficient of the fluid
phase is related to the effective region dispersion coefficient, by
D°A= e f D{ (21.138)
21.42 FLUID FLOW HANDBOOK
The correlation in Equation 21.137 works well in packed beds of regular geom
etry for which the bed depth, L, is fixed and the porosity is uniform over the depth
o f L. In a general case in which the porosity varies with position and hence the
local effective diffusivity varies as a function of position, it is not clear how Equa
tion 21.137 can be adapted for use due to the dependence of DL and D T on the
ratio of the particle size to the bed depth, given in Equations 21.134 and 21.135.
To overcome this limitation another approach is considered.
Wakao and Funazkri |34] correlate the effective dispersion coefficient in a form
that can be applied locally in the volume averaged continuum equations. In the axial
direction the fluid phase effective dispersion coefficient is related to the to the mol
ecular diffusion by:
(21.139)
which can be applied locally. In this correlation, parameter V has a value of be
tween 0.6 and 0.8 and Sc and Re are the Schmidt and Reynolds Numbers, respec
tively, given by:
(21.140)
and
(21.141)
The above correlations allow us to estimate the effective diffusion coefficient for
diffusion in the fluid phase.
Equation 21.139 gives the fluid phase dispersivity. For solid rigid particles in the
solid phase, there is no relative motion to cause dispersion, and the solid phase
effective diffusion coefficient equals the solid phase molecular diffusivity. If the
solid particles have an internal porosity, the diffusivity can be related to the inter
nal porosity [35]. Amiri and Vafai [36] model the effective thermal conductivity
of an internally porous solid as being the same as the solid phase intrinsic thermal
conductivity and let the internal solid phase volume fraction scale the solid phase
thermal conductivity. In effect, it is assumed that the solid phase particles are intrin
sically rigid and that there is no dispersion effect on the solid phase. The same
assumptions can be made in the case of species diffusion. When we model systems
with diffusion through both the solid and fluid phases then we must also consider
mass transfer between the phases.
TRANSPORT PHENOMENA IN POROUS MEDIA 21.43
The total species flux for the multiphase region is analogous to Equation 2 1. 119,
given by
j° =efjf + £ ' /
•La - a - a
(2 1.142)
which is the sum of the fluxes through each phase weighted by the volume frac
tions of the phase.
21.8 CONCLUSIONS___________________________________________
VARIABLES _______________________________________________
ACKNOWLEDGMENT
This work was partially supported by the National Science Foundation grant CTS
9900949.
REFERENCES
8. D ullien, F.A .L. 1992. Porous M edia: Fluid Transport and Pore Structure, 2nd ed.
A cadem ic Press, San Diego.
9. C hoo, W. 1988. “M ultiphase-M ulticom ponent Transport Equations Incorporating
Effects o f Interfacial Q uantities.” Dissertation. The University o f Akron, Akron,
Ohio.
10. Stephan, E.A., M.S. W illis, R. Vengim alla, and V.K.N. Yerra. 1997. “A M odern
D evelopm ent o f Friction Factors in Porous Media by Analogy to Turbulence,”
A IC hE Sym posium Series for Fluidization and Fluid-Particle System s, D. King,
Editor. AICH E.
11. C oulson, J.M ., J.F. Richardson, J.R. Backhurst, and J.H. Harker. 1991. Chem ical
E ngineering, Volume 2, Particle Technology and Separation Processes. 4th ed.
Pergam on, Oxford.
12. Eringen, A.C. 1980. M echanics o f Continua. Krieger, Huntington, New York .
13. W hitaker, S. 1969. “A dvances in Theory o f Fluid M otion in Porous M edia.” Indus
trial a n d Engineering Chem istry. 61(12): 14-28.
14. G uin, J.A ., D.P. Kessler and R.A. Greenkorn. 1971. “The Perm eability T ensor for
A nisotropic N on-uniform Porous M edia.” Chem. Engr. Sci. 26: 1475-1478.
15. Bird, R.B., W .E. Stewart, and E.N. Lightfoot. 1960. Transport Phenom ena, W iley,
New York.
16. Ergun, S. 1952. “Fluid Flow Through Packed Colum ns.” Chem. Eng. Prog. 48:
89-94.
17. M acdonald, I.F., El-Sayed, M .S., Mow, K. and Dullien, F.A.L. 1979 “Flow
T hrough Porous M edia- the Ergun Equation Revisited.” Ind. Chem. Fund. 18(3):
199-207.
18. Kyan, C.P., D.T. W asan, and R.C. Kintner. 1970. “ Flow o f Single-Phase Fluids
T hrough Fibrous Beds.” Ind. Eng. Chem. Fund. 9(4): 596-603.
19. C hristopher, R.H., and S. M iddlem an. 1965. “Pow erlaw Flow Through a Packed
T ube.” Ind. Eng. Chem Fund. 4(4): 422-426.
20. Park H.C., H aw ley, M .C., Blanks, R.F. 1975. “The Flow o f N on-N ew tonian S olu
tions T hrough Packed B eds” Polym er Engineering a n d Science. 15(11): 761-773.
21. M arshall, R.J. and A.B. M etzner. 1967. “Flow o f V iscoelastic Fluids Through
Porous M edia.” Ind. Eng. Chem. Fund. 6(3): 393-400.
22. H ayes, R.E., A. Afacan, B.B oulanger and A.V. Shenoy. 1996. “M odelling the Flow
o f Pow er Law Fluids in a Packed Bed llsing a V olum e-Averaged Equation o f
M otion.” Transport in P orous Media. 23: 175-196.
23. Shirato, M ., Aragaki, T., Iritani, E., and Funahashi, T. 1980. “Constant Rate and
V ariable Pressure-V ariable Rate Filtration o f Pow er-Law Non-N ew tonian Fluids.”
Journal o f C hem ical Engineering o f Japan. 13(6): 4 73-478.
24. D achavijit, P., and G.G. Chase. 2000. “Yield Stress Fluid Flow Through a Packed
Bed: A M odified Ergun E quation.” Los Angeles: A ICH E 2000 Annual M eeting.
25. M axw ell, J.C. 1891. A Treatise on Electricity and M agnetism , 3rd ed., C laredon
Press, (reprinted by D over Publications, New York, NY 1954).
26. B ruggem ann, D.A.G.. 1935. “The Calculation o f Various Physical C onstants o f
H eterogeneous Substances, I. The Dielectric Constants and C onductivities o f M ix
tures com posed o f Isotropic Substances.” Ann. Physik. 24: 636-664.
27. W eissberg, H.L. 1963. “Effective Diffusion C oefficients in Porous M edia.” J. Appl.
Phys. 34: 2636-2639.
28. C unningham , R.E., and R.J.J. W illiams. 1980. D iffusion in Gases and Porous
M edia, Plenum Press, New York.
TRANSPORT PHENOMENA IN POROUS MEDIA 21.47
29. Plumb, O.A. and S. W hitaker. 1990. “ Diffusion, Adsorption and Dispersion in Porous
Media: Sm all-Scale A veraging and Local Volume Averaging.” Dynam ics o f Fluids
in Hierarchical Porous M edia, J. H. Cushm an, Editor. Academic Press, London.
30. W akao, N. and K. Kato. 1968, “Effective Thermal Conductivity in Packed Beds.”
J. Chem. Eng. Japan. 2(1): 24-33.
31. Fand, R.M., M. V arahasam y, and L.S. Greer. 1993. “Empirical Correlation Equa
tions for Heat T ransfer by Forced Convection from Cylinders Em bedded in Porous
Media that A ccount for the Wall Effect and Dispersion." Int. J. Heat M ass Transfer.
36(18): 4407-4418.
32. Carbonell, R.G. and S. W hitaker. 1983. “ Dispersion in Pulsed System s - II: T heo
retical D evelopm ents for Passive Dispersion in Porous M edia." Chem. Eng. Sci.
38(11): 1795-1802.
33. Howie, L.E. and J.G. G eorgiadis. 1994. “Natural Convection in Porous M edia with
Anisotropic D ispersive Therm al Conductivity.” Int. J. Heat M ass Transfer. 37(7):
1081-1094.
34. W akao, N., and T. Funazkri. 1978. “Effect o f Fluid Dispersion C oefficients on Par-
ticle-to-Fluid M ass T ransfer Coefficients in Packed Beds: Correlation o f Sherwood
Num bers.” Chem. Eng. Sci. 33: 1375-1384.
35. W hitaker, S. 1988. “ D iffusion in Packed Beds o f Porous Particles," A IC hE J. 34(4):
679-683.
36. Amiri, A. and K. Vafai. 1994. “A nalysis o f Dispersion Effects and Non-Therm al
Equilibrium , N on-D arcian, Variable Porosity Incom pressible Flow Through Porous
M edia,” Int. J. H eat M ass Transfer. 37(6): 939-954.
CHAPTER 22
FLUID TRANSIENTS
2 2 .1 1NTRODUCTION
Previous chapters dealt mostly with steady flows, or flows that do not change with
time. Flows that change with time are called unsteady flows, in general. Transient
flow is usually referred to a closed conduit flow that changes from one steady
state condition to another steady state condition. The flow is unsteady during the
transient process. This chapter is concerned mainly with unsteady closed conduit
flows of small compressibility, except in Section 22.5 where simultaneous occur
rence of unsteady closed conduit flow and open channel flow will be considered.
The physical characteristics and the method of computation depend on the
degree of flow unsteadiness. When the flow is changing very slowly it may be con
sidered as a steady incompressible flow. As the rate of change of flow velocity
increases, acceleration begins to affect the flow and the Bernoulli equation for
steady flow is no longer applicable. This will require the time derivative term in
the equation of motion to be included in the flow analysis. Further increase of accel
eration makes the compressibility effect also important even when the Mach num
ber is small. Radiation due to pressure wave replaces convection as the major means
of information transfer. In this case the time derivative term in the equation of con
tinuity must also be included in the flow analysis. In a restricted sense, the term
transient flow often refers to the flow for which both compressibility and accel
eration effects are important.
22.1
22.2 FLUID FLOW HANDBOOK
General Equations
We shall first consider general unsteady flows in closed conduits for which both
acceleration and compressibility effects are important. The flow is assumed to be
one-dimensional, in which the longitudinal velocity and pressure (V, p)
are the
two main unknowns that change with time and location (f, x).Two equations, the
equations of motion and continuity, are needed to solve for the two unknowns.
Figure 22.1 shows a small segment of conduit, dx, isolated from a general system
to serve as a free body.
Forces to be considered are, the pressure on the two cross sections and the
pipe wall, the weight of fluid, and the shear on the pipe surface. By equating the
sum of all forces in the longitudinal direction with mass times acceleration of the
fluid, according to Newton’s second law of motion, the following equation of
motion is obtained.
dV x,dV
— + V— +
3(/? + Z) JV\V\ A
----- ^ + - ^ - ^ = 0(22.1)
dt dx dx HR
In the above equation h is the pressure head and Z is the elevation of the cen
terline of the pipe. The last term in the above equation represents the shear force
due to energy loss, which is assumed to be the same as that of the corresponding
steady flow. The coefficien t/is the Darcy-Weisbach friction factor and is the R
/
hydraulic radius, which is of the diameter for a circular pipe. For a more detailed
derivation, reference is made to Streeter and W ylie.1 An alternative form of the
equation is
at + V ox + S^r~
ox+ g{S, - S „ ) = 0 (22.2)
where
s = M A (2 2 3)
8 gR
is the energy slope and
AT
The ordinary derivatives in the above equation represent the usual material
derivative defined as follows.
( 22.6)
dt dt dx
By a thermodynamic relationship, the first term of Equation 22.5 can be trans
formed as follows.
dp _ p dp (22.7)
dt ~K dt
K
where is the bulk modulus of elasticity of the fluid. Assuming an elastic pipe,
dAin Equation 22.5 may be related to dp
in the following way.
dA= -----
eEdp
A D A ( 22 .8 )
e E
In the above equation, is the thickness of the pipe wall and is the modu
lus of elasticity of the pipe. By substituting Equations 22.7 and 22.8 into Equa
tion 22.5 and rearranging, the following equation is obtained.
dp idV
— + pa — - = 0 (22.9)
dt dx
a
where is the speed of pressure wave traveling in the pipe given by the follow
ing equation.
K
a= P
1+
KD (22.10)
eE
FLUID TRANSIENTS 22.5
dll dh a 2 dV
— +1/ — + — — = () (22.1 1)
at ox g ox
x)
Equations 22.2 and 22.1 1, with (/7, VO as functions of (/, are one form of the
basic equations representing general unsteady flows in closed conduits. These
equations can also be converted into equations in which the total head H and the
Q
discharge are the dependent variables.
The general governing equations can be simplified for some special cases.
a
For an example, if the fluid is incompressible so that is an infinity then the first
two terms in Equation 22.1 1 are negligible in comparison with the third term. In
V
this case the equation of continuity implies that is independent of x. For water
a
in a rigid pipe, is very large but not an infinity. In this more realistic case, we
may transform Equation 22.11 to the following dimensionless equation.
=0 1 2 2 121
In general the Mach number Mis very small and the second term can be ignored.
But the first term may not be ignored if the Strouhal number S, is large. A good
example, which will be analyzed later, is the transient flow generated by a rapid
valve closure (small T) in a long pipe (large L).
Equations for Idealized Case
To study the basic characteristics of the governing equations, consider a special case
of nonviscous flow in a horizontal pipe. Also assume that the Mach number is
small so that the nonlinear terms, the second term in the equations of motion and
continuity, can be neglected. Under these conditions the equations are simplified
to the following equations.
22.6 FLUID FLOW HANDBOOK
V
By eliminating from the above two equations the following second order par
tial differential equation is obtained.
d 2h
dt - ci~
2
(22.17)
J\ f
where and are arbitrary functions. The fact that Equation 22.8 satisfies Equa
2
tion 22.17 can easily be validated by substitution. Equation 22.18 indicates that any
disturbance in terms of pressure and velocity introduced into the flow will travel at
a
speed in both upstream and downstream directions without changing its magni
tude. Next, it is easily shown that the following coordinate transformation
x- at =C or —
dt - a (22.19)
d(gh+aV) d(gh+aV) ^
( 22 .20 )
dt dx
The general solution of Equation 22.20 is
a, h
In other words, to an observer moving at velocity the variables and v appear
to satisfy Equation 22.21. In a similar manner it can also be shown that
the pressure head and velocity are related to each other through the following
equation.
(22.24)
The solutions to the idealized case of hydraulic transient flow. Equations 22.23
and 22.24, can be applied to a simple reservoir-pipe-valve system as sketched in
Figure 22.3.
We assume that there is initially a steady flow of velocity V and negligibly
small head h = 0. At t = 0, the valve is instantaneously and completely closed so
that the velocity is reduced to zero at the valve. The pressure generated at the
valve, by applying Equation 22.24, is calculated to be
(22.25)
a
This pressure will propagate to the left at speed or velocity -a arriving at
t Lla.
the = At this moment the pressure drops to zero imposed by the reservoir
condition. Sudden drop in pressure creates a reflecting wave which moves to the
right at speed a. By applying Equation 22.24 following this wave, we find that
the velocity is-V hand = 0. This reflected wave arrives at the valve at t = 2 Lla,
which is reflected to the left again. Following this wave applying Equation 22.24,
the condition behind this wave is found to be = 0, =V h -aV/g. t Lla,
After = 3
the wave moving to the right carries behind it the original steady flow of veloc
V h
ity and = 0. At t = 4 Lla, t
the condition becomes identical to that of = 0, and
L
the process repeat itself thereafter. Thus, for a pipe of length the period of water
T
hammer is
(22.26)
V h
The changing and during one period of oscillation is shown in Figure 22.3.
It is instructive to compare the water hammer pressure head with the steady
flow stagnation pressure head. The ratio of the two heads is
/V = a_= _
2 2
(22.27)
hs V M
In the above equation, subscripts Wand Srepresent, respectively, water ham
mer and stagnation, and A/ is the Mach number of the original steady flow. For
an example if the original velocity is 10 fps and ais 4,000 fps, then Mis 0.0025
a
v, h=0
L L
—
a
<t<—
a
2
a a
a a
22.8
FLUID TRANSIENTS 22.9
and the water hammer head is 800 times greater than the corresponding stagna
tion pressure head.
Method of Characteristics
The ideal case of last section gives a simple solution of constant amplitude wave
that travels back and forth indefinitely. But for real flow cases, the amplitude may
be damped due to energy loss or magnified due to resonance and the governing
equations must be solved numerically. Among several available numerical meth
ods, the method of characteristics is the most popular method for solving tran
sient flows in closed conduits. This method is not only simple and accurate it also
describes the physical process of hydraulic transients well. First, we linearly com
bine the two governing equations to obtain
2 \
av
k)dx +g(Se-S,.) =0
d V + */
dt + V + k-
8
—
dx dt I
(22.28)
The above equation holds for any kvalue, but it can be transformed into an
ordinary differential equation
(22.29)
(22.31)
a
dx
= V±a (22.32)
~dt
Equations 22.32 and 22.33 are the characteristics form of the original partial
differential equations indicating that, along the trajectories of the pressure waves
V h
given by Equation 22.32, and varies according to Equation 22.33. The trajec
tories represented by Equation 22.32 are called the positive and negative charac
teristic lines. Equation 22.33 is called the compatibility equation.
t)
To illustrate a basic numerical procedure, reference is made to a fixed (x, grid
system sketched in Figure 22.4.
The computational domain is divided into a finite number of rectangular grids
P
of finite sizes. Let be a typical interior point where the How is unknown and
B
points A, , and C be the three neighboring points where the flow is known. Two
P
characteristic lines drawn from according to Equation 22.32 meet the line ACS
R S.
at and R
Note that will be within the line AC
if
(V+a)At (a-v)At
FIGURE 22.4 A typical fixed grid system.
FLU ID TRANSIENTS 22.11
(22.36)
Vp- VR+—
a(hP- hK)+g(Se- Sn) At - 0 (22.37)
and
VP-Vs-£(hP-hs)+g{Sr-Sn)sAt =0 (22.38)
The above two linear equations can be solved for the two unknowns to obtain
2 2 a
and
The right hand side of these two equations contain only known quantities so
that the unknown velocity and pressure head can be directly computed. The pro
cess is repeated for all interior points on the same time line. After the boundary
points on the same time line are also solved, as will be described later, the time is
advanced by one step and the process is repeated for a specified amount of time.
This time marching method is stable as long as the CFL stability condition rep
resented by Equation 22.34 is satisfied. If too large a time step is taken and Equa
R
tion 22.34 is violated, then may fall ousted of AC
, indicated as /?' in Figure 22.4.
In this case, Equation 22.34 becomes an extrapolation formula, which will lead
the numerical process to diverge. To keep the numerical process stable, it is nec
essary to interpolate the values at /?' between points andD A.
If the point is at an upstream boundary, P\
as shown in Figure 22.4, then Equa
tion 22.37 is not applicable because the positive characteristic line falls outside of
the system. Therefore, in addition to Equation 22.38, an equation representing the
22.12 FLUID FLOW HANDBOOK
upstream boundary condition must be supplied to solve for the two unknowns. Sim
ilarly, if the point is at a downstream boundary, P'\
Equation 22.37 is solved with
an appropriate downstream boundary condition. Other types of boundary conditions
are needed where the flow is locally far from one-dimensional. These include loca
tions such as junctions, dropshafts, pumps, gates and valves. Some of these bound
ary conditions will be described in more detail later.
(22.41)
The downstream boundary condition depends on the type of valve and the clo
sure time selected. Assuming that the flow through the valve to be an orifice type,
the following equation can be used as the downstream boundary condition.
(22.42)
o
where subscript represents initial value. For the sake of simplicity, a linear valve
closure as described by the following equation is considered first.
T = 1— —
trC if t<tc (22.43)
r =0 if t> tc
CNJ
o
o
;8 9 J ‘PB9H
22.13
o
o
o
o
I
CM
E
<D
CD
CO
O
FIGURE 22.5 Time variation of pressure head at the valve, linear valve closure in 10 seconds.
22.14 FLUID FLOW HANDBOOK
t
The flow is assumed to be steady at = 0 and subsequent change in and h V
at all 11 nodes are computed. A typical computed time variation of h at the valve,
for the valve closure time of 10 sec is plotted in Figure 22.5.
This figure shows that the pressure increases very slowly until about t = 8 sec.
It then rises sharply to a maximum value of 254 ft at / = 10 sec followed by a
t
sharp drop to - l 85 ft at = 10.5 sec. Thereafter, the pressure head continue to
oscillate at l sec period, the natural period of the system, while the amplitude con
tinue to decrease. The rate of damping is proportional to the friction factor/or the
n
Manning’s coefficient assumed. Eventually, the flow and the pressure oscilla
tion will stop completely. It is a typical behavior of the flow due to a linear valve
closure that the maximum head occurs at the end of the valve closure and the min
A
X
imum head occurs at about period later.
It is a simple matter to repeat the simulation, each time using different closure
time, to study the effect of closure time on the pressure peaks. Such exercise has
been carried out and the maximum and minimum heads at the valve are plotted as
functions the closure time in Figure 22.6. It is important to note that very large neg
ative pressure is shown in this figure when the closure time is short. This is not
realistic because water can’t remain in liquid state and cavitation or column separa
tion takes place when the pressure falls below the vapor pressure. Cavitation is a
very destructive phenomenon that must be avoided if possible. At room tempera
ture the vapor pressure is about -3 0 ft of water. Therefore, according to Figure 22.6,
the closure time should be greater than 40 sec if the linear valve closure is used.
A linear valve closure process as described above is not an efficient way of flow
control under emergency conditions. It may be desirable to close the valve rapidly,
while the maximum or minimum pressure is limited. The next example illustrates
how much the water hammer pressure can be reduced using a relatively simple non
linear valve closure process given by the following second order equation.
/ \2
T = 1 -2 — + ( — 0 <t<t. (22.44)
'c V c )
T = 0, t>t,
Compared with Equation 22.43, this second order equation closes the valve
twice as fast initially but very slowly at the end. The computed pressured at the valve
as the function of time for 10 sec closure time is plotted in Figure 22.7. Very sub
stantial improvement over the linear closure case is noted. The maximum pressure
/
is reduced to about 64 ft, which is about of the linear closure case. The ampli
tude of pressure oscillation is so small that no negative pressure is generated at any
time. A mild water hammer like this is acceptable in most practical applications.
A valve stroking method of Streeter and Wylie [3] may be used to optimize the
valve closure process. In performing an analysis of a real system, information on
valve closure for a specific valve must be obtained from the valve manufacturer.
800
22.15
Id ‘P«©H U|W * xeyy
F
E
o
C
13
CD
O
O
</)
/)
S>
<D
FIGURE 22.6 Maximum and minimum head at the valve, linear valve closure.
o
00
o
o
CO
o
to
o
22.16
o
U ‘pB9H ejnssejd
CO
o
CM
E
oQ)
d)
C/3
FIGURE 22.7 Time variation of pressure head at valve, non-linear closure in 10 seconds.
FLUID I'RANS I ENTS 22.17
Consider a case of water being pumped from a sump to a reservoir located at higher
elevation as sketched in Figure 22.8.
The flow within the pump is too complex to be treated as one dimensional. It
must be isolated from the rest of the system and replaced by boundary conditions
d u
applied at sections and as shown in the figure. Under steady operating condi
N Q
tions, a pump is to rotate at a constant speed and generate discharge and pres
H d u.
sure head between sections and Under a transient condition like that produced
by a power failure, all these three quantities and the torque generated by the pump
T may change with time. Therefore, the pump characteristics under transient con
ditions are represented by the four variables (Q, //, N, T).For numerical analysis
it is more convenient to use the dimensionless variables defined as follows.
Qr
v= n~’ h = Hr Nr
a =lT ' Tr
P = T (22A5)
R
In the above equation the subscript indicates a reference quantity. In the fol
lowing, the widely used method of Marchal et al [4] is adopted. Because hydraulic
similitude between various operating conditions exist for a given pump, only three
equations or curves are needed to determine the four variables. To describe the char
acteristics of a pump, it was suggested that the following two variables
Fh a2h
+v
,2 (22.46)
Fh=a2h,2
+v (22.47)
ol
fta =tan -i — (22.48)
T=-I 260kddN
t (22.49)
T I
where is the resisting torque and is the combined polar moment of inertia. In
dimensionless form the equation is rewritten as
0 .2kNr da
607; ~ch (22.50)
d u Ax
(22.51)
(22.52)
(22.53)
Step 4. Three linear equations, Equations 22.51, 22.52 and 22.53 are solved
for the new discharge and the two new pressure heads.
Step 5. The new torque is found from a characteristic curve or a table.
Usually the time step taken according to the CFL condition is small enough
for the explicit method described above to be stable. If necessary, an iterative
method can be used.
A system shown in Figure 22.8 needs to be controlled under a power failure con
dition to avoid reverse flow and reverse rotation. Most common approach is to pro
vide a flow control valve, which can be closed within a specified time after power
failure so that reverse flow and reverse rotation are avoided while the maximum
and minimum pressures are kept within specified limit. For an example, a control
valve is assumed to be located immediately downstream of the pump as shown in
Figure 22.10a. The corresponding grid system is shown in Figure 22.1Ob. The pump
and the valve are together isolated from the rest of the system and replaced by
boundary conditions. Compared with Figure 22.9, there is an additional section v
due to the valve shown as in Figure 2 2 .10. But, because Q is the same for all three
sections, only one additional variable, the pressure head at the section v needs to
be computed. The additional equation needed is the valve discharge equation
(22.54)
All other equations except Equation 22.52 remain the same. In Equation 22.52,
u
it is only necessary to replace the subscript by v.
22.20 FLU ID FLOW HANDBOOK
o
[a) Pump a n d V a l u e (b) G r i d S y s t e m
F IG U R E 22.10 A pum p and a control valve between two reservoirs, (a) pump and valve (b) grid
system.
Storm and combined sewers are usually designed to operate as open channels. But,
they frequently fill up and operate as closed conduit when the How rate exceeds
the design capacity. Underground tunnel used to convey and store water for later
treatment alternately operates as an open channel and a closed conduit. The flow
during change over between free-surface flow and pressurized flow is always
accompanied by highly dynamic surges, positive surge when filling and negative
surge when emptying. Figure 22.11 shows some photographs of surge waves, two
positive surges and two negative surges, generated in a laboratory. Note that water
is flowing from right to left. Case (a) is a positive surge initiated at the downstream
end and moving in upstream direction as moving hydraulic jump. This condition
is commonly generated when out flow is less than the inflow and water must be
stored in the pipe. Case (b) is a positive surge starting somewhere upstream and
moving in the downstream direction. This condition typically occurs when the
inflow exceeds the carrying capacity of the pipe. Case (c) shows a negative surge
generated at downstream end and moving upstream. This is a typical flow pattern
when an initially full pipe is being emptied from the downstream end by, for an
example opening the gate rapidly. Case (d) shows a negative surge which is gen
erated at the upstream end and moving in the downstream direction. This condi
tion occurs when the inflow rate is suddenly decreased to an initially pressurized
flow so that the pressurized flow is changing into an open channel flow. It is inter
esting to note that Case (d) contains a hydraulic jump like configuration while
Case (c) has no hydraulic jump.
A positive surge during the filling phase of a conveyance-storage facility can
be so severe to cause geysering and structural damage. A simple storage tunnel
with slope from left to right as sketched in Figure 22.12 explains why a pressur
ization surge can be severe. Starting from an empty condition, a free surface flow
FLUID TRANSIENTS 22.21
initially takes place from left to right as shown in (a). The downstream end of the
tunnel fills up first to generate a surge as shown in (b). The surge front or the
interface moves upstream at speed W.
Note that the tunnel is being filled from
right to left as the surge moves upstream. The How behind the surge is in reverse
direction, from right to left, and increases as the surge moves upstream as shown
in (c). When the surge arrives at the upstream end, water in the tunnel enters the
dropshaft causing the water level to rise rapidly. For a long tunnel, a very large
22.22 FLU ID FLOW HANDBOOK
(a) Q'
Q'
(b)
(c ) Q'
negative momentum has been established by this time, which requires a very
large impulse to bring the flow to an equilibrium condition.
Governing Equations
The governing equations for pressurized part of the flow are given by Equations
22.11 and 22-13. It can be easily shown that the same equations also govern the
h y
free surface part of the flow if is replaced by the flow depth and is replaceda
by the gravity wave speed C.
FLUID TRANSIENTS 22.23
T h e g ra v ity w a v e sp e e d is
C= (22.55)
The pressure wave speed is usually about two orders of magnitude greater than
the gravity wave speed. This large change in the wave speed is responsible for the
abrupt change in the flow condition at the interface, for both positive surge and neg
ative surge. To avoid the difficulty in dealing with the surge problem, Cunge and
Wagner [7] proposed a model with a slot along the crown of the pipe. In this way
the entire flow is computed as a free-surface flow. Although this model greatly
simplifies the computation, it may under represents the dynamics of a real sys
tem, because the model implies that ventilation is allowed along the entire length
of the tunnel.
Song et al, [8] proposed a shock wave tracking method to more accurately model the
two-phase flow phenomenon. Figure 2 2 .13a shows the flow near a positive surge
moving in the upstream direction at speed W
located at a distance / from the nearest
station downstream. To make the problem simpler, the flow is made to appear steady
by introducing a moving coordinate system as shown in Figure 22.13b. The surge
appears like a stationary internal hydraulic jump in the moving coordinate system.
A computational grid system near the positive surge front is shown in Figure 22.14.
According to the CFL stability condition, much smaller time steps are needed
on the pressurized side than the free-surface side. Therefore, within one time step
of free-surface flow computation, a number of time steps must be computed on the
pressure side of the flow. The dotted line represents the trajectory of the interface
P
and is where the 6 new values, velocity and head on each side of the interface,
the speed and the location of the interface, are to be computed. Because the flow
energy must decrease from left to right, according to the second law of thermody
namics, Cardie and Song [9] showed that the free-surface flow relative to the surge
is supercritical but the pressurized flow is subsonic. This condition assures the exis
tence of two characteristic lines on the free-surface side and only one character
istic line on the pressure side. Thus, we can write three characteristic equations at
P as follows.
V,+W v2+w
Finally, the kinematic relationship between the distance moved and the speed
is given as
AL=WAt (22.61)
INTERFACE
TRAJECTORY
The above equation indicates that the pressure in the “pressurized flow” near
the interface is negative. This negative pressure is responsible for the negative
surge that empties the pipe. Experimental data of Cardie and Song [9] confirms
the existence of negative pressure region.
The above analysis may not be exactly correct for a negative surge moving in
downstream direction as shown by the photograph of Figure 22.11 (d). This photo
FLU ID TRANSIENTS 22.27
graph appears to indicate that the flow first contract so much to produce a super
critical flow before becoming a subcritical flow through a hydraulic jump. The flow
in the contraction region appears highly three-dimensional that must be excluded
from the one dimensional analysis. By selecting sections 1 and 2 in such a way that
the hydraulic jump is excluded from the computational zone, Equations 22.56, 22.62
through 22.65 are still applicable. However, Equation 22.66 may have to be modi
fied to include the energy loss due to the hydraulic jump. It is difficult to estimate
the energy loss at the hydraulic jump at present time.
An interesting question is how a positive surge weakens and retreat when the
flow condition changes. Experiment and analysis of Cardie and Song [9] showed
that a positive surge doesn't gradually slow down and become a retreating positive
surge. For an example, if the outflow from the downstream end is suddenly increased
while there is a positive surge, then a negative surge is generate on top of the exist
ing positive surge as sketched in Figure 22.16. The negative surge moves to the right
while the positive surge continue to move to the left as a moving hydraulic jump.
The two surges have to be treated separately in this case.
As a surge moves in a system, it may interacts with various hydraulic structures such
as dropshafts, junctions, and gates. The flow near a structure is usually highly three
dimensional requiring a boundary condition to approximate its effect on the one
dimensional flow system. Only a relatively simple case of a positive surge passing
through a dropshaft will be described below. Figure 22.17 illustrates the flow con
figurations when a positive surge approaches and passes through a dropshaft.
j- 1 )
F IG U R E 22.16 A positive surge breaks up into a negative surge and a hydraulic jump.
22.28 FLU ID FLOW HANDBOOK
1. Assume that the flow quantities at sections 2 and 3 are the same, thus elim
inating 2 unknowns. Note that these conditions replace the negative charac
teristic equation at 2 and the positive characteristic equation at 3 that would
have existed if L was large.
2. Flow depths at sections 1 and 2 are assumed to be the same.
3. The positive characteristic equation at 1, the negative characteristic equa
tions at 3 and 4 together provide 3 equations.
4. The shock boundary condition between sections 3 and 4 provide two
equations.
5. The continuity condition between sections 1 and 2 that includes the inflow
at the dropshaft is used.
6. Finally, the kinematic condition relating the location and the speed of the
surge is used to determine the new position of the interface.
The above process is repeated at each time step until the new L becomes neg
ative indicating that the surge has passed the dropshaft and the flow configuration
FLUID TRANSIENTS 22.29
has changed to that shown in Figure 2 2 .17b. This case is treated essentially in the
same way as the case (a), except that the three characteristic equations to be used are
the positive and negative characteristic equations at 3 and the negative character
istic equation at 2. Shortly thereafter, as the dropshaft starts to fill and L exceeds the
grid size, the interface and the dropshaft become independent again.
Large and deep underground tunnels are being used in many cities for the purpose
of conveying and storing storm and combined sewer water for later treatment. A
tunnel system receives water from existing or new shallow sewer systems through
a number of strategically located dropshafts. Some of the dropshafts are equipped
with adjustable gates to control the inflow rate when necessary. A pump station,
usually located at the downstream end of the tunnel, is needed to pump water out
to a treatment plant. The tunnel is designed to store the water of a medium size
storm. But overflow structures are provided to spill excess water of larger storms.
A relatively simple conveyance-storage system is used to illustrate the highly
dynamic nature of the sewer filling process. A system to be analyzed consists of
a 14 ft diameter 4 miles long main tunnel to which 6 tributary sewer systems feed
the tunnel with wastewater and storm water. There are 8 dropshafts, 3 overflow
structures, and a pump station in the main tunnel. Synthetic hydrographs due to a
25 year/6 hour storm are used as inflows for the purpose of evaluating the surge
characteristics. The amount of water already stored in the tunnel at the beginning
of a storm is an important factor affecting the performance of the system. For this
reason, initially empty tunnel and initially 42 percent full cases were considered.
Figure 22.18 shows some computed instantaneous hydraulic grade lines, at
10 minutes intervals, after the start of the storm when the tunnel is initially empty.
Sharp surge front is generated near the downstream end shortly after the beginning
of the storm. The surge front moves upstream as the tunnel is being filled. More than
10 ft jump in pressure head across the surge can be observed. The speed as well the
size of the surge tend to increase as the net inflow rate increases. The amount of time
it takes for the surge to travel from the downstream end to the upstream end is
roughly equal to the amount of time it takes for the tunnel to fill up.
Figure 22.19a shows the time variation of water level at selected dropshafts.
Note that the water level at each dropshaft rises slowly before the surge arrives at
that particular shaft. When the surge arrives and passes a dropshaft, the water level
jumps up or rises very quickly to the head behind the surge. After the passage of the
surge, the water surface level rises according to a nearly single trajectory defined
by the slope of the tunnel and the speed of the surge.
Eventually, the water level stops rising and keeps the level defined by the eleva
tion of the overflow structures. The amount of jump is usually greatest at the upper
most dropshaft, unless it is designed to serve as a surge tank. Figure 22.19b shows
the corresponding time variation of velocity in the tunnel at the downstream side of
each dropshaft. The most prominent feature to be noted is the abrupt change in the
flow direction upon the arrival of the surge as indicated by sudden drop in velocity.
22.30 FLUID FLOW HANDBOOK
40
30
20
10
0
C
O
co
-1 0
>
0
LU -20 ------ Tunnel Invert
<D ■— • t=10 mins
O
CO
t: -30 •— ■ t=20 mins
D ♦— * t=30 mins
CO *— * t=40 mins
k_ -40 * t=50 mins
0) •«—
-60
-70
-80
j — I— i— i— i i,.j j i >
0 2 4 6 8 10 12 14 16 18 20 22
Main Tunnel (1000ft)
F IG U R E 22.18 Instantaneous hydraulic grade lines show ing surge movement.
Figure 22.20 shows the time variation of water surface elevations at three
locations computed using the same inflow hydrographs assuming the tunnel is 42
percent full at the beginning of the storm. Note that the water surface elevation at
the upstream end of the uppermost tributary rises very rapidly and exceeded the
ground surface by more than 30 ft. This geysering phenomenon needs to be con
trolled by various methods to be described in a later section.
In this example, surge is shown to be stronger when the tunnel is initially 42 per
cent full as compared with the case of initially empty tunnel. But, this doesn’t mean
an initially partially full condition is always worse than initially empty condition.
The magnitude of the positive surge is essentially proportional to the inflow rate
at the time, and therefore it is best to avoid large inflow rate at the time of com
plete filling.
FLUID TRANSIENTS 22.31
F IG U R E 22.19 Tim e variation o f w ater level and velocity at dropshafts, tunnel initially empty,
w ater surface elevation.
T=-1 =—
4/
(22.67)
f a
12
10
4
i. 2
s
0 0
CD
>
-2
•— • At upstream end of
-4 ■— ■ At Marquam
♦— * After Ankeny Drop
-6 a— * After Tanner Drop
*— « After Fremont Drop
»— ►Northwest CSO PS
-8
-10
~j P 1 .1 1 t -1 .1 I I 1 1 1 1 I I l 1-1 1 I I I I —L—t—I _ 1 1 i . i —L -l—i. I 1 1 I. 1 H . 1 . J t | I > L - l —t—1 I 1.1 1 I I 1 I > 1 1
known as the resonance phenomenon. A simple example is, in the case of a reser-
voir-pipe-valve system, when the valve is closed but the reservoir level varies with
time due to wind wave in a periodical manner given by the following equation.
The problem can be solved with the method of characteristics using Equation 22.68
as the upstream boundary condition and 0 as the downstream boundary condition. V=
FLUID TRANSIENTS 22.33
T im e (m in )
F IG U R E 22.20 Tim e variation o f water level in dropshafts, tunnel initially 42 percent full.
Assuming that there is initially no flow in the pipe and the water level in the
reservoir starts to oscillate at / = 0, the water in the pipe will be driven to oscil
late at the same period. When the period of the external excitation is close to that
of the natural period of the system, energy will accumulate within the pipe caus
ing the intensity of the oscillation and the related rate of energy loss to increase
with time. Eventually, the rate of energy loss balances with the rate of energy gain,
and permanent constant amplitude oscillation is achieved. The computed constant
H
amplitude oscillation of at the downstream end due to resonance when the period
of the forcing function is equal to the natural period of the system, 1 sec, is shown
in Figure 22.21.
22.34
U ‘peen ainssaid
o
FIGURE 22.21 Pressure head oscillation at valve due to resonance with water wave in reservoir.
FLUID TRANSIENTS 22.35
o n
In the above equation, subscripts and represent, respectively, the external
excitation and the system’s natural characteristics. For this particular example, the
ampkfication factor as a function of the period ratio is plotted in Figure 22.22. Note
that the maximum amplification factor of 102 occurs at the period ratio of 0.995,
which is very close to 1. The amplification factor decreases very sharply as the period
ratio deviates from 0.995.
Instability of Dropshaft
Next we consider a typical dropshaft connected to a large tunnel with a long con
necting pipe or a drift tube as sketched in Figure 22.23.
It is assumed that the inflow to the dropshaft and pressure in the main tunnel
are given functions of time. Detailed analysis of the system, including resonance
conditions, were carried out by Guo and Song [ 11 ]. A set of four equations of con
tinuity and motion, two each for the dropshaft and the drift-tube, were written and
analyzed. By linearising the equations and dropping terms of small quantities, the
T
natural frequency of the combined dropshaft-drift tube system was found to be
given by the following equation.
(22.70)
22.36
jopBj u0!iB0^!|diuv
z
5
o
03
<D
3
‘u
S
S
d)
■o
Q_
•C
O
Q_
"D
<D
DROPSHAFT COVER
(22.71)
Note that the stand alone dropshaft has its natural frequency represented by
Equation 22.71 when the inertia and gravity forces are both taken into account
and an appropriate boundary condition is applied to the exit of the shaft. The sec
ond case is obtained when the dropshaft is eliminated. The resulting period is
exactly the same as Equation 22.67.
7; = — (22.72)
a
a
Finally, if the water in the drift-tube is also incompressible so that is infin
ity, we obtain the natural surge frequency of a tank-pipe system as follows.
r3 = 2 k (22.73)
Any external excitation of period matching the overall natural period or the
natural period of any of the components can cause severe vibrations.
22.38 FLUID FLOW HANDBOOK
TME (SECO N D S)
TIME (SECONDS)
opening may change its size due to natural vibration related to the seal or the valve
itself. If the returning positive pressure wave coincides with the closing of the leak
opening, the positive pressure is magnified at the valve. This positive wave return
ing as a negative wave is further magnified by negative pressure generated by the
maximum leak opening. Thus, the pressure and leak oscillations amplify each other
to become a large amplitude oscillation.
An example of severe penstock vibration of a power plant induced by a leaky
seal was described and model tested by Abbott, Gibson, and McCaig [12|. The same
phenomenon was analyzed by Streeter and Wylie [ 13] using different methods,
including the method of characteristics. Good agreements between field measure
ments, model testing and analytical results were obtained.
Self excited vibration of a pipe fitted with a valve is possible when the natural
frequency of the valve matches that of water hammer. Saito [14] carried out a num
ber of experiments and demonstrated the possibility of the phenomenon. Fashbaugh
and Streeter [15] analyzed self-excited vibrations in a liquid rocket engine fuel feed
system with computer.
Water hammer and surge are closely related phenomena that can be controlled in
very similar ways. There are structural or passive control methods and operational or
active control methods. Some of better known control methods are described below.
22.40 FLUID FLOW HANDBOOK
Surge Tanks
Air Chambers
An air chamber is a modified surge tank, which is air light to allow the air pres
sure to change as the water surface moves. The volume of air required to produce
the required pressure change is much smaller than a surge tank, which relies on the
hydrostatic pressure change. Air chamber is an effective water hammer control
device because it takes only a small volume change to affect water hammer pres
sure. On the other hand, surge phenomenon is a slower phenomenon which involves
much larger volume change making an air chamber less effective in controlling
surge problems. Allievi [17] and Angus [18] discussed the use of air chambers in
pumping systems to control water hammer generated by power failure to pumps.
Control Valves
Overflow Structures
Inflow Control
REFERENCES
9. C ardie, J. A. and Song C.C.S. “M athem atical M odeling o f U nsteady Flows in Storm
Sew ers,” International Journal o f E ngineering F luid M echanics, 1 (4) pp. 495-518.
1988.
10. Benjam in, T. B. “G ravity C urrents and Related Phenom ena,” Journal o f Fluid
M echanics, Vol. 31, Part 2. 1968.
11. Guo, Q. and Song, C.C .S. “D ropshaft H ydrodynam ics U nder T ransient C onditions,”
Journal o f H ydraulic Engineering, Vol. 117, No. 8. Aug. 1991.
12. Abbott, H. F., Gibson, W . L., and M cCaig, I. W. “ M easurements o f Auto-Oscillation
in a H ydroelectric Supply T unnel and Penstock System ,” Journal o f Basic E n g i
neering, Vol. 85, pp.625-630, D ec., 1963.
13. Streeter, V. L. and W ylie, E. B. H ydraulic Transients, M cG raw -H ill, pp. 137-149.
1967.
14. Saito, T. “Self-excited V ibrations o f H ydraulic C ontrol Valve Pipelines,” Bull.,
Japan Soc. o f M ech. E n g r s Vol. 5„ No. 19. 1962, pp. 437-443.
15. Fashbaugh, R. H. and Streeter, V. L. “R esonance in Liquid Rocket Engine System .”
Jour., Basic Engineering, ASM E, Vol. 87, Dec. 1965, p. 1011.
16. C haudhry, M. H. A p p lied H ydraulic Transients, C hapter 11, Van Norstrand
R einhold Co. New York. 1979.
17. A llievi, L. “A ir C ham ber for D ischarge L ines,” Trans. ASM E, Vol. 59, Nov. 1937.
pp. 651-659.
18. A ngus, R. W. “A ir C ham bers and V alves in R elation to W ater ham m er,” Trans.
ASM E, Vol. 59, Nov. 1937, pp. 661-668.
__________ CHAPTER 23__________
FLOW HYDRODYNAMICS
IN CHEMICAL PROCESSING
UNITS
The design of chemical processing units depends to a large degree on the fluid
hydrodynamics. For example, reactors are used to bring two components into con
tact to produce a third one; the degree and efficiency of mixing (flow hydrody
namics) is an important factor in sizing and operating reactors. The design model
of any reactor is a function of the flow and mixing characteristics. The two well-
known reactor models, the plugflow model and thecontinuousstirredtank (CSTR)
are based on the degree of mixing (turbulence) between reactants along the reactor
volume. In plug flow, the reactants are mixed along the unit flow-path with no
back-mixing, resulting in a concentration and temperature profile, while in CSTR,
the entire reactant volume is well mixed, resulting in a uniform concentration in
all of the reactor volume. Tray and packed columns are used to increase the contact
area of two fluids to achieve a required degree of mass transfer. The flow regime,
flow rate, and many other flow characteristics have very important effects in the
overall design of the column since mass transfer rate is a function of the flow rate
and superficial velocities. Many design and operating problems may be attrib
uted to improper unit hydrodynamics such as partial wetting of catalysts in reac
tors, channeling, flooding, stagnation etc. The hydrodynamics, such as pressure
drop, flow regime, and gas and liquid holdups of selected chemical process units
will be considered in this chapter.
23.1
23.2 FLUID FLOW HANDBOOK
23.1 REACTORS
Many types of reactors exist in industry. It is not within the scope of this hand
book to cover the reactor topics in detail; only the fluid flow and unit hydrody
namics are of concern here. Reactors are classified, based on the number of distinct
fluid phases, into homogeneous and heterogeneous reactors. In homogeneous reac
tors, only one fluid phase exists, gas or liquid. For plug-flow reactor with liquid-
phase reactants, the pressure-drop correlations along with equations to calculate
Reynolds number presented in Chapter 8 may be used. However, one must be care
ful to use the correct physical properties, which may change significantly along
the reactor path especially for highly exothermic or endothermic reactions. Also,
although the mass conservation law applies, new components are being produced
which may posses different density and viscosity. The flow hydrodynamics of het
erogeneous reactors are more complex than homogeneous since at least 2 distinct
phases coexist in the reactor. Phases coexisting in heterogeneous reactors may be
gas-liquid, liquid-catalyst, gas-catalyst, or gas-liquid-catalyst. Catalyst may be fixed,
suspended, or fluidized. In the following sections, the flow hydrodynamics in
selected reactors will be covered briefly.
In homogeneous gas-phase plug flow reactors, the number of total moles may change
along the reactor length; this depends on the stoichiometry of the reaction, see Foglor
[11]. The temperature may also change due to heat of reaction and heat transfer to
or from the surroundings, the following simple equation, derived from ideal law, may
be used to find the pressure drop as a function of temperature and concentration:
(23.1)
' C, T,
where
Subscripts 1 and 2 refer to the inlet and outlet of a reactor length increment. Tz
is found with a heat balance for the current reactor increment and C2 is found from
the design model of the plug flow reactor.
In catalytic gas reactors, the species concentrations are directly affected by the
reactor pressure. The pressure drop in packed reactors depends on many of the
bed and catalyst characteristics and on fluid density and viscosity. The bed poros
FLOW HYDRODYNAM ICS IN CHEMICAL PROCESSING UNITS 23.3
ity, which is defined as the ratio of void volume to the reactor volume, plays a
major role. The Ergun equation may be used to estimate the pressure drop in a
packed porous bed. The Ergun equation may be expressed per unit length of the
reactor, per unit volume, or per unit weight of the catalyst. The later expression
is desired since it relates pressure drop and catalyst weight:
dP G ( \ ) m z h iE + u s G (23.2)
d..
where
Flow of gas or liquid through a fine bed of catalyst is characterized by the fol
lowing flow regimes based on the fluid flow rate: At low flow rate, the solid bed
is stationery and one may apply Equation 23.2 to find the pressure drop. As the
fluid flow rate increases, solid particle start to vibrate and move apart, this state
is known as expanded bed. A further increase in fluid flow rate results in a com
plete suspension of the solid particle or what is commonly known as fluidized-bed,
see Figure 23.1. In the fluidization state, the frictional force between particles and
fluid counterbalance the weight of the particles. Smooth and progressive bed flu
idization is observed when flowing fluid is liquid; however, bubbling fluidization
characterizes gas systems.
The fluid velocity for a minimum bed fluidization is found by the following
equation which was proposed by Wen and Yu 1966 [30]:
U. = dpp(pp-p)g
33.72 + 0.0408 —
3
dpp
where
i
Liquid
!
Low Flow High Flow
Rate Rate
which is a volume average of the fluid density and particle density and thus a
function of the bed void fraction.
t Gas
Ca ta ly sts
Gas
L ic ru ic i
B ub bl e Slurry-bed
Re, Litfuid
FIGURE 23.3 Flow regimes in co-current down-flow three-phase fixed-bed reactors [13].
where
Reactor pressure drop determines the energy required to move the fluid through
the catalyst bed and is therefore an important variable in the design and operation
of the trickle bed. Many correlations predict the frictional gas-liquid pressure gra
dient; Larkins et al [18) proposed the following correlation:
where
X= A^p,
P (23.5)
where
xc =—
XL
. 0.111 0.5
xL=^ Pc Hsl G,, Pa 'We,} (23.8)
G Cs ,p,- Vl G,Gs \P l ) WeG
Re^
(0.001 + Re i5) (2 3 .9 )
Gls Liquid-phase mass flux, g /n r-s
GCs Gas-phase mass flux, g/rrr-s
WeL Liquid-phase Weber number
WeG Gas-phase Weber number
The Weber number is defined as:
WeL=GU<P 1
(23.10)
P l°
and
(23.11)
Pc°
where
_ 6(1 ~ £ b)
<*, = (23.13)
ip*
FLOW HYDRODYNAM ICS IN CHEM ICAL PROCESSING UNITS 23.9
dp* Particle diameter modified to account for wall effects and is defined as:
(23.14)
' \+(4dp/6dT
— 4 —
(\-eB))
dr Reactor diameter
Th.‘ NANCY group (Ellman et al. 1990 [9]) suggested the following correla
tion fo total liquid holdup based on the results of 5000 hydrodynamic experiments:
eL=eB10* (23.15)
kis deined as:
0.42
k=0.001 - 0.48 (23.16)
4 = X0-5 Re^03
/
aA. \4
(23.17)
dh Hydraulic diameter:
0.33
' 16 e 3e '
dh= 9 7 t ( \ - £ B)2 (23.18)
In this type of three-phase catalytic reactors, the gas and liquid flow concurrently
upward through a fixed-bed of catalysts. Liquid-phase in continuous, catalyst is
completely immersed in liquid, while gas-phase is dispersed as bubbles. As the
gas flew rate increases, bubble coalescence will occur causing slug-flow regime
which :haracterized by a periodic fluctuation of gas and liquid phases through the
flow scheme. A further increase in gas flow rate changes the flow regime to spray
flow in which liquid is transferred as heavy mist. Bubble flow regime is the
desirec flow pattern for many industrial applications such as the Fischer-Tropsch
process and coal liquefaction. Fukushima and Kuasaka (1979) [14], reported the
followng different flow regimes for co-current up-flow fixed bed reactors:
Bubble (I) flow regime occurs at low liquid How rates, while bubble (II) occurs
at high liquid flow rates as can be noted from Figure 23.4. The hydrodynamic map
shown in Figure 23.4 was predicted using a particle diameter of l . 12cm and a ratio
of the particle diameter to the reactor diameter equivalent to 0 .128. Equations used
to predict the transition in flow boundaries are listed in Table 23.2. Since indus
trial co-current up-flow fixed beds are run under the bubble-flow regime, hydro-
dynamic correlations will be presented for the bubble flow only.
Trupin and Huntington (1967) [28] used the friction factor approach to cor
relate the two-phase pressure drop in co-current up-flow fixed-beds, as:
A Prl - (23.19)
dpe
The friction factor, / GL, is found by the following equation:
cl (23.22)
" 3 1-
Re, Liqu id
FIGURE 23.4 Flow regimes in co-current up-flow three-phase fixed-bed reactors [14].
FLOW HYDRODYNAMICS IN CHEM ICAL PROCESSING UNITS 23.11
£q + £L + £s =1.0 (23.23)
whtre
\chwal and Stepanek (1976) [ 1] proposed the following correlation for the gas-
pha;e holdup:
/ \ 0.423
— = 2[1.0 + (1.72
£n u ? * 9
U L
rr
0.5 i - l
(2 3 .2 4 )
V UC/
Liquid holdup can be estimated by the Fukushima and Kuasaka (1979) [14]
correlation as:
In three-phase slurry reactors, the catalysts are suspended in the liquid phase due
to the flow of gas bubbles through the liquid phase. Catalysts are normally fine
23.12 FLUID FLOW HANDBOOK
particles suspended in a deep pool of liquid. The reactor height to diameter ratio
ranges between 4 and 10. Deckwer et al ( l 979) [8], proposed the map shown in
Figure 23.5 to represent the different types of flow regimes for non-foaming sys
tems in bubble column slurry reactors. The flow regimes are distinguished by:
Reactor Diameter, cm
aimount of solid that can be kept in complete suspension for a given operating
conditions and reactor and particles characteristics as:
-.18
Wmax = 6.8x10 , C"ndTurpr °i.£g
P
- —
KugVlJ Y-3 (23.26)
l V Uc J
where
dr Reactor diameter
Wmax Catalysts loading, g/cm3
Cu Viscosity correction factor, defined as:
Cu= 0.232 - 0.1788 \og(/JL) + 0.10 2 6 (lo g (^ ))2
oL Surface tension, dyne/cm.
uv Terminal settling velocity of the particles and is defined as:
1. Stoke’s regime, (Re,; < 0.4)
sd2
A p„-pL)
(23.27)
18^.
2. Intermediate regime, (0.4 < ReP < 500)
1/2
31 g(p„-pL) (23.28)
Pl
(23.32)
m
(23.33)
ReP is the Reynolds number based on the particle settling velocity, defined as-:
Re,, = (23.30)
Pl
pp Catalyst-particle density
The wetting factor, y, can be assumed equal to 1.0 for most catalysts.
Yamashita and Inoue (1975) [31] have proposed the following gas holdup
correlation as a function of gas superficial velocity and bed diameter:
£cc = -----------
2.2 u c+—---------JT
0.3 (gdT
) (23.31)
Maselkar (1970) [21] suggested another correlation that agrees well with the
above, as:
0 30 + 2u„
Either of these two correlations may be used for preliminary design purposes.
l~uG Hl dp (Pp-PL)
1 , 0.436 . . 0.227 j 0.598 / * . \ - 0.305 t'1'1 'IA\
(23.34)
FLOW HYDRODYNAM ICS IN CHEMICAL PROCESSING UNITS 23.15
10 100 1000
Re (L ic ju id ) B a s e d on M in im u m V e l o c i t y
where
23.16 FLU ID FLOW HANDBOOK
The mechanical design of the reactor vessel, in addition to the agitation speed,
plays an important role in catalyst suspension. The following optimum ratios for
the diameter of the impeller to the diameter of the tank are recommended to
obtain a homogenous distribution of the catalyst particles in both central and
peripheral parts of the reactor:
where
dj Impeller diameter
dT Tank diameter
Flat-bottom
Spherical-bottom
/V™ = 'rp.—
f K ' \ p , - P L ) M1 g 042< l4w 0125
2------ £--------------------------------------
0 .5 8 j 0 . 8 9
(2337)
where
f N l.33
j8 = 2 | ^ (23.38)
Several correlations for the power consumption of agitated slurry tanks may
be found in literature. The power consumption for liquid-phase reactor (no gas or
solid) is found by:
P = N„ N 3d * p L (23.39)
23.18 FLUID FLOW HANDBOOK
where
The power number, Np, is function of the agitator Reynolds number, Rea:
Re..=^dLf\ (23.40)
(23.41)
Nd] G
where
P- 0.812
(
PlNd\_
r>2
\ 0.45
(23.42)
Qob
Superficial gas velocity and power consumption are important parameters
that affect gas holdup in agitated slurry reactors. Calderbank (1958) [5] proposed
the following correlation to predict the gas holdup in agitated gas-liquid tanks:
£c =
“g£ g
+
0.0216 (P/VL)UApaL2( uc V>5 (23.43)
lB G
0.6
\UBJ
u
The terminal bubble velocity in free rise, B, can be estimated by the follow
ing equation:
/ \ 0-25
uB = 1.53 Rec > 500 (23.44)
V P l j
FLOW H YDRO DYNAM ICS IN C HE M IC A L PROCESSING UNITS 23.19
An alternate correlation to estimate the gas holdup (Yung et al. 1979) [32] is:
(23.45)
Shell and tube heat exchangers are available in many forms of mechanical designs
to accommodate the need for different flow orientations such as co-current flow or
countercurrent flow. The thermal design of the shell and tube heat exchangers is
directly affected by the direction of fluid flow as well as other parameters. In co
current flow, the hot fluid can’t be cooled below the cold-fluid outlet temperature;
this limitation for the co-current flow makes the countercurrent flow advantageous.
Howiever, co-current flow may be advantageous for certain flow types such as vis
cose fluids or possibility of freezing conditions. In the actual mechanical design
of shtelland tube heat exchangers, multi-passes of tubes bundles are used. This tube-
side orentation results in co-current flow in sections of the heat exchanger and
23.20 FLUID FLOW HANDBOOK
AP=2f,G'Ln,
gpID (23.46)
where
The mass flux per tube is found by dividing the total tube-side mass rate by
tube flow area which defined as:
M D J _ 2 ) > ( 2 3 4 ? )
n
where N, Number of tubes per pass.
IV
G,=— (23.48)
where
045
ST Re, <1000
Re^
(23.49)
0.0028
Re, >1000
Re;
FLOW HYDRODYNAM ICS IN CHEMICAL PROCESSING UNITS 23.21
where V,is the fluid velocity inside the tube and the 2.5 numeric represents the
bene resistance coefficient, K.
Shell side maybe treated as a duct with a givenequivalentdiameter, free shell-
side area divided by thewetted perimeter. Pipe pressuredrop correlation may be
used to estimate the shell-side pressure drop. The shell-side equivalent diameter
maybe found by the following equations for the different pitch orientation:
Square Pitch:
A(Pl-nOD;IA)
nOD
Triangular Pitch:
1.72P;-0.5/r<9D,2)
De = (23.53)
0.5 nOD
where
Although air cooled heat exchangers are advantageous over water cooled shell
and tube heat exchangers in situations where water is scarce (or tough regulatory
laws on thermal pollution exist) and in colder climates; however, due to the fact
that air has a much lower heat transfer coefficient, the heat transfer surface area
has to be much larger in air cooled than water cooled for the same heat duty.
23.22 FLUID FLOW HANDBOOK
A larger surface area (15 to 20 times the bare pipe surface area) is achieved
by high concentration of fins, 7 to 11 fins per inch. Fins are 0.5 to 0.65 inch in
height. The flow of air is controlled by fans which may have motors with
adjustable speeds to accommodate the change in ambient temperature or fan
blades with variable angles. The mechanical structure of air cooled heat
exchanger consists of rectangular bundles of tube rows. Hot liquid flows inside
the tubes and air flow is induced or forced by means of fans. The air-flow area
decreases inside the tube bundles causing pressure drop. Air-side pressure drop
may be estimated by the following equation:
where
The term 0.074/p is an air density correction factor where 0.074 is the air den
sity in lbm/ft3 at standard atmospheric conditions.
Equation 23.54 may be rewritten as:
Many types of chemical process units are basically used to achieve mass transfer
between 2 fluids. Distillation and absorption columns are used to accomplish mass
transfer between vapor and liquid streams, extraction columns are used to transfer
mass between 2 liquids. The efficiency of mass transfer depends to large degree
on the fluid hydrodynamics. Serious malfunction of mass transfer units may result
due to improper flow and thus contact between the two fluids such as channeling
FLOW HYDRO DYNAM ICS IN C H E M IC A L PROCESSING UNITS 23.23
Distillation columns are very common in the chemical and refinery industrial com
plexes. Figure 23.10 shows schematics of tray and packed distillation columns. In
both tray and packed distillation columns it is desired to maximize the vapor-liquid
contact area and time. Also, it is required that liquid flows downward while vapor
flows downward in order to achieve a physical separation of vapor and liquid. There
fore, liquid entrainments and flooding severely reduce the column (mass transfer)
efficiency. Liquid and vapor flow in counter flow in packed columns, however, tray
columns may be design to achieve cross or counter flow on individual trays, but
the overall column flow will be counter flow. Figure 23.10 shows schematics for
counter and cross flow of vapor and liquid in tray and packed distillation columns.
The flow rates of the liquid- and vapor-phase and their ratio is an important fac
tor to operate the column within the stable condition with optimum mass transfer
rates. The following list of flow instability may occur in packed columns [ 15]:
• Flooding: Liquid-phase will fill the column and liquid will flow up the col
umn. Flooding occurs at high liquid or/and vapor flow rates.
Liquid
Liquid
Liquid
I
l\ /n
/\ r* i\
Ln=--
t
Vapor
Vapor
C- C o u n t e r f l o w
A- Cr o s s F l o w in B- C o u n t e r in P a c k e d
t r a y columns. F l o w in T r a y columns
columns
FIGURE 23.10 Counter and cross flow in tray and packed distillation columns.
23.24 FLUID FLOW HANDBOOK
Many correlations have been suggested to estimate the pressure drop, flooding,
and entrainment boundaries. The hydrodynamics in packed columns is a strong
function of the tower packing type such as rasching rings, partition ring, lessing
rings, or saddles. A typical pressure drop curve for counterflow of liquid and gas
in random packing is shown in Figure 23.12. Correlation to predict pressure drop,
flooding and loading may be found in detail in many packed-bed distillation books
such as Perry Handbook [24]. Flow of gas and liquid in tray columns must also be
within stable operating conditions to achieve the required area, time, and turbu
lence in vapor-liquid contact. Liquid flows across trays while vapor flows upward
through the tray perforations. Operating problems occur when the intended flow
path of liquid and vapor is not followed. A list of operating problems which may
occur due to improper flow rates of the vapor and liquid phases are summarized
below [15]. (See also Figure 23.13.)
• Flooding and foaming: Flooding occurs at high vapor flow rate or very high
liquid flow rates. Liquid may be carried by vapor to higher tray, while foam
ing may form due to impurities, and low pressure. Foaming tends to trap
vapor flowing upward and carry it to a lower stage thus decreasing efficiency.
• Entrainment: When the liquid from the upper tray contacts with the forth
layer, a back-mixing will result in lower mass transfer. The main cause for
entrainment is the improper tray spacing or downcomer sizing.
• Weeping: The fast flow of liquid through the tray perforations into the lower
tray will reduce the vapor-liquid contact and thus tray efficiency. Weeping
occurs at low vapor flow rate which results in lowering vapor-phase pressure
required to keep the liquid a longer time in the upper tray.
• Liquid maldistribution: A uniform distribution and height of liquid across the
tray is maintained by means of weirs. When a uniform height of liquid across
the tray is not maintained, the vapor-phase flow will follow the least resis
tance path thus reducing the liquid-vapor contact area. Liquid maldistribution
may occur in very large trays.
To avoid entrainment, a maximum vapor velocity for each liquid to vapor ratio
may be estimated based on the vapor and liquid physical properties, and other
column characteristics as [15].
FLOW HYDRO DYNAM ICS IN C HE M IC A L PROCESSING UNITS 23.25
■Vi • ■
’^
V Z / •>"» S A
% H i 'T / T ;
rV ///n \ ’7
•yVS S - W - # v
, v. /?• .VS %V Vi;
► f V\‘ '
S%V •■•<-^
L ' ',/v' >
► * / <.y ^
L iquid A dequate L iq u id
M a l-D istrih u tio n D istrib u tio n
F I G U R E 23.11 Gas pressure drop in gas-liquid flow packed columns.
Per
P r e s s u r e Drop
Unit H i g h t
Gas
DP/Z,
F I G U R E 23.12 Log-Log plot of a typical gas pressure drop in gas-liquid flow packed
columns.
23.26 FLUID FLOW HANDBOOK
where
C='*dpg'
M l
(23.57)
K j
Fair [10] proposed the following equation to find C:
C=FstFfFhaCf (23.58)
where
F lv - L I G ( p c / p L) ° 5
F I G U R E 23.14 Capacity factor for tray columns. (L, G: Liquid and gas mass flow rate)
23.4 MIXERS
Mixing and agitation operations are among the most common process operations in
the chemical, paint, food, and pharmaceutical industries. In a mixing process, two
fluids are blended by creating a high degree of hydrodynamic turbulence to bring
the two fluids into a required degree of product uniformity and homogeneity within
a given time and power. The required uniformity of the final product may be phys
ical, chemical, or both. In gas-liquid contact, liquid-solid, gas-solid, or gas-liquid-
solid agitation may be used to increase the interfacial contact area or to ensure a
homogenous distribution of the solid phase. Creation of the high degree of hydraulic
turbulence may be done by many devices, the most common is mechanically agi
tated mixers. Other devices may also be used such as jet mixes, in-line static mixers,
23.28 FLUID FLOW HANDBOOK
in-line dynamic mixers, or extruders. Different types of mixer devices, and differ
ent dimension and characteristics of the same mixer category will produce differ
ent flow pattern. The mixer power consumption, and mixing time, which are the
most important design parameter, depend on the How pattern, degree of turbulence,
and physical properties of the fluids. Mixing processes may be classified based
on the fluid phases as [7]:
1. Single-phase liquid mixing, which covers the mixing of two or more liquid
fluids which are completely miscible. The process may or may not be com
bined with heat transfer and chemical reaction.
2. Mixing of immiscible liquids. When the fluids are immiscible liquids, the
mixing process is intended to increase the interfacial contact area by break
ing one fluid into tiny droplets. Large interfacial areas will enhance the
mass transfer between the two fluids and when fluids are brought into stag
nation, the immiscible fluids will separate again.
3. Gas-solid, liquid-solid, gas-liquid and gas-liquid-solid. The mixing process
will increase the interfacial contact area between the different phases which
enhances mass transfer and thus reaction or separation. When a liquid-
phase exists, the gas-phase is most likely dispersed as small bubbles while
liquid is continuous. Solid-phase may be stationery as in packed columns
or suspended or fluidized as in slurry beds and fluidized reactors.
Mechanically agitated vessels are very common in the mixing technology. A wide
range of mechanical designs exists; however, a typical mixing vessel consists of
a tank, impeller, baffles, and a motor/gearbox drive unit (see Figure 23.15.) The
tank or vessel is often of a cylindrical shape mounted vertically. The dimension of
the vessel differs greatly with diameters ranging from 0.3 to 30 ft (0.1 to 10 m).
Vessel bottom may be flat, dished, or spherical. The vessel wall, internal or external,
may be fitted with heat transfer jacket. Baffles are mainly used to prevent the forma
tion of gross vortexing which is common in low viscosity mixing applications. Gross
vortex refers to the dramatic drop in the liquid level in the vicinity of the impeller
shaft; this phenomena results in a severe reduction of the mixing efficiency. Baffles
are thin strips mounted on the inside wall of the mixing vessel either flushed to the
vessel wall or a with a wall clearance. A typical baffle width is 1/10 of the vessel
diameter. Four to six baffles is a common number of baffles in a vessel. An impeller
is mounted on a shaft, a vertical central shaft in most applications, however, hori
zontal or inclined shafts have also been used, see Figure 23.16. The type and diam
eter of the impeller, the number of impellers mounted on a shaft, the location of
the impeller on the shaft, and whether the shaft is vertical or horizontal will greatly
affect the flow pattern produced and thus mixing characteristics. Many types of
impellers are available in the process technology; propellers, turbines, paddles,
anchors, and helical screw and ribbons are the most common ones. Figure 23.17
FLOW HYDRO DYNAM ICS IN C H E M IC A L PROCESSING UNITS 23.29
where
V e rtic a l sh a ft w ith
on e i m p e l l e r two i m p e l l e r s
In clin ed sh a ft
FIGURE 23.16 Orientations of different shaft and number of impellers.
s'
y j i
Three-blade ci Four-bladed
Propeller Turbine
1r~~n
Cuved blade
Paddle
Helical
Screw
Anchor
F I G U R E 23.17 Various impeller designs.
direction is toward the vessel walls, see Figure 23.18. Circular zones develop above
and below the impeller due to radial flow created by the turbine. Any slight change
in the turbine characteristics such as the number of blades, the depth or diameter
of the turbine, or the angle of the blade will change the flow pattern.
Large Blade Low-SpeedAgitators. Anchors, and helical screw and ribbons are
used for high viscosity systems. They are characterized by large diameters and
large blade area and relatively low tip speeds.
23.32 FLUID FLOW HANDBOOK
V( <
/
f l 'r ^ 'V s
•1 11
1
R a d i a l Flow , D isc
T urbine
F I G U R E 23.18 Flow patterns for different impeller types and positions.
The power consumption, rate of energy dissipation within the liquid, for a
mechanically agitated tank is a function of the fluid properties including density
and viscosity, tank dimension such as tank diameter, and impeller characteristics
such as impeller diameter, rotation speed, and impeller depth. Power consump
tion is related to the mixing Reynolds number and the mixing Froude number by
the following general-form equation:
P - CRexFry
1 () II I III (23.60)
FLOW HYDRODYNAMICS IN CHEMICAL PROCESSING UNITS 23.33
Cis a dimensionless shape factor and P0 is the power number which defined as:
Reynolds and Froude numbers for the mixing process represents the ratio of
the applied agitation forces to the viscose forces, and the ratio of the applied agi
tation force to the gravitational force and are expressed as a function of the im
peller diameter and impeller rotational speed as:
(23.62)
(23.63)
The Froude number in Equation 23.63 may be ignored for baffled tanks since
vortex formation does not develop or when Reynolds number is less than 300.
A typical power number curve as a function of Reynolds number is shown in
Figure 23.19. A unique power number curve exists for different agitated tanks;
however, the same curve may be used when geometrical similarity exists. The
power curve shown in Figure 23.19 is distinguished by three flow regimes: lam
inar flow at low Reynolds number, < 10, transition flow, and turbulent flow. In
the turbulent flow region, > 104, the power number is constant, while a negative-
sloped curve characterize the laminar area due to the dominant viscose forces.
Power curves for various impeller types are compared in Figure 23.20.
An overall mixer efficiency is defined as a function of the power input per
unit volume, mixing time as:
(23.64)
When a jet stream of the liquid, created by an outside pump, is introduced to a stag
nant liquid in a tank, a mixing effect is observed. The high speed of the pumped
liquid will cause the liquid in the tank to agitate and circulate. Jet mixers are used
in mixing of fluids with low viscosity. Figure 23.21 shows a schematic of a jet
mixers with multi-jet nozzles which is a typical orientation for large tanks. Gas
may also be used to agitate a liquid pool. The degree of agitation is a function of
23.34 FLUID FLOW HANDBOOK
Re
the gas flow rate and tank characterizes. Gas bubbles rising through a pool of liquid
will provide good agitation and when a draft tube is used, see Figure 23.22, an
efficient liquid circulation may also be achieved. Jet mixers are simpler in mechan
ical design than mechanically agitated tanks, however, their application to highly
viscose fluids is rare.
FLOW HYDRO DYNAM ICS IN C H E M IC A L PROCESSING UNITS 23.35
Pump
F I G U R E 23.21 Jet mixers in single-liquid phase mixing.
In-line mixers are advantageous over agitated tanks in continuous operations since
in-line mixers can be fitted in pipelines. In-line mixers are available in a variety of
geometrical configurations and sizes (% to 24 inch. 1 to 50 cm). The concept ot the
static in-line mixers depends on a repetitive splitting each fluid into two portions
and then mixing of the different portions. Dynamic mixers consist ot a rotor that
spins at high speed inside a casing and acting as fluid pump. The high degree of tur
bulence created by the rotor mixes the fluids.
REFERENCES
2 4 .1 1NTRODUCTION
24.1
24.2 FLUID FLOW HANDBOOK
In this chapter we are mainly concerned with the flow solver part of CFD. This
chapter is divided into five sections. In section two of this chapter we review the
general governing equations of the flow. In section three we discuss three stan
dard numerical solutions to the partial differential equations describing the flow.
In section four we introduce the methods for solving the discrete equations, how
ever, this section is mainly on the finite difference method. And in section five we
discuss various grid generation methods and mesh structures. Special problems aris
ing due to the numerical approximation of the flow equations are also discussed and
methods to resolve them are introduced in the following sections.
The equations governing the fluid motion are the three fundamental principles of
mass, momentum, and energy conservation.
Energy p — ■+ p(V.V) =
Dt dt V. q + O (24.3)
where p is the fluid density, Vis the fluid velocity vector, Tjj is the viscous stress ten
sor, pis the pressure, F is the body forces, e is the internal energy, Qis the heat
source term, t is time, O is the dissipation term, and V.q is the heat loss by conduc
tion. Fourier’s law for heat transfer by conduction can be used to describe q as:
q =-kWT (24.4)
k T
where is the coefficient of thermal conductivity, and is the temperature. Depend
ing on the nature of physics governing the fluid motion one or more terms might be
negligible. For example, if the fluid is incompressible and the coefficient of viscos
ity of the fluid,/J,
as well as, coefficient of thermal conductivity are constant, the
continuity, momentum, and energy equations reduce to the following equations:
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.3
Presence of each term and their combinations determines the appropriate solution
algorithm and the numerical procedure. There are three classifications of partial dif
ferential equations ; elliptic, parabolic and hyperbolic |6|. Equations belonging to
each of these classifications behave in different ways both physically and numer
ically. In particular, the direction along which any changes are transmitted is differ
ent for the three types. Here we describe each class of partial differential equations
through simple examples:
Elliptic
V2w= 0 (24.8)
u
where is the fluid velocity. The incompressible irrotational flow (potential flow)
of a fluid is represented by this type of equation. Another practical example of
this equation is the steady state pure heat conduction in a solid, i.e., W2T
= 0, as
can be readily obtained from Equation 24.7.
Parabolic
du j2
r—
— = vV2u (24.9)
at
v
where is the kinematic viscosity (v=/,Up ). Transient diffusion equation is rep
resented with a similar equation. In this type of equations, events propagate into
the future, and a monotone convergence to steady state is expected.
Hyperbolic
d2u ->d2u
( 2 4 - 1 0 )
c
where is the wave speed. In this case, values of solution depend locally on the
initial conditions. Continuous boundary and initial values can give rise to discon
tinuity. Solution is no more continuous and, therefore, shocks can be observed and
captured in this class of equations.
24.4 FLU ID FLOW HANDBOOK
Depending on the flow, the governing equations of fluid motion can exhibit
all three classifications.
The governing equation of fluid motion may result in a solution when the bound
ary conditions and the initial conditions are specified. The form of the boundary
conditions that is required by any partial differential equation depends on the equa
tion itself and the way that it has been discretized. Common boundary conditions
are classified either in terms of the numerical values that have to be set or in terms
of the physical type of the boundary condition. For steady state problems there are
three types of spatial boundary conditions that can be specified:
Here the values of the variable 0 on the boundary are known constants This j\.
allows a simple substitution to be made to fix the boundary value. For example,
u
if is the flow velocity, its value may be fixed at the boundary of the domain. For
instance, for no-slip and no-penetration conditions on the solid walls, the fluid
velocity is the same as the velocity of the wall.
Here the derivatives of the variable 0 on the boundary are know n/2, and this
gives an extra equation, which can be used to find the value at the boundary. For
example, if the velocity does not change downstream of the flow, we can assume
u
that the derivative of is zero at that boundary.
//)
III. Mixed type boundary condition: a(p+b——=/ 3(x,y, z)
on (24.13)
The physical boundary conditions that are commonly observed in the fluid
problems are as follows:
1. Solidwalls: Many boundaries within a fluid flow domain will be solid walls,
and these can be either stationary or moving walls. If the flow is laminar then
the velocity components can be set to be the velocity of the wall. When the
flow is turbulent, however, the situation is more complex.
2. Inlets: At an inlet, fluid enters the domain and, therefore, its fluid velocity or
pressure, or the mass flow rate may be known. Also, the fluid may have cer
tain characteristics, such as the turbulence characterizes which needs to be
specified.
AN INTRODUCTION TO COM PUTATIO NAL FLUID DYNAM ICS 24.5
Ps— O K (24.14)
where <7 represents the liquid-air surface tension and k the total curvature
of the interface[7-10J. A boundary condition is required at the contact line,
the line at which the solid, liquid and gas phases meet. It is this boundary
condition which introduces into the model information regarding the wet
tability of the solid surface.
In order to solve the governing equations of the fluid motion, first their numerical
analogue must be generated. This is done by a process referred to as discre-
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.7
Outlet
Periodic Boundaries
F I G U R E 24.2 Minimizing the
computational domain using periodic
boundary condition.
t
O p e n Boundary
I Solid Wall
Inflow 1
tization. In the discretization process, each term within the partial differential
equation describing the flow is written in such a manner that the computer can be
programmed to calculate. There are various techniques for numerical discreti
zation. Here we will introduce three of the most commonly used techniques,
namely: (1) the finite difference method, (2) the finite element method and (3) the
finite volume method. Spectral methods are also used in CFD, which will be
briefly discussed.
Finite difference method utilizes the Taylor series expansion to write the deriva
tives of a variable as the differences between values of the variable at various points
in space or time. Utilization of the Taylor series to discretize the derivative of depen
dent variable, e.g., velocity w, with respect to the independent variable, e.g., spe
cial coordinate jc, is shown in Figure 24.4. Consider the curve in Figure 24.4 which
u
represents the variation of with x, i.e., u(x). After discretization, the curve u(x)
can be represented by a set of discrete points, s. These discrete points can be
related to each other using a Taylor series expansion. Consider two points, (/' + l )
and (/ - I), a small distance A* from the central point, (/). Thus velocity ut can
be expressed in terms of Taylor series expansion about point (z) as:
and
These equations are mathematically exact if number of terms are infinite and
Ax is small. Note that ignoring these terms leads to a source of error in the numer
ical calculations as the equation for the derivatives is truncated. This error is
referred to as the truncation error. For the second order accurate expression, the
truncation error is:
(Ax)-'
h W l n!
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.9
By subtracting or adding these two equations, new equations can be found for the
first and second derivatives at the central position /. These derivatives are
_ -",-1 (a 3« ] (A*)2
(24.17)
(d x ) 2Ax vdx ' , 6
and
Equations 24.17 and 24.18 are referred to as the central difference equations
for the first and the second derivatives, respectively. Further derivatives can also
be formed by considering Equations 24.15 and 24.16 in isolation. Looking at
Equation 24.15, the first-order derivative can be formed as
24.10 FLUID FLOW HANDBOOK
This is referred to as the Forward difference. Similarly, from Equation 24.16 another
first-order derivative can be formed, i.e.,
(a32a> (Ax)
, Ax- U 2J 2
This is referred to as the Backward difference. As noted by the expressions, dif
ference formulae are classified in two ways: (l) by the geometrical relationship
of the points, namely, central, forward, and backward differencing; or (2) by the
accuracy of the expressions, for instance, central difference is second-order accu
rate, whereas, both forward and backward differences are first-order accurate, as
the higher order terms are neglected.
We can obtain higher order approximations by applying the Taylor series ex
pansion for more points. For instance, a 3-point cluster would result in a second
order approximation for the forward and backward differencing, rather than a first
order approximation:
Forward difference:
'du'
2Ax(-3ui+ 4w,+l + ui+2) + O(Ax)2 (24.21)
Backward difference:
fdi 2- 4 + u, 0(Ax)~
3 )+ (24.22)
// 2 \x
Similarly a 4-point cluster results in a third order approximation for the for
ward and backward differencing:
Forward difference:
'duN - 3m, - 6 uM- uM ) + 0(Ax)3 (24.23)
<dXy 6Ax
Backward difference:
/ d~\u\
(M|._2 + 6w-_j + 3 U j + 2w(+| ) + O (A a') (24.24)
dx 6Aa
The above difference equations are used to produce the numerical analogue of
the partial differential equations describing the flow. In order to apply this discre
tization method to the whole flow field, many points are placed in the domain to be
simulated. Then, at each of these points the derivatives of the flow variables are writ
ten in the difference form, relating the values of the variable at each point to its
neighboring points. Once this process is applied to all the points in the domain, a
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.11
set of equations are obtained which are solved numerically. For more discussion on
this topic refer to text books on numerical analysis such as Hildebrand, and Chapra
and Canale 11 1. 12].
In the finite element method, the fluid domain under consideration is divided into
finite number of sub-domains, known as elements. A simple function is assumed
for the variation of each variable inside each element. The summation of variation
of the variable in each element is used to describe the whole flow field. Consider
u
the two-nodded element shown in Figure 24.5, in which variable varies linearly
inside the element. The end points of the element are called the nodes of the ele
u, u
ment. For a linear variation of the first derivative of with respect to is simplyx
u
a constant. If is assumed to vary linearly inside an element, we cannot define a
second derivative for it. Since most fluid problems include second derivative, the
following technique is designed to overcome this problem. First, the partial dif
ferential equation is multiplied by an unknown function, and then the whole equa
tion can be integrated over the domain in which it applies. Finally the terms that
need to have the order of their derivatives reduced are integrated by parts. This is
known as producing a variational formulation.
As an example, we will develop the finite element formulation of the Laplace’s
equation in one dimension:
(24.25)
+■ 1 - + > X
Xi-I *i+/
u
where velocity is a function of the spatial coordinate We multiply Equation x.
W
24.25 by some function and integrate it over the domain of interest denoted by Q:
f d~u
Jw dx2 dQ=0 (24.26)
r dWdu dQ+Ir W—
dun </T = 0 (24.27)
_ dx dx J dx
where T denotes the boundary of the domain Q and nx
is the unit outward normal
vector to the boundary T. The second order derivative in Equation 24.26 is now
transformed into products of first order derivatives. Equation 24.27 is used to pro
duce the discrete form of the partial differential equation for the elements in the
domain. Equation 24.27 is known as the variational form of the partial differen
tial Equation 24.25. Although this technique reduces the order of the derivatives,
it introduces the terms corresponding to the boundary of the domain into the gov
erning Equation 24.27.
We will now divide the domain into several elements and assume a function for
u
the variation of the variable in each element. If a two-noded linear element is
u
assumed (see Figure 24.5), the variation of in each element can be represented by
(24.28)
or
a. - u i-1
XM ~ Xi + «i+l
X; ~ JC.i-l
(24.29)
X i+\ ~ X i -1
The terms in the brackets are called the shape functions and are denoted as
AVs. Ui- u
| and w, + j are the nodal values of the variable and are denoted as w,’s.
u
Therefore, the variable can be written in the following form
Thus, the shape functions corresponding to the two-nodal linear element, repre
sented by Equation 24.28 are
N = (24.31)
*,-+i - * , - i
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.13
and
(24.32)
We can now determine the derivatives of the variable m, using Equation 24.30:
du m
dN. (24.33)
where m is the number of nodes on the element. Note that m/s are nodal values
of m and they are not variables, therefore, they are not differentiated.
In order to solve Equation 24.27 we still need to describe the function W.There
are several methods, which are used for the specification of the variable W.How
ever, the most common method is the Galerkin method in which Wis assumed to
be the same as the shape function for each element. Therefore, Equation 24.27 is
discretized by using equations similar to Equation 24.33 for the derivatives of the
variable, and equations similar to Equations 24.31 and 24.32 for W.For every ele
ment there can be several equations depending on the number of the nodes in that
element. The set of equations generated in this form are then solved together to
find the solution.
The above formulation was based on a linear variation of the variable in each
element. If higher order variations are used, e.g., quadratic or cubic, second deriv
atives will appear which require more points to describer them. This makes the com
putation more cumbersome. References [3] and [4] are recommended for more
detailed discussion of FEM.
The finite volume method is currently the most popular method in CFD. The main
reason is that it can resolve some of the difficulties that the other two methods have.
Generally, the finite volume method is a special case of finite element, when the
function W is equal to l everywhere in the domain. This technique is discussed
in detail by Patankar [34].
A typical finite volume, or cell, is shown in Figure 24.6. In this figure the cen
troid of the volume, point P, is the reference point at which we want to discretize
the partial differential equation.
The neighboring volumes are denoted as, W, volume to the west side, and E,
the volume to the east side of the volume P. For the one-dimensional finite vol
ume shown in Figure 24.6, the volume with centroid P, has two boundary faces
at w and e.
24.14 FLUID FLOW HANDBOOK
du du
d2u dx dx (24.34)
dx2
The first derivatives at the volume faces can be written as to be the differences
in the values of the variable at the neighboring volume centroids:
du _ uE- up (24.35)
dx XE - X P
and
du Up (24.36)
dx Xp—xw
We can apply this technique to Equation 24.25 to obtain its finite volume
formulation. The above method is also referred to as the Cell Centered (CC)
Method, where the flow variables are allocated at the center of the computational
cell. The CC variable arrangement is the most popular, since it leads to consid
erably simpler implementations than other arrangements. On the other hand, the
CC arrangement is more susceptible to truncation errors, when the mesh departs
from uniform rectangles.
Traditionally the finite volume methods have used regular grids for the effi
ciency of the computations. However, recently, irregular grids have become
more popular for simulating flows in complex geometries. Obviously, the com
putational effort is more when irregular grids are used, since the algorithm
should use a table to lookup the geometrical relationships between the volumes
or element faces. This involves finding data from a disk store of the computer,
which increases the computational time.
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.15
The main differences between the above three techniques include the following.
The finite difference method and the finite volume method both produce the numer
ical equations at a given point based on the values at neighboring points, whereas the
finite element method produces equations for each element independently of all the
other elements. It is only when the finite element equations are collected together
and assembled into the global matrices that the interaction between elements is
taken into account.
Both FDM and FVM can apply the fixed-value boundary conditions by insert
ing the values into the solution, but must modify the equations to take account of
any derivative boundary conditions. However, the finite element method takes
care of derivative boundary conditions when the element equations are formed
and then the fixed values of variables must be applied to the global matrices.
One advantage that the finite element method has is that the programs are writ
ten to create matrices for each element, which are then assembled to form the global
equations before the whole problem is solved. Finite volume and finite difference
programs, on the other hand, are written to combine the setting up of the equations
and their solution. The decoupling of these two phases, in finite element programs,
allows the programmer to keep the organization of the program very clear and the
addition of new element types is not a major problem. Adding new cell types to
a finite volume program can, however, be a major task involving a rewrite of the
program and so some finite volume programs can exhibit problems if they have
multiple cell types. The differences between the three techniques become more
pronounced once they are applied to two- and three-dimensional problems.
As it was stated in section two, CFD provides the solution to the governing equa
tions of the flow subject to particular initial and boundary conditions. Equations 24.1
24.16 FLUID FLOW HANDBOOK
to 24.3 plus the equations of the state or the property relations are the general form
of the governing equations. These equations are highly non-linear and very difficult
to solve even numerically. In applying these equations to a particular problem, some
of the terms may disappear or be negligible which makes the solution much simpler.
Various numerical techniques are developed for each of the particular application
of the general flow equations and their simplified forms. In order to introduce vari
ous computational techniques we will first consider a simple form of the momen
tum equation, and then discretize various forms of that equation. The momentum
equation (24.2) for a one-dimensional, incompressible flow with no body force,
and constant properties reduces to
du du d'u l dp
dt + dx=ox ------
V p ox- f (2437)
The first term in Equation 24.37 is the transient term, the second is the con
vective term, the third is the diffusive term, and the fourth is the pressure term.
We will consider various combinations of the terms in this equation and discuss
the methods to solve them.
Consider only the first and the third terms in the above equation and, to further
simplify, assume v = 1:
(24.38)
dt dx
This is the transient diffusion equation which consists of a first derivative in the
time direction / and a second derivative in the space direction x.This is a para
bolic partial differential equation that can be used to model the temporal changes
in the diffusion of some quantity through a medium. For instance, the transient
diffusion of heat (conduction) in a solid. We will solve this equation using both
a finite difference and a finite element approach.
/!+/-
At
> 1-1
This procedure is referred to as the grid generation. Once the grid is generated one
of the differencing scheme can be used to discretize the governing equation of the
problem, Equation 24.38. The type of differencing scheme used depends on the par
ticular problem. It is mainly through testing that one may find the accuracy and effi
ciency of one scheme over another. One simple method to discretized the diffusion
equation is to use a forward difference formula for the time derivative and a central
difference formula for the spatial derivative. The discretized equation will then be
(24.39)
At Ax2
This can be written in the following form:
At At U;+' At
Ax2“i-1+
ri ,
1-2 (24.40)
Ax2 'Ax2 * «+ i
i n
Note that the velocity at position and time + 1 depends on the three values
u
at the time level /?. Thus by knowing the values of at time level , its value at n
n
the next time level + 1 can be calculated. Therefore, to start the calculation, val
u x
ues of in all the domain, e.g., all the locations, should be known. These known
values at t=
0 are known as the initial conditions.
24.18 FLUID FLOW HANDBOOK
We can generate other differencing equations. For instance, the right hand
side of Equation 24.38 can be discretized based on the next time level + I : n
;i+ l
u.
. r,
- a. C' - 2U; M+l
+ U.
. ,,» + l
*/+l (24.41)
At AjT
Equations 24.40 and 24.41 define an explicit and implicit form of equations,
respectively. In Equation 24.40, an unknown variable is directly related to a set
of known variables. When a direct computation of the dependent variables can be
made in terms of known quantities, the computation is said to be explicit. Some
common explicit methods for parabolic partial differential equations (e.g.. Equa
tion 24.38) are:
In Equation 24.41, several unknown variables are related to the several known
variables. This is referred to as an implicit equation. When the dependent variables
are defined by coupled sets of equations, and either a matrix or iterative technique
is needed to obtain the solution, the numerical method is said to be implicit. Some
common implicit methods for parabolic partial differential equations are:
ul\'-2u "+i +
T Mi+1
At A .r
This scheme has first-order accuracy with a truncation error of 0[Ar, (Ax)~]
and is unconditionally stable.
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.19
11, -11,
r
. "+I II 1 . /I+l /I+l . /l+ l n
1
rN
1
+
_
1
/-I ~ 2 “ i + " ,+ . U i- 1
+—
Ax2
1
1
2
<1
A.v2
(24.46)
At 3 jt
24.20 I I A ID FLO W HANDBOOK
J IV
u/I+l -uI dn = J w d2u da (24.47)
At dx2
We now integrate the second derivative on the right hand side by parts to obtain:
u
Note that the continuity requirement for is reduced from second to first deriva
tives, therefore, we say it is weakened. We will now divide the domain into a series
of linear elements and use the Galerkin method to derive the finite element formu
u
lation 13). On each element the variation of is described by:
in
u=J^Niui (24.49)
1=1
where m is the number of nodes on the element and the /V, terms are the shape
u
functions. After substituting for the multiplier W, for the values of at the two
u
time levels and for the spatial derivatives of at the /?’th time level, an explicit
form of Equation 24.49 is obtained:
ij
where the indices refer to the summ ation. In many problems the boundary term
is not discretized. Usually, this so-called flux can be taken to be a known value that
needs to be added later. On the faces of most elements the flux term is ignored, as
we assume that the fluxes cancel out across those faces that are internal to the
domain. This is an equilibrium condition. It is only on the boundaries of the domain
that the flux terms need to be added. If the fluxes are not added, they will be cal
culated by the method as being zero, and because of this they are known as nat
u
ural boundary conditions. If we specify the value of at a boundary then the flux
term is not required, just as with the finite difference method, and this is known
as an essential boundary condition.
For simple elements the shape functions /V, are simple functions of the coordi
nates, say jc, and so Equation 24.50 can be integrated exactly over each element, but
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.21
*..
(W,"+l
#22 j n +1 (24.51)
V*2I V W2 /
where the terms atjare functions of position derived from the integration of the
first term on the left hand side of Equation 24.50, and the terms come from all f
the other terms in Equation 24.50. This matrix equation is, in fact, part of a larger
matrix equation for all the unknown values of Once all the equations for each u.
element, the so-called element equations, are known then the full set of equations
for the whole problem has to be produced. This is shown in Figure 24.8 where two
elements are shown. An expanded version of the element equations can be formed
by relating the local node on an element to its global node number. For example,
on element 2 the local node numbered 1 is global node number 2. Combining the
two expanded equations produces a global matrix equation, and the process of com
bination is known as assembling the equations. This is done by adding all the ele
ment equations together as follows:
o '
'* 1 1 *12 ( f i)
= (24.52)
*21 *22 0 «r* f2
0 0
V u3? ' /
l o 0 ,
\ /
This gives:
/ \
an *12 'ur N ' /. '
0
«2I *22 ^11 b\2 < • —fi+Rt (24.54)
,0 K ^22 / Vm3"+i < <?’ y
where terms and g, come from the terms on the right side of Equation 24.50.
The matrix on the left hand side is called the stiffness matrix and the matrix on
the right hand side is called the load vector.
24.22 FLUID FLOW HANDBOOK
Element 1
•—
Local node 1 Local node 2
Element 1 Element 2
(b) Tw o Elements
Once these global matrices have been created, the fixed value boundary con
ditions are imposed on the matrices and the equations can be solved. Again the
solution of the original partial differential Equation 24.38 has been reduced to the
solution of a set of simultaneous equations. Finite elements produce the numerical
equations for each element from data at known points on the element and nowhere
else. Consequently, there is no restriction on how the elements are connected so
long as the faces of neighboring elements are aligned correctly. By this we mean
that the faces between elements should have the same nodes for each of the adjoin
ing elements. This flexibility of element placement allows a group of elements to
model very complex geometry.
By considering the first and the second terms in Equation 24.37 we obtain the main
parts of the equation representing the inviscid flows. If we further assume that the
u
velocity in the convection term is a constant (we differ the discussion of the non
linear terms to the next section), we obtain the wave equation. Thus, the first order
wave equation becomes:
(24.55)
c
where is the wave speed propagating in the x-direction. For an initial condition
given by
u(x,0) = f ( x ) (24.56)
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.23
where /(.v) is monotonic in x, the exact solution for a wave of constant shape is
u- f(x-ct) (24.57)
This is an explicit equation since only one unknown, 1 appears in the u”+
equation. Thismethod is refereed to as Euler ExplicitMethod and, unfortu
nately, it is unconditionally unstable and will not workfor solving the wave
equation. This method is first-order since the lowest-order term in the trun
cation error is first order, i.e., A and t, Ax.
2. First-Order UpwindMethod: The Euler method can be made stable by
using a backward difference instead of a forward difference for a positive
wave speed:*
'
A~t ' + Ax = 0 (24.59)
This method is stable for 0 < cAt/Ax< 1, where cAt/Axis referred to as the
CFL (Courant-Friedrichs-Lewy) number. This method is referred to as the
First-Order Upwind Method.
For discretized transport problems, the CFL number determines how
many mesh cells, a fluid element passes during a timestep. For compress
ible flow, the definition is different. Here, the CFL number determines how
many cells are passed by a propagating perturbation. Hence, the wave-
speed, i.e., fluid speed plus the sound speed, is employed. For explicit
time-stepping schemes, such as Runge-Kutta, the CFL number must be less
than the stability limit for the actual scheme to converge. For implicit and
semi-implicit schemes, the CFL limit does not constitute a stability limit.
On the other hand, the range of parameters in which these schemes con
verge may often be characterized by the CFL number.
3. LaxMethod:Another method of making the Euler equation stable is by
using an average value for u-1based on the two neighboring points:
. f l+ l
u i - - u L= 0
/ . . 'i i
( “ ,-M + ) / 2 ^ „
+ c _!±!-----U
n \ / 'I
UM
(24.60)
H
At 2Ax
This is referred to the Lax Method which is stable for CFL < 1. [18|
"■
2At 2Ax : -0 (24.62)
6. Lax-WendroffMethod[19]
, <,-<1 2 At ( n ^ n t n\ . _
At A#-2Ax 2(Ax) '
+C~ T 7 -----= c o/A (24.63
u
Here, w” + 1 is the predicted value for at point and time leveli 1. The n+
forward and backward differencing used in the above equations can be
changed depending on the particular problem.
8. Second-OrderUpwindMethod: This is a modification of the MacCormack
method where upwind (backward) differences are used in both predictor
and corrector.
Corrector:
= i j + ( « r * r - « r ) * ] - c £ ( « r - 2 « * . + < 2)}
(24.67)
The fluid dynamics of inviscid flows are governed by Euler equations. These
equations may have different character for various flow regimes. For time-depen
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.25
dent flows, the equations are hyperbolic for all Mach numbers. Therefore, a time-
marching method can be used to obtain the solution. In steady inviscid flows, the
Euler equations are elliptic for subsonic conditions, and hyperbolic for supersonic
conditions. Several simplified form of the Euler equations are used for inviscid
flows. For instance, if the flow is incompressible, by considering the flow is irrota-
tional as well; a solution to the Laplace’s equation for the velocity potential or stream
function can describe the flow field. The traditional method of solving hyperbolic
PDEs are by the method of characteristics. Alternatively, there are numerous FDM
based solution schemes for such flows.
For flows with shocks, several shock-capturing techniques are developed. Godunov
schemes have been particularly efficient for shock problems. Godunov supposed
that the initial data could be replaced by a set of piecewise constant data with discon
tinuities and used exact solutions of Riemann problems to advance the piecewise
constant data. One of the key points in Godunov schemes is to calculate the flux at
each interface of numerical cells through a Riemann problem. A major extension to
the Godunov’s scheme was made by Van Leer in his MUSCL scheme (Monotone
Upstream-centered Scheme for Conservation Laws) which used a Riemann solver
to advance piecewise linear data [22, 23]. Other examples of Godunov schemes
include Roe’s method, the piecewise parabolic method (PPM), and the TVD (Total
Variation Diminishing) methods [24-27].
Godunov schemes for hydrodynamical equations may be second-order accu
rate in time, but they are explicit. The time step in an explicit scheme is restricted
by the largest CFL number, which may not be larger than unity for a stable calcu
lation. The stability limit in an explicit scheme is imposed by the local conditions
in the regions, where wave speeds are high, regardless of the significance of spa
tial variations prevailing in the problems. The regions drastically reduce the time
step possible from explicit schemes. Implicit schemes for hydrodynamical equa
tions are favored over their explicit counterparts for some problems, in which the
time-step size necessary for procuring a required temporal accuracy may be signifi
cantly larger than that dictated by the explicit stability condition. Implicit-explicit
hybrid schemes are useful when a flow attains different wave speeds either in differ
ent regions or at different instants, and the time accuracy is important in some parts
of simulation domains.
Hybrid implicit-explicit schemes have also been developed, which use a com
bination of both schemes. The difference approximation in time is either implicit or
explicit, separately for each family of characteristics and for each cell in the finite
difference grid, depending on whether the local CFL number for that family is
greater than or less than one. To the extent possible, the hybridization is continu
ous at CFL number equal to one, and the scheme for the explicit modes is a second-
order Godunov method of a type discussed by Colella [28, 29].
24.26 FLUID FLOW HANDBOOK
Consider the convective term (second term) and the diffusive term (third term) in
Equation 24.37.
du d2u (24.68)
U~^~vd7
The first term contains a nonlinearity due to the convective term. These factors
increase the complexity of the solution. The nonlinearity of the equations will make
the iteration procedure very complex. Therefore, the equations are linearized at each
time step. Linearization is achieved by using the current value of the velocity at
a point or in a volume or element as the velocity multiplier. For example, the con
vective term (first term) can be written as
“i+1 (24.69)
2Ax
u
where a central difference is used for the derivative and is found from the cur
rent solution for u: u=u”. This linearization technique is conducted on all the
nonlinear terms in the equations before solving the set of simultaneous equations.
The solution procedure for this type of equations is shown in Figure 24.9. As shown
in this figure, there are several levels of iteration before a solution is obtained.
One other problem that needs to be addressed is that of producing numerical
forms of the convection operator. Problems occur when this operator is discretized
using central differences for the first derivative of the velocity. For example, take
the linearized form of the Equation 24.68:
u,
«/ + i - *-_ i
=v
Uj+ i —2Uj+ W/_ j (24.71)
2Ajc (Axj5
Re =
uAx (24.73)
The value of the cell Reynolds number has an important effect on the numer
ical equation. When the Reynolds number is less than two both terms on the right
hand side have positive coefficients but when the Reynolds number is greater
than two the first term on the right hand side becomes negative. This negative
term result in very poor results. Therefore, a restriction is put on the cell Reynolds
24.28 FLUID FLOW HANDBOOK
number. In a two-dimensional problem, each mesh cell has one cell Reynolds num
ber for each of its directions, defined by the cell dimension and the flow speed in
that direction.
One way around this limitation is to use a first-order accurate difference equa
tion to model the first derivative in Equation 24.70 instead of the second-order
accurate difference equation used above. However, the reduction in accuracy can
lead to a poor solution. Typically the use of lower-order accuracy schemes gives
results, which are the results for a flow which has more viscosity than the one we
are trying to model. Such schemes are in common use together with more accu
rate schemes. A good review of this topic is given by Abbott and Basco . The fol
lowing are options for the discretization of the convection operator [30]:
When considering all the terms in Equation 24.37 a special difficulty arises due
to the weak coupling of the velocity and pressure fields. For the incompressible
fluids, the continuity equation is only a function of velocity and not a function of
pressure. Only the momentum equations contain pressure terms. Since most of
the terms in the momentum equations are functions of the velocity components it
is natural to use these equations to produce the solutions for the velocity compo
nents. Then, the problem is how to obtain the pressure solution, since continuity
does not contain pressure. A direct method is to discretize all the equations, i.e.,
continuity and momentum, and solve them simultaneously. This results in a very
large solution vector that contains all variables and consequently very large com
putational effort. There are two commonly used methods to resolve this problem:
(1) pressure-based methods, and (2) methods based on the concept of artificial
compressibility (also known as pseudo-compressibility). We will only discuss the
former one.
the continuity equation. In this method the velocity and pressure are written in the
following form:
u= u*+ u
v = v* + v (24.74)
p = p*+ p'
u, p
where v, and are the actual velocity components and pressure, w*, v*, and p*
are the guessed or the intermediate values, and uf, p'
v', and are the corrections for
the velocity components and pressure. After substitution in the continuity equation:
(24.75)
dy
we obtain
dii dv du dv (24.76)
dx dy dx dy
In this equation the derivatives of the correction velocity components depend on
the derivatives of the velocity components that satisfy the momentum equations.
An approximate form of the momentum Equation 24.6 is used to relate the pres
sure correction to the velocity corrections. In order to obtain this approximate
form, we start with the momentum equations:
and
where A, B,C D uj Vj p}
and are matrices, and , and are vectors of the variables at
grid points or nodes. These equations can be rewritten if the variables are split
using Equation 24.74, to give
and
Since in each step of calculation we are solving for the estimated values or the
intermediate values, then
Au*=Bp]
and
cv; = DP]
Therefore, these terms can be subtracted from the matrix equations 24.79 and
24.80 giving
A u / = Bpj
and
Cv; = D p ;
These are the approximate forms of the momentum equation and can be written as
Uj'=A'Bp; (24.81)
v; =C-'DP; (24.82)
Using these two forms of the equations we can find the pressure from the continu
ity equation. This is done by substituting them into the modified continuity Equa
tion 24.76, to produce an equation for the correction pressure pj which has on its
right hand side the imbalance in the continuity of the flow after the momentum equa
tions have been solved. Once the correction pressure / / has been found, so andu
v' are readily calculated from Equations 24.81 and 24.82 . FinallyEquations 24.74
are used to find the corrected velocity components and pressure. At this stage in the
solution the velocity components satisfy the continuity equation and a new value of
pressure has been calculated, but the velocity components do not satisfy the momen
tum equations. To resolve both the solution of the momentum equations and the non-
linearity, the momentum equations are used again to produce further simultaneous
equations, which are solved, followed by the calculation of the correction pressure
and the correction velocities.
In the momentum Equation 24.6 the pressure variable appear in a first-order
spatial derivative. If a central difference is used to discretize the pressure deriva
tives, pressure solution may oscillate in what is known as a checkerboard pattern.
24.32 FLUID FLOW HANDBOOK
In the implicit set of equations each equation has several unknowns, therefore,
they should be solved simultaneously. There are many different methods for solv
ing such a set of equations. Here, we will describe only the general procedure for
solving a set of equations simultaneously.
Au= b (24.83)
u A
where vector represents the unknown variables, is a matrix which acts as an
b
operator on the vector of variables w, and is a vector of known values. The solu
tion to the above equation is written as follows:
u=A~lb (24.84)
A=LU (24.85)
where Lis a lower triangular matrix and Uis an upper triangular matrix. The inverse
can be easily found once matrix Ais decomposed into Land U . This is referred
to as the direct method. Direct methods are commonly used in the finite element
methods. However, the problem associated with the direct methods is that it requires
significant amount of computational times for large matrices. Many iterative meth
ods are developed to resolve this problem and reduce the computational effort.
AN INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS 24.33
Iterative methods obtain the solution by iteratively guessing the solution until the
correct one is found. In computational fluid dynamics, the governing equations
are nonlinear and the number of unknown variables is typically very large. There
fore, iterative techniques are more suitable for solving such equations, especially
when the equations are formulated implicitly. Iterations are used to advance a solu
tion through a sequence of steps from a starting state to a final converged state. This
is true whether the solution sought is either one step in a transient problem or a final
steady-state result. In either case, the iteration steps resemble a time-like process.
Various iterative schemes are designed and used in the numerical analysis [3,4].
We will introduce the more commonly used ones in CFD applications. Consider
a system of three equations as:
auul+anu2+al3u3=/?,
a2\U\-I- a22u2+ anu3=b2 (24.86)
=— [bx-anu, -a^uA
“u
u, = — \b2-a2]u{-a23u3] (24.87)
a 22
w3 = —
tf33
\b3- aM
ux- a32u?]
Note that this method can only work if in Equation 24.87 the diagonal terms of
matrix A, i.e., the terms aih
are not zero. The Jacobi method takes the right hand
k
sideof Equation24.87 to be the known values at the "th iteration and the left
k
hand side to bethe new values at the + l ’th iteration:
24.34 FLUID FLOW HANDBOOK
The Gauss-Seidel method uses the new values at the k+ l'th iteration on the
right hand side of the equations giving:
H, *+l =
- — \bt ]
«n
u2m= -L\b2-a2tU
iM-a2iu,k] (24.89)
a22
u,M = — Ml‘ + l - f l r « 2*+ l l
«3 3
Both of these methods require that an initial guess to the solution be made which
can then be used during the first iteration. Then the numerical equations are used
to produce a more accurate approximation to the numerically correct solution, where
all the variables satisfy the governing equations. This new approximation, the up
dated solution, is then used as the new starting solution and the process is repeated
until the error in the solution is sufficiently small. Each repetition of the solution
process is known as an iteration.
Sometimes during an iterative process the updated solution at the end of one
iteration can be very different from the solution at the start of the iteration. Con
sider Figure 24.10 which represents variation of velocity against time. The numer
ical scheme predicts the velocity uk+1 at some time At ahead of the current time
by using values of the current acceleration ak
and the current velocity in the uk
following way:
* + l _ k
-------
At —=ak (24.90)
or
uk+l=uk+akAt (24.91)
There are several other parameters that can be used to control the convergence
of the solution. These are: (i) the number of time steps to run; (ii) the number of
iterations within each time for solving the non-linearity of the problem; (iii) the
number of internal iterations required in solving the simultaneous equations; and
(iv) limits on the residual errors.
r =b- Au (24.92)
As the solution process progresses from one iteration to another, the residual errors
reduce in which case the solution is said to be converging. If the residuals become
larger, then the solution is said to be diverging. This process can be accelerated
in various ways.
24.36 FLU ID H OW HANDBOOK
k+\ k .
U-, — U-, + (24.93)
U3 = M , +
The terms in the brackets represent the residuals. We can multiply the residual by
a factor in order to accelerate the iteration process:
(0
Uyk —k .
+1
W| +
11
^ r»
------- [ 2 211
k+1 a
~ a 22U 2 ^23
*i
3 J (24.94)
CO T/ 4+1 a+i „ *1
w* —1/3 + |^l?3 ^ 31^1 32^2 33 3 J
33
The factor O) is called the relaxation factor and, for most systems, it is set
co
between one and two. If is unity the method becomes the original Gauss-Sei
del method. The proper selection of the relaxation factor depends on the particu
lar problem under consideration and it is generally obtained by trial-and-error.
U * 1! " - 1 4 * 1 -2 u*1 + u * n
M
At ~ Ax2
Let us assume that the (round off*) error between the numerical solution u*and
the exact solution to the finite-difference equation, w, is £, i.e.,
u*—
u= e
Since both uand u* satisfy Equation 39, so does the error:
€» pn_^c,2 p"T+fct>|£n
cici —
£',+l _ c,-|
(24.95)
At Ax2
24.38 FLU ID FLO W HANDBOOK
£j
If decreases as the solution is progressed in time, then the solution is sta
ble. In other words, for stability, we need to have the following conditions:
Considering that the solution to the transient diffusion is often exponential in time
and its spatial dependence is in the form of a Fourier series, it is reasonable to
assume that the error has a similar dependence on time and space, hence:
£=ea,JjejklX
i
kt
where is the wave number, j =\-\ a
and can be a real or a complex number.
For the stability condition to be satisfied, it is sufficient that each term in the above
series satisfy the stability condition, i.e.,
^a(f+Ar)^;Vi
,ceA i
<1 <1
eaeTk,x'"
Substituting one term of the above series in Equation 24.95 and rearranging, yields:
_ g - ^ - 2 -fg7^
At " (Ax)2
But,
= 1 — ^ s i n 2[l/2*,A x] = G
{Ax)2
where G is referred to as the amplificationfactor. For a stable solution, the mag
nitude of Gshould be less than unity;
-1 < G < 1
AN INTRODUCTION TO COM PUTATIONAL FLU ID DYNAM ICS 24.39
_A ^<!
(Ax)2 2
Application of the Von Neuman stability analysis to Equation 24.58 (or 24.59)
will result in the Courant-Friedrichs-Lewy stability criterion, i.e.,
CFL=—
Av
<1
Hence for stable solution of the wave equation, the time-step should be.
At<Ax/c
There are several methods to predict the convergence of the Jacobi and Gauss-
Seidel methods. These are discussed below.
Let us consider Equation 24.83:
An= b
A
The coefficient matrix can be split into three parts, B, 7, and D, which con
tains only the diagonal elements, i.e.,
A=B+D+T
For example, for the system of Equations 24.86, B, 7\ and Dare:
/
f0 0 o' a\\ 0 0 ' '0 *12 *13 '
B= *2 1 0 0 ; d= 0 a 22 0 ; T= 0 0 *23
Duk+'=b-(B+T)uk
or,
alternatively,
«f+l = — ( * i - X a X )
Clii j*i
P
The Jacobi method is convergent if the iteration matrix is convergent. A con
vergent matrix is a matrix whose spectral radius is less than unity. Spectral radius
is the largest eigenvalue of a square matrix, in this case P.
A more practical method to study the convergence of the method is to study the
norm of the solution vector change, <5, from one iteration to another. 5 is given as:
s * = x |« *
/
h
l>Xk
l j*i
The convergence rate increases if the left hand side is much bigger than the right
hand side. The above condition is indeed the simplest. As an example, when applied
to the Laasonen method Equation 24.41, this condition requires that:
(M 1 > 0
At
that is, the time-step must be positive. For a given grid spacing, smaller time-steps
will result in faster convergence. Similarly, for a Fixed time-step, larger grid spacing
results in a faster convergence. It should be stressed that a converged numerical
solution is not necessarily an accurate solution. Some typical graphs of the resid
ual value for one of the flow equations plotted against iteration number are shown
in Figure 24.11.
One of the most important parts of the CFD is the mesh generation. Although tor
very simple flows, mesh generation is easy, it becomes very complex when the
AN INTRODUCTION TO COM PUTATIONAL FLU ID D YNAM ICS 24.41
problem has many cavities and passages. Mesh generation is basically the dis
cretization of the computational domain. The mesh in finite difference methods
consists of a set of points, which are called nodes. The finite volume method con
siders points that form a set of volumes which are called cells. The finite element
methods use sub-volumes called elements which have nodes where the variables are
defined. Values of the dependent variables, such as velocity, pressure, tempera
ture, etc. will be described for each element.
Figure 24.12 shows some of the commonly used elements. The top row shows
two-dimensional elements (3-noded triangle, 4-noded quadrilateral, and a square),
whereas the bottom row shows the three-dimensional elements (4-noded tetrahedral,
6-noded prism, and 8-noded hexahedral). The most common type in CFD programs
is a hexahedron with eight nodes, one at each corner, and this is known as a brick
element or volume. For two-dimensional applications the equivalent element is a
four-noded quadrilateral.
Before generating the mesh, we should know something about the flow behav
ior. For instance, where in the flow field we have boundary layers, vortices, large
gradients in pressure or velocity, etc. The mesh size and shape should be such that
it can capture the proper physical conditions that occur in the flow. For regions
where large gradients exist, large number of points within the mesh is needed. This
is due to using very simple variation of the parameter, usually, linear, within the
each element. Thus the mesh should be small enough so that a linear approxima
tion between two points is valid.
u
Consider, for example, the variation of a function along the coordinate as x
shown in Figure 24.13. We will use a linear variation between the points of a
numerical solution. If a coarse mesh (Ax/J is used for the numerical calculation
of the curve, the results would be far from the actual variation as observed by
24.42 FLU ID FLOW HANDBOOK
A
3-noded triangle 4-noded quadrilateral square
connecting the open circles in Figure 24.13. However, a fine mesh (Ax$) can pro
duce much better prediction of the actual curve, as observed by connecting closed
circles. It is also clear that the errors are greater in the regions of large gradients.
This is due to the linear approximation used in the difference equations.
One of the main difficulties of mesh generation is that, in many cases we do
not know where the large gradients are. Some physical condition which result in
large gradients of velocity are the boundary layers near the walls, shear layers, and
shock waves. Large temperature and concentration gradients occur close to the
flame fronts and reaction zones among others. Grid refinement is needed to resolve
important flow details. Adaptive grid generation is the solution for complex phys
ical and geometrical problems in which the location of large gradients is not pre
dictable or varies with time, but this is beyond the scope of this text. Generally,
refinement is needed near walls, stagnation points, in separation regions, and in
wake regions. By increasing number of nodes better accuracy is achieved. Solu
tion should always (if possible) be based on grid independence tests with same
style and mesh arrangement.
Grid generation can be assigned to two distinct categories, structured or unstruc
tured grids. Relating the mesh structure to the numerical method; finite difference
programs require a mesh to have a regular structure and finite element programs can
use a mesh with an irregular structure. In theory finite volume programs could use
a mesh with an irregular structure, but many implementations insist that the mesh
has a regular structure.
When a mesh with a regular structure is used there is an advantage in that the
solver program should run faster than if a mesh with an irregular structure is used.
This is due to the implicit relationship that exists between the number of a cell or a
point and the number of its neighbors in a regular mesh, which enables data to be
found easily. No such relationship occurs for meshes that have an irregular structure
AN INTRODUCTION TO COM PUTATIONAL FLU ID DYNAM ICS 24.43
and so when trying to find the values of flow variables in neighboring volumes
there must be a computational overhead. This often takes the form of a look-up
table which relates the faces to the cells or the nodes to the elements.
stretching of grid elements using algebraic method can be done by different func
tions such as: polynomial, trigonometric, logarithmic, and geometric functions.
Using the algebraic grid generation results in a good control over the grid struc
ture and is relatively simple to apply.
Unstructured grids have the advantage of generality in that they can be made to
conform to nearly any desired geometry. This generality, however, comes with a
price. The grid generation process is not completely automatic and may require con
siderable user interaction to produce grids with acceptable degrees of local resolu
tion while at the same time having a minimum of element distortion. Unstructured
grids require more information to be stored and recovered than structured grids
(e.g., the neighbor connectivity list), and changing element types and sizes can
increase numerical approximation errors. A popular type of unstructured grid con
sists of tetrahedral elements (Figure 24.15b). These grids tend to be easier to gen
erate than those composed of hexahedral elements, but they generally have poorer
numerical accuracy. For example, it is difficult to construct approximations that
maintain an accurate propagation of one-dimensional flow disturbances because
tetrahedral grid elements have no parallel faces.
In summary, the best choice for a grid system depends on several factors: con
venience in generation, memory requirements, numerical accuracy, flexibility to
conform to complex geometries and flexibility for localized regions of high or low
resolution.
eo
-a
n
c
o
*—<
c3
3
Q-
E
o
( J
c
'5
E
o
-a
1G5
O
3a.
E
o
u
•a
a>
n
E
o
T3
’rt w
o x:
H
<S)
>,
_c
CU
<N
w
as
o
24.45
24.46 FLU ID FLO W HANDBOOK
24.7 REFERENCES
1. Hirsch, C. N um erical Computation o f Internal and External Flow s, John W iley &
Sons. 1992.
2. T annehill, J.C ., Anderson, D.A., and Pletcher, R.H. C om putational Fluid M echanics
an d H eat Transfer, T aylor & Francis. 1997.
3. Z ienkiew icz, O.C. and Taylor, R.L. The Finite Elem ent M ethod— Volume 2: Solid
and F luid M echanics, M cGraw -H ill, New York. 1991.
4. Reddy, J.N. A n Introduction to the Finite Elem ent M ethod, M cGraw -H ill, New
York. 1993.
5. Sm ith, G.D . N um erical Solution o f Partial D ifferential Equations: Finite Difference
M ethods, 3rd ed., Clarendon Press, Oxford. 1985.
6 . Tyn M yint-U. Partial D ifferential Equations o f M athem atical P hysics, Elsevier
North H olland. 1980.
7. Poo, J.Y ., and Ashgriz, N. ‘"Curvature Calculation in Interfaces,” J. Compt. Phys.,
84, no. 2, pp. 483-491. 1989.
8 . Brackbill, J.U ., Kothe, D.B. and Zem ach, C. “A C ontinuum M ethod for M odeling
Surface T ension,” J. Compt. Phys., 100, 335. 1992.
9. A shgriz, N., and Poo, J. Y. “ FLAIR: Flux Line-segm ent Advection and Interface
R econstruction,” J. Compt. Phys., 93, 2, 449-46. 1991.
10. M ashayek, F., and Ashgriz, N. “A Hybrid Finite Elem ent— Volum e o f Fluid
M ethod for Sim ulating Free Surface Flows and Interfaces,” Int. Journal o f N um eri
cal M ethods in Fluids, 20, 10, 1363-1380. 1995.
11. H ildebrand, F.B. Introduction to Num erical Analysis, M cG raw -H ill, New York.
1956.
12. Chapra, S.C., C anale, R.P. N um erical M ethods fo r Engineers, M cG raw -H ill, New
York. 1988.
AN INTRODUCTION TO COM PUTATIONAL FLU ID DYNAM ICS 24.47
13. G ottlieb, D. and O rzag, S.A. “ Num erical Analysis o f Spectral M ethods: Theory and
A pplications,” SIAM . Philadelphia. 1977.
14. Richardson, L.F. “The A pproxim ate Arithm etical Solution by Finite D ifferences o f
Physical Problems Involving Differential Equations, with an Application to the Stresses
in a M asonry D am ,” Philos. Trans. R. Soc. London, Ser. A, 210, 307-357. 1910.
15. DuFort, E.C., and Frankel. S.P. “Stability Conditions in the Num erical Treatm ent of
Parabolic E quations,” M ath. Tables O ther Aids Com pute 7, 135-152. 1953.
16. Laasonen, P. U ber eine M ethode zur Losung der W arm eleitungsgleichung, Acta
M ath., 81, 3 09-317. 1949.
17. Crank, J. and N icolson, P. “Practical M ethod for Numerical Evaluation o f Solutions
o f Partial D ifferential Equations o f the Heat-Conduction T ype,” Proc. Cam bridge
Philos. Soc., 43, 5 0 -6 7 . 1947.
18. Lax, P.D. “W eak Solution of N onlinear Hyperbolic Equations and their Num erical
Com putations., Comnuin. Pure Appl. M ath., 7, 159-193. 1954.
19. Lax, P.D., and W endroff, B. “ System s o f Conservation Law,” Commun. Pure Appl.
M ath., 13, 2 17-237. 1960.
20. M acC orm ack, R.W . “The Effect o f Viscosity in Hypervelocity Impact C ratering,”
AIAA paper 69-354, C incinnati, Ohio. 1969.
21. Godunov, S.K. “Finite Difference M ethod for Num erical C om putation o f D iscontin
uous Solutions o f the Equations o f Fluid Dynam ics,” Mat. Sb., 47, 271-306. 1959.
22. Van Leer, B. “T ow ards the Ultim ate Conservative D ifference Schem e, V: A Second
O rder Sequel to G o d u n o v ’s M ethod,” J. Comput. Phys., 32, 101-136. 1979.
23. Van Leer, B. “T ow ards the ultim ate conservative difference schem e. I. The quest o f
m onotinicity,” in Lecture N otes in P hysics, vol. 18, Springer Verlag, Berlin, p. 163.
1973.
24. Roe, P.L. “The Use o f the Reim ann Problem in Finite-D ifference Schem es,” Lect.
Note Phys., vol. 141, Springer-V erlag, New York, pp. 354-359. 1980.
25. Roe, P.L. “A pproxim ate Reim ann Solvers, Param eter Vectors and D ifference
Schem es,” 7. Comput. Phys., 43, 357-372. 1981.
26. W oodward, P.R., and P. Colella, P. Lecture N otes in Physics, vol. 141, Springer-
Verlag, New Y ork/B erlin, p. 434. 1981.
27. Harten, A. “H igh-R esolution Schem es for Hyperbolic Conservation Law s,” J. Corn-
put. Phys., 49, 3 57-385. 1983.
28. Colella, P., and H.M . G laz. “Efficient Solution A lgorithm s for the R iem ann Prob
lem for Real G ases,” J. Comp. Phys., 59, 264. 1985.
29. Colella, P., and P.R. W oodw ard. J. Comp. Phys., 54, 174. 1984.
30. Abbott, M. B. and Basco, D. R. “C om putational fluid dynamics: An introduction for
engineers,” Longm an Scientific & Technical, Harlow, Essex, England; W iley, New
York. 1989.
31. Leonard, B.P. “A Stable and A ccurate Convective M odeling Procedure Based on
Q uadratic U pstream Interpolation,” Comput. M ethods Appl. Mech. Eng., 19, 5 9-98.
1979.
32. Harlow, F. H., and J. E. W elch. “A Calculation Procedure for Heat, Mass and M om en
tum Transfer in T hree-D im ensional Parabolic Flow ,” Phys. Fluids, 8, 2182. 1965.
33. Patankar, S.V ., and D. B. Spalding. “ Num erical Calculation o f T im e-D ependent
Incom pressible Flow o f Fluid with Free Surface,” Int. J. Heat M ass Transfer, 15,
1787. 1972.
34. Patankar, S.V. N um erical H eat Transfer and Fluid Flow, Hem isphere, W ashington,
DC. 1980.
24.48 FLU ID FLO W HANDBOOK
35. Harlow , F.H. and W elch, J.E. “Num erical C alculation o f Tim e— D ependent Viscous
Incom pressible Flow o f Fluids with Free Surface,” Phys. Fluids, 8, 2182-2189.
1965.
36. Chorin, A.J. “Num erical Solution to the N avier-Stokes Equations” . Math. Comput.,
22, 7 45-762. 1968.
37. Issa, R.I., A. G osm an, A.D., and W atkins, A.P. “The Com putation o f Com pressible
and Incom pressible Flows by a Non-iterative Im plicit Schem e,” J. Comput. Phys.,
6 2 ,6 6 .1 9 8 6 .
38. Issa, R.I. “Solution o f the Im plicitly Discretized Fluid Flow Equations by Operator-
splitting,” J. Comput. Phys., 62, 40. 1986.
39. A rakaw a, A. “Com putational Design for Long-Term Numerical Integration o f Equa
tions o f Fluid Motion: Tw o-D im ensional Incom pressible Flow. Part I,” J. Comput.
Phys., 1, 119. 1966.
40. Rhie, C.M . and Chow, W .L. “Num erical Study o f the Turbulent Flow Past an Air
foil with Trailing Edge Separation,” A lA A J ., 11, 1525. 1983.
41. Rhie. C.M . Comput. Fluids, 13,443. 1985.
42. M elaaen, M .C. Int. J. Numer. M ethods F luids, 15, 895. 1991.
43. C oelho, P.J., and J. C. F. Pereira. Int. J. Numer. M ethods Fluids, 14, 423. 1992.
44. Frankel, S.P. “Convergence Rates o f Iterative T reatm ents o f Partial Differential
E quations,” Math. Tables O ther Aids Comput., 4, 6 5 -7 5 . 1950.
45. Pecem an, D.W ., and Rachford, H.H. “The Num erical Solution o f Parabolic and
Elliptic D ifferential Equations., J. Soc. Ind. Appl. M ath., 3, 2 8 ^ 1 . 1955.
46. Briley, W .R. and M cDonald, H. “Solution o f the Three-dim ensional Com pressible
N avier-Stokes Equations by an Im plicit T echnique,” Proc. Fourth Int. Conf. Num.
M ethods Fluid Dyn., Boulder Colorado, Lecture N otes in Physics, vol. 35, Springer-
V erlag, New York, pp. 105-110. 1974.
47. Beam , R.M ., and W arm ing, R.F. “ An Implicit Finite Difference A lgorithm for
H yperbolic System s in Conservation Law Form ,” J. Comput. Physics., 22, 87-110.
1976.
48. Hirt, C.W . “ Heuristic Stability Theory for Finite D ifference Equations,” ./. Comput.
Phys., 2, 339—355. 1968.
49. Atta, E.H., and Vadyak, J. AlAA J., 21, 1271. 1983.
__________ CHAPTER 25__________
CORROSION AND EROSION
IN PIPES
VARIABLE DEFINITION
25.1
25.2 F LU ID FLO W HANDBOOK
25.1 INTRODUCTION
Corrosion and erosion result from interactions between the pipe surface and flu
ids travelling along that surface. Flow dynamics at surfaces have a profound
influence on the processes that take place. The location, rate and extent of metal
loss depend strongly on the nature of the flow regime.
Three flow regimes are illustrated in Figure 25.1 for single-phase fluid flow in
a pipe [l]. Laminar flow shown in Case l has a parabolic profile of flow velocities
across the pipe diameter due to drag experienced by the fluid travelling along the
pipe wall. Turbulent flow, Case 2, has a logarithmic velocity profile with large gra
dients in velocity occurring close to the pipe wall and is more often associated with
erosion-corrosion problems. Events in this boundary layer are of key importance to
corrosion or erosion of the pipe surface. In Case 3, a step change in pipe diameter
results in a complex velocity field with reverse flow occurring locally at the discon
tinuity. Conditions in this disturbed domain focus corrosion and erosion effects
in the affected region.
CORROSION AND EROSION IN PIPES 25.3
FIGURE 25.1 Three main types of pipe flow in a one phase liquid system. Case 1: laminar flow
showing a parabolic velocity profile across the pipe diameter. Case 2: turbulent flow showing a log
arithmic velocity profile across the pipe diameter. Case 3: disturbance of turbulent flow following
a step change in pipe diameter [ 1].
For continuous straight piping, the flow regime can be readily determined. The
transition from laminar flow to turbulent flow occurs over a velocity range that
depends on the geometry of the pipe, the viscosity of the liquid and the roughness
of the surface. A dimensionless Reynolds number (Re)
is used to take account of
such effects. The higher the Reynolds numbers, the greater the tendency for tur
bulent flow.
In turbulent How a boundary layer separates the flow of the bulk of the fluid trav
elling through a pipe from events associated with the pipe wall. The basic structure
and energy exchange in the turbulent boundary layer are illustrated in Figure 25.2
12]. Surface effects including corrosion and erosion of the pipe wall occur in the
viscous sub layer. Flow in this zone is essentially laminar. Viscous stresses asso
ciated with laminar flow are the principal mechanical forces acting on the solid
surface. While there is a transfer of energy between the outer layer and the lami
nar sub layer, this energy exchange is generally balanced in established flow.
Disruption of the turbulent boundary layer, particularly within the viscous region,
can have a significant influence on local events at the pipe surface. Movement of
corroding species to the pipe wall and of corrosion products from the pipe wall as
well as chemical reactions occurring at the surface can be upset by flow distur
bances [3-8]. Equilibrium conditions established in the hydrodynamic boundary
layer can be upset to produce a steady-state situation controlled by the kinetics of
local transport processes and chemical reactions. This can enhance local erosion and
corrosion damage resulting in localized areas of pitting or larger areas of metal loss
that reflect the pattern of local flow directions at the pipe surface [5]. Disruption of
the boundary layer in turbulent flow occurs primarily by the formation of turbulent
25.4 FLU ID FLO W HANDBOOK
; | D if& is*on:of;: j : ; : j : j : •: j : | : j : •: | | | : j : •
::::::::::: :::::: : by::::: $j turbiiliejitf w ierg y j: j : : :: j
s •s.’•1’•l ' :':’;ReyiioIds :StFess : j from ;sublayer j : | : j: j : j : j : j ::: j : j : | : j : | |
;: O a ^ r : Laiyeur: i::
bursts and sweeps that move fluid to and from the wall as a result of the formation
of local vortices 19—12].
Most severe corrosion and erosion problems under conditions of disturbed turbu
lent flow [3-8] occur at sudden changes in the flow system geometry, such as bends,
heat-exchanger-tube inlets, orifice plates, valves, fittings, and in turbo-machinery
including pumps, compressors, turbines, and propellers. Surface defects in the form
of small protrusions or depressions such as corrosion pits, deposits, and weld beads
can also give rise to disturbed flow on a smaller scale but can be sufficient to initi
ate local erosion-corrosion damage [13, 14]. The presence o f suspended solid par
ticles, gas bubbles in liquid flow, or liquid droplets in high-speed gas flow can be
especially damaging.
Oil
droplets
Water
Bubble
flow
Oil
Water
Stratified
flow
Mixed
flow
FIGURE 25.3 Two-phase flow patterns for a liquid/liquid system in
horizontal pipe, e.g. water and oil [15J.
Bubble
Plug
flow
flow
Smooth
Stratified
Wavy Pseudo
stratified slug
- ^ ’ ■
Rolling Annular
- -------- " flow
FIGURE 25.4a Two-phase How patterns for a liquid/gas system in horizontal pipe, e.g. oil (LVT
200 and C 0 2)[16].
25.6 FLU ID FLO W HANDBOOK
1.0 10.0
Superficial Gas Velocity, m/s
FIGURE 25.4b Flow regime map for various gas and liquid flow velocities [ 16).
through microbially induced corrosion (MIC) with the oil serving as a nutrient
source to sustain the action of a mixed microbial community including anaerobic
sulfate reducing bacteria [I7|.
Stratified flow occurs at increased oil velocity. It consists of water flowing along
the bottom of the pipe and oil traveling above with an interface between the two
phases. Again severe internal corrosion can occur in the aqueous zone along the
bottom of the pipe especially where the aqueous phase is a brine, contains corro
sive agents or where MIC occurs.
Smooth Slug
Stratified flow
W avy Pseudo
stratified slug
Rolling
w ave A n n u lar
flow
P lug
flow
FIGURE 25.5a Three-phase flow patterns for a water/oil/gas system in horizontal pipe [16].
CORROSION AND EROSION IN PIPES 25.7
c/)
E
o
o
CD
>
■g
’3
c r
<5
o
Q)
Q.
13
CO
F IG U R E 25.5b F low reg im e m ap to r various gas and liquid flow v elocities f 16].
In pipes sustaining higher velocities, turbulence can entrain water in the oil flow
and limit its contact with the metal wall (mixed flow, Figure 25.3). A more exten
sive discussion of flow patterns in oil-water mixtures can be found in reference [16].
For horizontal flow in gas-liquid systems, the How patterns observed include
bubble flow, stratified flow (smooth, wavy and rolling wave), intermittent flow
(plug, slug and pseudo slug) and annular flow. Figure 25.4. In stratified flow, the
boundary separating gas and liquid can be smooth, wavy or rolling wave, depend
ing on liquid and gas velocities. At increased liquid velocity, the waves grow and
bridge the pipe resulting in slug flow, which consists of four zones; a stratified liq
uid film with gas pockets above it, the mixing zone, the slug body, and slug tail. The
degree of gas entrainment in pseudo-slug flow is increased by higher flow veloc
ity. Annular flow occurs at high gas velocity and gas-liquid ratio. A liquid film flows
along the entire pipe circumference while a continuous gas phase bearing drops of
liquid moves through the core.
For multiphase oil/water/gas phase flow in horizontal pipe [16], three main
flow patterns are identified: stratified flow, intermittent flow and annular flow, Fig
ure 25.5. In stratified flow, water flows at the bottom of the pipe, oil flows above
the water and gas flows along the top of the pipe. The phase boundaries may change
from a smooth surface at low gas velocity to a wavy surface at higher oil/gas veloc
ity. With further increase in gas velocity, waves take the shape of rolling waves at
the gas-oil interface. At very high gas velocities, the oil-water interface tends to
break up, and the oil and water stream becomes well mixed. The oil-water mixture
flows as an apparently homogeneous liquid phase, with wave or rolling wave flow
at the gas-liquid interface. Intermittent flow includes three types of flow patterns;
plug, slug and pseudo slug flow. In plug flow, the oil flows over a water layer. The
oil bridges the pipe and the gas pushes the blockage along the pipe. The oil plug
flows at a faster velocity than the water layer. For slug flow, at low gas and low
liquid flow rates, the oil flows above the water. When either liquid or gas flow
rate is increased, the oil will break into the water layer and produce an oil/water
mixture. This mixture is not homogeneous at low relatively low' velocities. Oil
tends to the top of the pipe while water settles to the bottom. At high gas veloci
CORROSION AND EROSION IN PIPES 25.9
ties, the liquid phases become well mixed. The pseudo slug pattern resembles the
pseudo slug regime for gas-liquid two-phase flow. Annular flow resembles two-
phase gas/liquid annular flow but in this case the liquid is of an oil/water mixture.
Many of the examples noted for internal corrosion occur in multi-phase flow.
Where solids are present in low velocity flow patterns, significant scouring
can occur along the bottom of the pipe. For higher velocity regimes where the
solids remain suspended more general erosion effects can be expected especially
at points where flow is disturbed by changes in pipe dimensions, bends and the
like where particle impact is focused at particular locations on the pipe wall.
Slug flow has been highlighted as potentially the most aggressive regime for
general corrosion because of its very high turbulence [18]. Slug flow in oil and gas
pipelines results in enhanced corrosion damage where high shear force and flow tur
bulence strip off surface deposits or corrosion inhibitors that may have protected
the pipe wall. Slug flow may also result in the formation and impact of gas bubbles
on the pipe wall in the mixing zone along the six o’clock position on the internal
pipe surface [19]. Cavitation that produces impinging jets near the pipe surface [20]
can also lead to high local mass-transport rates or shear stress. Flow regimes are
significantly altered in inclined flow, Figure 25.6 [21 ]. Increased damage to the
pipe wall can result due to the dramatic effect of slug flow on corrosion [22].
Inclined flow leads to an increased probability of slug flow accompanied by an
increase in both the frequency and the intensity of slug movement [23].
A generalized summary of the types of corrosion expected for the flow regimes,
along with the distribution of the water phase and the degree of water turbulence,
is outlined in Table 25.1 [23 |. The occurrence of flow-induced corrosion is noted.
In the real world, few systems experience purely corrosion or purely erosion dam
age. Table 25.2 describes a spectrum of consequences arising from the progres
sive combination of corrosion and erosion processes ranging from predominantly
metal dissolution to predominately mechanical damage [24].
Many investigators use the term erosion-corrosion to denote environments where
the mechanical aspects of deterioration are more profound than chemical corrosion
aspects. Conversely, the term corrosion-erosion is applied where the corrosion con
tribution is greater than the mechanical. In actual practice it is quite difficult to sep
arate the contribution from each form of degradation; therefore, erosion-corrosion
has become a general term used to identify any situation in which corrosion is affect
ed by flow or vice versa.
For most systems, a critical or breakaway velocity can be observed at which the
corrosion rate abruptly accelerates due to the stripping of protective films from the
metal surface, Figure 25.7 [25]. At very high flow velocities a high rate of steady-
state material loss can be sustained by erosion alone in susceptible systems.
U U U
i s i
53
"7i<v
oQ. to
<u
c
o
c
o
£
S
O
C/3
<u ’3?
c
3
o
'c/3
c^ W
O W o
c—
a. to C/3 C/3
Cl E/3 CL C
OJ u P P
■O M t t ■o C/3 ■o
C/3 q •rt
^ 5y/.3
u- OJ c c I— JU i— c i_ <U
<u
3 .£RJ U o
<u —
TC OJ o o —o o
<u -3
c C3 Taj3 "O
Q. T3 c *3 ■a .£
>. T3 "3
UJ o> c •O
<u
H 3 to 30 o 3 2 3
c/3 o
3
3 3
o o T
C <*_
3 T3J C
_o
i*- 3
o _o "O
e •
—
o o ■a *a
c c
.£ .3
If u- tu
Zn OD 'c/3 •—
O
t 11 1 £ t
C
S
o _c c o c o .3
U U c l u- u. O £ U UL U a-
c £
JU o
3 x
JO Ofc
x J3
£ og £ •?
f- _o o o
>» X >. x:
(D £ .*,J 4)
g ob u-
a>
> > X > hJ X > > X
£
o.
•a 8-
H “O *o
<u
a.
c 1) (U
o X x X
’(75 •a
o
§ c £ -2 £
c3
§a. 5= .2
U
-J
i— S
2 =
<L*
o
OJ
o
3
c
U
w o C/3
c
o C3 £ 5 X) 5 5
!U
(J £ 3
O
^ s
13 3
3
o to «
-a £
o - 3
3
w
<U
o s a o O ■- o s I
G
3
CQ £ u CQ S 2 CQ S G
c
<u
_£
'5b
<U OJ
k. c3
w. £
£ «
_o cd
E $
£ 2> z
a o £ ~3 o
in
CM
<u
6
'ob
oij G
«u -o
■j'. 5J
s
* *S
O
0)
c/3
cd
X
o
L-
73
O
c:
T3
OJ
1c c ©
<U
a 53
^ 2 - 0
-*= ">»d
X
^ 'r
UJ
o
OC X •~
Q.
X v_,
G s X) 73
<L> a. *= 1
Oij c
-J 2 j> i
Q. -a r3 X5 X
<u 'g * s
aj C3 J3
_3 c
CQ
< o
Z G C?5 CO <
o
X
>>
CQ § Of) CO on < O
I- u. H H
25.10
C O R R O SIO N A N D E R O SIO N IN PIPES 25.11
Flow thins protective film to steady-state thickness, w hich is a function o f both m ass transfer rate and
grow th kinetics. E rosion-corrosion rate is controlled by the dissolution rate o f the protective film.
Film is locally rem oved by d issolution, surface sh ear stress o r p article/b u b b le im pact; but can
repassivate. E rosio n -co rro sio n rate is a function o f the frequency o f film rem oval, bare m etal
dissolution rate and su b seq u en t repassivation rate.
Film is rem oved and d o es not reform . E rosio n -co rro sio n rate is the rate the bare m etal can dissolve.
Film is rem oved and underlying m etal surface is m echanically d am ag ed w hich contributes to
overall metal loss i.e. ero sio n -co rro sio n rate is equal to bare m etal disso lu tio n rate plus possibly
synergistic effect o f m ech an ical dam age.
Film is rem oved and m ech an ical dam age to underlying m etal is the d o m in an t dam age m echanism .
Erosion-
corrosion
rate
V elocity
F IG U R E 25.7 An ideal rep resen tatio n o f the abrupt ch an g e in co rro sio n rate
at the critical o r breakaw ay v elocity co rresp o n d in g to the loss o f protective
film s from the m etal surface [25].
interest where the metal loss rate is controlled by convective mass transfer. Ellison
and Schmeal [30] demonstrated the validity of a flow-mechanical mass transfer cor
relation for prediction of metal loss rates for carbon steel in concentrated sulfuric
acid. In this system, the corrosion rate is limited by the convective transport of fer
rous ions away from the ferrous sulfate/liquid interface into the bulk acid. Another
interesting system that is controlled by convective mass transfer is the loss of car
bon steel in high purity boiler feed water at temperatures o f-l5 0 °C [31]. For this
system the anodic reaction is the rate-determining step and is limited by the disso
lution rate of iron oxide/hydroxide formed as a continuous layer on the corroding
surface. This dissolution rate can be directly related to the rate of convective mass
transfer from the oxide/liquid interface into the bulk liquid.
In many situations, mass transport considerations are more complex. In the
case of copper alloys exposed over time to flowing seawater, the rate of arrival of
oxygen, carbon dioxide and complexing agents from the bulk solution can all
play a role in determining the corrosion rate. To reach the surface, each species
must transit the turbulent boundary layer, laminar sub layer, diffusion boundary
layer, and the Helmholtz double layer, Figure 25.8. The slow step is usually pas
sage through the diffusion boundary layer. This sub layer is relatively static unlike
overlying zones that experience turbulent or laminar flow. As suggested by its name,
transport through this layer takes place by diffusion. Above the diffusion layer
key chemical species can move not only by diffusion, but also by other processes
such as turbulent mixing, mechanical stirring, convection, and thermal currents.
Stirring can decrease the thickness of the diffusion layer, and thus decrease the
time taken for travel through it to and from the solid surface.
CORROSION AND EROSION IN PIPES 25.13
AIR
TURBULENT ZONE
LAMINAR ZONE
(Slow diffusion, depending on
conc. gradient & mobility)
'c.oj$LO$i£>KniM
The rate of mass transfer Jat the solid/fluid interface can be described by
Fick’s law as [28]:
J=Dc(Ch-Cmlf)/8 (25.1)
where Dc is the diffusion coefficient for the diffusing species, and Cb Cm,f
are the
concentrations of thediffusing species in the bulk fluid and at the solid/fluid inter
8
face respectively, and is the thickness of the diffusion boundary layer [3-8]. For
8 V
a given flow geometry, is affected by the flow velocity and the kinematic viscos
ity. It follows that, for any flow system:
JocVa (25.2)
a
where the velocity exponent retains a value of -0 .5 or -0 .9 for fully developed
laminar and turbulent flow respectively. In the case of disturbed flow a local increase
25.14 FLU ID FLO W HANDBOOK
in turbulent intensity enhances the mass transfer rate, but the velocity exponent wil l
become smaller than 0.9. Presence of a second phase in a liquid, e.g. solid parti
cles or entrained gas bubbles that penetrate the viscous boundary layer, will increase
the mass transfer rate. Heat transfer across the metal/liquid interface can also have
this effect.
Increased mass transfer may or may not increase corrosion. For example, the
initiation of pitting corrosion on stainless steel is closely related to the local con
centration of corrosive species at the site of pit nucleation. Accordingly, an increase
in flow velocity that promotes convective transport of corrosive species away from
the nucleation site will retard formation of corrosion pits. On the other hand, decreas
ing the mass transfer rate can increase corrosion in systems where corrosion inhib
itors from the bulk phase are controlling the rate of corrosion at the metal surface.
The pattern of metal loss produced in a mass transport controlled corrosion pro
cess reflects the pattern of local flow distribution at the metal/liquid interface. Pro
gressive modification of the metal surface can in turn affect its interaction with the
fluid phase and further direct local attack. These effects vary depending on whether
the corrosion process is controlled by the rate of metal oxidation (the anodic pro
cess) or by the rate of the complementary reduction reaction (the cathodic process).
In the case where the anodic process (i.e. metal dissolution) is rate determining,
the corrosion morphology reflects the pattern of local mass transfer efficiency at the
metal/fluid interface. In the case of fully developed flow on a smooth surface, uni
form metal loss can be expected. Where flow disturbances are created by surface
irregularities increased mass transfer efficiency causes shallow pits or grooves to
form locally. A “horseshoe” morphology [28] is typically seen tor flow-induced cor
rosion in copper based alloys. This effect can be reproduced on the flat surface ot a
salt block dissolving in a liquid flowing over the block surface. The initiation of
horseshoe shaped washouts occurs around small surface irregularities due to local
turbulence increasing the efficiency of the dissolution process.
Where the cathodic process is rate determining, corrosion attack may not reflect
local increases in mass transfer efficiency [28]. For iron based alloys corroding in
aerated aqueous solutions with high electrolytic conductivity maximum metal loss
tends to be focused in areas with a low mass transfer rate. This is due to the fact
that anodic activity is stimulated by the accumulation of certain dissolved species.
The cathodic reaction (reduction of oxygen from the bulk solution) is favored else
where on the metal surface where faster mass transfer rates prevail.
flow velocity and mass transfer at the steel surface may therefore he a decrease
in corrosion rate due to (he formation of a more protective surface film [28]. In
other corrosion systems a high mass transfer rate at the metal/fluid interface may
prevent the deposition of solid corrosion products that would retard metal loss. It
is also possible that a corrosion product layer already present on the metal surface
could be dissolved or displaced as result of enhanced mass transfer exposing the
underlying metal surface to enhanced corrosion.
In phase transport controlled corrosion, the distribution and nature of attack de
pends on the wetting of the metal surface by the corrosive phase. In multiphase sys
tems, regions of wall contact by the water phase are subject to corrosion. The
degree of damage is related to turbulence in a liquid/liquid system and, therefore,
can be directly correlated with superficial flow velocities. In contrast to the smooth
streamlined attacks seen for convective flow dependent corrosion, damage in phase
transport controlled corrosion directly reflects the distribution of the corrosive
phase. Damaged areas are often covered by porous and voluminous corrosion
products [28].
Corrosion of this type is commonly seen in the production, processing, and trans
portation of oil and gas. The location and severity of damage are controlled by
the flow pattern, geometry of the piping and general flow parameters. For exam
ple, in a pipeline transporting crude oil, water droplets can settle out at low spots
below some critical flow velocity [28]. Attack can then proceed on pipe surfaces
contacted by a corrosive water phase with consequent formation of a layer of por
ous corrosion products on the pipe surface. These hold water and promote under
deposit attack. In systems that will support microbial growth, colonization of inter
nal pipe surfaces by a mixed population of microorganisms can result in the for
mation of a continuous biofilm in water wet areas. The biofilm is a complex domain
held together by a matrix or “slime” of extracellular biopolymers produced by the
organisms. The biofilm recruits material from the bulk How and entraps corrosion
products as well as organisms [32]. Greatly enhanced corrosion rates can occur in
certain cases. The best known mechanism for MIC involves anaerobic sulfate reduc
ing bacteria that link the oxidation of metallic iron to the reduction of sulfate in
their metabolism to produce and sustain corrosive iron sulfides on the metal sur
face. Biofilms also influence the nature of the flow regime near the pipe wall. Where
the biofilm thickness exceeds that of the viscous sublayer, a notable increase occurs
in the Fanning friction factor characterizing the drag exerted on fluid flow on the
pipe wall [33]. McCoy et al., [34] related this increase to the growth of filamen
tous bacterial cells that extended into the flow regime beyond the biofilm domain.
Experiments at Reynolds numbers of 25,000 and 47,000 showed that higher flow
velocities delayed the establishment of a thick biofilm but ultimately led to the
formation of a more tenacious deposit.
25.16 FLU ID FLOW HANDBOOK
Biofilms and corrosion deposits can hold water, increase drag, enable corrosion
and promote further wetting of the pipe surface. Because water is entrained in the
surface deposit it cannot be readily removed. Thus water collected in low spots on
a pipeline system during periods of static or stratified flow may not be displaced
by modifying the flow regime. Corrosion initiated in low spots during stagnant or
low flow conditions may persist under later flow conditions where water would
not be expected to collect in the pipeline. In this case corrosion can proceed unde
tected in what appears to be a flow of relatively benign oil. The usual approach to
controlling this kind of corrosion is to run cleaning tools through the line periodi
cally to disrupt surface deposits and remove water and debris.
The removal of corrosion films depends on the nature of surface film and the
nature of fluid flow. The erosion-corrosion resistance of alloys often falls into
three categories, depending on their film forming characteristics [25J. Alloys with
very tenacious protective films, such as titanium alloys and some Ni/Cr/Mo alloys,
exhibit excellent erosion-corrosion resistance in seawater. Alloys such as steels
and copper-base alloys that form protective or semi-protective films that are easily
removed are resistant at low velocities but seriously degraded at medium and high
CORROSION AND EROSION IN PIPES 25.17
where
AP is pressure drop
x is the distance along tube
V is the fluid velocity
ff is known as the Fanning friction factor which is a function of Reand the
roughness of the tube relative to its diameter.
Efird [35] has presented a good correlation between critical velocity and the
shear stress created at the film/diffusion layer interface for several copper-based
alloys in seawater. Shear stress must also be considered in the selection of a film
ing corrosion inhibitor for a given system to ensure retention of inhibitor on the
metal surface.
Removal of a surface film can also be described as a mass transfer process.
Because the film is simultaneously forming and dissolving at the metal/film inter
face there can be a critical rate of mass transfer above which the film is removed.
For corrosion products this rate depends on the bare metal dissolution rate [36-37]
responsible for formation of the corrosion product. Mass transfer is in fact inti
mately linked to wall shear stress through the Chilton-Colburn analogy:
/ \ 0-5
T
(25.4)
where kd
is the mass transfer coefficient, defined as the proportionality between
the mass flux density and the concentration gradient.
Cavitation-Erosion
Cavitation is defined as the repeated nucleation, growth, and violent collapse of cav
ities, or bubbles, in a liquid. The driving force for cavity collapse is the difference
between the hydrostatic pressure and the vapor pressure of the liquid. If, in a par
ticular location in a liquid flow system, the local pressure falls below the vapor
pressure of the liquid, then cavities may be nucleated, grow to a stable size, and
25.18 FLU ID FLO W HANDBOOK
be transported downstream with the flow. When they reach a higher pressure
region, they become unstable and collapse, usually violently. In practice, cavitation
can occur in any liquid in which the pressure fluctuates either because of flow
effects or vibrations in the system.
The collapse velocity, Vc,
of a cavity is a function of the hydrostatic pressure,
P, under which the cavity collapses, the volume, Cv, of the initial cavity, and the
density, p, of the liquid:
(25.5)
V P
Cavity radii for flow cavitation are typically from -0 .2 5 to l.O mm, and for
vibratory cavitation, -5 0 um. The cavity collapse velocity typically ranges from
100 to 150 m/s.
The collapse time, /, of the cavity is -1 0 0 ns for the above case. The estima
tion is related to the initial radius,R(„
of the cavity, the liquid density and the hydro
static pressure at collapse, P, as follows [ 1]:
(25.6)
Despite much research, the mechanism of damage has not been fully explained.
It is, however, clearly related to the conversion of the potential energy associated
with the size of individual bubbles prior to collapse into kinetic energy in the sur
rounding liquid [39-44J. Two mechanisms have been proposed [45-46]. The first
involves the formation of a shock wave radiated towards the surface and the other
suggests that damage results from the impact of a high-velocity micro jet, which
emerges from the collapsing bubble. When a cavity collapses within the body of
the liquid, away from any solid boundary, it does so symmetrically and emits a shock
wave into the surrounding liquid. On the other hand, those cavities that are either
in contact with or very close to a solid surface collapse asymmetrically, forming
a micro jet of liquid directed toward the solid surface.
In all practical situations involving cavitation, a large number of cavities form at
the same time in clusters. With increased external hydrostatic pressure, they col
lapse in a concerted manner, starting with those at the outer perimeter of the clus
ter and proceeding inward toward the central ones. In this sequence, much of the
energy generated by the collapse of the outer cavities is transferred to the cavities
in the inner part of the cluster through increased local hydrostatic pressure. This
results in a significant increase in the intensity of collapse of the central cavities [9].
Because of the localized nature of the cavitation process, the energy dissipation
results in a significant temperature increase associated with the collapse [381.
CORROSION AND EROSION IN PIPES 25.19
The local mechanical impact from the collapse of cavity clusters can he
severe. The impact energy is absorbed by the pipe surface through elastic defor
mation, plastic deformation, or fracture of the pipe material; the latter two pro
cesses lead to erosion of the pipe wall. The more elastic or plastic deformation
energy that the material can absorb, the greater will be the cavitation-erosion resis
tance of the material. Erosion for most materials is preceded by an incubation
period [47), during which little or no material loss occurs, although roughening
(plastic deformation) of the affected surface may take place. The incubation stage
is followed by an acceleration stage, during which the erosion rate increases
rapidly to a maximum. At this stage, the erosion rate remains constant or nearly
so. This is the rate most commonly quoted as the single-number result of an ero
sion test. The maximum rate stage is then usually followed by a deceleration (or
attenuation) stage, during which the erosion rate declines to some fraction of the
maximum rate.
It has been calculated that shock pressures generated in cluster collapse can
be comparable to the tensile strength of many commonly used materials [48]. How
ever, this level of pressure is probably insufficient to produce a pit on a metal sur
face in a manner analogous to the formation of an indent by a hardness indentor,
since the stress required for this would be roughly three times the tensile strength.
Damage by collapse shock pressure probably requires repeated impacts generat
ing a high frequency fatigue stress below the tensile strength. The incubation and
acceleration stages can be explained by a fatigue like failure mechanism, in which
numerous impacts are required to cause a local failure. No clear explanation exists
for the subsequent decrease in erosion rate, although factors such as reduced impact
due to topographical changes in the surface, or changes in surface properties may
be involved.
Attempts have been made to correlate erosion resistance to material proper
ties such as hardness, tensile strength or “ultimate resilience” defined as tensile
strength1(2 X Young’s modulus) |48]. It appears that a good correlation is found
for engineering alloys with a power of hardness in the range 2.0 to 2.5.
25.3.4 Erosion-Corrosion
or scale deposits from a susceptible metal surface by fluid shear stress under high
turbulence conditions, or by other erosion mechanisms. Even in the presence of
impinging solid particles, as in a coal slurry line, metal loss can be primarily due
to enhanced corrosion, not abrasion.
Enhanced metal loss due to erosion-corrosion in a system involving cavita
tion is illustrated in Figure 25.9 [59]. The collapse of gas bubbles damages the
surface film, causing increased corrosion in areas where the underlying metal sur
face is exposed. Increasing the rate of film breaks sustains an enhanced corrosion
rate (compare curves b and c, Figure 25.9). The overall rate of degradation is a func
tion of both the mechanical damage of the film and the corrosion of the metal pipe
through chemical oxidation.
The relative roles of corrosion and erosion following damage to a protective
film can be estimated on the basis of the exponent y in the relationship between
the metal loss rate and the bulk flow velocity, [21]: V
| Erosion + corrosion] Vy (25.7)
The value of y, listed in Table 25.3, depends on the relative contribution of cor
rosion and erosion to the total metal loss. In disturbed flow it is the flow charac
teristics in the direct vicinity of the wall rather than the bulk flow velocity that is
important. In practice, however, the superficial flow velocity is the flow parame
ter that is readily measured and controlled [60].
Erosion-corrosion is generally a form of phase transfer controlled corrosion
when solid particles are absent; however, in a fluid/solid particle system abrasive
panicles play a key role by impinging on or sliding along the metal surface causing
fatigue or shear stress on protective surface films. Whether fatigue or shear stress
represents the main mode of destructive action depends on the angle of impact of
the solid particles. Often, these two forms of damage are referred to as deforma
tion damage and cutting wear for impact angles around 90° and 30° respectively.
Erosion-corrosion may also be found in pure liquid systems where protective surface
F IG U R E 25.9 E rosion a n d co rro sio n fo r a surface w ith (a) grow th o f a p ro tectiv e integral surface
film , (b) a film broken o cc asio n ally o v er tim e and (c) a film broken freq u en tly [59].
25.22 FLU ID FLO W HANDBOOK
TA B LE 2 5.3 F low V elocity as a D iag n o stic Tool fo r E ro sio n -C o rro sio n R ates
F o llo w in g D am ag e to a P ro tectiv e S urface Film [60]
C orrosio n
L iq u id -p h ase m ass transfer control 0 .8-1
C h a rg e -tra n sfe r (activation) control 0
M ixed (ch a rg e /m a ss tran sfer) control 0 -1
A ctiv atio n /re p assiv atio n (passive film s) 1
E rosion
S o lid -p artic le im pingem ent 2 -3
L iq u id d ro p let im pingem ent in high -sp eed g as flow 5 -8
C a v ita tio n attack 5 -8
scales are susceptible to damage by shear stress. In some cases, even relatively low
shear stress can damage susceptible deposits.
There is sometimes a problem in differentiating erosion-corrosion and mass
transport controlled corrosion based on flow characteristics. This confusion is due
to the fact that flow hydrodynamics influence momentum transport (mechanical
action) and mass transport in a similar way (Chilton-Colburn analogy). Generally,
these processes can be distinguished phenomenologically by the appearance of
the damage. Erosion-corrosion occurs in the form of shallow pits, horseshoes or
other patterns of localized metal loss reflecting the direction of flow. In contrast,
sharp-edged deep pits with no obvious mechanical damage to the surface layer can
be attributed to mass transport controlled corrosion.
The classification of cavitation corrosion and liquid droplet impingement cor
rosion can also be ambiguous. From a mechanistic point of view these processes
might be classified as erosion-corrosion in cases where a protective film is removed
from a metal surface by mechanical action; however, because different fluid phases
are involved, these processes might also be classified as phase transfer controlled
corrosion. Usually, the flow threshold value for the occurrence of cavitation is
directly related to the superficial flow velocity. The intensity of attack is determined
by parameters such as bubble size (interfacial tension), number of cavitation nuclei
(water quality) and microscopic jet velocity— al! of which are relatively ill defined.
In liquid droplet impingement the droplet velocity is an appropriate quantity to
relate to corrosion.
A usten itic stain less S ulfuric acid slurry Process plant, p um p im p eller
steels W ater N u clear fuel cans
A lum inum allo y s C o n cen trated fum ing HNO^ Process plant
For copper alloys both laboratory data and practical experience suggest that for a
given tube diameter there will be a critical velocity above which erosion-corrosion
can occur [35]. This critical velocity is a function of the environment and is higher
tor more resistant alloys [60] (Table 25.5). A good correlation between critical veloc
ity and the shear stress created at the film/diffusion layer interface has been found
tor several copper-based alloys in seawater; however, the basis of this correlation
may be questionable. The critical shear stress values seem too low to disrupt any but
loosely adhered surface films [36-37] and the correlation was not established over
a range of geometries or environmental conditions. It may be that the protective film
was being dissolved in this case by the increased rate of mass transfer to and from
the surface rather than being mechanically removed [60J.
The relative resistance of copper alloys to erosion-corrosion can be judged
from the critical velocity values shown in Table 25.5 [60]. Copper and high copper
alloys have excellent resistance to seawater corrosion, but are susceptible to erosion-
corrosion at high How velocities. The addition of zinc to copper does not improve
corrosion in aqueous solution markedly as long as the zinc content is below 15 per
cent; above 15 percent dezincification can occur. Low levels of aluminum in a cop
per zinc alloy markedly increase its resistance to impingement attack in seawater
in turbulent high-velocity flow. Aluminum brasses are susceptible to dezincification,
but this can be inhibited by adding 0.02 to 0.10 percent arsenic. Addition of tin and
phosphorus to copper also produces good resistance to flowing seawater and alloys
containing 8 to 10 percent tin have high resistance to impingement attack. In gen
eral copper nickel alloys have the best general resistance to aqueous corrosion of
all the commercially important copper alloys.
■a E sC 00
0 .6 -0 .9
<u 3 c sC
E. E w> f^i r- T
<D T
o * •s in <N (N n
CJ ■a
< E
c
CJJ SO O' O'
*275 ‘o
1.3
c oc (N
U 13 c
Q >
C 3 »tn L—
O3 u. 00 ) O'
9.6
•s r- »- E
— O'
rI
5 ^ * £
•— Al
u
o E sC sO
00
0.6-0.9
D •— u O'
> T3 O' rn r-
<N
-T
r^i T
l/~i (N
Critical Velocity of Copper Alloys on Seawater [60]
03 E
.a E
•c «
u 7
C/5
•n u.
?
© »n
N <
OO )D yl/"l c/5in
0 .0 2 /P
99.9/Cu
>o u © 5 ? s r j O
O 3
(U 3 O
a. U 00 —
E
o y ° £
*r )
SS
o
I (J c
~ c/3
O'
U CO
oo
00
00
oi CN r- —
Phosphorus deoxidized copper
B
a>
E
U <u u >,
jr. j* <0
o o N
C 03
l—
E e
£ -C iu
N
o X! E c
a.' 2 -a
TABLE 25.5
o
a.
3
a.
3 E 03 2
3 03 ■o X>
E u u C O <D
o o ’c c/S
2
C
m E ID £
l _3 sz
3 <i c c
u i-' O' < J
25.24
CORROSION AND EROSION IN PIPES 25.25
The addition of iron to some copper alloys produces beneficial effects with
respect to erosion-corrosion due to an increase in the amount of nickel in the corro
sion product film on the metal surface 163]. Increased levels of nickel may increase
the electronic or ionic resistivity of the corrosion product film, impede anodic [64]
or cathodic [65] corrosion reactions or improve the mechanical strength of the
film [66].
The important detrimental role that trace soluble sulfide can play in damaging
piping systems has been highlighted in recent years [67]. The addition of ferrous
sulfate has proved extremely beneficial [68]. Surface pretreatments have also been
suggested as means to prevent problems especially during pre-service testing of
systems with contaminated water where the probability of sulfide damage can be
high. The detrimental effects of sulfides have been attributed to: the cathodic oxy
gen reduction reaction being catalyzed [69], the evolution of hydrogen being pro
moted or the reduction of the mechanical durability of protective surface films due
to incorporation of copper sulfide [70].
Figure 25.10 [71 ] summarizes the factors controlling the erosion-corrosion of steels
in water. The most important environmental variables are pH, oxygen content and
temperature of the water. These parameters can influence the nature of the corrosion
product films formed on the surface of exposed steel. Additions of small amounts of
oxygen [72] to the water can be beneficial since this increases the electrochemical
potential of the metal surface and promotes the formation of less soluble hematite
surface films. This beneficial effect, however, may be difficult to achieve in two-
phase water/steam systems where partitioning effects between water and steam
generally interfere with the effective distribution of additives. The effects of both
temperature and pH influence the solubility of magnetite films [73]. Additives con
taining chromium have proven to be effective in preventing erosion-corrosion but
higher amounts must be added in systems with high flow velocity or in two-phase
systems. The beneficial effect of these additives is attributed to segregation of
chromium into the magnetite oxide layer formed on the steel surface, which low
ers its solubility.
s
0 100 200 7 8 9 10
Temperature °C PH
Ic
o
<2 §o
§ U
.1 co
•
§o C£
O £
•§^C/* .01
•S
Hi
S ed
w
F IG U R E 25.10 Sum m ary o f factors controlling the erosion-corrosion o f steel in flow ing w ater (71
may not be primary corrosion products. a-Fe20 3 can be formed through a sec
ondary reaction involving air oxidation of siderite after removal of samples from
the operating system, while Fe3C is likely to be a residual component from the
steel [74].
C 0 2 corrosion is influenced by several factors such as temperature, pressure,
fluid composition, and flow parameters. DeWaard and Williams found that as tem
perature increased, corrosion rates rose until they hit a maximum between 60°C
and 70°C, after which they declined. [75]. Vuppu also observed this trend in flow-
loop studies of stratified flow [78]. This was attributed to the formation of a pro
tective film on the steel surface [76-77]. In more turbulent flow conditions, no
maximum corrosion rate was observed between 40°C and 80°C [79]. Instead, cor
rosion rates continued to increase with temperature. More turbulent flow tends to
CORROSION AND EROSION IN PIPES 25.27
leave the pipe surface less protected. It may prevent passive F eC 03 films from
forming, remove existing films, or retard the growth of such films by enhancing
mass transfer near the steel surface 174].
The corrosion rate increases with C 0 2 partial pressure [75]. Presumably, more
C 0 2 can dissolve in solution at higher pressures allowing greater quantities of
H2C 0 3 to form to drive the corrosion process. In oil-salt water systems, oil per
centages between 0% and 20% cause little change in corrosion rates. Oil cuts above
60% were found to cause a dramatic decline in corrosion rate probably due to the
exclusion of water from the steel surface by a non-corrosive oil film on the pipe
interior [77].
Wet Gas. Wet gas is a term applied to gas compositions that can separate liquids
as a second phase in handling and transport. In natural gas pipelines, changes in
temperature and pressure can result in condensation of moisture and liquid hydro
carbons along the length of a pipeline anywhere the relevant dew point is exceeded.
Liquid droplets can pose erosion concerns and even relatively low levels of C 0 2,
H2S and organic acids in combination with free water can generate potentially very
corrosive environments inside the pipe. Stringent gas specifications are therefore
commonly used in major gas transmission and distribution systems to avoid these
problems. Liquids and undesirable gas components are removed in upstream facil
ities prior to pipeline transportation with only sweet dry gas entering the transmis
sion distribution network. This strategy places high costs for dehydration, H2S
scrubbing units and the like on upstream facilities. Because of the costs involved,
gas is transported to centralized facilities for treatment where this is practical and
gas specifications are under constant challenge. In recent years the concept of raw
gas transport has received increasing attention to minimize treatment facilities and
associated operating costs.
The flow patterns that are encountered in gas systems with water present as a
second phase are mist flow, annular mist flow, and stratified flow, each of which
contributes differently to the corrosion rate [80]. In annular flow, a sheath of liq
uid travels along the pipewall with gas moving through the core of the pipe. This
flow pattern usually occurs when the gas velocity is high. At very high gas veloc
ity and low liquid content, a mist flow can be achieved in which water droplets
are suspended in the gas phase. The stratified flow regime is common at low gas
velocities. Typically, the gas phase flows in the upper portion of the pipe and the
liquid phase is transported in a film covering the bottom portion of the pipe. Some
liquid is transported as droplets entrained in the vapor phase. These can deposit
on the top of the pipe. Consequently corrosion can occur at both the top and bot
tom of the pipe in this flow regime.
Corrosion protection can be achieved by addition of corrosion inhibitors pro
vided these can be distributed to all parts of the pipeline under all transport condi
tions [81]. In principle, there are corrosion inhibitor packages commercially available
that can effectively protect the inner walls of carbon steel pipelines from sweet or
sour corrosion under water wet conditions or even in the presence of a water phase
at the bottom of the pipeline (BOL).
25.28 FLU ID FLO W HANDBOOK
In a pipeline at high gas flow rates, injected inhibitor liquids are dispersed as
a spray, and the droplets are carried along the line with the gas flow to produce a
protective film over the internal pipe surface even where the pipeline passes through
uneven terrain. However, in the case of stratified flow or stagnant conditions this
mechanism of inhibitor distribution is not possible [82]. Injected inhibitors do not
remain suspended but collect and travel along the BOL. In this situation the upper
part of the pipe is not sufficiently supplied with corrosion inhibitor and remains
susceptible to local corrosion [83]. This phenomenon is generally termed top-of-
the-line (TOL) corrosion (TLC). TLC increases with increasing water condensation
rates and is most difficult to control in pipelines laid in hilly terrain. Conditions
promoting TLC can be summarized as follows [84].
• Flow rate and flow regime: The extent of damage depends on the gas flow
rate, the water condensation rate and the length of pipe affected by condensa
tion. Water condensation may be localized where disruptions in flow lead to
conditions that exceed the dew point for the gas mixture. In more turbulent
flow regimes, a larger area of pipe surface is wetted leading to increased heat
transfer through the pipe wall from the fluids inside the pipe to the local
environment around the pipe. Improved heat loss generally increases the rate
for water condensation and consequent corrosion. In annular or slug flow
regimes, a corrosion inhibitor present in the liquid phase can inhibit damage;
however, stratified flow leaves the TOL exposed to corrosion.
• Temperature factor: High heat transfer rates through the pipe wall increase the
rate of water condensation where the water dew point for the gas being trans
ported is exceeded. Cooling of the pipe by river water, seawater or cold air out
side the pipe can promote water condensation and potential corrosion or cracking
problems on inside pipe surfaces. Unburied pipelines without concrete cas
ings or heat insulation or uninsulated pipelines passing through wet soil or
water are most susceptible. High gas temperatures along the first kilometers
of the line following a compressor station or just downstream of wellheads
for hot production zones favor corrosion problems in water wet systems.
• Gas content: Pipeline systems enforcing stringent specifications on incoming
gas quality do not experience significant erosion or corrosion problems in the
transport of clean sweet dry gas. Water if present provides the electrolyte
needed to sustain corrosion. TLC is also strongly dependent on the content of
corrosive agents such as C 0 2 or organic acids in the gas flow. Corrosion rates
up in excess of 1 mm/year have been reported. Contamination of the trans
ported gas by hydrogen sulfide can lead to special corrosion and cracking
problems especially in harder higher strength steels.
Prevention of erosion and corrosion damage begins with system design and mate
rials selection. Management of operating conditions and use of mitigative measures
CORROSION AND EROSION IN PIPES 25.29
might seem a simple matter to choose an alloy with a hard surface and eliminate
the erosion problem. This can be the case if the hard surface possesses good cor
rosion resistance and the hard material is easy to weld and to handle [92, 93).
The behavior and range of materials utilized to solve high-speed liquid droplet
impingement problems are similar to those chosen for resistance to cavitation attack.
Figure 25.11 provides an approximate ranking of materials to liquid droplet impinge
ment and cavitation attack [49]. Because erosion test data used to create this fig
ure were not very consistent this ranking should only be used as a rough guide.
Materials made from a wide range of polymers show good resistance to cav-
itation-erosion and excellent resistance to corrosion. For example, high density
polyethylene has a cavitation-erosion resistance similar to that of nickel-base and
titanium alloys [94]. Use of internal polymer coatings from materials such as
fusion bonded epoxy or liners made from polyethylene can protect low cost car
bon steel pipe from erosion and corrosion damage.
To control erosion-corrosion, the microstructure of a metal is very important.
It has been found for example that iron-chromium alloys are especially resistant
Ti alio vs
Stellite weld overlay (...)
Inconel wmmmmm (150-380)
pH st.St., 603, 631 ■— (320-460)
Cupro-nickel (70-200)
^ (...) Hardness HB or HV
Cast iron ■ " ■ (1 4 0 -2 3 0 ) v 7
I i -4------
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10 20 50 100
FIGURE 25.11 Relative erosion resistance of various metals and alloys according to ASTM G73 [491.
25.32 FLU ID FLO W HANDBOOK
to damage it the carbide in the matrix is finely distributed 195]. Flow-enhanced film
dissolution attack can be controlled by replacing carbon steel with low-alloy chro
mium > 0.1 percent, low alloy Cr-Mo, 304 stainless steel or in very severe con
ditions Inconel, or by duplex piping with a thin inner layer of stainless steel or
other high alloy [96].
SurfaceCoatings. Heat insulation should be even over the length of the pipe to
prevent preferential condensation of water with consequent TLC. This is espe
cially important for unburied piping. Joints should be considered as well as the
body of the pipe in the application of insulation.
Internal cladding of critical sections with corrosion resistant alloy can help
prevent TLC [88]. Impingement plates can be used to shield critical areas and suit
able internal coatings and liners can be used in aggressive conditions.
25.5 SUMMARY
Erosion and corrosion pose a significant threat to the safe and profitable opera
tion of many industrial systems. Understanding the mechanisms of damage can
25.34 FLU ID FLO W HANDBOOK
lead to the identification of design and operating options that will prevent or mini
mize the chance of failure. Prediction of the rate and location of damage will allow
targeted inspection and maintenance of system integrity.
REFERENCES
19. Gopal M., Kaul A. and Jepson W .P. 1995. Corrosion/95, Paper No. 105. Houston,
TX. N A CE International.
20. Tom ita Y. and Shina A J. 1986. Fluid Mech. Vol. 69. p 535.
21. Lotz, U. 1991. “V elocity effects in flow induced corrosion,” Proceedings o f Sym po
sium on F low -induced Corrosion; Fundam ental Studies and Industry Experience.
Kennelley. K.H.; H ausler, R.H. and Silverman, D.C. (Eds.). Houston, TX. NACE.
p 8 : 1- 8: 22 .
22. Kang, C., W ilkins, R., and Jepson, W.P. 1996. “The effect of slug frequency on
corrosion in high pressure inclined pipelines.” Corrosion/96. Denver, CO. NACE
International. M arch. Paper No. 96020.
23. Zhou, X. and Jepson, W .P. 1994. “Corrosion in three-phase oil/w ater/gas slug flow
in horizontal pipes.” C orrosion/94, LA. NACE International. March. Paper No.
94026.
24. Poulson B. 1987. P lant C orrosion, Predictions o f M aterials Performance. Strutt, J.
E. and Nickolls, J. (eds.). Ellis Horwood, Chichester.
25. Graig, B.D. 1991. Fundam ental A spects o f Corrosion Films in Corrosion Science;
Plenum Press, New York, p 167.
26. Ellison, B.T., W en, C.J. 1981. A lC hE Symposium Series. No. 204. Vol. 77.
27. Lotz, U., Heitz, E. 1983. W erkstoffe u. Korrosion. Vol. 34. p 454.
28. Sydberger, T. 1987. Br. Corros. J.; Vol. 22 (no. 2). p 83.
29. Syrett, B.C. 1976. Corrosion. Vol. 32. p 242.
30. Ellison, B.T. and Schm eal, W .R. 1978. J. Electrochem. Soc/, Vol. 125. p 524.
31. Bignold, C.J. et al. 1981. Proc. 8th Inter. C o n f on M etallic Corrosion. 1548.
Frankfurt am M ain Dechema.
32. Ferris, F.G., Jack, T.R ., and Bram hill, B.J. 1992. “Corrosion Products Associated
with Attached B acteria at an Oil Field W ater Injection Plant.” Canadian Journal o f
M icrobiology. Vol. 38. p 1320-1324.
33. Characklis, W .G. 1980. “ Biofilm developm ent and destruction.” Electric. Power
Res. Inst.; RP912-1, Palo Alto, CA.
34. M cCoy, W .F., Bryers, J.D ., Robbins, J., and Costerton, J.W .C. 1981. “Observations
o f Fouling Biofilm Form ation.” Canadian Journal o f M icrobiology; Vol. 27
(no. 9). p 910-917.
35. Efird K.D. 1977. Corrosion. Vol. 33. p 3.
36. Syrett, B.C. 1976. Corrosion, Vol. 32. p 242.
37. Coney M. 1980. CERL Report RD /L/N /97/80
38. Flint, E.B. and Suslick, K.S. 1991. “The tem perature o f cavitation.” Science, Vol.
253. p 1397-1399.
39. Preece, C.M . 1979. “C avitation-E rosion.” Treatise on M aterial Science and Tech
nology. Vol 16. London: Erosion A cadem ic Press.
40. Hammitt, F.G. 1980. C avitation and M ultiphase Flow Phenomena. M cGraw-Hill,
New York.
41. Arndt R.E.A. m \ . A n n u . Rev. Fluid M ech.; Vol. 13. p 273-327.
42. Plesset, M .S. and Prosperetti, A. 1977. Annu. Rev. Fluid M ech/, Vol. 9. p 145-185.
43. Morch, K.A. 1979. “D ynam ics o f cavitation bubbles and cavitation liquids” . Trea
tise on M aterials Science a n d Technology. Vol. 16. p 309-353. Erosion. Academic
Press, London.
44. Travena, D.H. 1984. J. Physics D; A pplied Physics. Vol. 17. p 2139-2164.
45. Lush, P.A. 1983../. F luid M ech/, Vol. 13. p 373-387.
25.36 FLU ID FLO W HANDBOOK
46. G rant, M. McD and Lush. P.A. 1987. J. F luid M e c h .; Vol. 176. p 237-252.
47. C.M . Preece. 1979. “Cavitation C orrosion.” Erosion, C. M. Preece, (Ed.). Acade
mic Press, p 249.
48. Lush P.A. 1987. Chartered M echanical Engineer. October.
49. H eym ann, F.J. 1992. “ Liquid Im pingem ent E ro sio n .” A SM M etals Handbook.
Vol. 18, Friction, W ear and Lubrication Technology. ASM M etals Park, OH.
p 221-232.
50. F.J. Heym ann. 1970. M achine Design. Dec. p 118.
51. C entury Brass. 1977. The Century' H eat Exchanger Tube M anual. Century Brass
Products, W aterbury, CT.
52. H eym ann, F.J. 1992. “Liquid Im pingem ent E rosion.” A SM H andbook, Vol 18. The
M aterials Inform ation Society, p 222.
53. Finnie, I. 1960. “Erosion o f surfaces by solid particles.” Wear. Vol. 3. p 87-103.
54. Levy, A.V. Solid Particle Erosion and Erosion-C orrosion o f M aterials. ASM Inter
national, M aterials Park, OH. p 195.
55. W asp, E.J. Kenny, J.P., and Gandhi, R.L. 1977. “Solid-liquid flow: Slurry pipeline
transportation.” Series on Bulk M aterials Handling. Vol. 1 (1975/77)
No. 4, Clausthal, G erm any. Trans. Tech. Publications, p 144.
56. Postlethw aite, J. and Nesic, S. 1993. Corrosion. Vol. 49. p 850.
57. M adsen, W. and Blickensderfer, R. 1987. “A new flow-through slurry erosion wear
test.” Slurry Erosion: Uses Applications and Test Methods. M iller, J.E.; Schmidt, F.E
(Eds.). A S T M S T P 946. W est Conshohocken, PA. ASTM . p 169-184
58. W ilson, G. 1972. “The design aspects o f centrifugal pum ps for abrasive slurries;”
Proceedings o f the 2nd International Conference on H ydraulic Transport o f Solids
in Pipes. Cranfield, U. K.: BHRA Fluid Engineering, p H2: 25-H2: 52
59. Evans, U.R. 1960. The Corrosion and Oxidation o f Metals; Edward Arnold. London.
60. Postlethw aite, J. and Nesic, S. 2000. “Erosion-C orrosion in Single and M ultiple
Phase Flow.” U hlig's Corrosion H andbook. 2nd edition. Revie, R. W inston; W iley.
John (ed.). Sons Inco. P 249.
61. Poulson, B.S. 1993. “Erosion Corrosion.” Corrosion. 3rd Edition. Shreir L.L.,
Jarm an, R.A., and Burstein, G.T. (Ed.), B utterw orth-H einem ann. p 1:295
62. Polan, Ned. W. 1992. “Corrosion o f C opper and C opper A lloys.” A SM H andbook.
Vol 13. The M aterials Inform ation Society, p 611.
63. Efird, K.D. 1977. Corrosion. Vol. 33. p 347.
64. Bailey, G.L. 1951. J. Jrn. Inst, o f M etals. Vol. 79. p 243.
65. Kato, C. et al. 1980. J. Electrochem. Soc. Vol. 127. p 1890.
66 . C astle, J.E. and Parvizi, M.S. 1986. Corros. Prev. Control. Feb.
67. G ilber, P.T. 1954. Trans. Institute o f M arine Engineers, p 66 .
68 . H ack, H.P. and Gudes, J.P. 1979. Corrosion/79, paper No. 234.
69. Eiselstein, L.E., et al. 1983. Corros. Sci. Vol. 23. p 223.
70. Little, B., W agner, P., and M ansfeld, F. 1991. Int. Met. Reviews. Vol. 36. p 253.
71. Duereux, J. 1983. Proc o f Specialists M eeting held at ‘Les R en a rd iers” M ay 1982
on Erosion Corrosion o f Steels in High Tem perature W ater and Wet Steam .
Berge, P. and Khan, E. (Eds). Electricite de France, Paris.
72. Penfold, D. et al. 1986. N uclear Energy. Vol. 25. p 257.
73. Bignold, G.J. et al. 1983. Proc. o f UK Corr. Vol. 83. Birm ingham : Inst. C orrosion
Sci. and Tech.
CORROSION AND EROSION IN PIPES 25.37
74. Heuer, J.K. and Stubbins, J.F. Corrosion. Vol. 54 (no. 7). p 566-575.
75. W aard, C. de, W illiam s, D.E. 1975. Corrosion. Vol. 31 (no. 5). p 177
76. Dugstad, A. 1992. “ Im portance o f F e C 0 3 supersaturation on the C 0 2 corrosion of
carbon steels;” Corrosion/92. Paper No. 14. Houston, TX. NACE.
77. Palacios, C.A. Shadley, J.R. 1991. Corrosion. Vol. 47 (no. 2). p 122.
78. Vuppu, A. 1994. Study o f Carbon D ioxide Corrosion o f Carbon Steel Pipes in M ul
tiphase Systems. M aster’s thesis. Ohio University.
79. Bhongale, S. 1996. E ffect o f Pressure, Temperature and Froude Num ber on C orro
sion Rates in H orizontal M ultiphase Slug Flow. M aster’s thesis. Ohio University.
80. G unaltun, Y.M. 1996. “Use o f laboratory and field data for corrosion rate predic
tion.” Corrosion/96. Paper No. 26.
81. Dougherty, J.A. 1998. “C ontrolling C 0 2 corrosion with inhibitors.” Corrosion/98.
Paper no. 15. H ouston, TX. NA CE International.
82. Bonis, M .R., Crolet, J. L. 1998. “An inhibitor policy based on laboratory and field
experience.” Corrosion/98. Paper No. 103. Houston, TX. NACE International.
83. Attwood, P.A. G elder, K. van, Charnley, C.D. 1996. “C O 2 corrosion in wet gas
system s.” Corrosion/96. Paper No. 32. Houston, TX. NACE International.
84. Sun, Y.H., Hong, T., Jepson, W .P. 2001. “Corrosion under wet gas conditions.”
N ACE/2001. Paper No. 01034.
85. M cIntyre, D. (Ed.). 1997. “Form s o f Corrosion, Recognition and Prevention.”
N A C E H andbook 1. H ouston, TX. Vol. 2.
86 . Postlethw aite, J., Tinker, E.B., and Haw rylak, M.W. 1974. Corrosion. Vol. 30.
p 285.
87. Schm itt, G., Scheepers, M., Siegm und, G. 2001. “ Inhibition o f the top-of-the-line
corrosion under stratified flow .” Corrosion/2001. Paper No. 01032.
88. Gunaltun, Y.M. and Larrey, D. 2000. “Correlation o f Cases o f Top o f Line Corrosion
with Calculated W ater Condensation Rates.” Corrosion/2000, paper No. 00071.
89. W hite, R.A. 1965. M aterials Protection. Vol. 4. p 48.
90. Hansson, C.M . and H ansson, I.L.H. 1992. “Cavitation Erosion.” A SM Handbook.
Vol 18. The M aterials Inform ation Society, p 218.
91. Shreir, L.L, Jarm an, R.A ., and Burstein, G.T. (Ed.). 1993. Corrosion. 3rd
Edition. Butterw orth-H einem ann.
92. Tischner, H., Pieger, B., Ziegler, H. 1989. W erkstoffe u. Korrosion. Vol. 40. p 575.
93. Kunze, E., Now ak, J. 1982. W erkstoffe u. Korrosion. Vol. 33. p 14 and 262.
94. Heym ann, F.J. 1992. “Liquid im pingem ent erosion.” A SM M etals Handbook. Vol.
18, Friction, W ear and Lubrication Technology. ASM M etals Park, OH. p 214-220.
95. Kohley, T., Heitz, E. 1989. M aterialw issenschaft u. W erkstofftechnik. Vol. 20. p 27.
96. 1996. EPR I (E lectric P ow er Research Institute) Flow A ccelerated Corrosion in
Pow er Plants. T R -106611. Pleasant Hill, CA. p 4:2, 6 : 25.
97. Bom berger, D.R. 1965. “ H exavalent chrom ium reduced corrosion in a coal-w ater
slurry pipeline.” M aterials Protection. Vol. 4, Jan. p 43-49.
98. Coale, R.D., T hom pson, T.L., and Ehrlich, R.P. 1976. Trans. SME, AIME,
Vol. 260, Dec. p 289.
99. Jennings, M.E. 1981. M ining Eng. Feb. p 178.
100. Ostroff, A.G. 1979. Introduction to O ilfield W ater Technology. Houston, TX.
NACE. p 293.
25.38 FLU ID FLO W HANDBOOK
101. Jones, D.G., Daw son, S J., Clyne, A.J. 1998. “ Reliability o f internally corroding
pipelines.” Corrosion/98. Houston, TX. NA CE International. Paper No. 80.
102. Sm ith, S.N ., Duvivier, J.P., Lefevre, D.E., Robb, G.A. 1996. “Field experience with
intelligent pigs.” Corrosion/96. H ouston/ TX. Paper No. 37.
103. G ipson, F.W . (Conoco Inc.). 1987. M ethod f o r distributing corrosion inhibitor with
fo a m ; Patent: GB 2185196, N L 8700011, 1987-07-15
104. M artin, R.L. 1997. “Inhibition o f vapor phase corrosion in gas pipelines.”
Corrosion/97. Houston, TX. NACE International. Paper No. 337.
105. A ndreev, N .N ., Kuznetsov, Yu. I. “ Volatile inhibitors for C 0 2 corrosion;” Corro
sion/98. H ouston, TX. NACE International. Paper No. 241.
106. Heeg, B., M oros, T., and Klenerm an, D. 1998. Corrosion Science. Vol. 40 (no. 8).
p 1304.
_________ CHAPTER 26_________
BLOOD FLOW DYNAMICS
26.1 INTRODUCTION
Biofluid mechanics is the field that studies the behavior of fluids in living organ
isms. The human body contains a number of fluids, each of which has a different
function. Blood is a fluid that demonstrates considerable motion and it is linked to
the deadliest disease in the U.S. and internationally, namely cardiovascular (CV)
disease. This chapter will focus on reviewing the main characteristics of blood
flow dynamics in the CV system.
The main function of the CV system is to provide oxygen and nutrients to the
cells and remove C 0 2 and wastes from them. The arteries and veins are not a pas
sive conduit system for blood transportation, but an active component of blood cir
culation. Blood flow mechanics play an important role in CV function. The nature
of blood flow patterns is often related to the cause and development of CV dis
ease. Knowledge of the way in which blood flows can provide invaluable diagnos
tic information about CV pathology.
This chapter will focus on the principles of blood flow. The same laws and prin
ciples that govern flow in simple tubes are valid for blood flow in the heart, arter
ies, and veins, although the conditions invivo are more complex. This is due to
interactions between many regulatory factors that affect CV function, such as the
pulsatility of blood flow, the elasticity and complex geometry of the arteries, the
peripheral resistance, and the non-Newtonian nature of blood. Therefore, discus
sion will begin by briefly reviewing the principles of simple steady flow in straight
tubes, and will evolve to describe more physiologic flow environments in the cir
culatory system. Blood flow characteristics in physiological geometries, such as
curved tubes and bifurcations, will be reviewed. Changes in blood flow patterns
26.1
26.2 FLU ID FLO W HANDBOOK
The heart consists of two sides: the left, responsible to pump oxygenated blood
to the body, and the right, responsible to pump deoxygenated blood to the lungs.
Each side consists of two chambers: a ventricle and an atrium. In a normal heart,
there is no communication between the left and right heart sides. Flow through the
heart is regulated by four one-way valves. Two of these valves, the mitral valve on
the left side and the tricuspid valve on the right side, are located between the atria
and the ventricles. The other two valves, the aortic valve on the left side and the
pulmonic valve in the right side, are located between the ventricles and the large
arteries leaving the heart [1].
The cardiac cycle consists of two major functional periods: diastole (ventricu
lar filling) and systole (ventricle contraction). Deoxygenated blood from the body
enters the right atrium through the venae cavae, and oxygenated blood from the
lungs enters into the left atrium through the pulmonary veins. During diastole, blood
from the right and left atria enters into the right and left ventricles through the tri
cuspid and the mitral valves, respectively. Then, during systole, the ventricles con
tract and blood is ejected from the right ventricle into the main pulmonary artery
through the pulmonic valve, and from the left ventricle into the aorta through the
BLOOD FLO W DYNAM ICS 26.3
aortic valve. In a normal resting heart, systole covers approximately one third of
the cardiac cycle, with diastole covering the remaining two thirds.
The frequency of ventricular contraction is the heart rate (HR) [beats/min]. The
amount of blood that is ejected from the left ventricle during one contraction is the
stroke volume (SV) [ml/beat|. Therefore, the cardiac output (CO) IL/min 1 is the
product of SV and HR.
P
Blood flows as a result of the pressure difference A [Pa] between the arter
ial and venous parts of the circulation. This pressure difference is related to the
Q
blood flow rate [m 3/s] and the peripheral resistance Rs
[Pa • s/m3] through Equa
tion 26.1:
A P=QR, (26.1)
This model was an early attempt to explain the pressure and flow behavior of blood
in the arterial system (see Figure 26.1). The compliant vessel represents the aorta
with its elasticity. This elasticity is responsible for storing some of the blood (ejected
from the left ventricle) during systole and releasing it to circulation during dias
tole. The continuity equation in this model is:
—
dt =Qin• K(t)-Q
' ^ o uM
t K)
} v(26.2)7
clVdP_ 1 dP
dP dt ~E' dt
where Ef is the modulus of volume elasticity. The right side of Equation 26.2 can
be written as (see Equation 26.1):
Qiny( 0'- —r
Then, Equation 26.2 becomes:
26.4 FLU ID FLO W HANDBOOK
(26.4)
Blood flow is unsteady, and, actually, in most regions pulsatile. To clarify, the term
“unsteady” is used to characterize any type of non-steady flow. If the unsteady flow
is periodic and oscillates back and forth without a net forward or reverse output it
is calledoscillatory flow, whereas if it has a net output throughout the cycle it is
called pulsatile.Although blood flow in the heart and large vessels is pulsatile and
three-directional, the principles of fluid flow under simplified conditions will be
reviewed first, before moving to more physiologic situations.
The simplest case is steady flow of a Newtonian fluid through a straight, rigid, cylin
drical tube. Detailed description and mathematical derivations for steady flow in a
straight rigid tube is given earlier in this text. This section provides only a brief review.
The nature of flow of a Newtonian fluid in a straight rigid tube is controlled by
the viscous and the inertial forces applied to the fluid elements. When viscous forces
Peripheral
Vessels
Compliant
Vessel
FIGURE 26.1 The Windkessel model.
BLOOD FLO W DYNAM ICS 26.5
dominate, the flow is laminar. When inertial forces dominate, the flow is turbulent.
Laminar flow is well organized and smooth, whereas turbulent tlow is disturbed
and can be accompanied by considerable energy losses. Therefore, turbulent flow
is undesirable in blood circulation.
For an incompressible laminar flow, the motion of a fluid is described by the
Navier-Stokes equations [2-4]:
3
—
V+(V■V)V = — 1 VP+g+v(V~
7
V) (26.5)
dt p
where
By considering a control volume, Equation 26.5 shows the balance between the
accumulation of momentum, the transport of momentum by convection (bulk flow),
the generation of momentum by a pressure difference and by gravity, and the trans
port of momentum with molecular (viscous) mechanisms. Using Equation 26.5, the
velocity profile can be derived as [2-4]:
it - umax (26.6)
where umax R
[m/s] is the maximum (centerline) velocity, and [m] is the radius of
the tube. Equation 26.6 shows the main characteristic of the fully developed lam
inar flow in a tube, that is the parabolic shape of the velocity profile.
Among the variables involved in the study of steady (but also unsteady) flow
is the shear stress, r [N/m2]. This parameter has great clinical relevance, as it pro
vides information about the magnitude of the force that the fluid exerts on the
vessel walls (wall shear stress) or the force from one fluid layer to an adjacent layer.
High levels of shear stress (caused, for example, by atherosclerotic lesions, artificial
heart valves, or similar flow-obstructing structures), as well as strongly oscillatory
shear stress can cause vascular wall damage or red blood cell damage (hemolysis),
and change the biological behavior of platelets and endothelial cells.
For laminar flow, the shear stress is linearly related to the shear rate(du/dr) by
Newton’s law of viscosity:
T=-V—
dr (26.7)
26.6 FLU ID FLO W HANDBOOK
/u
where r [N/m2] is the shear stress and |kg/m s] is the dynamic viscosity of the
fluid. From the force balance in a control volume, the wall shear stress tw
can be
P L
determined if the pressure drop A is known along a tube of length [m] and inside
d
diameter [m], by the following equation:
tw=-l^L (26.8)
4 L
The velocity profile of turbulent flow is flatter than that of laminar flow, but
the flow is not plug (see Figure 26.2). The combination of a higher inertia in the flow
(flatter profile) with the non-slip conditions at the wall results in higher wall shear
stresses for turbulent flow as compared to those for laminar flow.
It should be noted that normal flow in the human circulatory system is generally
laminar. Although direct measurements have shown that flow in the ascending aorta
has a tendency to destabilize during the deceleration phase in late systole, there is
not enough time for flow to become turbulent. Blood flow becomes turbulent down
stream of vascular stenosis or distal to defective heart valves. Therefore, the detec
tion of turbulence in blood flow is a diagnostic indication of the presence of disease.
The above considerations are valid only for fully-developed flow in a straight
smooth-surfaced tube, distal enough from any upstream disturbances caused by cur
vatures, bifurcations, anastomoses, aneurysms, occlusions, etc. The effects of flow
disturbances are significant in a region located immediately downstream of a distur
bance (flow development region). A description of a flow development region can
be seen in Figure 26.3, which shows the flow field downstream of a sudden contrac
tion. Immediately after the contraction the flow profile is essentially flat. Then, the
layer of fluid close to the wall decelerates and slows other adjacent fluid layers
through viscous forces. At the same time, the fluid layers near the center of the tube
accelerate to satisfy the mass conservation principle. The region of flow in which
viscous forces affect the velocity profile is the boundary layer. In a laminar bound
ary layer, viscous effects dominate, while inertial forces are insignificant and can
be neglected [5].
• ••• Laminar
— Turbulent
F I G U R E 2 6 .2 L a m in a r (d o tte d
line) and tu rb u len t (so lid line) flow
velocity p ro files in a straight circu lar
tube.
BLOOD FLO W DYNAM ICS 26.7
Free
Stream
The length of the flow development region (entrance length), can be estimated
as [2 ]:
So, for flow in the aorta and under a hypothesized steady flow case, the entrance
length should be in the order of 3 meters. In other words, in humans, the entire length
of the aorta is the entrance length. Therefore, using these traditional fluid mechan
ics rules, the arterial flow should not be fully developed.
Assuming one-directional inviscid flow, the continuity and the ^-component momen
tum balance equations become respectively:
dA+ ——
— d{uA)= —dA+ w—dA+A.—
du= 0_ (26.11)
dt dx dt dx dx «
du du dP
p ¥ +p" a T - a 7 (26,2)
where Aa P Pn.
is the area at = If, initially, the tube contained stationary fluid, and
u
flow has just started, the velocity and the change (deformation) of the area (dA/dx)
can be assumed to be small. Then, Equation 26.11 becomes:
M.+A—=0 (26.14)
dt dx
du^ \ dP
^ +! ^ =0 (26.15)
dt p dx
Also, from Equation 26.13:
^
dt = dt (26.16,
A— + A ^ = 0 (26.17)
dt dx
By taking the derivative of both sides of Equation 26.20 with respect to time /,
we have:
dx{dl ) p dx^
Combining Equations 26.18 and 26.19 leads to the waveequation:
A ^ _ ! ^ =0 (26.20)
Adt2 Pdx1
Since blood flow in the heart and arteries is pulsatile, the pressure and the veloc
ity profiles vary periodically with time. When the heart contracts (systole), a pres
sure wave starts from the left ventricle and travels down the arterial tree. When
BLOOD FLO W DYNAM ICS 26.9
the pressure in the ventricle becomes larger than the pressure in the aorta, blood
is ejected out from the heart to the arteries. Due to the compliant nature of the arter
ies, the pressure wave travels faster than the How wave. The pulsatility in the flow
affects the velocity profile and the pressure distribution in the vessel. It also affects
the transition of flow from the laminar to the turbulent regime. In the first part of sys
tole, flow is accelerating, whereas in the second part of systole flow decelerates.
If we use the Reynolds number to characterize the flow in a vessel like the ascend
ing aorta, the instantaneous Re during systole well exceeds the critical point for
the transition from laminar to turbulent flow. Nevertheless, in the first part of sys
tole, the flow remains laminar and well-streamlined, because acceleration tends to
stabilize flow. But even during the decelerating part of systole, flow remains fairly
laminar, although experiments show a tendency for destabilization of flow and tran
sition towards turbulence, mainly because deceleration destabilizes the flow due to
the developed adverse pressure gradient. If there were enough time, flow would
become turbulent. However, in a healthy vessel, turbulence is generally absent, as
the flow destabilization period is too short for turbulence to develop. While flow
decelerates and tends to become turbulent, the velocity drops in diastole, and the Re
becomes low. Then, a new acceleration phase in the next systole further restabilizes
the flow. It should be pointed out that although complex flows are present in regions
like bifurcations or anastomoses, a complex vortex-like flow is not necessarily a
turbulent flow, in the way turbulence is defined in fluid mechanics.
Clearly,Re is not enough to characterize pulsatile flow. To fill this void, another
dimensionless parameter, the Womersley number a, is used to characterize the
periodic nature of blood flow. The definition of is: a
(26.21)
d
where [m] is the inside diameter of the vessel, p [kg/m3] is the density of blood, w
ju
is the frequency of pulsation (heart rate), and [kg/m s] is the dynamic viscosity
of blood. The Womersley number in unsteady flow has a similar role to the Re
in
steady flow, providing a comparison between unsteady inertial forces and viscous
forces. In the human circulatory system, a ranges from 1(T3 in capillaries to 18.0
a
in the ascending aorta at rest. For < 1.0, the viscous forces dominate in every
region in the tube (quasi-steady a
flow). As increases, the inertial forces become
significant and start to dominate initially in the center of the tube. As a result, a
delay— with respect to the driving pressure gradient— can be observed in the bulk
flow, and the velocity profile becomes flat in the central region of the tube [6 , 7].
Solution of the physiologic pulsatile flow problem would provide the relation
between velocity, pressure, and wall motion/deformation. The Stokes’ problems of
pulsatile flow provide some insight for the mathematical formulation of the case
of pulsatile blood flow in arteries. The first problem involves a sudden accelera
tion of an infinite flat plate, and applies only to early systolic flow. The second
problem involves a sinusoidal oscillation of a semi-infinite plate, and is closer to
26.10 FLU ID FLO W HANDBOOK
(but still distant from) the physiologic situation. By considering pulsatile flow in
an elastic tube, we need to consider five main equations: the continuity equation, two
equations for the axial and radial m omentum balances, and tw o equations for the
axial and radial wall m otion. Sim ultaneous solution o f these equations would pro
vide com plete inform ation about the relation between blood pressure, velocity,
and vessel wall motion.
Solving the N avier-Stokes equations (along with the equations for wall motion) is
a difficult task— if at all possible. To find an adequate solution, assum ptions are
m ade to sim plify the N avier-Stokes equations. If, for exam ple, flow is assum ed to
be inviscid (viscous forces negligible), the Navier-Stokes equation is sim plified to
E uler’s equation o f motion (Equation 26.5, without the last right-hand side viscous
term ). To further simplify the situation o f motion, we could focus on flow along a
stream line. Integration o f this equation leads to the Bernoulli equation, which is a
statem ent o f the conservation o f the mechanical energy betw een tw o points along
a stream line:
(26.22)
w here P, f Pa], V, [m /s], and Z\ [m] are the pressure, velocity and height at the up
stream location, and P2, V2, and z2 are the pressure, velocity and height at the dow n
stream location. The integral term accounts for the flow acceleration or deceleration
betw een the tw o locations. If this equation is applied at peak systole, this term be
com es zero, as the derivative becom es zero at the maximum point on the waveform .
The Bernoulli equation states that the mechanical energy at any point rem ains
constant. So, betw een any tw o points, the form o f m echanical energy may change
(from kinetic to potential and vice versa), but the total m echanical energy does not
change. The Bernoulli equation also states that the pressure drop betw een point l
and point 2 located along a streamline is a function of the velocity at these tw o points,
assum ing that these tw o points are at sim ilar heights (z\ = £2) and at peak systole
W hen the dow nstream velocity is much greater than the upstream velocity (V2 »
V\)> ^i2 can be neglected. Equation 26.23 can then be sim plified further to:
p, - p 2 = 4 V i (26.24)
BLOOD FLO W D YNAM ICS 26.11
where P, is in mm Hg, V2 in m/s, and the blood density is assum ed to be 1.07 g/cm 3
which is incorporated into the “constant” 4. Equation 26.24 is called the simplified
Bernoulli equation, and it has found wide application clinically in determining pres
sure drops (P| - P2) across severe stenosis from non-invasive velocity (V2) m ea
surem ents with D oppler ultrasound and M RPV M (Figure 26.4).
The Bernoulli equation is not valid in cases in which viscous forces are signif
icant, such as in long constrictions with a small diam eter, or in flow separation
regions. In addition, it does not account for turbulent energy losses in cases o f
severe stenosis. Such losses should be taken into account, because they reduce the
energy content o f the fluid.
The geom etry o f a vessel greatly affects hem odynam ics. C onsidering the co m
plexity o f the circulatory system , vessel geom etry becom es a m ajor determ inant
o f blood flow patterns and wall shear stress distribution [8, 9]. This section will
describe the effect o f curvature and branching on the flow patterns. Due to the
difficulty o f assessing, in detail, the fluid m echanics o f blood in vivo, m ost o f our
know ledge is based on in vitro experim ental and com putational sim ulations, per
form ed under a variety o f conditions ranging from very sim plistic to alm ost phys
iological. W ith the advancem ent o f clinical im aging techniques, such as M RPV M
and color D oppler ultrasound, there have been a num ber o f anim al and human stud
ies assessing blood flow patterns.
26.5.1 Curvature
In contrast to the straight tube case, in w hich flow is one- o r tw o-directional, cur
vature causes a three-directional flow field. It has been show n that curvature sta
bilizes the flow, that is, flow becom es turbulent at higher Re com pared to that in
1 2
Pu V1 P2, V2
F I G U R E 26.4 Application of the Ber
noulli equation. T h e pressure drop
between points I and 2 is calculated
from velocity measurements at point
2 (assuming that it is m u c h greater
than the velocity at point 1).
26.12 FLU ID FLO W HANDBOOK
the case o f a straight tube [ 10]. The effect o f the curvature on the How is described
by the D ean num ber (De) [ l l ] which includes the effects o f both curvature and Re
on the flow profile:
(26.25)
w here R [m] is the cross-sectional radius and rc lm ] is the radius o f the curvature.
If the inflow velocity profile is parabolic (fully-developed lam inar flow), as in
Figure 26.5a, curvature results in skewing the profile towards the outer wall, because
o f the centrifugal force (due to inertia tending to move the fluid along a straight line).
B ecause o f this centrifugal force, a radial pressure gradient is developed from the
inner wall to the outer wall o f the vessel, resulting in a secondary— to the main axial
flow — radial flow [6, 7].
If the inflow velocity profile is uniform (non-fully developed flow), as in Fig
ure 26.5b, curvature initially results in skew ing the profile tow ards the inner wall,
because the centripetal force (F = m V 2/rc, where F [N] is the centripetal force, m [kg]
is the m ass, V [m/s] is the velocity, and rc [m] is the radius o f curvature) is greater
on the side o f the wall with the sm aller radius o f curvature (inner side). The curva
ture has a greater effect on the inner side o f the flow field than on the outer side,
and the flow is skewed tow ards the inner side. Then, as the flow develops, the pro
file becom es skew ed tow ards the outer wall as described in the previous paragraph
[6, 7]. A secondary radial flow is developed, and com bination o f this secondary
radial flow with the main axial flow results in the form ation of tw o counter-rotat-
ing helices, as seen in Figure 26.5b.
26.5.2 Branching
In the presence o f branching (Figure 26.6), the flow in the parent vessel has to
split into the tw o branches. Since the velocity has to be zero at the wall o f the
inner side o f the branches (non-slip condition), there is a high level o f shear stress
on the inner side o f the wall (profiles skew ed tow ards the inner wall). At the flow
divider, there is a high-pressure stagnation point. As How enters from the parent
vessel to the branches, it also has to follow the curvature due to the angle betw een
the branches. As a result, and in accordance to the descriptions in the previous sec
tion, there will be secondary flows developed in the branches.
In pulsatile flow, the flow patterns vary depending on the level o f inertial forces.
At peak systole, the flow is well-streamlined, without regions o f flow separation and
recirculation. H ow ever, in diastole, during which flow generally decelerates, flow
recirculation can be observed at the location where the branches start. The wall
shear stress will be higher at the inner w alls near the flow divider and low er at the
outer w alls, w here the separation zones are present.
The angle 0 between the parent vessel and the branches (Figure 26.6) and the ratio
o f cross-sectional areas o f the branches to the parent vessel are important parameters
for the nature o f the developed How patterns. A decrease in the angle betw een the
parent vessel and one o f the branches causes more intense flow separation and recir
culation at the outer wall of that branch and at the outer wall o f the other branch, while
the velocity profiles in the branches becom e m ore skewed. The ratio o f cross-sec -
tional areas of the branches to the parent vessel affects the stability o f the flow. W hen
the ratio is less than one, flow accelerates, and thus becomes more stabilized, whereas
w hen the ratio is greater than one, the flow decelerates (destabilization) [12 - 2 1].
In addition, the w all elasticity, in com bination with flow pulsatility, has m ajor
effects on blood flow and pressure distribution. This can be appreciated in patho
logic situations, such as hypertension. V essel geom etry and wall elasticity change
as a result o f aging and pathology (atherosclerosis, aneurysm s, etc.). Finally, co n
sidering the non-N ew tonian characteristics o f blood is in some cases very im por
tant to correctly study and understand hem odynam ics [22,23].
26.14 FLU ID FLO W HANDBOOK
The aortic valve is a tri-leaflet valve located betw een the left ventricle and the aorta.
The diam eter o f the aortic “annulus” is 2 -2 .5 cm. Imm ediately distal to the valve is
the aortic root w ith its three sinuses o f V alsalva. T w o o f these sinuses are the o ri
gins o f the coronary arteries that supply blood to the m yocardium .
W ith the onset o f systole, the left ventricle contracts and thus the left ventricu
lar pressure becom es higher than that in the ascending aorta. T his pressure differ
ence between the proximal and the distal sides o f the valve forces the leaflets to open
and blood accelerates through them. At this stage, the pressure in the annular side
o f the leaflets is higher than that inside the sinuses, causing blood to circulate inside
the sinuses and form vortices (Figure 26.7). These vortices probably contribute to
the valve closure, w hich starts during end-systole [24, 25], w hile an adverse pres
sure difference is developed in the aortic root forcing a flow deceleration through
the valve. The pressure drops in the annular side o f the aortic root. The higher pres
sure in the sinuses starts pushing the leaflets to close.
D uring diastole, the pressure difference betw een the ascending aorta and the left
ventricle causes the leaflets to com pletely block the valve “annulus,” preventing
backflow o f blood into the left ventricle.
The advancem ent o f Doppler ultrasound and, especially, M RPV M has enabled reli
able characterization and quantification o f blood flow in the aorta clinically. In early-
systole, the flow velocity profile in the ascending aorta is alm ost flat, with some
skew ing along the posterior left wall (higher velocities along the inner curvature).
In m id-systole, the high velocities m ove tow ards the outer side o f the c u rv a
ture. A clockw ise-m oving helical flow is developed in the ascending aortic arch,
w hereas an anti-clockw ise rotation is developed in the descending arch. T here is
flow separation in the inner wall o f the descending arch.
Valsalva
In late-systole, there is retrograde flow along the inner wall o f curvature (Fig
ure 26.5). Paired, counter-rotating helices are often developed in the beginning o f
the descending aorta, with How separation along the inner wall o f the curvature.
Eventually, the helix close to the inner wall dom inates, resulting in anti-clockw ise
rotation in the descending aorta. This transition, from a clockw ise helical flow , in
the ascending arch, to an anti-clockw ise helical flow in the descending arch, starts
in m id-systole and is clearly developed at end-systole [25].
In diastole, both forw ard flow (anterior right wall) and reverse flow (posterior
left w all) can be observed in the ascending aorta. This reverse flow, w hich starts in
m id-systole, is probably necessary to assist coronary inflow circulation. In general,
arteries that supply blood to m uscles show net retrograde flow during diastole. On
the other hand, arteries that supply blood to internal organs do not show retro
grade flow [26). T urbulence is essentially absent during the entire cycle [25, 27].
Flow in this case is sim ilar to that in a bifurcation (described in Section 26.5.2). W e
review this case, how ever, because o f the presence o f a sinus in the internal carotid
artery, which affects blood flow patterns, causing flow separation and recirculation.
As seen in Figure 26.8, the sinus has a divergent and a convergent part. If we assume
one-directional inviscid flow along the axial direction o f the internal carotid artery,
the N avier-Stokes equation is sim plified to:
du du l dP
— + --------- — (26.26)
dt dx p dx
The volum etric flow rate is Q = u(x)A(x), where A(x) is the cross-sectional area,
w hich is a function o f x. A ssum ing rigid vessel walls, A is not a function o f t. Then:
dQ d(uA) du
JT = - \ - 1 = A- r (26.27)
dt dt dt
and
du d (Q ! A) d ( \ ! A) Q dA
i j » 9 j \2.0.Zo)
dx dx dx dx
Flow separation takes places under the presence of an adverse pressure gradient
(dP/dx > 0). From Equation 26.29, accelerating flow (dQldt > 0) in the converging
part o f the sinus (dAldx < 0) results in absence o f flow separation (dPfdx < 0). How
ever, deceleration o f flow (dQldt < 0) in the diverging part o f the sinus (dAldx > 0)
results in possible flow separation (dP/dx > 0). In the case o f acceleration o f flow
(dQldt > 0) in the diverging part o f the sinus (dAldx > 0), the relative m agnitude
o f dQ ldt and dAldx determ ine w hether or not flow separation takes place.
M easurements have demonstrated that the velocity profile in the common carotid
artery is parabolic-like during the entire cardiac cycle [28]. A lthough the entrance
length in a straight tube for a fully-developed parabolic velocity profile is approxi
mately equal to 100 tim es the tube diameter, in reality even in 30 to 40 times the tube
diam eter the profile is parabolic-like. A ctually, after 10 tim es the tube diam eter, we
m ay have 90 percent parabolic profile (depending on the W om ersley num ber). So,
there is enough length for a parabolic-like profile to be present in the com m on
carotid artery.
High shear stresses are observed tow ards the flow divider and low shear
stresses are present at the outer wall. No flow separation is observed in the exter
nal carotid artery. Secondary flows develop dow nstream the sinus in the internal
carotid artery [14, 20, 22, 29-31].
26.6.4 Stenosis
In arterial stenosis, the cross-sectional area o f the vessel is reduced locally (Fig
ure 26.9) due to the developm ent o f vascular disease. This occlusion results in severe
flow changes, including the developm ent o f turbulence dow nstream of the steno
sis, which have been used in diagnostic im aging to assess the presence and sever
ity o f the stenosis [32].
BLOOD FLO W DYNAM ICS 26.17
upstream post-stenotic
stenosis
F IG U R E 26.9 The effect of stenosis on the velocity profile.
In the converging part o f the stenosis, the average velocity across the artery
increases. The velocity profile becom es flat at the point o f sm allest diam eter
(throat). This leads to a pressure decrease, according to the Bernoulli equation
[Equation 26.23].
In the diverging (downstream ) part o f the stenosis (immediately after the throat),
the diam eter increases again, the average velocity decreases and the pressure in
creases. Because o f the adverse pressure gradient causing this flow deceleration,
the flow field destabilizes and flow separation may develop. This, in com bination
w ith turbulence, causes energy loss. Therefore, the pressure is not fully recovered
dow nstream o f the stenosis (see graph in Figure 26.10).
The flow field dow nstream to the stenosis is characterized by the presence o f
je ts, flow separation zones, and turbulence [33]. Blood exits the stenosis in a je t
form , which separates from the vessel wall, leaving a flow recirculation region
betw een the vessel wall and the je t (see Figure 26.9). In cases o f mild or m oder
ate stenoses, the spatial and tem poral characteristics o f the post-stenotic flow dis
turbances (such as vortices, which result from the shear layer between the je t and
flow recirculation zone [34]) can be determ ined during the cardiac cycle. In cases
A
pressure
pressure
loss
stenosis
F IG U R E 26.10 Pressure changes in a stenosis.The pressure decreases, as flow
accelerates through the stenosis, and becomes least at the vena contracta. Recovery
is not complete due to energy dissipation.
26.18 FLU ID FLO W HANDBOOK
of severe stenoses, the dominance o f inertia over viscous forces results in turbulence.
Although the flow becomes lam inar farther dow nstream , there is alw ays an irrevers
ible energy loss (due to viscous forces and turbulence).
Clinically, the severity o f a stenosis is determ ined by the reduction in the cross-
sectional area o f the artery. Although hem odynam ic changes are observed even for
less than 50 percent area reduction, it has been clinically accepted that blood flow
through a stenosed artery is not significantly disturbed for cross-sectional area reduc
tions o f less than 75 percent, w hich is the critical degree o f stenosis beyond which
small changes in the cross-sectional area o f the vessel cause significant post-stenotic
pressure and flow changes. The relationship between the post-stenotic flow field and
the severity o f stenosis is com plicated, because the post-stenotic flow field changes
throughout the cardiac cycle and is non-linearly related to the stenosis severity. Sev
eral geom etric (axisym m etric versus asym m etric stenosis, m ultiple stenoses, etc.)
135-40], hem odynam ic (lam inar versus turbulent upstream flow, flow pulsatility,
etc.) [41,42], and biological (tissue perfusion, vascular resistance) [42] factors can
predict the developm ent o f post-stenotic jets, flow recirculation zones, vortices,
and turbulence.
26.6.5 Atherosclerosis
A_
A. + A
+ —
w here \A J and \A+\ are the areas o f the negative and positive, respectively, part
o f the “wall shear stress vs. tim e” curve. The greater the OSI at the wall, the more
vulnerable that region is for plaque developm ent.
26.6.6 Aneurysm
A neurysm s are local expansions o f the vessel that affect blood flow m echanics.
The forces exerted on the wall are altered, and this may result in rupture of the
BLOOD FLO W DYNAM ICS 26.19
aneurysm , leading to death. The main flow features are flow separation and recircu
lation inside the aneurysm . A num ber o f numerical and in vitro studies have shown
the form ation o f flow recirculation and vortices, with their intensity depending on
the intensity o f the upstream flow that varies throughout the cardiac cycle and on the
size o f the aneurysm. The form ed vortices inside the aneurysm seem to shift towards
the distal end o f the aneurysm (Figure 2 6 .11), although their location changes dur
ing the cardiac cycle [47,48]. The highest shear and pressures occur at the distal end
o t the aneurysm [4 7 -4 9 ]. T urbulence breaks the vortex and causes an increase in
the shear stress.
Placem ent o f a stent in the aneurysm site has shown promise to prevent rupture,
because it alters the flow patterns inside the aneurysm cavity [50, 51]. The flow
activity in the aneurism al cavity and the pressure at the distal end o f the aneurysm
are significantly decreased after stenting [51]. An overall reduction o f shear stress in
the distal end region (by 90 percent) has also been observed, although local increases
have been observed. Even in the latter case, however, shear stresses appear to be uni
directional in contrast to the non-stented case [50].
V eins have thinner w alls than arteries and the pressure inside a vein is low er than
the pressure o f an artery at the same level. To assist in blood circulation, some veins
have valves to prevent backflow o f blood.
O ne o f the prim ary factors that affect the function o f veins is the difference
betw een the internal and the external pressure (transm ural pressure, P, - P()). W hen
the transm ural pressure is positive (region 1 in Figure 26.12), the cross-sectional
area is circular and flow is sim ilar to that in an artery (subcritical flow), although
the flow w aveform is different from that in an artery. If the velocity inside the
vein increases to the point at which the transm ural pressure becom es negative
(supercritical flow ), the greater external pressure will force the wall o f the vein to
contract (region II) and the vein collapses (region III). The result o f the oscilla
tory transition betw een subcritical flow , supercritical flow, and wall collapse is a
pulsation in the flow caused by the pressure difference betw een an upstream non
collapsed part o f the vein and the collapsed part o f it [6,7].
V enous flow is m ore com plex than arterial flow (w hich is mostly determ ined
by le ft ventricular function and vascular resistance), because it can be affected by
Transm ural
Pressure
(PrPo)
Area (A)
a num ber o f factors, such as the arterial pulse, the right heart function, the respira
tory m otion, and the function o f m uscles located next to the vein [52].
T o study and understand hem odynam ics, the param eters that characterize blood
flow m ust be measured. In vitro, there are several techniques to measure fluid veloc
ity and pressure (DP1V, D oppler ultrasound, LD V , pressure transducers, etc.).
H ow ever, these techniques are neither proper nor desirable in vivo. Clinically, there
is need for non-invasive, accurate techniques w hich will provide m easurem ents
w ithout risks to harm the subject. Currently, the main non-invasive techniques for
blood flow velocity m easurem ents are D oppler ultrasound and M R PV M .
D oppler ultrasound techniques can be applied com pletely non-invasively (trans-
thoracically), o r partially invasively (transesophageally and intravascularly). The
m ost com m only used Doppler modality is the color Doppler technique [53], although
continuous-w ave and the traditional pulsed-w ave D oppler techniques are still use
ful in particular applications. The technique is based on the fact that, w hen an ultra
sonic beam is scattered on a m oving particle, the transm itted and scattered waves
have different frequencies. This frequency difference is the D oppler shift, which is
proportional to the velocity o f the particle. The relationship betw een the velocity
and the D oppler shift is given Equation 26.34:
v = V _ c _
(26.31)
2/„ cos e
BLOOD FLO W DYNAM ICS 26.21
where
W ith color D oppler, the velocity o f blood cells (the scattering particles o f the ultra
sonic beam ) is encoded into color, which is displayed on the screen superim posed
on the traditional tw o-dim ensional echocardiographic image. C olor D oppler ultra
sound is extensively used clinically to m easure blood velocity through heart valves
(for diagnosis o f valvular stenosis and regurgitation) [5 4 -5 8 ], and the velocity in
the aorta, large vessels, and peripheral arteries (carotid, iliac, etc.) [5 9 -6 2 ].
A m ore recently developed velocim etric technique is M RPVM [63]. It is based
on the principles o f traditional magnetic resonance im aging (M RI), but in this case
both the m agnitude and the phase o f the detected signal are used to reconstruct two
im ages: a m agnitude im age with anatom ical inform ation and a phase im age with
velocity inform ation. M RPV M is based on the fact that by using proper m agnetic
field gradients during data acquisition, the phase o f the signal and the velocity o f
blood are linearly related as shown in Equation 26.32:
The velocity can be m easured in all three spatial directions in an im aging slice, and
quantitative results about the blood velocity, flow rate, and flow volume are available
through im age data analysis. In vitro M RPV M studies [6 4 -6 7 ] have show n high
accuracy and precision (errors o f less than 10%). The technique has been success
fully validated in hum ans, and is currently being used clinically for quantification
o f blood How inside the heart, through heart valves, in the aorta and other large
and peripheral arteries, and in the coronary arteries [2 5 -2 7 , 6 8 -7 3 ].
26.8 CONCLUSION
In this chapter, the basic principles o f blood flow dynam ics w ere discussed in
order to provide a basic understanding o f how blood flows in the vascular system .
26.22 FLU ID FLO W HANDBOOK
W e started w ith the traditional W indkessel m odel, and continued with flow in
arteries and veins, review ing the effects o f elasticity, flow pulsatility, and vessel
geom etry on fluid dynam ics. W e review ed a num ber o f com m on cases o f disease
that affect blood flow patterns. Finally, the principles o f the m ost m odern clinical
im aging techniques for the assessm ent o f hem odynam ics w ere described.
The inform ation in this chapter was provided in general term s. M ore specific
inform ation can be found in the references provided and the reader is urged to
review this literature.
REFERENCES
1. Alexander R.W., Schlant R.C., Fuster V., O ’ R o u r k e R.A., Roberts R., Sonnenblick
E.H. 1998. H u r s t's The H eart. 9th Edition. McGr aw - Hi ll , N e w York.
2. W h i t e F.M. 1999. F lu id M echanics. 4th Edition. M c Gr aw -H i ll , N e w York.
3. Bird R.B., Stewart W.E., Lightfoot E.N. 1960. T ransport P henom ena. Wiley, N e w
York.
4. Streeter V.L., W y l i e E.B., Bedford K . W . 1998. F lu id M echanics. 9th Edition.
Mc Gr a w - H i l l , N e w York.
5. Schlichting H. 1978. B o u n d a ry -L a y e r Theory. McGraw -H il l, N e w York.
6. C a r o C.G., Pedley Ti, Schroter R.C., Seed, W A . 1978. The M e ch a n ics o f th e C irc u
lation. O x f o r d University Press, Oxford.
7. Fung, Y.C. 1984. B io d yn a m ic s — C irculation. Springer-Verlag, N e w York.
8. F r i e dm a n M.U., Deters O.J., M a r k F.F., Bargeron C.B., Hutchins G . M . 1983. “Arte
rial g eo me t r y affects h e m o d y n a m i c s — potential risk factor for atherosclerosis.”A th
ero sclerosis. 46: 2 2 5 - 23 1.
9. Bargeron B.C., Hutchings G .M., M o o r e G . W . , Deters O.J., M a r k F.F., F ri ed m a n
M . H . 1986. “Distribution of geometric parameters of h u m a n aortic bifurcations.”
A rterio sclero sis 6: 109-113.
10. M c D o n a l d D. A. 1974. B lo o d F lo w in A rteries. Arnold, London.
11. D e a n W . P . 1927/1928. “T h e streamline motion of fluid in a curved pipe.”P h il M ag.
4: 2 0 8 a n d 5: 673.
12. Feuerstein I.A., El M a s r y O.A., R o u n d G.F. 1976. “Arterial bifurcation H o w s —
effects of flow rate a n d area ratio.”C an J P h ysio l P harm acol. 54: 795-807.
13. Siouffi M., Pelissier R., Farahifar D., Rieu R. 1984. “T h e effect of unsteadiness o n
the flow through stenoses and bifurcations.”J B iom ech. 17: 299-315.
14. K u D.N., G i d d e n s R.P. 1983. “Pulsatile flow in a m o d e l carotid bifurcation.”
A rterio sclero sis. 3: 31-39.
15. Lutz R.J., H s u L., M c n a w a t I., Z r u b c k I., E d w a r d s K. 1983. “C o m p a r i s o n of steady
a n d pulsatile flow in a double branching arterial model.”J B iom ech. 16: 753-766.
16. El M a s r y O.A., Feuerstein I.A., P o u n d G.F. 1978. “Experimental evaluation of
streamline patterns an d separated flows in a series of branching vessels with impli
cations for atherosclerosis a n d thrombosis.”C irc R es. 43: 608-617.
17. C h o Y.I., B a c k L.U., C r a w f o r d D . W . 1985. “Experimental investigation of branch
flow ration, angle, and R e y n o l d s n u m b e r effects o n the pressure and flow fields in
arterial branch models.”J B io m ed E ng. 107: 257-267.
18. W a l b u r n F.I., Stein P.D. 1980. “F l o w in a symmetrically branched tube simulating
the aortic bifurcation: the effects of unevenly distributed flow.”A n n B io m e d Eng.
8: 159-173.
19. Perktold K., Peter R.O. 1990. “Nu merical 3D-simulation of pulsatile wall shear
stress in an arterial T-bifurcation model.”J B iom ech Eng. 12: 2-12.
26.24 FLU ID FLO W HANDBOOK
20. Perktold K., R e sc h M . 1991. “Numerical flow studies in h u m a n carotid artery bifur
cations: basic discussion of the geometric factor in atherogenesis.”J B iom ech
E ngng. 12: 111-123.
21. Fabregues S., Baijens K., Rieu R., Bergeron P. 1998. “H e m o d y n a m i c s of endovas-
cular prostheses.”J B iom ech. 31: 45-54.
22. Gijsen F.J.H., van de V o s s e F.N., Janssen J.D. 1999. “T h e influence of the non-
N e w t o n i a n properties of blood on the flow in large arteries: steady flow in a carotid
bifurcation model.”J B iom ech. 32: 601-608.
23. Gijsen F.J.H., Allanic E., van de V o s s e F.N., Janssen J.D. 1999. “T h e influence of
the n on - N e w t o n i a n properties of blood o n the flow in large arteries: unsteady flow
in a 9 0 degree curved tube.”J B iom ech. 32: 705-713.
24. Bellhouse B.J., Talbot L. 1969. “T h e fluid me c hanics of the aortic valve.”J Fluid
M ech. 35: 721-735.
25. Kilner P.J., Y a n g G.Z., M o h i a d d i n R.H., Firmin D.N., L o n g m o r e D.B. 1993. “Heli
cal and retrograde secondary flow patterns in the aortic arch studied b y three-direc
tional magnetic resonance velocity mappi ng .”C irculation. 88: 2 2 3 5 - 2 2 4 7 .
26. B o g r e n H.G., B u o n o c o r e M . H . 1994. “B l o o d flow me as u r e m e n t s in the aorta and
major arteries with M R velocity mapp in g. ”JM R l. 4: 119-130.
27. Klipstein R.H., Firmin D.N., U n d e r w o o d S.R., Re es R.S.O., L o n g m o r e D.B. 1987.
“B l o o d flow patterns in the h u m a n aorta studied by magnetic resonance.”B r H eart J.
58:316-323.
28. Moller H.E., Klocke H.K., Bongartz G.M., Peters P.E. 1996. “M R flow quantifica
tion using R A C E : clinical application to the carotid arteries.”JM Rl. 6: 503-512.
29. Zarins C.K., Gi ddens D.P., Bharadvaj B.K., Sottiural V.S., M a b o n R.F., G l a g o v S.
1983. “Carotid bifurcation atherosclerosis: quantitative correlation of plaque local
ization with flow velocity profiles and wall shear stress.”C irc R es. 53: 502-514.
30. K u D.N., G i d de ns D.P., Zarins C.K., G l a g o v S. 1985. “Pulsatile flow a n d athero
sclerosis in the h u m a n carotid bifurcation: positive correlation b e t w e e n plaque loca
tion and l o w and oscillating shear stress.”A rterio sclero sis. 5: 293-302.
31. L o n g Q , X u X.Y., Ariff B., T h o m S.A., H u g h e s A.D., Stanton A . V . 2000. “R e c o n
struction of blood flow patterns in a h u m a n carotid bifurcation: a c o m b i n e d C F D
a n d M R I study.”JM R l. 11: 299-311.
32. Khalifa A.M.A., G i d d e n s D.P. 1981. “Characterization an d evolution of poststenotic
flow disturbances.”J B iom ech. 14: 279-296.
33. Talukder N., Fulenwider J.T., M a b o n R.F., G i d d e n s D.P. 1986. “Post-stenotic flow
disturbance in the d o g aorta as me as u r e d with pulsed Doppler ultrasound.”J B io
m ech Eng. 108: 259-265.
34. Solzbach U., Wollschlager H., Zeiher A., Just H. 1987. “Effect of stenotic geometry
o n flow behavior across stenotic models.”M e d B iol E ng C om p. 25: 543-550.
35. Y o u n g D.P., Tsai F.Y. 1973. “F l o w characteristics in m od e l s of arterial stenoses. I.
Steady flow.”J B iom ech. 6: 3 9 5 — 410.
36. Y o n g c h a r e o n W., Y o u n g D.F. 1979. “Initiation of turbulence in m o d e l s of arterial
stenosis.”J B iom ech. 12: 185-196.
37. F e l d m a n R.L., Nichols W . W . , Pepine C.J., Conetta D.A., Conti C.R. 1979. “T h e
coronary h e m o d y n a m i c s of left m a i n and branch coronary stenoses: the effects of
reduction in stenosis diameter, stenosis length, and n u m b e r of stenoses.”J T hor
C ardiovasc Surg. 77: 377-388.
38. Ka rayannacos P.E., Takulder N., N e r e m R.M., R o s h o n S., V a s k o J.S. 1977. “T h e
role of multiple noncritical arterial stenosis in the pathogenesis of ischemia.”J Thor
C ardiovasc Surg. 73: 458-4 69 .
BLOOD FLOW DYNAM ICS 26.25
39. Kilpatrick D., W e b b e r S.D., Colie J-P. 1990. “T h e vascular resistance of arterial
stenoses in series.”A n g io lo g y . 41: 278-285.
40. D e B r u y n e B., Pijls N.H.J., Heyndrickx G.R., H o de ig e D., Kirkeeide R., G o u l d K.L.
2000. “Pressure-derived fractional flow reserve to assess serial epicardial stenoses:
theoretical basis a n d animal validation." C irculation. 101: 1840-1847.
41. A h m e d S.A., G i d d e n s D.P. 1983. “Velocity m ea su r e m e n t s in steady flow through
axisymmetric stenoses at moderate Reynolds numbers." J B iom ech. 16: 505-516.
42. E v a n s D.H., Barrie M.J., Bentley S., Bell P.R.F. 1980. “T h e relationship between
ultrasonic pulsatility index and proximal arterial stenosis in a canine model.”C irc
R es. 46: 470- 47 5 .
43. A s a k u r a T., Karino T. 1990. “F l o w patterns and spatial distribution of atheroscle
rotic lesions in h u m a n coronary arteries.”C irc Res. 66: 1045-1066.
44. F r i e d m a n M.U., Bargcron C.B., Hutchins G .M., M a r k F.F., Deters O.J. 1980.
“H e m o d y n a m i c m e a s u r e m e n t s in h u m a n arterial casts, and their correlation with
histology a n d luminal area.”J B iom ech Eng. 102: 247-251.
45. F o x I.A. H u g h A.E. 1966. “Localization of atheroma: a theory based o n boundary
layer separation.”B r H e a rt J. 28: 388-399.
46. M o o r e J.E., X u C., G l a g o v S., Zarins C.K., K u D.N. 1994. “Fluid wall shear stress
m e a s u r e m e n t s in a m o d e l of the h u m a n abdominal aorta: oscillatory behavior and
relationship to atherosclerosis.”A th ero sclero sis. 110: 225-240.
47. B u d w i g R., Eiger D., H o o p e r H., Slippy J. 1993. “Steady flow in abdominal
a n e u r y s m models.”J B io m ec h Eng. 115: 418-423.
48. Taylor T.W., Y a m a g u c h i T. 1994. “Three-dimensional simulation of blood in an
a b do mi na l aortic aneurysm: steady and unsteady flow cases.”J B iom ech Eng.
116:89-97.
49. V i s w a n a t h N., R o d k i e w i c z C.M., Zajac S. 1997. “O n the abdominal aortic
aneurysms: pulsatile state considerations.”M e d Eng Phys. 19: 343-351.
50. Aenis M., S t a n c a m p i a n o A.P., W a k h l o o A.K., Lieber B.B. 1997. “M o d e l i n g of flow
in a straight stented a n d non-stented side wall a n e u r y s m model.”J B iom ech Eng.
1 1 9 :2 06 - 21 2.
51. Y u S.C.M., Z h a o J.B. 2000. “A steady flow analysis o n the stented and non-stented
side wall a n e u r y s m models.”M e d E ng P hys. 21: 133-141.
52. L u n d b r o o k J. 1962. “Functional aspects of the veins of the leg.”A m H e a rt J. 64:
706-713.
53. Kisslo J., A d a m s D.B., Belkin R.N. 1988. D o p p ler C o lo r F low Im aging. Churchill
Livingstone, N e w York.
54. Perry G.J., H e l m c k e F., Nand a, N.C., B y a r d C., Soto B. 1987. “Evaluation of aortic
insufficiency by D o p p l e r color flow map pi ng . ”JA C C . 9: 952-959.
55. B a u m g a r t n e r H., Kratzer H., He l mreich G., K u h n P. 1988. “Quantitation of aortic
regurgitation b y colour c o d e d cross-sectional Do ppler echocardiography.”E ur
H eart J. 9: 380-387.
56. Taylor A.L., Eichhorn E.J., Brickner M.E., Eberhart R.C., G r a y b u m P.A. 1990.
“Aortic valve mo r ph ology: an important in vitro determinant of the proximal regur
gitant jet width b y D o p p l e r color flow map pi ng .”JA C C . 16: 4 0 5 - 4 12 .
57. R e i m o l d S.C., Atkinson C.M., L u n a F.B., L e e R.T. 1993. “Influence of jet i m pinge
m e n t o n color D o p p l e r parameters of aortic regurgitation.”E chocardiography.
1 0 :1 13-119.
26.26 FLU ID FLO W HANDBOOK
58. Recusani F., Bargiggia G.S., Y o g a n a t h a n A.P., Raisaro A., Valdes-Cruz L.M., S u n g
H - W . , Bertucci C., Gallati M., M o i s e s V.A., S i m p s o n I.A., Tronconi L., S a h n D.J.
1991. “A n e w m e t h o d for quantification of regurgitant flow rate using color
D op pl er flow imaging of the flow convergence region proximal to a discrete ori
fice.”C ircu la tio n . 83: 594-6 0 4.
59. S i m p s o n I.A., S a h n D.J., Valdes-Cruz L.M., C h u n g K.J., S h e r m a n F.S., S w e n s s o n
R.E. 1988. “Color D o ppler flow m a p p i n g in patients with coarctation of the aorta:
n e w observations an d im pr o v e d evaluation with color flow diameter and proximal
acceleration as predictors of severity.”C irculation. 77: 736-744.
60. D e e g K.H., H o f b e c k M., Singer H. 1993. “Diagnosis of subclavian steal in infants
with coarctation of the aorta and interruption of the aortic arch b y color-coded
D o pp le r sonography.”J U ltrasound M ed. 12: 713-718.
61. Steinke W . , Kloetzsch C., Hennerici M . 1990. “Carotid artery disease assessed by
color D o p p l e r flow imaging: correlation with standard D oppler sonography a n d
angiography.”A JN R . 11: 259-266.
62. Bluth E.I., Merritt C.R. 1992. “Do ppler color imaging. Carotid a n d vertebral arter
ies.”C lin D ia g n U ltrasound. 27: 61-96.
63. M o r a n P.R. 1982. “A flow velocity zeugmatographic interlace for N M R im aging in
h u m a n s . ”M a g n R es Im ag. 1: 197-203.
64. Bryant D.J., P a y n e J.A., Firmin D.N., L o n g m o r e D.B. 1984. “M e a s u r e m e n t of flow
with N M R im aging using a gradient pulse an d phase difference technique.”JC A T .
8: 588-593.
65. M e i e r D., M a i e r S., Bosiger P. 1988. “Quantitative flow me as u r e m e n t s o n p h a n t o m s
a n d o n blood vessels with M R . ”M a g n R es M ed. 8: 25-34.
66. Bendel P., B u o n o c o r e E., Bockisch A., Besozzi M . C . 1989. “B l o o d flow in the
carotid arteries: quantification b y using phase-sensitive M R imaging.”A JR . 152:
1307-1310.
67. Chatzimavroudis G.P., W a l k e r P.G., Oshinski J.N., Franch R H , Pettigrew R.I.,
Y o g a n a t h a n A.P. 1997. “T h e importance of slice location o n the accuracy of aortic
regurgitation m e a s u r e m e n t s with magnetic resonance phase velocity mapping: an in
vitro investigation.’ M w ? B io m ed Eng. 25: 644-652.
68. U n d e r w o o d S.R., Firmin D.N., Klipstein R.H., R e e s R.S.O., L o n g m o r e D.B. 1987.
“M a g n et ic resonance velocity mapping: clinical application of a n e w technique.”B r
H e a rt J. 57: 4 04 - 4 1 2 .
69. M o h i a d d i n R.H., W a n n S.L., U n d e r w o o d R., Firmin D.N., R e e s S., L o n g m o r e D.B.
1990. “V e n a caval flow: assessment with cine M R velocity map p in g. ”R a d io lo g y.
I l l : 537 -5 4 1.
70. K o n d o C., C a p u t o G.R., S e m e l k a R., Foster E., S h i m a k a w a A., Higgins C.B. 1991.
“Right a n d left ventricular stroke v o l u m e m e a s u r e m e n t s with velocity-encoded cine
M R I : in -vitro a n d in-vivo validation.”A JR . 157: 9-16.
71. Pelc L.R., Pelc N.J., Rayhill S.C., Castro L.J., Glo ve r G.H., Herfkens R.J., Miller
D.C., Jeffrey R.B. 1992. “Arterial and venous blood flow: noninvasive quantifica
tion with M R imaging.”R adiology. 185: 809-812.
72. Chatzimavroudis G.P., W a l k e r P.G., Oshinski J.N., Franch R.H., Pettigrew R.I.,
Y o g a n a t h a n A.P. 1997. “Slice location d e p en de n ce of aortic regurgitation m e a s u r e
m e n t s with M R phase velocity mappi n g. ”M agn R es M ed. 37: 545-551.
73. Chatzimavroudis G.P, Oshinski J.N., Franch R.H., Pettigrew R.I., W a l k e r P.G.,
Y o g a n a t h a n A.P. 1998. “Quantification of aortic regurgitation with magnetic reso
n ance phase velocity mapping: a clinical investigation of the importance of slice
location.”J H e a rt Valve D is. 7: 94-101.
CHAPTER 27
HEAT TRANSFER
IN PIPE FLOW
Heat transfer analysis provides a means to estim ate the fluid tem perature along the
entire flow path o f the fluid, and to estim ate the tem perature o f outside pipe sur
faces, which is a health and safety concern. Heat transfer analysis must be com bined
with fluid flow analysis to achieve an accurate estimation o f the fluid physical prop
erties pertinent to fluid flow such as density, viscosity, and liquid surface tension
in 2-phase flow . A lso, fluid flow therm odynam ics, such as vapor-to-liquid ratio,
which is an im portant param eter in 2-phase flow o f gases and liquids, is a func
tion o f the fluid tem perature. Flow o f fluids in pipes is subjected to heat transfer,
energy loss or gain to or from the surroundings, by the three know n heat transfer
processes: conduction, convection, and radiation. This chapter will briefly cover
heat transfer analysis pertinent to the flow o f fluids in pipes.
T herm al energy crosses the pipe boundary by one or a com bination o f the fol
low ing processes: conduction, convection, and radiation. Each w ill be consid
ered in turn.
27.1
27.2 FLU ID FLO W HANDBOOK
27.1.1 Conduction
where
L = Pipe length
q = H eat transfer rate
k = Conductivity
T = Tem perature
r = Radius
Equation 27.1 can be sim plified into the follow ing equation by elim inating T2
and Ty.
2 k ( T .- T 4)
(27.2)
L In (r2/r,) In (r3/r 2) ln(r4/ r 3)
Pipe
Insulation
Insulation (k3 )
The last term in the denom inator o f the above equation is the soil heat trans
fer resistance term , w here ks is the soil conductivity and S is the shape factor. For
very long horizontal pipes, the follow ing shape factors may be used:
s . 4
2 - L » r
cosh ( z / r )
(27.4)
2 nL
z>3
ln(2z / r)
where
O nce the heat transfer rate, q, is calculated from Equation 27.2 or 27.3, the
tem perature at any surface may be calculated using Equation 27.1.
27.1.2 Convection
C onvection refers to heat transfer due to displacem ent o f fluids inside and outside
the pipe (and insulation). The convection heat transfer is a function o f the fluid
physical properties, fluid velocity (turbulence), and the system geom etry, such as
pipe diam eter. Fluid flow in pipes is normally subjected to sim ultaneous heat trans
fer by convection and conduction. C onvection heat transfer is either free or forced.
Free (or natural) convection heat transfer occurs due to the natural m ovem ent of
fluid caused by a density difference induced by a tem perature difference. Forced
convection refers to fluid m ovem ent im posed by external m echanical means such
as fans. Equation 27.5 estim ates heat transfer per unit length o f the pipe due to
convection and conduction (refer to Figure 27.2 for nom enclature):
Inside Boundary
layer, (hi) Temp, of Inside Fluid
1^3
Outside Boundary
layer; (h0 )
T5 (Fluid Temp.
Pipe (k:1
Insulation
Insulation (k3 )
F I G U R E 27.2 Schematic and variables of an insulated pipe with fluids inside and outside.
Although the purpose o f insulation is to reduce the heat transfer rate, under
certain conditions the addition o f insulation may cause an increase in the heat
transfer rate. Although adding insulation to a pipe will increase the conductive heat
resistance, the outside area o f the pipe will also increases and hence the convec
tive resistance will decrease. The critical insulation radius rc is the insulation radius
that m axim izes the overall heat transfer rate and is given by
where
The above equation is derived assum ing constant pipe wall tem perature and
constant outside heat transfer coefficient. W hen the pipe radius, /?, is larger than
rc, any additional insulation will increase therm al resistance and thus decrease the
heat transfer rate. H ow ever, when pipe radius is sm aller then rc, initial addition of
insulation will promote heat transfer by convection and thus increase the heat trans-
HEAT TRA N SFER IN PIPE FLOW 27.5
(a) (b)
F I G U R E 27.3 Effect of insulation thickness on overall heat transfer rate, (a) Pipe radius < rc;
(b) Pipe radius > rc. (R is pipe radius.)
fer rate. Figure 27.3 show s the heat transfer rate as a function o f the insulation
thickness when pipe radius, /?, is larger and sm aller than the critical insulation
radius, rc.
27.1.3 Radiation
Therm al radiation is the electrom agnetic transfer o f energy from a hotter surface
(therm ally excited body) to a colder surface, and thus represents heat transfer with
out a physical m edium . H eat transfer from the sun to the earth is an exam ple of
heat transfer by therm al radiation. A lthough heat transfer by radiation is minimal
for short lengths o f pipes, it is an im portant elem ent in the heat transfer analysis
o f onshore oil and gas transm ission lines. The m axim um energy that can be radi
ated from an ideal black body is found by the Stefan-B oltzm ann law as:
qr = o A T 4 (27.7)
w here
qr = Radiant heat
A = Surface area
T = A bsolute tem perature
a = C onstant = 5.670 X 10"8 W /m 2 • K4 or = 0.1714 X 1O '8 B tu/hr • ft2 • R4
For real bodies the above equation must be multiplied by the surface emissivity, £.
27.6 FLU ID FLO W HANDBOOK
Therm al radiation between two bodies is a function o f the bodies' absolute tem
peratures, bodies’ em issivities, and a shape factor F \2 such as:
W hen a fluid flow s at uniform velocity into a pipe, a hydrodynam ic boundary layer
begins to develop at the tube wall. The thickness o f this hydrodynam ic boundary
layer grow s with distance jc from the entrance. At a certain distance, Xfclyh, the hydro-
dynam ic boundary layer thickness becom es equal to the radius o f the tube. The
region from the entrance to Xjd>h is called the “hydrodynam ic entrance region” and
the distance xfdj x is term ed the “hydrodynam ic entry length” . The region after xfdth
is called the “fully-developed region” for velocity distribution. In the fully devel
oped region, a parabolic velocity distribution is developed for a N ew tonian fluid.
T he velocity profiles are shown in Figure 27.4.
For lam inar flow, the hydrodynam ic entry length m ay be obtained from Equa
tion 27.9:
(27.9)
For turbulent flow , Equation 27.10 may be used to estim ate the hydrodynam ic
entry length:
(27.10)
W hen fluid enters a tube with a uniform tem perature that differs from the surface
tem perature, convective heat transfer occurs and a therm al boundary layer begins
to develop as can be seen from Figure 27.5.
H EA T TRA N SFER IN PIPE FLOW 27.7
V el oc ity
d, h
Fully
E ntrance Dev e lo pe d
Region Region
For lam inar flow, the thermal entry length may be obtained from Equation 27.11:
fxfd ,
= 0.05 Re/:>Pr (27.11)
D lam
For turbulent flow, the entry length is independent o f both R eynolds and Prandtl
num bers, and Equation 27.10 can be used to estim ate the thermal entry length.
FIGURE 27.5 Thermal boundary layer development in a heated pipe. Ts > T(r,0).
27.8 FLU ID FLO W HANDBOOK
2 7 .3 MEAN TEMPERATURE
The tem perature profile inside a tube is not uniform as long as heat transfer occurs.
T hus, in general the fluid tem perature will be a function o f both the axial position
x and the radial position r. How ever, it is frequently useful to define the m ixed
tem perature, which represents the average tem perature o f the fluid at a given
axial position. The m ixed tem perature, 7m, is defined as
pucvTdAc
Tm = - --------------- (27.12)
mcv
w here u and T are the velocity and tem perature distributions, respectively, and p
and cv are the density and heat capacity o f the fluid, respectively.
The m ean tem perature o f a fluid that enters the tube at m ixed tem perature Tm ]
and leaves at a m ixed tem perature o f Tm l is given sim ply as the arithm etic m ean
o f these tw o tem peratures; thus:
Finally, the overall heat flux from the tube wall (at tem perature Ts) to the fluid
is given by N ew ton’s law o f cooling:
There are tw o types o f heating system used industrially and in the laboratory: steam
heating trace and electric heating tap. For the steam heating system, the surface tem
perature may be assumed to be constant (CST), while for the electric heating system,
the heat flux through the tube surface is a constant (CH F). The equations used to
estim ate the heat transfer coefficients are different for the heat transfer system s of
C ST and CHF.
H EAT TRA N SFER IN PIPE FLOW 27.9
The hydrodynam ic behavior o f lam inar flow is quite different from that o f turbu
lent flow and, naturally, the heat transfer equation for lam inar flow will be differ
ent from that o f turbulent flow.
For flow in a pipe, laminar flow usually can be expected when the Reynolds num
ber, Re, is less than 2300. A fter this number, the flow will pass through a transition
zone until the Re num ber is higher than 4000, then, turbulent flow is established.
where
Q uite often, the equation for the heat transfer coefficient is expressed as
a
(
- j (27.16)
where a, m, n, and a are constants which can be determined from experim ental data.
In a therm al entry region, the heat transfer coefficient is a function o f the axial
location, x , as can be seen from Figure 27.6.
The local heat transfer coefficient, hx, is called the “local” because it applies
only at a given axial position x. An “average” heat transfer coefficient, hL, may be
27.10 FLU ID FLO W HANDBOOK
F I G U R E 27.6 Local and average Nusselt numbers for a pipe in thermal entrance
regions with fully-developed laminar flow. (From Holman, 1990.)
obtained by integrating the local heat transfer coefficient over the entire length o f
the pipe,
hL = — [ hxdx (27.17)
L Jo
w here L is the length o f the pipe. Both the local and the average heat transfer
coefficients in therm al entrance regions are shown in Figure 27.6.
The equations collected in Table 27.1 can be used to calculate heat transfer coef
ficients in pipe flow . The user should pay special attention to the conditions spec
ified for each equation. U sing the equation to predict the heat transfer coefficient
outside the specified conditions will result in some uncertainties. In these correla
tions, jjs is the fluid viscosity at the surface tem perature o f the w all, and all o f the
H EAT TRA N SFER IN PIPE FLO W 27.11
Equation Conditions
N u = 0.027Re°'8 Pr1'-’
^ ! " U Fully developed turbulent flow
/ n \°055
N u = 0.036ReO8Prl/'1|y-J Entrance region, turbulent flow
0.0668(D/L)RePr
N u - 3.66 + ™ Fully developed laminar flow.
1 + 0.04|(D/L|RePrJ
At constant surface temperature
10 < ( ^ - j < 4 0 0
other physical properties o f the fluid are evaluated at the mean tem perature o f the
fluid, Tb.
As an exam ple o f how to use these correlations, consider the following problem:
EXAMPLE 27.1
A ir at P, = 147 psia (1013 kPaa) and Tm ] = 648°R (360 K) enters a 1.64 f t (0.5
m)-long tube with an inner diameter o f D = 0.0164ft (0.005 m) and a
constant surface temperature o fT s = 576°R (320 K). The average velocity o f
the a ir in the tube is 32.8 ft/s (10.0 m/s). The a ir exits at Tm 2 = 612°R (340 K).
What are the heat transfer coefficient h, the heat flu x q", and the total heat
transfer rate q?
Solution
The mean temperature o f the air is
At this tem perature and 14.7 psia (101.3 kPaa) air has the follow ing physi
cal properties:
ft - s
Since Re > 4000, the flow is turbulent, so using the first correlation from
T able 27.1 gives
Nu = — = 0.023 R e °8 P r(U
k
H ere, the exponent on the Prandtl num ber is 0.3 because the fluid is being
cooled. Solving for h gives
(0 .0 2 3 (2 4 0 4 0 )°s (0 .6 9 7 )"’)
0.0164 ft
REFERENCES:
28.1 INTRODUCTION
R ecent years have seen rapid developm ent in technologies for micro scale devices,
such as MEM S (M icro Electro-M echanical Systems). M EM S are devices that inte
grate electrical and m echanical com ponents on single silicon substrates fabricated
by IC fabrication processes, or they may be fabricated by using other materials and
micro machining processes at micron scale. This is the so-called top-down approach.
One the other hand, N ano is a prefix which literally m eans one billionth. Thus
nanotechnology can be defined as the technology based on the m anipulation o f
individual atom s and m olecules to build structures to atom ic specifications at a
scale o f billionths o f a meter. O pposite to the fabrication m ethod o f m icro system
technology, this is the so-called bottom-up approach. It was said to be first proposed
by R ichard Feynm an [32], a N obel Laureate in Physics, in his keynote speech in
the A m erican Physical Society annual m eeting held in 1959. Since the 1980s,
D rexler has been developing and prom oting the ideas o f m olecular nanotechnol
ogy [23, 24, 25] and the F oresight Institute [w w w .foresight.org] was specifically
set up by him to guide this em erging technology to im prove the human condition.
R ecent years have seen a strong interest in the technology as exem plified by the
announcem ent o f a nanotechnology initiative by the A m erican President in 2000.
Sim ilar initiatives have also been announced by many other countries recently.
C urrently m icro and nano flow system s, being parts o f the m icro and nano tech
nology, are being very actively developed for possible applications in the areas of
biom edical engineering, inform ation technology, m aterials engineering, energy
and environm ental engineering, where m iniaturization has been a driving force
for developm ent for decades.
28.1
28.2 FL U ID FLO W H ANDBOO K
Recently various m icrofluidics for biom edical applications have been developed.
For exam ple, a m icrofabricated fluorescence-activated cell sorter developed by
Fu et al. [3 4 ] can be used to separate cells. The device is a silicone elastom er chip
with fluid flow channels 100 (I wide joined at a 3 fi wide T-junction (see Figure 28.1).
By using electrodes to control the electro-osm otic flow , depending on the fluores
cence em ission o f the cells to be collected, the flow can be directed to the collec
tion channel if the fluorescence em ission exceeds a certain level. Based on sim ilar
technique, C astro et al. [16] w ere able to detect a single D N A m olecule and m ea
sure its length.
B hatia et al. [8] have successfully sustained rat-liver cells for a m onth on a
BioM EM S chip. It is full o f m icro pores that let nutrients and chemicals flow through
but block larger bacteria and viruses. Figure 28.2 displays the pores etched in the
porous silicone substrate and Figure 28.3 show s close-up view s o f the liver cells
in silicone substrates.
Epstein et al. [27,28] have been w orking on a pow er M EM S project to develop
a M EM S-based m icro gas turbine-generator system . The m icro-gas turbine has a
diam eter o f 1 cm and a thickness o f 3 m m , w hich is expected to produce 10 to 20
w atts o f power. The turbine w heel show n in Figure 28.4 m easures ju st 4 m illim e
ters in diam eter and this radial inflow turbine w heel was m anufactured from sil
icon using deep reactive ion etching. R esearch is being undertaken to study the
surface effects, gas bearings and fabrication m ethods for the system.
//iV:-.
j j J r 4' Y ; vaV' J ..
SO
1U0
X 5 0 . 0 0 Cl uw/ d iv
Ib 'J 2 bOO i.V U* rsM /aiv
UM
F I G U R E 28.2 Atomic force micrograph of electrochemically etched holes in a silicon substrate [8J.
•# \
(b)
D ro plet A
ejection Nozzle Plate
0 = 3 u n a n n e b a rrie r
O v e r c o a t s —{
<o=a c o n d u c to r
K e s is to r
•<2=3 Thermal Barrier
s u b s tra te £=e£ s>
A nother im portant device in micro fluidics systems is the m icro pump, many of
w hich have been developed in recent years for chem ical, m edical and biom edical
applications [39,90]. M ost m icro pum ps rely on positive displacement to move fluid
due to low Reynolds num bers, and they are mostly o f the reciprocating type. Valves
used in these pumps can be passive membranes or actively controlled devices. Forster
et al., [33] have developed a micro pump with fixed valves, which is especially suited
for delivering fluid containing cellular materials. Shown in Figure 28.6(a) is the fixed
valve conduit w hich was etched in a silicone substrate using reactive ion etching
(RIE). It was found that this valve has higher volum etric efficiency than other con
ventional design because forw ard flow is mainly dom inated by viscous loss while
reverse flow is m ainly influenced by dynam ic pressure loss due to strong m om en
tum interactions at the junctions. Figure 28.6(b) illustrates the piezoelectric disk used
for actuation and Figure 28.6(c) is a schem atic o f the pum p-valve assembly. Shown
in Figure 28.7 is another valveless m icro pum p actuated by a piezo-electric disk [92,
93]. H ere diffusers and nozzles are used as flow rectifying elem ents for the inlet and
outlet o f the pum p, w here the forw ard flow is in the diffuser direction and reverse
flow in nozzle direction. Figure 28.7(a) is a schematic of the pump and Figure 28.7(b)
illustrates a double pum p configuration etched on a silicon substrate with a diffuser
thickness o f 80 ji. It w as dem onstrated that the pum p could achieve a m aximum
pressure o f 7.6 m H 20 and a m axim um flow rate o f 2.3 ml/min. Rahm an et al. [84]
have also proposed and dem onstrated a simple design for a valveless pump with dif
fusers and nozzles, w hich is actuated by therm al bubbles.
A novel application o f m icrofluidics system s was recently proposed by Chiu
et al, [21 ] to use three-dim ensional m icrofluidic channel netw orks for solving com
putationally hard m athem atical problem s. The netw orks are designed such that the
fluid flow in the m icro channel netw ork provide a parallel algorithm for searching
all the potential solutions and parallel optical readout o f all the solutions.
C arbon nanotube is a nano structure discovered in recent years and is being
intensively studied. For exam ple, carbon nanotube based gear system s rotating in
a coolant gas have been studied by Han et al. [108] as potential com ponents o f
FLU ID FLO W HANDBOOK
MW
Piezoelectric
disc
In let .Outlet
TE S LA T 4 5 -4
OUTLET
VALVE
G lass
(b )
nano-m achinery (see Figure 28.8 for the visualization o f the system ) using m olecu
lar dynam ics (M D ) sim ulation. The M D sim ulation used the B renner’s potential for
the bonded interactions [13] and Lennard-Jones (6 -1 2 ) potential for non-bonded
interactions. It was found that the gears would work well if temperature is lower than
600K. The surrounding gas can cool the gears to a low tem perature so that they can
work. The higher the kinetic energy o f the gas is, which is cooler than the gears,
the faster the cooling becom es, but also resulting in higher drag on the gears.
Gerstein and Levitt [38] reported some o f the earliest M D simulations performed
in the 1980’s to m odel w ater at m olecular scale and study how w ater affects the
structures and dynam ics o f biological m olecules (m ainly protein m olecules and
DN A ) in living organism s. They confirm ed that w ater m olecules play a critical
role in successfully sim ulating the alpha helix’s behaviors o f protein m olecules
and double helix structures o f DNA. A recent sim ulation o f the binding o f an
estrogen receptor (a protein m olecule) to a DNA [61, 87] was conducted to study
the m echanism underlying DNA sequence recognition by protein. Show n in Fig
ure 28.9 is an estrogen receptor (ER ) hom odim er (protein m olecule) interacting
28.8 FLU ID FLO W HANDBOOK
with an asym m etric DNA segm ent in w ater (w ater m olecules are denoted by red
and white colors and the yellow particles are sodium ions). The estrogen receptor
is able to recognize and bind to specific DNA sequences to control the expression or
switch-on o f specific genes. It was found in the sim ulation that w ater molecules
betw een the DNA and protein are instrum ental in the recognition o f the target
D NA sequences.
The above exam ples are used to dem onstrate the potential applications of micro
and nano flow system s, and also to illustrate the im portance o f understanding how
the different com ponents and system s interact with each other and with its sur
rounding fluids in order to perform their tasks. O bviously a better understanding
o f the forces involved in the interactions will lead to better and even innovative
designs o f such system s.
H ow ever, it should be noted that free-m olecular flow may not exist for liquid
flow because liquid m olecules are alw ays in constant interaction. As show n in a
num ber o f experim ents (see the following section), when Kn is between 0 .0 1 and 0 .1,
although the N avier-Stokes equations for gas flows are still valid, a slip boundary
condition is necessary for the solid wall due to the slight deviation from equilibrium
there. H ow ever, it should be noted that the average distance betw een m olecules in
the liquid is com parable to the diam eter o f its m olecules and the m om entum trans
port is dom inated by inter m olecular interactions instead o f the random motion and
collisions as found in gases. Thus it is necessary to separate the studies o f m icro
and nano flow s in dilute gases and liquids (including dense gases). It is im possible
to derive the slip boundary conditions for liquid flow s due to a lack o f theoretical
guidance. As a result, we can only use either the continuum m odel or the m olecu
lar m odel to describe micro and nano liquid flows. It is expected that experim ental
study o f m icro and nano flows will be difficult if m icro and nano system technol
ogy has not been fully developed. Therefore theoretical and com putational m eth
ods w ill play a relatively dom inant role at the present time.
R ecently a num ber o f experim ents have been perform ed to study gas flow s in micro
channels. Tison [96] conducted a study o f the rarefied gas flow in a pipe with a diam
eter o f 2m m and length o f 400m m . The inlet and outlet flow conditions were varied
to obtain a w ide range o f Kn, ranging from 0 to 200. Figure 2 8 .10 show s the m ass
flow rate versus ( p i - p 2). Three characteristically different flow regim es were iden
tified, corresponding to:
A rkilic [3] conducted experim ents to investigate the im pacts o f low -Reynolds num
bers and com pressibility on m icro gas flow in m icrofabricated channels in silicon
substrates. In particular, the slip flow boundary conditions w ere studied in detail by
m easuring a param eter called Tangential M om entum A ccom m odation Coefficient
(T M A C ) using a precise m ass flow rate m easuring technique. T he TM A C can be
defined as the fractional m om entum exchange betw een the fluid and the surface:
TM A C = T' — r
T ,~ T W
w here r i s the tangential m om entum flux with subscripts /, r and vv denoting inci
dent, reflected and wall conditions. A schematic representation o f the primary wafer
M ICRO AND NANO FLOW S 28.11
<p)-p\> fPa2]
fabrication steps are illustrated in Figure 28.11. The nominal height o f the channel
is 1.33 m icrons and the length 7490 m icrons. Three types o f gases w ere em ployed
in the test: argon, nitrogen and carbon dioxide. Inlet and outlet pressures w ere
varied to generate different flow conditions with Kn ranging from 0.03 to 0.44. It
was found that T M A C w as independent o f the Knudsen num ber and gas species
for the cases tested, and it was less than unity, ranging from 0.75 to 0.85. This
indicates that the flow is indeed in the slip flow regime. A rkilic also perform ed a
theoretical study on the com pressibility effect in the m icro channel and the effect
o f K nudsen num bers on com pressibility. These are shown in Figure 28.12. The
im portant features o f the flow is that pressure distribution along the channel is
nonlinear due to the com pressibility effect although the inlet M ach num ber is
very low. A nd w hen the K nudsen num ber (rarefaction effect) is increased, the
non-linearity is dim inished, w hich m eans that the rarefaction effect counteracts
the com pressibility effect. T he experim ental m easurem ent by Pong e t al, [76J
and Sreekanth [91J seem s to agree with this observation.
Harley et al, [45] also conducted a sim ilar experim ent to study m icro gas flow s
in channels fabricated in silicon substrates. Low R eynolds num ber, com pressible
(high subsonic) gas flow s o f nitrogen, helium and argon in channels o f different
sizes w ith various K nudsen num bers w ere studied experim entally and num erically.
The channels w ere typically 100 m icrons wide, 10,000 m icrons long and 0.5 to
20 m icrons in depth. The K nudsen num ber ranged from 0.001 to 0.4. It w as found
that the m easured friction factor was in good agreem ent with theoretical predictions
28.12 FLU ID FLO W HANDBOOK
based on the assum ptions o f isotherm al, locally fully developed and first-order
slip flow.
On the other hand, liquid flows at m icro/nano scale are less well studied and dif
ferent experim ents have contradictory results due to the lack o f m olecular-based
theory, as pointed out by G ad-el-H ak [36]. For exam ple, Israelachvili [57] found
that the apparent viscosity is equal to the fluid viscosity for thin-film flows if the
film thickness exceeds 5 nm. If the thickness is less than 5 nm, the apparent vis
cosity can be as high as 100,000 tim es larger than the fluid viscosity. H ow ever,
Pfahler [77] and Pfahler et al. [78, 79] found that apparent viscosity is always sm aller
than the fluid viscosity for both liquid and gas flows in m icrochannels with a depth
ranging from 0.5 m icrons to 50 m icrons, and it decreases with the m icrochannel
depth. O thers reported that the apparent viscosity is the sam e as the fluid viscos
ity [70,98]. These observations suggest the failure o f the classical laws for the study
o f m icro liquid flow s. A nd it seem s that the com m only accepted approach for the
study o f m icro and nano liquid flow s is m olecular dynam ic sim ulation (M D S).
Finally, Peng et al. [73], Peng and Peterson [74] and W ang and Peng [103] ex p er
im entally studied the heat and liquid flow in m icrochannels, and the transition to
turbulence in heated m icrochannels. They found that lam inar flows can only exist
for R eynolds num ber (R e) less than 400, beyond which the flow s are in transition
up to Re = 1000. W hen Re is higher than 1000 the flows are fully turbulent.
Having exam ined som e o f the recent experim ental studies and their m ajor find
ings, w e will discuss the theories and sim ulation tools available for the study o f
m icro and nano gas and liquid flow s separately in the following sections.
28.5.1 K inetic Theory. In kinetic theory, it is assum ed that the gas is dilute and
thus there are only binary collisions betw een gas molecules. The kinetic theory for
dilute gases is based on the B oltzm ann equation. It is an equation for the d istrib u
tion function f(r,V ,t), w hich provides statistical description o f gases on m olecular
level. This equation takes into account o f tw o-body collisions only and also assum es
that the m olecular dim ensions are sm all in com parison with the mean distances
betw een m olecules. T hus it can only be applicable to dilute gases.
T he fundam ental statistical m echanics equation for /V particles is the Liouville
equation w hich expresses the conservation o f the particle distribution function in
6/V dim ensional phase space (three positions and three velocities for every particle).
A com plete description o f the /V-particle distribution function in the equation is
impossible for gas flows with realistic num ber o f molecules. Based on assum ptions
of dilute m onatom ic gas and m olecular chaos, it is possible to derive the Boltzm ann
equation for the velocity distribution function (or single particle distribution func
tion) (See [109] for detailed derivation). The Liouville equation can be w ritten as
28.14 FLUID FLO W HANDBOOK
w here n is the num ber density a n d / the norm alized particle distribution function.
The right hand side (RHS) o f the equation is the nonlinear collision integral which
represents the rate o f increase o f molecules as a result o f inter-m olecular collisions.
The variable / , denotes the value o f / a t C] and the asterisk superscript represents
post-collision values.
A nalytical solution to the Boltzmann equation for flows with com plex geom e
tries or large disturbances is impossible. Direct numerical solution o f the Boltzm ann
equation is also very difficult and im practical, because the num ber o f dim ensions
in phase space is six for spatially three dim ensional problem s and the num ber o f
m esh points is thus enorm ous. In addition, the range o f velocity is infinite and this
m akes it difficult to set the bounds in velocity space. Finally very large num ber
o f operations is required to num erically calculate the integrals in the nonlinear
collision term.
The Chapm an-Enskog theory provides an approximate solution to the Boltzmann
equation when the distribution is only slightly perturbed from the equilibrium M ax
w ellian distribution. And the distribution function can be expressed as
W ith the use o f the zero-order term and the Boltzmann equation, the Euler equa
tions o f continuum gas dynam ics can be derived. If the first-order term is included,
the linear transport term s for m om entum and energy can be derived and these will
lead to the m onatom ic gas form o f the N avier-Stokes equations o f continuum gas
dynam ics [18]. The inclusion o f the second-order term in the distribution will result
in a set o f very com plicated higher-order continuum equations, w hich are the so-
called B urnett equations.
28.5.2 Conservation Laws f o r Gaseous Flows. The Knudsen num ber can be related
to tw o dim ensionless param eters in fluid mechanics: the Reynolds num ber Re and
the M ach num ber Ma. The Reynolds num ber is defined as
v
MICRO AND NANO FLOW S 28.15
w here V is a characteristic velocity and v is the kinem atic viscosity o f the fluid.
The M ach num ber is
V
Ma = -
a
(28.4)
(28.5)
V 2 Re
For K n —>0 (R e—>°c), collisions becom e dom inant in the flow and it approaches
the continuum regim e and equilibrium state. Using the zero-order approxim ation
in the C hapm an-E nskog theory and substitutes it into the B oltzm ann equation re
sults in the E uler equations, which can be w ritten in conservation form in three
dim ensions as:
dW
+ V •F = 0 (28.6)
9/
w here W is the state vector, Fc is the convective flux vector. The state and convec
tive flux vector are given below:
pV
p
pu puV + pi
W = pv F, = pvV + pj (28.7)
pw pw V + pk
PE _ pH V
w here E is the total internal energy o f the gas and H the total enthalpy.
For flow with 0.001 < Kn < 0.1, i.e. the slip flow regime, the first-order term has
to be included in the C hapm an-Enskog approxim ate distribution function. W ith this
28.16 FLU ID FLOW HANDBOOK
function and the B oltzm ann equation the N avier-Stokes equations can be derived.
The N avier-Stokes equations in three dim ensions can be w ritten in conservation
form as
— + V -F C= V -F v (28.7)
ot
The inviscid flux vector Fc is the same as in the E uler equations and the vis
cous flux vector Fv is defined as
f.
V r +q
w here r is the viscous stress tensor, f v, f y and tz are its com ponents in a*, y and z
directions. is the heat conduction flux vector determ ined from the Fourier law.
The stress tensor is related to the rate o f strain based on the Stokes hypothesis:
28.5.3 Velocity Slip and Temperature Jump at the Wall. The slip velocity at the
wall was first studied by Maxwell [691 using the kinetic theory for dilute, monatomic
gases under isotherm al conditions. It can be expressed as
(28.9)
MICRO AND NANO FLOW S 28.17
where u, and ur are the incident tangential velocity at a distance to the wall equal
to the m ean free path and the reflected tangential velocity from the wall. And the
reflected velocity can be calculated by
«r = ^ A „ I / + ( 1- ^ ) M , (28.10)
< 7 .= ^ (28.11)
r, - r w
where r is the tangential m om entum flux with subscripts /, / and w denoting incident,
reflected and wall conditions. It defines the portion o f m olecules that are reflected
diffusively. Based on E quations 28.10 and 28.11, we have
(28.12)
| 02^
Uslip = - ( ( 2 - C J , ; )[uslip + A( — )wall + “y +*••] + ® \Mwall ) (2 ^ -13)
Uslip ~ UW
a!l + (28.14)
a, dn A,■all
Sim ilarly a tem perature ju m p on the wall can be derived and this phenom enon
was first studied by von Sm oluchow ski [101 J. The non-dim ensionalized gas tem
perature on the wall is
27 Kn, f d T '
T
1gas = T + “
1wall T ° v (28.15)
<7. y + 1 P r 1y 3/1 , vail
w here Pr is the Prandtl num ber and y is the specific heat ratio.
The m odified slip velocity with tem perature ju m p becom es:
28.18 FLU ID FLO W H ANDBOO K
w here the derivative in the second term is with respect to the curvilinear coordi
nate in the w all tangential direction.
G as flow s in the slip regim e have been studied by m any researchers using N avier-
Stokes solvers (com putational fluid dynam ics m ethods) for com pressible gas flows
w ith the slip boundary m odification. A m ong them , B eskok [6], Beskok et al. [5, 6]
have used a spectral elem ent m ethod for com pressible viscous flows with the first-
order slip model. A second-order slip m odel was also developed by them, which is
sim ilar to the first-order model in term s o f com plexity, but requiring an em pirical
coefficient. The solver was em ployed to study pressure-driven m icrochannel flows,
shear-driven m icroflow s, thermal creep and separated m icroflow s in m icrochannels
w ith steps. Som e o f the results were com pared w ith those o f the D SM C sim ula
tions for validation o f the m odels. In particular, it w as found that com pressibility
w as dom inant in pressure-drive m icroflow s in m icrochannels, which was negated
by rarefaction effect. This is consistent with the experim ental observations in low
pressures presented in [91] and [76]. On the other hand, in shear-driven flow s, rar
efaction w as found to be a more dom inant effect than com pressibility.
M aureau et al. [68] developed an analytical infinite-series solution to the Stokes
equation with slip boundary condition for the study o f m icro flow in a m icro jo u r
nal bearing. And the infinite-series solution was com pared with num erical solution
for verification. It w as found that the m icrobearing is characterized by the possi
bility o f a recirculation zone which can be a dom inant feature in the flow field. This
is a very different phenom enon from that o f m acrobearings. It was also found that
the load-bearing and the frictional torque are both affected by eccentricity and
slip factor: both increase with the eccentricity but decrease with the slip factor.
Z hao et al. [105] have developed a m ethod w hich com bines a new dynam ic
unstructured m esh m ethod with a high-order upw ind finite-volum e solver with a
im plicit dual tim e stepping schem e and a first-order slip w all boundary to study
com plex m icro gas flow s, in particular, the surface effects w ith m oving surfaces
in M EM S devices. T his new m ethod w as found to be very flexible in dealing with
com plex boundaries, and efficient for solving the fluid-structure interaction prob
lem s at m icro scale in the slip regim e. A fter the validation o f the m ethod, m icro
flow around an oscillating cylinder and flow in a com plex m icro pum p w ere ana
lyzed by the m ethod under various operating conditions. It was noted that there
w as significant reduction in drag due to the use o f the slip boundary condition in
all cases studied, w hich may affect the natural frequency o f devices w ith an oscil
lating elem ent o r the perform ance o f m icro pumps.
T he com pressible N avier-Stokes equations w ith slip w all conditions can also
be used to derive the generalized com pressible R eynolds equation, w hich is used
to describe the m ovem ent o f a gas thin film betw een tw o solid surfaces. B ased on
the assum ptions o f negligible inertial and body forces, lam inar flow, N ew tonian
M ICRO AND NANO FLOW S 28.19
viscosity and small distance betw een the tw o surfaces, the non-dim ensionalized
form o f the generalized R eynolds equation can be written as [86)
_a_ 7 )P cIP
Q P H ’ ------- A PH +• Q P H 3------- A PH = a — [P H J
dx dX dY dY v dTl J
(28.16)
where P = p/pa<P is the non-dim ensional pressure and pa the am bient pressure; H
= h/hm, h is the local surface separation and hm the m inim um separation; T = t!tr
is the non-dim ensional tim e and tr the reference time; X = x!L and Y = y /L , x and
v are the x and y coordinates and L is a reference length; A v = 6nU LI(pah2m) and
A y = 6/uVL/(pahm) are the so-called bearing num bers in x and v directions and U
and V are the sliding Avelocities in x and *y directions; <JV = \2 jjL 2/(Pah2tltr) is the
squeeze num ber and Q is the Poiseuille flow factor, which can have different val
ues corresponding to different types o f slip flow conditions:
Kn
4. Q = f , Fukui-K aneko slip m odel [35].
PH
The above-m entioned equations have m ainly been used to study thin film gas
bearings, in particular slider air bearings found in data storage devices, such as hard
disk drives. Alexander et al. [110] showed that the Fukui-K aneko slip model, which
is based on the linearized B oltzm ann equation, is applicable to air bearings with
nanom eter spacing by com paring the solution o f the generalized R eynolds equation
with the slip model with that o f D SM C m ethod. A nother study o f nano gas flow s
between a gas bearing slider and a rotating disk was also conducted by Huang et al.
[54] and the predictions o f a 3D D SM C m ethod and the generalized Reynolds equa
tion with Fukui-Kaneko slip model were com pared. It was also found that the results
from the tw o methods agree well with each other for Knudsen num bers as large as
35, which corresponds to a m inim um spacing o f 2 nm. Hu [52J and Hu and Bogy
[53] have developed a com putationally efficient A dditive C orrection (A C ) based
structured multigrid control-volum e m ethod for num erically solving the general
ized Reynolds equations with the Fukui-K aneko slip model. The m ethod was used
to study air bearing in a R ead-R ite tripad slider and a Headway ABB slider, w hose
nominal trailing edge center fly-heights are only 40 nm.
Figure 28.13 illustrates the bearing surface and the corresponding calculated air
bearing pressure o f the H eadw ay ABB slider using the m ethod developed. C om -
28.20 FLU ID FLO W HANDBOOK
W i d th (m m ) Length (mm)
pared with its single-grid counterpart, the m ultigrid solver could lead to saving o f
C PU tim e by a factor o f 3.9 to 39.7, depending on the type o f slider studied, grid
size and bearing num ber. The larger the grid size and bearing num ber, the better
the perform ance o f the m ultigrid solver. Later, Wu et al, [104] em ployed the Full
A pproxim ation Storage (FAS) M ultigrid method o f Brant [14] and an unstructured
grid upw ind finite volum e schem e to solve the air bearing problem o f hard disk
MICRO AND NANO FLOW S 28.21
drives, also based on the generalized Reynolds equation. The non-nested m ultigrid
technique was used to speed up the convergence rate o f the num erical solution,
w hile a high-order upw ind scheme was adopted for higher accuracy and to prevent
instability in high pressure gradient regions.
U pw inding was introduced using the R oe’s flux difference splitting and high-
order accuracy was achieved by linear reconstruction with flux limiting. The unstruc
tured finite volum e schem e ensures that the schem e is conservative and the grid
can easily handle com plex slider design with com plicated geom etrical features and
can be adapted according to the gradients o f the flow field variables for high accu
racy. T he m ethod has been used to study slider air bearing in the IBM Travel star
25 GB hard disk drive. This method was shown to be even better than the structured
grid (or rectangular m esh) solver in term s o f perform ance when the slider design
w as com plicated.
28.5.4 The Probabilistic Method f o r Gaseous Flows. Direct Sim ulation M onte
C arlo is a probabilistic num erical m ethod which uses sim ulated m olecules to rep
resent a large num ber o f real m olecules and com bines the direct calculation o f the
trajectories o f the sim ulated m olecules with a probabilistic approach for sam pling
and selecting the m olecules for collision calculations. It was first developed by
Bird [9] using the classical kinetic theory o f gases for collision calculations, which
is adequate for m onatom ic gases. It has been mostly used to calculate rarefied
atm ospheric gas flows. With the recent developm ent o f M EM S and NEM S (nano
electro-m echanical system s), it has been em ployed as a sim ulation tool to calcu
late rarefied gas flow s in these system s, ranging from slip, to transitional and free
m olecular flows. The classical elastic model, how ever, is less adequate for diatom
ic and polyatom ic gases and phenom enological m odels, such as the one proposed
by B orgnakke and Larsen [64], can be em ployed. The m ajor assum ptions o f DSM C
are m olecular chaos and dilute gases. The com putational approxim ations intro
duced by the D SM C m ethod include the use o f sim ulated m olecules to represent
large num bers o f real m olecules, uncoupling o f the m olecular m otion and colli
sions o v er small and finite time steps, probabilistic collision m odels, and the use
o f finite cells and subcells for sam pling and m olecular collision calculations.
A lthough progress has been made in the direct numerical solution o f the Boltz
m ann equation, direct sim ulation (D SM C ) is still the preferred and the only prac
tical choice for alm ost all problem s o f engineering interest. By com parison, the two
are based on alm ost the same assum ptions and physical reasoning: m olecular chaos
and dilute gases except that the D SM C does not depend on the existence o f inverse
collisions. For m icro and nano gas flow s, statistical scatter is a prom inent feature
and very large sam pling size may be required to elim inate it due to the low M ach
num ber o f the flow s com pared with large therm al velocity. In fact, the scatter is
directly related to the inverse square root o f the sam ple size. Thus it w ould require
excessive am ount o f com puting tim e to get rid o f the scatter. R ecently a so-called
inform ation preservation (IP) technique was proposed [29, 30, 31] to address this
issue for micro gas flow s at low M ach num bers. The technique assigns a m olecu
lar velocity as in norm al DSM C com putation and a stream velocity w hich can be
28.22 FLU ID FLO W HANDBOOK
considered as the collective velocity o f a large num ber o f real m olecules. And a
separate procedure was proposed to com pute the stream velocity and it was used
to calculate all the m acroscopic properties. This has resulted in significant
reduction in com puting tim e as a result o f reduced sam ple size.
Collision Models
D uring each tim e step the particles are allow ed to move throughout the domain and
collide with other particles. These collisions are handled on a probabilistic basis.
From w ithin each com putational cell or sub-cell, potential collision partners are
random ly selected w ithout regard to their relative positions. A collision is simulated
if their collision probability exceeds some random fraction. This process is repeated
until the correct collision rate is obtained in each cell. O nly binary collisions are
considered as the rarefaction o f the gas m akes any three-body collisions highly
im probable. O ne o f the most efficient m ethods for selecting the collision pairs is
B ird’s N ew T im e C ounter (N TC) schem e [9, 10]. O nce tw o particles have been
selected for a collision, the collision is calculated by m odifying the particle veloc
ities. The post-collision velocities are determ ined on the basis o f the collision model
chosen. The H ard-Sphere (HS) model essentially treats m olecules as a fixed diam
eter sphere w ithout any force field beyond its physical boundary. The force field
becom es effective only when the m olecules com e into physical contact. This hap
pens when the distance betw een the tw o is less or equal to
r = - ( d x + d 2) = dn (28.17)
w here d\ and d2 are the diam eters o f the m olecules. The collision cross-section
can thus be set by the diam eters o f the tw o m olecules
(28.18)
The im portant point to note here is the independence o f the collision cross sec
tion from the relative velocity Vr. Also, all directions o f the post-collision veloc
ity V; are equally likely (see Figure 28.14 for visual illustration o f the relations
am ong the param eters). The post-collision velocity can thus be calculated as
w' = c o s ( £ )w r - rr wr c o s£ - u rw r s i n e ) / ( v 2 + w 2 ) 2 (28.21)
MICRO AND NANO FLOW S 28.23
where £ is the angle betw een collision and some arbitrary reference plane, x *s the
scatter angle. C ollectively, they are known as the impact param eters and are given
by Equations 28.22 and 28.23. For details, readers are referred to [9|.
£ = 2 k - R[ (28.22)
* = 2 cos-'(/?*) (28.23)
where /?* and /?* are two random num bers uniform ly distributed betw een 0 and 1.
The HS model is sim ple but its isotropic scattering law is not realistic and its
cross-section is independent o f the relative velocity in the collision, w hich, in real
gases, should decrease as the relative velocity increases. This is related to the change
o f the coefficient o f viscosity with tem perature. This led to the introduction o f the
V ariable Hard Sphere (V H S) model by Bird [11]:
\v
( V r.ref
d=d
V K (28.24)
T he deflection angle is calculated using the sam e form ula as in the HS model
w hile the cross section is now a function o f relative velocity. H ow ever, there is
still a deficiency in the m odel because the ratio o f the m om entum to viscosity cross-
section varies w ith the inverse pow er law m odel, w hich is different from the real
gas values. To overcom e this problem , K oura and M atsum oto [62,63] proposed the
V ariable Soft Sphere (VSS) model, in which the diam eter varies with relative veloc
ity as in the VHS m odel, but the deflection angle o f the collision is different from
the HS model:
w here bid is represented by a random num ber uniform ly distributed betw een 0 and
1 in actual num erical sim ulation.
For polyatom ic gases, internal states have to be considered and the total energy
o f a m olecule is reassigned betw een the translational and internal m odes. And the
collision betw een the m olecules are thus considered as inelastic. The m ost widely
used and efficient m ethod o f dealing w ith internal states is the phenom enological
m odel proposed by B orgnakke and Larsen [64]. The m odel considers the energy
exchange between the internal energy and translational energy by artificially includ
ing the energy exchange in a m onatom ic elastic model. The relaxation rate is mainly
controlled through a fraction o f collisions that are considered to have energy ex
change w hile the rest are considered as purely elastic.
Gas-Surface Interaction
Particles colliding w ith a w all can be reflected diffusely with a reflected velocity
that is related to the wall tem perature or they can be reflected specularly. In the spec
ular reflection m odel, the collision o f the gas m olecules with the surface is co n
sidered to be perfectly elastic. The m olecular velocity com ponent norm al to the
surface is com pletely reversed, w hile that parallel to the surface rem ains unchanged.
T herefore there is no energy exchange betw een the m olecule and the surface. This
m odel, although m athem atically sim ple, does not accurately describe the interac
tion. In the diffuse reflection m odel, the velocity o f each m olecule after reflection
is independent o f its initial velocity. It is possible to use the therm al accom m oda
tion coefficient to determ ine the reflected tem perature and then the post-collision
velocity is sam pled from the equilibrium velocity distribution at this tem perature.
The therm al accom m odation coefficient can be estim ated from theories such as the
B aule m odel [44]. It is generally accepted that any physical m odel for the inter
action o f a m olecule w ith a surface should satisfy the reciprocity condition. This
m eans that the probability distribution o f a particular set o f incident and reflected
m olecules is the sam e as that o f the inverse interaction. C ercignani and Lam pis
[17] produced a gas-surface interaction m odel that satisfied this condition. L ater
this m odel w as further developed by Lord [66, 67].
M ICRO AND NANO FLO W S 28.25
A com prehensive review o f recent advances and applications o f the DSM C has been
given by Oran et al. [72]. Som e o f these include sim ulation o f thin film growth and
etching processes over m icro structures on silicon substrates in rarefied gas envi
ronm ent, typically found in sem i-conductor (m icrochip) m anufacturing processes
[ 12, 55, 56, 71, 88, 89], as well as sim ulation o f gas flows in or around microsystems
[6, 75]. The sem iconductor fabrication involves tw o im portant processes, which are
thin film deposition and etching on silicon or oth er substrates.
As T able 28.1 illustrates, typical m icro and nano gas flow s over a m icron-w ide
trench in sputter deposition, plasm a chem ical vapor deposition (PC V D ) and low
pressure chem ical vapor deposition (L PC V D ) are free m olecular (collisionless),
while flow s in atm ospheric-pressure chem ical vapor deposition (A PCV D ) are transi
tional. Thus the D SM C method is really an ideal tool for sim ulating the gas flows in
these processes. Ikegaw a et al., [55, 56] em ployed a dynam ic structured grid DSM C
approach to sim ulate film grow th o v er a trench in sputter deposition, A PC V D and
PC V D processes, w here the incident m olecules com ing from upper region o f the
trench is assum ed to have a M axwell velocity distribution and the film grow th is cal
culated with a string model [114]. D eposition profiles were predicted and the effects
o f sticking coefficient and trench aspect ration on profiles w ere investigated. The
results generated are useful in the optim ization o f reactor operating param eters
and im provem ent o f production yield and device reliability. Singh et al. [88, 89]
have developed a novel dynam ic unstructured grid D SM C m ethod for the study o f
gas flow s and thin film grow th in sim ilar deposition processes. Such an approach
is found to be very flexible in dealing com plex structures with arbitrary shapes com
pared w ith the traditional D SM C m ethod. A nother developm ent is the paralleliza-
tion o f D SM C codes for reduced C PU tim e using task parallel and data parallel
techniques [22, 85, 88].
The most com m on parallelization strategy is to divide the flow field into zones,
using the so-called dom ain decom position m ethod, and assign different zones to dif
ferent processors on a m assively parallel com puter. The sam e copy o f code is used
TABLE 28.1 Flow Regimes for Different Deposition Processes for a Trench of O n e Micro
Meter [55, 56]
Kn 00 10 10"2 0
by the processors but they only calculate their own zones, using the so-called Single
Program M ultiple D ata (SPM D) approach. W hen particles cross the boundaries o f
the zones, the processors can call the com m unication subroutines in a library, such
as the M PI library [41] and PVM [43] libraries, and use them to exchange data and
inform ation about these particles. The D SM C m ethod has also been extended to
m ake use o f unstructured [22, 85] and moving grids [88] in order to tackle com plex
flows with com plicated boundary conditions, such as those encountered in thin film
deposition and etching which lead to changes in the shapes o f the boundaries o f the
flow fields.
28.6 Theory and Numerical Methods for Dense Gases and Liquids
By com parison with the study o f dilute gas flows, it is m ore difficult to classify
incom pressible flows in liquids and dense gases into different flow regimes. This is
because the m ean free path can be difficult to define and there is no rigorous th e
oretical foundation for the study o f liquids at m olecular scale apart from m olecu
lar sim ulation (M S). In addition, experim ental m easurem ents are difficult and solid
conclusions have yet to be drawn on the major characteristics o f m icro and nano liq
uid flows. As a result, m olecular simulation is the most com monly used tool although
it is still not efficient enough for analyzing flow regions o f realistic sizes.
M olecular simulation includes Monte Carlo (M C) and m olecular dynamic (M D)
sim ulation. It involves the determ ination o f m olecular coordinates and the calcula
tion o f interm olecular forces or energies. Thus it provides num erical results based
on a theoretical m odel o f m olecular behavior using num erical com putation on co m
puters. And the results can then be used by statistical mechanics to determine m acro
scopic properties o f fluids. M etropolis et al. [1] (1953) introduced the concept o f
the M C sim ulation. Alder [1] (1957) was the first to develop the M D sim ulation
m ethod. In a M C sim ulation, trial configurations o f m olecular system s are ran
dom ly generated. A nd the interm olecular interactions and their energies in the g en
erated configurations are evaluated. If the energy o f a new configuration is low than
that o f the old one, the new configuration will be accepted. If not, the probability
based on the energy will be calculated and com pared with a random number. The
configuration is rejected if the random num ber is greater than the probability, other
w ise it will be accepted. This m ethod is not used for the sim ulation o f m icro and
nano flow s because it can not provide the trajectories o f m olecules in the flows. In
a M D sim ulation, atom s in molecules are treated as point mass and their motions are
calculated determ inistically based on the N ew ton’s second law and inter atom ic
force field. Thus the coordinates and m om enta o f atoms and m olecules are obtained
determ inistically. M D simulation is often used to evaluate dynam ic properties (tim e
dependent) and M C sim ulation is mainly used for equilibrium properties, although
the form er can also be used to calculate equilibrium properties.
28.6.1 Force Fields. The m icroscopic model for dense gases and liquids is based
on spherical particles or atom s and groups o f atom s or m olecules interacting w ith
MICRO AND NANO FLO W S 28.27
each other. The interactions can be described by potential functions. The sim plest
model is the m onatom ic model which treats every atom as a spherical particle.
And the best known potential function for such model is the L ennard-Jones (LJ)
potential for the van der W aals interaction, which has been used for the study o f
liquid argon and other m onatom ic liquids. For a pair o f atom s i andy, the potential
energy due to the van der W aals force is
/ \ ,z
G a
V( r ij) = 4 e (28.26)
\ ro y •J 7
where = f( - ?j is the distance betw een the tw o atom s, a is the collision diam
eter (the separation with zero energy) and e is the well depth w hich represents the
m inim um energy o f the potential. Thus the van der W aals force that atom j exerts
on atom i is
( \ 14 ( \ 8“
o 1 o
fij = - V V ( r ij) = \ (28.27)
U 2 J < r ‘j >
2
For polyatomic molecules, valence bonds exist. For small molecules with strong
valence bonds, they are norm ally considered rigid m olecules and intra-m olecular
interactions need not be calculated. On the other hand, if the m olecules are long
and flexible, then these internal bonding forces will have to be calculated and
included in the N ew tonian dynam ical equations.
If the m olecules being sim ulated is considered as rigid m olecules, we only need
to calculate the totalforce acting on the m olecule, the translational m otion o f the
center o f m ass, the torque about the center o f m ass and its angular velocity. If the
force exerted on atom a o f m olecule i by atom f i o f m olecule j is f ajp* then the
total force acting on m olecule i is
^ = <2 8 -2 8 )
j p a
f , = ' L ( f ia - R , ) x L (28.29)
a
1 ^
w here /?, = — > mia ria is the center o f m ass o f m olecule i.
Mi a
28.28 FLU ID FLO W HANDBOOK
MfR. = (28.30)
Flexible Molecules
F or flexible m olecules, the bonded interactions betw een the atom s o f a m olecule
are considered and theses include bond stretching, angle rotation or torsion, angle
bending and out-of-plane bending.
In M D sim ulation, bonds in m olecules are defined by the types o f atom s they
link. B onds resist both stretching and com pression that lead to deviations from
their equilibrium lengths. The M orse potential is an exam ple o f bond potentials,
that accounts for the effect o f bond stretching:
w here A ^is the stretching stiffness, / 0 is the equilibrium bond length, / is the actual
bond length and a is a constant with a invariant value o f 2 x 10,() m '1. V arious
bond-stretching param eters for some com m on bond types can be found in [25].
Bond torsion is rotation about a bond w hich involves four atom s. O ne o f its
potentials is
w here cois the torsion angle. The param eters (v„ i = 1,2,3) in this expression for
som e com m on torsion types and bonds are given in [25].
B ond angle-bending is due to the forced change in the angle o f tw o bonds that
share the sam e atom . The potential for bond angle-bending can be w ritten as
Ve = K- ( e 2 - [' + a 9( f l- f l„ ) 4] (28.34)
M ICRO AND NANO FLO W S 28.29
In this expression, G{) is the equilibrium bond angle, 6 is the current bond angle,
Ke is the angular spring constant, and a 6 is a constant taken to be 0.754rad-4. Bond
angle-bending param eters for some com m on bond types can also be found in [25].
O ut-of-plane bending occurs in planar m olecules with three bonds sharing an
atom. The energy o f out-of-plane blending com es from the displacem ent o f the
trigonal atom above or below the m olecular plane. This is given by
Vx = \ K x( X - X o ) 2 (28.35)
where %0 is the reference angle for the bending and Kx is a constant, x is the angle
between the bond and the plane that represents the displacem ent.
If the atoms and m olecules are not electrically neutral or if they show electric polar
ity, then the static electric interactions will be calculated by using an arrangem ent
o f fractional point charges throughout the m olecule. These charges are designed
to reproduce the electrostatic properties o f the molecule. Thus the electrostatic inter
action betw een tw o m olecules or betw een different parts o f the sam e m olecule is
then calculated as a sum o f interactions betw een pairs o f point charges according
to C oulom b’s law:
N' N> q q
Klecros.anc = X X (28.36)
w here N | and N 2 are the num bers o f point charges in the tw o m olecules. Som e
force fields also use a m odified L ennard-Jones term s betw een hydrogen-bonding
atom s, w hich can be described using a 10-12 Lennard-Jones potential [65]:
I , / - x A 8
VH(nj) = — - — ■ (28.37)
rij rij
The force com putations can be broadly grouped into tw o categories: bonded force
com putations and non-bonded force com putations. The form er includes the calcu
lations o f bond stretching, angle bending, bond torsion and out-of-plane bending
which involve tw o to four bonded atoms. The com putational com plexity o f bonded
force com putations is O (N ) where N is the num ber o f atoms. The latter involves
the calculations o f van der W aals and electrostatic interactions w hich constitutes
betw een 80 to 95 percent o f the overall com putation. If direct calculations are
28.30 FLU ID FLO W HANDBOOK
em ployed, its com putational com plexity is 0(/V 2). As a result, many m ethods have
been devised to achieve linear scaling with N in com putational com plexity for non
bonded force com putations. N on-bonded forces decrease as the distances between
atom s increases. It is possible to use a cutoff radius and a neighbor list o f the inter
acting atom s based on the cutoff radius to calculate forces in order to save com pu
tation time. H ow ever, the saving may not be significant for large system s because
it is still necessary to calculate the distance between every pair o f atom s in the sys
tem , in order to determ ine if they are close enough to be w ithin the radius. Thus
calculating all the N(N - 1) distances is alm ost as tim e consum ing as calculating
the energy o r the forces.
Verlet [100] suggested an improved method which m aintains a list o f the neigh
bors o f every atom and the list is updated at intervals only. Such m ethod is still not
efficient for large system s. A more efficient method for keeping track of the neigh
bors is the cell index method, first suggested in [50, 80]. The cubic sim ulation box
is subdivided into M x M x M cubic cells and the length o f the cells is greater than
the cu to ff distance. Thus the calculation o f forces only involves the m olecules in
the neighboring cells which is a fast process. The above m ethod is norm ally used
for calculating van der W aals interaction which is considered as a short range force.
A long range force is defined as one in which the interaction decreases no faster
than r d w here d is the dim ensionality o f the system . By this definition, the elec
trostatic interaction is a long range one and the above m ethod is not applicable.
Ewald Sum
The Ewald sum w as originally designed for the study of ionic crystals [ 111 ]. It is a
technique for efficiently calculating the interactions betw een atom s w ith a charge
in a sim ulation box, and their interactions with all o f their im ages in an infinite
array o f periodic cells as well. The total interaction energy can be calculated as
QiVj
(28.38)
ru + n\
w here n = (nxL,nyL,nzL) and L is the length o f the box and nx, ny and nz are inte
gers. The Ew ald sum is to reorganize this sum into sums o ver concentric spheri
cal shells, assum ing charge neutrality (i.e.: X/<7/ = 0 ):
erfc(a I r.xj + Lh l) \ K 2 1n I2 2m _ _
2 r 2 •+ -----n • r;;
nj + Lfi kL n*0 a L L
1 erfc(aL In I) | 1 La
+— ^ rre x p 2 r2 1/2
(28.39)
2 n *0 V
L\n kL In h a L K j* i
MICRO AND NANO FLOWS 28.31
Other methods tor calculating long range forces include the reaction field method,
particle-particle and particle-mesh (PPPM) algorithm which have been summarized
by Allen and Tildesley 11].
N
m ,n = X % (28.40)
j=\
( j *i )
Boundary Conditions
each other closely. In the worst case scenario, the integration of the dynamic equa
tions breaks down and the conservation laws are violated. The easiest way to avoid
the above problem is to adopt a known configuration obtained from experiment,
which is known to be close to equilibrium. However, for liquids, this is normally
not available at all. Usually a lattice method can be adopted. For example, mole
cules can be placed at the nodes and the face centers of cubic structures in a lattice,
resulting in 4M 3 molecules with A/3 being the number of cubes or lattice nodes.
The lattice spacing should be so chosen that the initial configuration results in the
desired liquid density.
The initial velocities of all the molecules can be specified from the Maxwell-
Boltzmann distribution (it is a Gaussian distribution) at the temperature given tor the
simulation. This can be realized by using random numbers that are uniform in the
range of 0 to 1 to obtain those with the Gaussian distribution. And the velocities can
then be chosen from the latter set of random numbers so that the probability den
sity p(v) of the three velocity components vik (i = 1,2 and 3) for molecule k is
1/2
m, "hv>k
P(vik) = exp (28.41)
2 k k hT 2k J
Given a random number Rik from the Gaussian distribution with zero mean and
unit variance, each random velocity component is set to
1/2
exp (28.43)
2K K j j 2k"b T j
where /,* is angular inertia for molecule k in i direction. Similarly each random
angular velocity component is calculated as
\ K ,T
©a = Rik (28.44)
r ( t - A t ) - r ( t ) - M ? ( t ) + ( A r I 2 )r (t) + ■■■
MICRO AND NANO FLOWS 28.33
Thus adding the above two equations and neglecting the higher-order terms gives
Subtracting the two equations and neglecting the higher-order terms we obtain
the velocity as
It is obvious that it is difficult to obtain the velocity in the Verlet scheme due to the
lack of an explicit velocity term and it can only be calculated after the new positions
at the next step are known. In addition, the first two terms in Equation 28.45 are
large while the last term is usually small, which may result in loss of precision.
The leapfrog scheme uses the following relationships based on the central dif
ference in time:
In the Leapfrog algorithm, velocities leap-frog over positions and positions also
leap-frog over velocities and the two are not synchronized. Thus it is difficult to
calculate total energy because the potential energy and the kinetic energy can not
be calculated at the same time.
The Beem an’s method [4] uses a more accurate formula than the above two
schemes for calculating velocity (it is accurate to 0(A f3) invelocity) and avoids
the difficulties faced by the above methods:
28.6.4 Parallel Computation. To simulate large systems of micro and nano flows,
the most viable approach is to employ parallel computers which have multiproces
sors with local or distributed memory. And the processors can com municate with
each other over a network, using the message-passing method. M essage-passing
can be performed using standard communication libraries, such as MPI [41] and
PVM [43]. For this type of simulation, the Single Program Multiple Data (SPMD)
programming model is also adopted. Input/output and certain process dependent cal
culations can be assigned to certain processors only using conditional statements.
28.34 FLUID FLOW HANDBOOK
28.6.5 Statistical Mechanics. One of the objectives of micro and nano flow sim
ulation is to derive its average macroscopic physical quantities and their relations,
such as velocity, pressure, temperature and density. Statistical mechanics is the
theory that bridges the microscopic random motions of fluids with their macro
scopic behavior.
Average velocity distribution in the flow field can be calculated by construct
ing a background grid in the field and sampling the particle velocities in the grid
cells at fixed time intervals.
Pressure is calculated by the virial theorem of Clausius, which states that the
virial is equal to - 3 N k bT. And the virial is defined as the sum of the products of
the coordinates of the particles and the forces acting on them. For an ideal gas,
the only forces are those between the gas and the wall of the container. It is well
known that the virial is -3 p V = - 3 N kb T, thus, p V = N kbT. For real gases and liq
uids, the forces between the particles affect the virial which is the sum of the ideal
gas part and the contribution due to the interactions between them:
Virial = - 3 p V + ^ f i rv - f v (28.51)
1=1 7=i'+I
N -1 N
P= (28.44)
V 3 ;= | j - j +1
N
=I
The internal energy can be calculated as the ensemble average of the energies
of the states in the course of a simulation:
u = (E ) = 4 j ' L Ei (28.54)
( lr ( f ) - r ( 0 ) l2)
D = lim -------- -------- (28.55)
6/
f N N
X X ( 0 - X m^ ( ° K <w
1 \x<y V i=\ 1=1 y .
V = 7} ™*im ------------------------- ----------------------------L (28.56)
Vk„T <-»“ 61
where ^ denotes the summation of the three pairs of the vector components: xy,
x < y
{ Y ( 5 e a( t ) - 8 e am 2\
A = — — l i m - ^ ------- ---------L (28.57)
k bT '->°° 61
where
V i=I
E = i L + i y Vfr)
2m, 2 £
w h ere a denotes the x, y, and z directions.
28.36 FLUID FLOW HANDBOOK
28.6.6 Recent Applications and Developments. Simple models for liquid water use
electrostatic and van der Waals interactions to reproduce the hydrogen bonding and
the molecules are normally assumed rigid. The TIP3P [58] and SPC [112] models
have three sites for the electrostatic interaction: negative partial charge at the oxygen
atom and positive partial charge at the two hydrogen atoms which balance the neg
ative charge at the oxygen atom. The van der Waals interaction between two water
molecules is only calculated between the two oxygen atoms only. The TIP4P [58]
model shifts the negative partial charge away from the oxygen atom to a point along
the bisector of the HOH angle towards the hydrogens. (See Figure 28.15 for a sche
matic representation of the models). Parameters for the above models are summa
rized in the following table (Table 28.2) for comparison. Further discussion of other
models can be found in [65].
Some force fields adopt explicit hydrogen-bonding terms to account for hydro
gen bonding in water. This can be done using a 10-12 Lennard-Jones potential [65]:
(28.58)
(28.59)
where 0 is the angle subtended at the hydrogen by the two bonds that link the hydro
gen atom to the donor and acceptor atoms. Recently Tuzun et al, [99] performed
molecular dynamics simulation of helium and argon flows in nanotubes of various
sizes. The fluid molecules were given some initial velocity at the start of the sim
ulations. It was shown that argon would slow down more quickly than helium. The
characteristics of fluid flows in the nanotubes was also found to be dependent on the
rigidity and diam eter of the tubes, as well as fluid density. A less rigid tube would
result in faster slowdown of the fluid flow in the tube. On the other hand, Walther
et al, [102] studied the hydrophobic/hydrophilic behavior of carbon nanotubes in
q(0)
water using molecular dynamics simulations. Similar to the study of Tuzun et al.,
the carbon nanotube was modeled by a Morse bond, a harmonic cosine angle bend
ing and a two-fold torsion potentials. The water molecules were described by a flex
ible water model ot Teleman et al. [951, which includes harmonic bonds between
the oxygen and hydrogen atoms and their partial charges. Non-bonded interactions
between the water molecules were modeled by a Lennard-Jones potential between
the oxygen atoms only. The carbon-water interactions were described by a Lennard-
Jones term between the carbon and the oxygen atoms and a quadrupole interaction
between the carbon atoms and the partial charges on both the oxygen and hydrogen
atoms. The nanotubes were immersed in water subject to periodic boundary condi
tions. The governing Newton’s equations were integrated in time using the leap frog
scheme. The simulation of nanotubes in water at 300 K revealed a continuous wet
ting and drying of the space between the tubes, which was strongly affected by the
initial spacing between them. A spacing of 7 and 8A resulted in a drying while a
spacing o f 9 A resulted in a permanent wetting of the interstice.
Greenspan [ 4 0 ] used molecular dynamics simulation to study turbulent water
flow at 15° Celsius in a three-dimensional cavity on molecular scale. The model for
water molecules was a simple Lennard-Jones potential with the molecules treated as
point particles. The particle positions and velocities were obtained using the leap
frog scheme for integration of the Newton’s equations. The molecules were initially
placed at lattice points at an equilibrium interval of 3 .0 6 A for water at 15°C for a
cavity o f 1 0 4 .4 A x 1 0 4 . 4 A x l 0 4 . 4 A . This resulted in the use of 4 2 , 8 7 5 molecules
in the simulation. The top of the cavity was moved at a constant speed to drive the
fluid flow inside the cavity while the five other sides were considered solid walls.
This system was simulated at a Reynolds number of 2 3 . Higher Reynolds numbers
of 3 2 0 0 and 6 0 0 0 were also used to study the turbulent flow with smaller numbers
of molecules in order to reduce computational time. At such Reynolds numbers, it
was shown, by using the velocity vectors and particle trajectories, that turbulence
developed very rapidly in the cavity.
28.38 FLU ID FLO W HANDBOOK
A series of study has been conducted to study fluid flow and heat transfer
phenomena and hydrodynamic instability from the atomistic viewpoint using two-
dimensional molecular dynamics simulation based on a Lennard-Jones fluid model
[47, 81, 82, 83]. One of the conclusions from these simulations is that they can
remarkably reproduce the flow structures that resemble their macroscopic coun
terparts. For example, the problem of thermal convection within a square cavity
enclosed by solid walls was simulated using the MD method [83]. The fluid was
heated from below and cooled at the top with an estimated Rayleigh number of
78,000, and 60,000 particles were used in the simulation. The resulting interaction
between the buoyant upward flow and the downward flow o f the cooled fluid due
to gravity leads to the formation of time-dependent roll patterns inside the cavity,
as shown in Figure 28.16. It is noted that when atoms collided with the walls, they
were assigned a new velocity with magnitude determined by the wall temperature
whilst they were specularly or diffusely (with random reflected velocity direction)
reflected from the wall, depending on whether the slip or nonslip condition was used.
It was found that the two conditions did not have significantly different results. To
generate the averaged flow pattern, a background grid was generated in the flow
field and measurements at grid cells were recorded at fixed time intervals.
The second example is fluid flow over a circular obstacle shown in Figure 28.17.
The initial flow field was specified by a stream velocity superimposed on a random
thermal velocity and the obstacle was represented as a ring of fixed atoms identi
cal to those in the flow field. Figure 12.17(a) shows the streamlines at t = 80 when
a pair of counter rotating vortices have formed. Figure 2 8 .17(b) shows the veloc
ity vectors at a later stage (t = 2400) when vortex shedding and oscillatory wake
regions appear. A large number of particles (160,000) was used in order to pro
vide for adequate space in the flow field and to achieve a high enough Reynolds
num ber of about 25. The last example is a MD simulation of Taylor-Couette vor
tex formation within two concentric cylinders rotating at different speeds [47].
The Lennard-Jones potential was also used to calculate the interactions between
particles and a nonslip wall was employed and for the rotating wall the local wall
velocity was added to the reflected velocity. The number of particles was 160,000.
Figure 28.18 gives a sequence of the snapshots of the development ot the vortices
in terms of streamlines. A set of four counter-rotating Taylor vortices are formed,
(a)
(b)
F IG U R E 28.17 Flow over a obstacle [81. 8 2 1.
in agreement with theory and experiment on macroscopic scale. Due to the lack
of experimental measurements at micro and nano scales, validation o f the MD
methods by directly comparing their results with the corresponding measurements
is still impossible for most cases.
Recently Schlick et al, [87] gave a review of the algorithmic challenges in MD
simulation applied to molecular biophysics, such as structural prediction of protein
molecules in solvents (water). Included in the review are recent algorithmic work
28.40 FLU ID FLOW HANDBOOK
I-350 Ia45U
• = 630 t « 700 t ■ MO
a Speed of sound
A, Slider bearing number in x direction
A, Slider bearing number in y direction
a Constant in bond stretching potential
a 0 Constant in the angle-bending potential
d Diameter of an atom
E Total gas internal energy
f Normalized distribution function
fij Force exerted on atom i by atom j
Fk Force on particle k due to all other particles
h Local slider spacing
H Non-dimensional slider spacing or total gas enthalpy
F Force
Kn Knudsen number
*7; Boltzmann constant
L Length
I Bond length
Ma Mach number
N Number of particles
n Number density
Nc Number of degrees of constraint
CO Bond torsion angle
P Pressure or momentum
Pi Momentum of molecule i or atomic site relative position
Pr Prantl number
/v
Q Poiseuille flow factor
R Heat transfer rate
28.42 FLUID FLOW HANDBOOK
p Density
Re Reynolds number
2 8 .7 REFERENCES
1. Allen M.P. and Tildesley D.J. 1986. Com puter Sim ulation o f Liquids. C larendon
Press, Oxford.
2. Allen R.R., M eyer J.D ., and Knight W.R. 1985. “Therm odynam ics and H ydrody
namics o f Therm al Ink Jets.’’ Hewlett-Packard Journal. 36(5):21-24.
M ICRO AND NANO FLOW S 28.43
3. Arkilic E.B. 1997. “M easurem ent o f the Mass Flow and Tangential M om entum
A ccom m odation Coefficient in Silicon M icrom achined C hannels.” PhD Thesis,
M IT. C am bridge, MA.
4. Beem an D. 1976. “Som e M ultistep M ethods for Use in M olecular Dynam ics C al
culations." Journal o f C om putational Physics. 52:24-34.
5. Beskok A. and K arniadakis G.E. 1994. “Simulation o f Heat and M om entum T rans
fer in C om plex M icrogeom etries.” J. Therm ophy.H eat Transfer. 8:647-655.
6. Beskok A. 1996. “Sim ulations and M odels for Gas Flows in M icrogeom etries.”
PhD Thesis, Princeton U niversity.
7. Beskok A., K arniadakis G .E., and Trim m er W. 1996. “Rarefaction and C om press
ibility Effects in Gas M icroflow s.” A SM E J.F luid Eng. 118:448—455.
8. Bhatia S.N. and Chen C. 1999. “Tissue Engineering at the M icro-Scale.” B iom ed
ical M icrodevices. 2(2): 131 -1 4 4 .
9. Bird G.A . 1994. M olecular G as D ynam ics and the D irect Simulation o f G as Flows.
Caledron Press, Oxford.
10. Bird G.A. 1989. “ Perception o f Num erical M ethods in Rarefied Gas D ynam ics.”
Progr. Astro. Aero. 118:211 -226.
11. Bird G.A. 1981. “ M onte C arlo Sim ulation in an Engineering C ontext.” Progr.
Astro. Aero. 74:239-255.
12. Boyd I.D. and Chen G. 1996. “Com putation o f Supersonic Flow for Thin Film
D eposition Using Expansion T hrough a Skim m er.” A SM E DSC-M EM S.
59:169-176.
13. Brenner D.W . 1990. “E m pirical Potential for H ydrocarbons for Use in Sim ulating
the Chem ical Vapor Deposition o f Diamond Films.” Phys. Rev. B. 42(15):9458-9471.
14. Brant A. 1977. “ M ulti-level A daptive Solutions to Boundary Value Problem s.”
M ath.C om put. 31:333-390.
15. Car R. and Parinello M. 1985. “ U nified Approach for M olecular Dynam ics and
D ensity-Functional T heory.” Phys. Rev. Lett. 55:2471-2474.
16. Castro A., Fairfield F.R. and Schera E.B. 1993. “Fluorescent Detection and Size
M easurem ent o f Single DNA M olecules.” /Iw//. Chem.. 65:849-852.
17. Cercignani C. and Lam pis M. 1974. R arefied Gas D ynam ics (ed. K. K aram cheti).
Academ ic Press, New York.
18. Chapm an and Cow ling. 1952. The M athem atical Theory' o f Non-uniform G ases
(2nd ed). C am bridge U niversity Press, Cam bridge.
19. Cooke M.J. and Harris G. 1989. “ M onte Carlo Sim ulation o f Thin-film deposition
in a R ectangular G roove.” J. Vac. Technol. A. 7 (6 ):3 2 17-3221.
20. Coronell D.G. 1999. “Sim ulation and Analysis o f Rarefied Gas Flows in Chem ical
Vapor D eposition Processes.” Ph.D Thesis, M IT, Cam bridge, MA.
21. Chiu D.T., Pezzoli E., W u H .K ., Stroock A.D., and W hitesides G.M. 2001. “Using
Three-dim ensional M icrofluidic N etw orks for Solving Com putationally Hard Prob
lems.” PNAS. 9 8 (6 ):2 9 6 1-2 9 6 6 .
22. Dietrich S. and Boyd I.D. 1996. “Scalar and Parallel O ptim ized Im plem entation o f
the Direct Sim ulation M onte C arlo M ethod.” Journal o f Com putational Physics.
126:328-342.
23. Drexler K.E. 1986. E ngines o f Creation. A nchor Books.
24. Drexler K.E., Peterson C. and Pergam it G. 1991. Unbounding the Future: the
N anotechnology Revolution. W illiam M orrow and Com pany, New York.
25. Drexler K.E. 1992. N anosystem s: M olecular M achinery, M anufacturing and
Computation. W iley-Interscience.
28.44 FLU ID FLOW HANDBOOK
26. Elrod S.A . 1988. “Capillary W ave Controllers for Nozzleless D roplet Ejectors."
Xerox C orporation, U.S. Patent No. 4748461.
27. Epstein A.H. et al. 1997. “Pow er M EMS and M icroengines.” IE E E Transducers
’9 7 :7 5 3 -7 5 6 , C hicago.
28. Epstein A.H. et al. 1997. “M icro-heat Engine. G as Turbines and Rocket Engines—
the M IT M icroengine Project." AIAA paper 97-1773, 28th A IA A Fluid Dynamics
C onference.
29. Fan J. and Shen C. 1999. “ Statistical Sim ulation o f Low -speed Unidirectional
Flow s in T ransition R egim e." In Rarefied Gas D ynam ics, Edited by R. Brun et al.
2:245. C epadus, T oulouse, France.
30. Fan J. and Shen C. 2001. “Statistical Sim ulation o f Low -speed Rarefied Gas
Flow s." Journal o f C om putational Physics. 167:393-^412.
31. Fan J., Boyd I.D., Cai C.P., Hennighausen K., and Candler G.V . 2001. “C om puta
tion o f Rarefied G as Flows Around a NACA0012 Airfoil." AIAA Journal.
3 9 (4 ):6 18-625.
32. Feynm an R.P. 1959. “T h ere's Plenty of Room at the Bottom ." Annual M eeting o f
APS, C altech, CA , USA.
33. Forster F., Bardell R.L., A from ow itz M .A., Sharm a N.R. and Blanchard A. 1995.
“D esign, Fabrication and Testing o f Fixed-V alve M icro-Pum ps." Proceedings o f
the A S M E Fluids Engineering Division. FED-Vol. 2 3 4 :3 9 ^ 4 , 1995, 1MECE.
34. Fu A .Y ., Spence C., Scherer A.. Arnold F.H., and Quake S.R. 1999. “A M icrotab-
ricated Fluorescence-activated Cell Sorter." Nature Biotechnology. 10:1109-1 111.
35. Fukui S. and Kaneko R. 1988. “Analysis of Ultra-Thin Gas Film Lubrication Based
on Linearized Boltzm ann Equation." ASM E Journal of Tribology. 110(2):253—262.
36. G ad-el-H ak M. 1999. “The Fluid M echanics o f M icrodevices— The Freeman
Scholar L ecture." A SM E Journal o f F luid Engineering. 121:5-33.
37. G arcia A.L. 1997. “ Direct Sim ulation M onte Carlo: Novel A pplications and Exten
sions." P roceedings o f the 3rd W orkshop on the M odeling o f Chem ical Reaction
S ystem s, H eidelberg, Germany.
38. G erstein M. and Levitt M. 1998. “Sim ulating W ater and the M olecules of Life."
Scientific Am erican. N ovem ber 1998:101-105.
39. G ravesen P., Branebjerg J., and Jensen O.S. 1993. “ M icrofluidics, A Review."
Proc. M M E 93 (M icro M echanics Europe), Neuchaiel, pp. 143-164.
40. G reenspan D. 2001. “M olecular Study o f Turbulence in Three-dim ensional Cavity
Flow ." C om put.M ethods Appl.M ech.Engrg. 190:4231-4244.
41. G ropp W ., L usk E., and Skjelluin A. 1994. Using M Pl: Portable Parallel Pro
gram m ing with the M essage Passing Interface. M IT Press, Cam bridge, MA..
42. G im elshein S., M arkelov G. and Rieffel M. 1996. “Collision M odels in the Hawk
D SM C Im plem entation." Caltech Technical Report CIT-TR-96-16.
43. G eist A., Beguelin A., D ongarra J., Jiang W., M anchek R. and Sunderam V. 1994.
PVM : P arallel Virtual M achine— A U sers' Guide and Tutorial f o r Netw orked P ar
allel C om puting. M IT Press, Cam bridge, MA..
44. G oodm an F. and W achm an H. 1976. D ynam ics o f G as-Surface Scattering. A cade
m ic Press. N ew York.
45. Harley J.C ., H uang Y., Bau H.H. and Zemel J.N. 1995. “Gas Flow in M icro-chan
nels." Journal o f Fluid M echanics. 284:257-274.
46. H irata S., Ishii Y., M atoba H. and Inui T. 1996. “ An Ink-jet Head Using
D iaphragm M icro-actuator." Proc. IEEE Micro. Electro. M echanical Systems
W orkshop. San Diego, CA , pp. 418—423.
MICRO AND NANO FLOW S 28.45
47. H irshfeld D. and Rapaport D.C. 1998. “ M olecular Dynam ics Sim ulation o f Taylor-
Couette Vortex Form ation.” Phys.Rev.Lett. 80(24):5337-5340.
48. Ho C.M . and Tai Y.C. 1996. “ Review: MEMS and Its A pplications for Flow C on
trol.'* Journal o f Fluids Engineering. 118:437—447.
49. Ho C.M . and Tai Y.C. 1998. “M icro-Electro-M echanical System s (M EM S) and
Fluid Flow s.” Annu.Rev.Flui M ech. 30:579-612.
50. Hockney R.W. and Eastw ood J.W . 1981. Com puter Sim ulation Using Particles.
M cG raw -H ill, New York.
51. Hohenberg P. and Kohn W. 1964. “ Inhom ogeneous Electron G as." Phys.Rev. B.
136:864-871.
52. Hu Y. 1996. H ead-D isk-Suspension Dynam ics. PhD Thesis. U niversity o f C alifor
nia, Berkeley, CA.
53. Hu Y. and Bogy D.B. 1998. “Effects o f Laser Textured Disk Surfaces on a S lider’s
Flying C haracteristics.” A SM E Journal o f Fluids Engineering. 120:266-271.
54. Huang W .D., Bogy D. and Garcia A.L. 1997. “Three-dim ensional Direct Sim ula
tion M onte C arlo M ethod for Slider Air Bearings.” Physics o f Fluids.
9(6): 1764-1769.
55. Ikegaw a M. and Kobayashi J. 1989. “ Deposition Profile Sim ulation Using the
Direct Sim ulation M onte C arlo M ethod.” J. Electrochem. Soc. 136( 10):2982—2986.
56. Ikegaw a M., Kobayashi J. and M aruko M. 1998. “Study on the D eposition Profile
Characteristics in the M icron-Scale Trench Using Direct Sim ulation M onte Carlo
M ethod.” A SM E Journal o f Fluids Engineering. 120:296-302.
57. Israelachvili J.N. 1986. “ M easurem ent o f the Viscosity o f Liquids in V ery Thin
Film s.” Journal o f Colloid and Interface Science. 110:263-271.
58. Jorgensen W .L., C handrasekhar I., M adura J.D., Impey R.W . and Klein M.L. 1983.
“Com parison o f Simple Potential Functions for Sim ulating Liquid W ater.” The
Journal o f Chem ical Physics. 79:926-935.
59. Kale L., Skeel R., Bhandarkar M., Brunner R., Gursoy A., K raw etz N., Phillips J.,
Shinozaki A., Varadarajan K. and Schulten K. 1999. “NA M D2: G reater Scalability
for Parallel M olecular D ynam ics.” Journal o f Com putational Physics.
151:283-312.
60. Kam isuki S., Hagata T., Tezuka C., Nose Y., Fujii M. and Atobe M. 1998. “A Low
Power, Small, Electrostatically-driven Com m ercial Inkjet H ead.” Proc. IEEE
M icro. Electro. M echanical System s, H eilderberg, G erm any, pp. 6 3-68.
61. Kosztin D., Bishop T.C. and Schulten K. 1997. “ Binding o f the Estrogen R eceptor
to DNA: The Role o f W ater.” Biophys. J. 73:557-570.
62. Koura K. and M atsum oto H. 1991. “ Variable Soft Sphere M olecular M odel for
Inverse Pow er Law or Lennard-Jones Potential.” Phy.Fluids A. 3:2459-2465.
63. Koura K. and M atsum oto H. 1992. “ Variable Soft Sphere M olecular M odel for Air
Species.” Phy.Fluids A. 4:1083-1085.
64. Borgnakke C. and Larsen P.S. 1975. “Statistical Collision M odel for M onte Carlo
Sim ulation o f Polyatom ic Gas m ixture.” J.Com pt.Phys. 18:405-420.
65. Leach A.R. 1996. M olecular M odelling, Principles and Applications. Addison
W esley Longman.
66. Lord R.G. 1991. In Rarefied G as D ynam ics (ed. AE Beylich), W einheim , G erm any.
67. Lord R.G. 1991. “Some Extensions to the Cercignani-Lam pis G as Scattering K er
nel.” Phys. Fluids A. 3:706-710.
68. M aureau J., Sharatchandra M .C., Sen M. and G ad-el-H ak M. 1997. “ Flow and
Load C haracteristics o f M icrobearings with Slip.” J. M icromech. M icroeng.
7:55-64.
28.46 FLUID FLO W HANDBOOK
69. M axwell J.C. 1897. “On Stresses in Rarefied G ases A rising From Inequalities of
T em perature." Philosophical Transactions o f the Royal Society/ Part 1. 170:231-256.
70. N akagaw a S., Shoji S., Esashi M. 1990. “ A M icro-C hem ical Analyzing System
Integrated on Silicon C hip.” Proceedings IEEE: M icro Electro M echanical Sys
tem s, IEEE 90C H 2832-4, Napa V alley, CA.
71. Nanbu K., Suetani M. and Sasaki H. 1999. “Direct Sim ulation Monte C arlo
(D SM C) M odeling o f Silicon Etching In Radio-Frequency Chlorine D ischarge.”
CFD Journal. 8(2):257-267.
72. Oran E.S., Oh C.K . and Cybyk B.Z. 1998. “ Direct Sim ulation M onte Carlo: Recent
A dvances and A pplications.” Annu. Rev. Fluid M ech. 30:403-441.
73. Peng X.F., W ang B.X., Peterson G.P. and M a H.B. 1995. “Experim ental Investiga
tion o f H eat T ransfer in Flat Plates w ith R ectangular M icrochannels.” Int. J. Heat
M ass Transfer. 38:127-137.
74. Peng X.F. and Peterson G.P. 1995. “The Effect o f T herm ofluid and Geom etrical
Param eters on Convection o f Liquids Through Rectangular M icrochannels.” Int. J.
H eat M ass Transfer. 38:755-758.
75. Piekos E.S. and Breuer K.S. 1996. “ Num erical M odeling of M icrom echanical
Devices Using the Direct Sim ulation M onte C arlo M ethod.” A SM E Journal o f F lu
ids Engineering. 118:464^169.
76. Pong K.C., Ho C .M ., Liu J. and Tai Y.C. 1994. “N on-linear pressure distribution in
uniform m icrochannels.” In A pplication o f M icrofabrication to Fluid M echanics,
A SM E W inter Annual meeting, pages 5 7 -6 5 , C hicago, 111., Nov.
77. Pfahler J. 1992. “Liquid Transport in M icron and Subm icron Size C hannels.” PhD
Thesis. U niversity o f Pennsylvania, Philadelphia, Pennsylvania.
78. Pfahler J., Harley J., Bau H. and Zam el J.N. 1990. “Liquid Transport in Micron
and Subm icron Size C hannels.” Sensors and Actuators. A 21-A 23:431-434.
79. Pfahler J., Harley J., Bau H. and Zam el J.N. 1991. “G as and Liquid Flow in Small
C hannels.” Sym posium on M icrom echanical Sensors, A ctuators, and System s, D
C ho et al. Eds., A SM E DSC-Vol. 32:49-60, A SM E, New York.
80. Q uentrec B. and Brot C.. 1975. New m ethod for searching for neighbors in m olec
ular dynam ics com putation, J. Comput. Phys. 13:430-432.
81. Rapaport D.C. 1995. “ The A n o f M olecular D ynam ics Sim ulation.” C am bridge
U niversity Press.
82. Rapaport D.C. and Clementi E. 1986. “ Eddy Form ation in O bstructed Fluid Flow:
A M olecular-D ynam ics Study.” Phys. Rev. Lett. 57(6):695-698.
83. Rapaport D.C. 1991. “Tim e-dependent Patterns in A tom istically Sim ulated C on
vection.” Phys. Rev. A 43(12):7046-7048.
84. Rahm an A., F. Ahm ed, Zhao Y. and Gong H.Q. 2000. “ A Novel Diffuser-Nozzle
M icropum p A ctuated by Therm al Bubble.” Proceedings o f SPIE Vol. 3990, S PIE ’s
7th Annual International Sym posium on Sm art Structures and M aterials: Smart
Electronics and M EM S, March 2000, N ew port B each, California, U.S.
85. Robinson C.D. and Harvey J.K. 1996. “A daptive D om ain Decom position for
U nstructured M eshes Applied to the Direct Sim ulation M onte Carlo M ethod.” P ro
ceedings o f Parallel CFD 96.
86. Ruiz O.J. and D.B. Bogy. 1990. “A Numerical Sim ulation of the Head-Disk Assem
bly in M agnetic Hard Disk: I. C om ponent M odels.” A SM E Journal ofT ribology.
112:593-602.
87. Schlick T., Skeel R.D., Brunger A.T., Kale L.V ., Board Jr. J.A., Herm ans J. and
Schulten K. 1999. “Algorithm ic Challenges in C om putational M olecular Bio
physics.” Journal o f Com putational Physics. 151:9-48.
MICRO AND NANO FLOWS 28.47
88. Singh A. and Zhao Y. 2001. “ Parallel U nstructured Dynam ic Grid D irect Monte
C arlo Sim ulation o f M olecular Gas Dynam ics and the Associated Thin Film D epo
sition.” Paper M IT -1/01. Proceedings o f The fir s t M IT conference on C om puta
tional F luid and Solid M echanics. June 11-13, 2001, MIT, Cam bridge, MA.
89. Singh A. and Zhao Y. 2000. “Parallel U nstructured Grid DSM C for the Study o f
M olecular G as D ynam ics in Sem i-conductor M anufacturing.” Paper No. 4228-37,
S P IE ’s International Sym posium on M icroelectronics and Assem bly (ISM A 2000),
27 N ovem ber-2 D ecem ber 2000, Singapore.
90. Shoji S. and Esashi M. 1994. “ M icroflow D evices and System s.” J. M icromech.
M icroeng. 4 : 157-171.
91. Sreekanth A.K. 1969. “Slip Flow Through Long Circular Tubes.” Proceedings o f
the Sixth International Sym posium on R arefied G as D ynam ics, Edited by Trilling,
L and Y. W achm an. 1:667-680.
92. Stem m e E. and Stem m e G. 1993. “A V alveless D iffuser/N ozzle-based Fluid
Pum p.” Sensor a n d A ctuators A. 39:159-167.
93. O lsson A., Enoksson P., Stem m e G. and Stem m e E. 1997. “M icrom achined Flat-
W alled V alveless D iffuser Pum ps.” Journal o f M EM S. 6(2): 161-166.
94. T orpey P.A. and M arkham R.G. 1987. “Therm al Inkjet Printhead, Xerox C orpora
tion.” U.S. Patent No. 4638337.
95. Telem an O., Jonsson B. and Engstrom S. 1987. “A M olecular Dynam ics Sim ula
tion o f a W ater M odel with Intram olecular Degrees o f Freedom .” Mol. Phys.
60(1): 193-203.
96. T ison S. A. 1993. “Experim ental Data and Theoretical M odeling o f Gas Flows
Through M etal C apillary Leaks.” Vacuum. 44:1171-1175.
97. T seng F.G., Kim C J . and Ho C.M . 1998. “ A M icro Injector Free o f Satellite Drops
and C haracterization o f the Ejected D roplets.” Sym posium on Application o f M icro-
Fabrication to Fluid M echanics, A SM E International M echanical Engineering
Congress a n d Expositions.
98. Tuckerm ann D.B. 1984. “ Heat T ransfer M icrostructures for Integrated C ircuits.”
P hD Thesis. Stanford U niversity, Stanford, CA.
99. Tuzun R., Noid D.W ., Sum pter B.G. and M erkle R.C. 1996. “D ynam ics o f Fluid
Flow Inside C arbon N anotubes.” Nanotechnology. 7 (3 ):2 4 1-246.
100. Verlet L. 1967. “C om puter ‘E xperim ents’ on Classical Fluids. I. Therm odynam ical
Properties O f Lennard-Jones M olecules.” Phys. Rev. 159:98-103.
101. Von Sm oluchow ski M. 1898. “ Ueber W drm eleitung in Verdiinnten Gasen
A nnalen der Physik und Chem ie. 64:101-130.
102. W alther J.H ., Jaffe R., H alicioglu T. and K oum outsakos P. 2000. “ M olecular
Dynam ics Sim ulations o f Carbon N anotubes in W ater.” Proceedings o f the Sum m er
P rogram 2000, C enter for T urbulence Research, NASA A m es/Stanford University.
103. W ang B.X. and Peng X.F. 1994. “Experim ental Investigation on Liquid Forced C on
vection Heat T ransfer T hrough M icrochannels.” Int. J. H eat M ass Transfer.
37:73-82.
104. W u L. and Bogy D.B. 2000. “Use o f An Upwind Finite Volume Method to Solve the
Air Bearing Problem o f Hard Disk Drives.” Computational M echanics. 26:592-600.
105. Zhao Y., Tai C.H . and A hm ed F. 2001. “Sim ulation o f M icro Flows with M oving
Boundaries U sing H igh-order Upw ind FV M ethod on U nstructured G rids.” Sub
m itted to C om putational M echanics.
106. Zhu X., Tran E., W ang W ., Kim E.S. and Lee S.Y. 1996. “M icro M achined
Acoustic W ave Liquid E jector.” Tech. Dig. Solid-State Sensor a n d A ctuator W ork
sho p, Hilton H ead Island, SC, pp. 280-282.
28.48 FLUID FLOW HANDBOOK
29.1
29.2 FLUID FLOW HANDBOOK
Subscript CD Cooldown
Subscript CP Cloud point
Subscript fl Flowline
Subscript HF Hydrate formation
Subscript HD Hydrate dissociation
Subscript i Inlet
Subscript M Methanol
Subscript ML Mud Line
Subscript NT No-touch
Subscript o Outlet
Subscript pig Pig
Subscript PP Pour point
Subscript s Sea
Subscript SS Subsea
Subscript treat Treat as necessary
Subscript W Water
Subscript WH Wellhead
Subscript WU Warmup
Superscript ’ New condition
29.2 INTRODUCTION
Flow assurance is the process by which uninterrupted oil and gas production is
assured by the management of solids deposition and processes that threaten sys
tem integrity, including corrosion and erosion, thus assuring economical petro
leum fluid production over the project design life. This involves consideration of
solids, integrity, and safety (and the implications any decisions have on econom
ics and operability). It is applied during the stages o f system selection, detailed
design, surveillance, retrofitting, troubleshooting operation problems, etc., to the
petroleum flow path (well tubing, subsea equipment, flowlines, initial processing,
and export lines). Flow assurance issues for refinery operations are related, but
will not be summarized here.
It might be said that flow assurance is a new field that has been around for
many years. W hile the key flow assurance issues have always been present, tra
ditional approaches are inappropriate for deepwater production due to extreme
distances, depths, temperatures, or economic constraints. The term flow assur
ance was coined by Margaret Santamaria of Texaco. It was adopted by DEEP-
STAR (an organization formed to attack challenges associated with deepwater
G ulf of M exico production) in the early 1990s.
Flow assurance is not simply predicting conditions where problems might
occur; rather, it includes strategy development, system selection, and establishing
safe operating conditions (not unlike a HAZOP analysis). At the end of the
process, an operating envelope is created. This work will focus primarily on
blockage prevention (solids) for deepwater operations (1000 to 5000 ft water
FLOW ASSURANCE 29.3
29.3 HYDRATES
F IG U R E 29.1 Picture o f a hydrate that has formed inside a pipe dem onstrating the snow-like
appearance. (DEEPSTAR, 1998)
T Hd - T hf + 5°F (29.1)
Gas Hydrates
No H y d ra te s
100% Methane
99% Methsne, 1% Ethane
1)8% Methane. 1% Ethane. 1% Propane
Temperature: (*P
FIGURE 29.3 Illustration of the effects of pressure and propane concentration on the
hydrate temperature. (DEEPSTAR, 1998)
29.6 FLUID FLOW HANDBOOK
Full modeling o f these hydrate processes is beyond the scope o f this work.
Detailed calculations can be found in the literature (e.g., Sloan, 1998) although it
is often more practical to use a commercial simulation package. Good simulation
software suppliers document the models used and validate predicted values with
both field and laboratory data. Less understood are the processes by which indi
vidual hydrate crystals agglomerate and form plugs that either partially or totally
block production.
The primary strategy for hydrate management in oil production systems is to stay
out of the pressure-temperature envelope where hydrates are stable. For example,
insulating the flowpath (or using a heat source) can keep the fluid temperature
sufficiently above the predicted hydrate temperature during steady state operation
and allow sufficient response time to reduce the hydrate temperature by reducing
pressure in the event o f a shut-in. The purpose of reducing pressure is to get the
hydrate formation temperature below the ambient temperature to allow an
extended shut-in. (A more conservative design is to use the dissociation tem per
ature rather than the formation temperature.)
The system can also be designed to operate at a low enough pressure that
hydrate formation is not a concern. The design strategy needs to balance steady
state requirements with transient operations. In addition to remaining above the
hydrate temperature, there must be sufficient cooldown time before the hydrate
tem perature is reached such that depressurization or other intervention proce
dures may be performed to stabilize the system for long-term shut-in.
G a s H y d r a te s
Mo Hydrates
P u re W a te r
25 NaCI
S w t % NaCI
7.5 NaCI
I 0 \ s t % N sC l
Temperature (°F)
Hydrate
Formation
Curve
Hydrate
Free
Region
Cautions associated with this technique are: (a) difficulties of processing, (b)
too much inhibitor can cause salt precipitation, (c) weaknesses in inhibitor parti
tioning model. From a processing standpoint, the presence of the inhibitor may
enhance the formation of oil-water emulsions. Removal of the inhibitor also
increases the processing complexity. As will be discussed in Section 29.6.5,
water can only dissolve a certain amount of salt. Addition o f a substance such as
methanol can reduce the capacity of water to keep the salt dissolved. Of addi
tional consideration is the fact that when methanol is added to the production
stream, it thermodynamically partitions into the available phases (oil, water, and
gas). The inhibition effect only occurs for methanol in the liquid water phase.
The partitioning effect is a mature area of modeling, but the implementation
is not trivial. Commercial simulation software packages exist with the appropri
ate models and are ready to be used directly or as DLLs (dynamic link libraries)
for use by other software programs. A DLL allows subroutines and functions to
be used by other software programs without requiring detailed knowledge o f the
specific algorithm s used by the DLL developer. If appropriate models are not
used, it is best to proceed with additional conservatism. The use of excess
methanol will likely lead to costs which exceed the cost of obtaining proper pre
dictive models. Glycol partitioning is not modeled as well as methanol partition
ing, but is of less concern since the partitioning to a hydrocarbon phase is much
lower (especially for gas).
Once phase partitioning is considered, the inhibition modeling itself results in a
combination of salt and methanol effects when considering the thermodynamic sta
bility of the hydrate. A rule-of-thumb is to assume methanol inhibition levels of:
FLO W ASSURANCE 29.9
Q// - Qw (29.3)
29.4 WAXES
Wax is paraffin that has solidified. The paraffin molecules that form wax at stan
dard conditions starts with octadecane (C 18H38) and go higher in carbon numbers.
Wax that crystallizes in the bulk oil is not a deposition problem. Crystallization
in the bulk affects viscosity until its concentration is such that the oil can no
longer flow. The temperature at which this condition occurs is referred to as the
pour point o f the oil. If wax crystallizes on a wall then it becomes a deposit. This
occurs if the wall temperature is lower than both the bulk temperature and the
wax deposition temperature. If sufficient wax is deposited over time, portions of
the production system, such as wellbores and flowlines, can become partially or
totally blocked, thus having a significant impact on production efficiency.
Figure 29.6 is a picture of a cutout section of pipe where significant wax
deposition has occurred.
29.10 FLU ID FLO W HANDBOOK
Thus waxes generate two flow assurance issues: pour point and wax deposi
tion (related to the cloud point). With wax deposition, the primary flow assurance
issue that must be addressed is when it will occur and at what rate. A solution
algorithm has been posed by Mueller, et al. (2001).
Caution must be used since the cloud point measured by one procedure will
not match that measured by another.
Pressure has a significant impact on the cloud point (or wax appearance) tem
perature for a live crude. Additional gas dissolved into the oil decreases the cloud
point temperature. This additional gas can either be monotonically increasing
with pressure (as in a black oil) or go through a minimum (thus local maximum
cloud point temperature) in a retrograde condensate. That is, increasing pressure
for a black oil will decrease the cloud point temperature until the bubble point
pressure is reached. Increasing the pressure beyond the bubble point increases the
cloud point temperature. For a retrograde condensate, increasing pressure could
increase or decrease the cloud point temperature. Gas-phase wax deposition is not
known to exist.
Pour point is often controlled with the use of a pour point depressant. A pour point
depressant is simply a solvent that lowers solidification temperatures. A pour point
depressant can also work by disturbing the crystallization process, thus limiting the
amount of solid formed. If a pour point depressant is required, it must be used con
tinuously, since interruptions in production due to mechanical problems cannot
29.12 FLUID FLOW HANDBOOK
The most common solution to wax deposition is to allow the deposit to form and
then remediate (chemically or mechanically). This approach is used whenever
physically practical. Wax is typically removed from DVA wells (wells that have
direct vertical access to the production tubing) during a workover using a device
such as a paraffin knife to physically scrape the walls of the well tubing.
Export line (both ends accessible) wax deposits can be scraped from the pipe
wall using a pig. Pigs are roughly spherical or bullet shaped devices that can be
pushed through pipelines using the pressure of the flow. They are typically
equipped with hard edges or steel knives to effectively scrape wax deposits from
the pipeline wall. This is the most commonly used technique. However, there is
a risk o f plugging the line during pigging operations if too much wax has been
allowed to accumulate. An alternative that is occasionally used is to flow hot oil
to melt the deposit.
Subsea flowlines (topsides-accessible dual flowlines) can have wax removed
with round-trip pigging. This technique is common. As an alternative, hot oil can
be circulated. Pigging of single subsea flowlines can be accomplished utilizing
subsea pig launchers but this incorporates increased risk. The implication o f a
stuck pig for a single flowline is a total loss of production. Additionally, if the
subsea pig launcher can only be loaded prior to installation the risks also include:
multiple launches, inability to change pig style, inability to service or perform
routine maintenance, inability to retrofit.
Part of the flow assurance strategy decision process is to consider the fre
quency o f remediation. The frequency of remediation involves estimation of
deposition rate along with some intervention criterion. Typical criteria are reduc
tion o f hydraulic capacity and piggability of the deposit. If the hydraulic capacity
has been significantly reduced, then the piggability is in question. The rate of
deposition predicted is based on an experimental evaluation o f the oil samples at
various conditions. The driving force can be modeled from the concentration gra
dient or from temperature subcooling.
There are several other remediation techniques that are being investigated by
the offshore industry. One alternative is the use of exothermically reacting chem
icals. The concept is that heat and gas generated from the reaction combine to
FLO W ASSURANCE 29.13
melt the deposit and provide pressure to push the wax downstream. Another alter
native is to use extended reach coiled tubing from the platform to scrape the flow-
line. Specialized self-propelled pig designs are also being considered.
Oftentimes, wax deposition occurs at a rate that will not affect the useful life of
flow path. As wax deposits, it forms an insulating layer on the inside of the pipe.
This acts to increase the inside wall temperature through additional heat transfer
resistance, thus reducing the deposition rate. If the deposit does not need to be
scraped away, a complete blockage could take tens of years to form for some sys
tems. It this significantly exceeds the design life of the flow path then there is no
reason to prevent deposition.
Wax deposition modeling is a mature field of research for single-phase dead oil.
Ongoing efforts in this field include modeling multiphase effects, depletion, melt
ing, and deposit roughness (e.g., Chen, et al., 1997). Research on wax formation
in the bulk fluid is active in the area of effects on viscosity.
29.5 ASPHALTENES
Asphaltene is a general term for the dark brown to black organic matter that
occurs either dissolved or dispersed in crude oil or as sediment. These are solids
that do not melt and have a molecular weight in the 500 to 1500 range with a
com plex ring structure. Asphaltene molecules are often defined as the component
o f oil that is not soluble in paraffinic solvents such as n-heptane but are soluble
in aromatic solvents such as toluene.
Figure 29.7 is an artistic interpretation of an asphaltenic structure. This is not
a fixed structure as it can vary from molecule to molecule.
The flow assurance issue for asphaltenes is associated with stability. When
asphaltenes become unstable in the flow path, they precipitate from the crude oil
and form a sticky tar-like or coal-like deposit. The stability of the crude oil is not
simply a function of asphaltene content; rather, it is a function of asphaltene con
tent and the stable amount that the crude oil can suspend or hold in solution. In
general, the com ponents in crude oil that aid stability are resins, which promote
asphaltene suspension (micelle structure with a head group that is attracted to the
asphaltene and a tail group that is attracted to paraffins), and aromatics, which act
to make asphaltenes soluble. On the other hand, paraffins act to destabilize the
asphaltene. Total asphaltene content is determined by the addition of heptane to
the oil (ASTM D6560-00, IP 143/96). As noted in Section 29.5.1, proximity to
the flocculation point can be an estimate o f asphaltene stability.
29.14 FLUID FLOW HANDBOOK
Stability is a strong function of pressure. Figure 29.8 illustrates the effect that
pressure has on asphaltene stability. As a crude oil is depressurized from reser
voir conditions to the bubble point, the asphaltene stability limit decreases. This
is essentially due to changes in the concentration (chemical activity) o f the com
ponents in the crude oil that stabilize asphaltene solubility. Once the pressure
drops below the bubble point, the asphaltene stability limit increases dramatically
with a decrease in pressure (the minimum solubility does not always occur at the
bubble point). This is caused by changes in the composition of the liquid phase.
It is thought that this is a liquid density driven effect (de Boer, et al., 1995). As
density increases, the resin’s desire to work as a micelle decreases and the resin
goes out into solution.
Another critical consideration is that two stable oils can be mixed in such a
way as to lead to an unstable oil. In particular, an oil with a high asphaltene con
tent and an oil with a high paraffin content (waxy crude) individually may have no
asphaltene problems. However, upon mixing, significant asphaltene precipitation
may occur. Additional factors that may destabilize asphaltenes include changes in
pH (acids can destabilize micelles), addition of certain types ot production treat
ing chemicals, temperature changes, and addition of light gases such as methane
or carbon dioxide to the oil, as occurs in gas lift and enhanced production opera
tions. Com m ingling can occur in a wellbore, flowline, export line, during top-
sides processing, etc.
FLOW ASSURANCE 29.15
Estimation of the stability of the asphaltenes for a given crude is termed screen
ing. If asphaltene stability is expected to be an issue, a common solution is to not
develop the prospect. Thus screening could generate a showstopper for a poten
tial production zone. Despite this awesome power to influence field development,
the screening techniques commonly used are somewhat primitive.
One method considers n-heptane solubility. In this test, n-heptane (a de-sta-
bilizing agent) is added drop wise to the crude sample until the asphaltenes begin
to precipitate. The amount of n-heptane required is considered a proximity to
instability (de Boer, et al., 1995).
Another method is based on composition and PVT prediction (density and
undersaturation). In this method, the behavior is predicted by attempting to repro-
29.16 FLU ID FLO W HANDBOOK
duce a solubility curve like that in Figure 29.8 for a given oil using data obtained
from a PVT report and an equation of state (de Boer, et al., 1995).
Also used is a method that considers relative amounts of the key components
of crude oil. These include paraffins, resins, aromatics, and asphaltenes. This
information is then compared to an experience-based correlation (like a stability
map). Such correlations are highly proprietary. Leontaritis and Mansoori (1988)
compare the merits of the differing approaches to screening based on field obser
vations. Often, all methods are considered knowing that each has a limitation.
An expensive alternative when screening techniques give conflicting results
is live oil depressurization. In this screening technique, a sample at reservoir con
ditions is slowly depressurized while monitoring turbidity. This can be used to
accurately determ ine if there is a stability problem. The drawbacks are expense,
loss of a valuable sample, and dependence on high quality sampling techniques.
An active area of screening work involves thermodynamic stability modeling
(e.g., Wu, et al., 2000).
29.6.1 Sampling
This is the first step of flow assurance. It is the basis on which design decisions
are made. Sampling and analysis techniques are largely proprietary and vary from
producer to producer. From collected samples producers need to consider which
FLOW ASSURANCE 29.17
tests are critical and which are a luxury in the goal of properly evaluating the
crude and designing a production system. These tests can affect how the samples
w ill be collected and be handled as some properties are pressure/temperature his
tory dependent. Kaczmarski and Lorimer (2001) discuss how the timing o f sam
pling and analysis fit into the project development schedule.
Desired information includes: black oil properties (PVT analysis, com posi
tional data such as sulfur, nitrogen, and presence of metals), wax content, pour
point, cloud point, asphaltene content, asphaltene stability, composition, viscos
ity, result through likely separator train, resin content, aromatic content, contam
ination level (e.g., from drilling mud). In addition, flow assurance analysis of
hydrate risk, corrosion prediction, and scaling tendency are dependent upon an
assessment of the salinity and composition of the water that will be produced with
the oil or gas. However, often the producer’s drilling programs are focused upon
penetrating and sampling hydrocarbon bearing zones only. In this case, the flow
assurance specialist must often be content with rough estimates of salinity and
compositions based upon analogs, such as samples taken from nearby fields that
have already been developed and put into production.
Samples retrieved from exploratory wells can have contamination levels that
can significantly affect analysis. While the contamination should always be ther
modynamically removed from testing results, a natural variation exists in the
sampling and testing. Contamination greater than 10-wt% is typically considered
to be excessive. Sampling variation should be approximately 5 percent for gas-
oil ratio and °API density. Sampling variation for black oil bubble point is about
200 psi. The bubble point variation increases for condensates to about 500 psi.
Flow assurance simulations require that the PVT behavior of the produced fluids
be matched through a tuning procedure (equation of state regression with lumped
and pseudo-components, modified binary interaction parameters, etc.). Wilkens
and Flach (2 0 0 1) demonstrate one technique for determining the implications of
sampling variation with respect to flow assurance modeling.
Often considered a second step in flow assurance, the flow path must be modeled
thermally and hydraulically. This information is used in conjunction with all
models to determine potential areas of flow assurance concern. This information
is also used for system sizing.
Two and three phase modeling is challenging for both thermal and hydraulic
aspects. Commercially available simulation software packages offer common
models from literature for steady state production or proprietary models for tran
sient operations. The effectiveness and validity of these models vary. For a full
discussion of gas-liquid flow, see Chapter 11.
The size o f slugs is also predicted during thermal and hydraulic modeling.
Knowledge of slug size, frequency, and velocity is used to verify sufficient pro
duction handling capacity. It also affects the design of bends to handle forces.
The predicted slug properties include those for slugs that are produced during
29.18 FLUID FLOW HANDBOOK
steady state operations and during transient operations. The complex bathymetry
of subsea flowlines has a significant impact on the slugging.
One objective o f steady state thermal modeling is to verify that the use of
excessive insulation is avoided. Not only is a high arrival temperature a safety
issue (temperatures above 140EF may require the use of personal protective
equipm ent or the use of a physical barrier), but it also becomes a materials issue
(e.g., thermal limitation on flex joints).
Too little insulation must also be avoided. Too low an arrival temperature
affects the wax deposition characteristics. Also, a typical objective is to maintain
the temperature above the hydrate temperature plus an appropriate amount of
time for cooldown to the hydrate temperature.
During shut-in, the pressure near the entrance to the flowline decreases while
the pressure near the exit from the flowline increases (due to the elimination of
the frictional pressure gradient). The critical location (from a hydrate formation
perspective) is often at the riser base (high pressure due to hydrostatic head, low
am bient tem perature plus a current, low initial temperature of fluids), which often
cannot be insulated as well as the rest of the flowline.
FLOW ASSURANCE 29.19
St(-)
FIGURE 29.9 Tool forestimating the steady state thermal effect of switching insulation.
29.20 FLU ID FLO W HANDBOOK
a x is la b e l in itia l t e m p e r a t u r e (°F)
FIGURE 29.10 General cooldown performance approximation fora typical 8 inch insulated riser.
FLOW ASSURANCE 29.21
The above method estimates the cooldown time available. The cooldown time
required is the total time required to safely prepare the system for a long term
shut-in condition. The minimum cooldown time required consists of three parts:
The no-touch time, tNT, is the time period during which a restart will be
attempted and no special flow assurance actions will be attempted. Lorimer and
Ellison (l 998) use a no-touch time of about three hours based on Shell’s operat
ing experience. This approximate value has been validated elsewhere; Berger and
McMullen (2 0 0 1) summarized operations experience with Troika (deepwater Gulf
ot M exico subsea tieback flowline). They found that over the initial 3-year period
of operation, 90 percent o f all shutdowns were shorter than 5 hours, and only one
event lasted over 24 hours. They also noted that the duration o f the shutdown
decreased significantly with experience. Once a reasonable value is obtained, shut
downs in excess of the no-touch time lead directly into flow assurance action.
The time to treat the wells and subsea equipment that will not be depressur
ized, tlreal, needs to be included. This can be estimated from hydrate inhibitor
delivery rates and volumes required to bullhead to fluids to a safe location such
as the SSSV (subsurface safety valve, located in a non-hydrate forming zone due
to geothermal temperature).
The time required to blowdown, tBD, is simply the time required to safely
depressurize the system. Adjusting the topsides choke can vary the rate of
depressurization. The appropriate settings are found through operations experi
ence, however they can be estimated by using a transient multiphase flow sim u
lator. The objective is to balance practical time constraints with a minimum and
a maximum liquid carryover. The minimum liquid carryover is the amount
which will ensure that sufficient hydrostatic head is eliminated from the highest
pressure point in the flowline (new critical location) such that its hydrate tem
perature is below the ambient temperature. That is, to make sure the objective of
depressurization is accomplished in all parts of the flowline. The maximum liq
uid carryover applies to the scenario of handling liquids topsides with no (or lim
ited) power. That is, if you cannot fully process the incoming liquids then you
must be able to put them somewhere. A faster blowdown will lead to more liq
uid carryover.
The design considerations leading to this point have been flow assurance
strategy for ongoing production and for a shutdown (planned and unplanned).
Additional transient situations need to be addressed for start-up.
One fortunate aspect of start-up is that an unplanned start-up needs not be
addressed. When considering startup, however, the initial condition must be con
sidered. In a well, the need for tubing insulation must also be evaluated (e.g., VIT
— vacuum insulated tubing). Insulating the tubing will allow the wellhead tem
perature to increase faster, reducing or eliminating hydrate inhibitor requirements
29.22 FLU ID FLO W HANDBOOK
during restart. However, the surrounding earth will not heat and the cooldown is
much faster than for bare tubing.
Consider startup from a cold condition. Figure 29.11 illustrates warmup (and
cooldown) performance of a component. The time tj—1() represents the time to
reach the hydrate dissociation temperature (inhibition typically required). If the
time t3- t 2 matches the required cooldown time for the com ponent, then t2- t 0 is
the time required to reach a safe condition (inhibition optional). Production
beyond the safe condition time does not require inhibition.
When restarting from a previous condition (i.e., not fully cooled), a similar
approach can be taken. It is best to represent the typical shutdown time with the
no-touch time limit as the amount of time that the system has been allowed to
cooldown.
The flowline initial condition can be assumed to be ambient unless it has been
buried. For cold dual flowlines, one method to heat them is to circulate hot dead
oil until sufficient heat has been added to the flowline. With sufficiently hot oil,
this process takes 2.5 to 3.5 line passes to fully heat unburied flowlines. A rule of
thumb is that for dual unburied but insulated flowlines:
FIGURE 29.11 Cold system startup and shut-in demonstrating the time to reach a safe condition.
FLOW ASSURANCE 29.23
Thus for 3 mph pigging, a 15-mile offset would require about 3 0 hours to
heat. Extended offsets may require boosting the velocity (e.g.. push oil with gas)
to reduce warmup time. Cooling implications of adding gas to the oil need to be
fully simulated.
All processes need to be fully modeled with appropriate simulation software
(e.g., actual shut-in procedure, line packing). From a hydrate perspective, special
locations (e.g., insulation weak links) need to be considered during shut-ins (e.g.,
flowline jumpers, sleds, wellheads, valves). Although these locations do not largely
affect the steady state performance, they become a hydrate formation liability.
29.6.3 Ice
It is not often that the ambient conditions are such that ice formation is a problem.
It can be a problem, for example, with extensive Joule-Thomson cooling. Care
must be used near chokes during steady state and transient (re-start and depres-
surization) conditions. In addition, there are some locations around the world
(deepwater North Sea, for example) in which the ambient water temperature in
deep ocean waters is below the normal freezing point of water. In general, flow
assurance solutions for ice formation closely follow those of hydrate formation.
29.6.4 Scale
The potential for scale does not have a significant impact on the system develop
ment scenarios. Precipitation of scale (e.g., NaCl, B aC 0 3, S rS 0 4, C a S 0 4 • 2H20 ,
C a S 0 4 • H 20 , C a S 0 4, C a C 0 3) from water forms a hard deposit on the walls of
the flow path. Unlike wax and asphaltene deposits, scale deposit is not organic.
The scaling process is driven by changes in temperature and pressure of the pro
duced fluids as it moves from the reservoir, through the production system, to the
processing facilities. In addition, the deposition of calcium carbonate is driven by
pH changes, which occur as the level of dissolved carbon dioxide in the produced
water changes during production. Problems can also arise from the commingling
o f waters.
When scale is expected to be an issue, an inhibitor is continuously injected
with the produced fluids. Alternatively, squeeze (batch chemical) treatments can
be applied to control scale buildup (Dyer, et al., 1999).
An alternative for scale deposit control is to attempt controlled crystallization
in the bulk (e.g., m agneto-hydrodynam ics). Excessive amounts of these solids,
however, can cause other operational problems.
Remediation of scale deposits is highly dependent on the chemistry of the scale.
Calcium carbonate can be easily removed by acid treatment. Calcium sulfate, and
especially barium sulfate, prove much more difficult to remove chemically. Oper
ators may have to resort to extremely aggressive mechanical measures, such as
drilling, to remove the most adherent and chemically resistant scale deposits.
29.24 FLUID FLOW HANDBOOK
There are many material integrity issues that are not of flow assurance concern
(e.g., stress, fatigue). Those that do become flow assurance issues are: need for
cryogenic piping, corrosion ( C 0 2. H2S, microbial, external, erosional), chemical
compatibility.
The need for cryogenic piping is not directly considered in flow assurance. It
is implied from thermal analysis. The ductile-to-brittle transition temperature
(DBTT) is dependent upon the specific alloy used. For typical subsea flowlines,
the DBTT is 0°F (-1 8°C). Such temperatures can exist due to transient operations.
Potential locations are downstream of the topsides boarding valve and downstream
of subsea tree valving. If a high pressure, cool gas bubble exists upstream of a
valve or choke (with low pressure downstream) during shut-in, Joule-Thomson
cooling can be significant enough to require cryogenic pipe as flow starts.
Internal corrosion in the production system is often curtailed by use of
inhibitors, coatings, and/or cathodic protection (applicable to processing vessels
only, not to wellbores and flowlines). There is a strong link between the thermal-
hydraulic modeling generated by the flow assurance analysis process and corro
sion rate prediction. The tem perature-pressure profiles generated by the
thermal-hydraulic models are used as a direct input into the corrosion models. As
a result, corrosion modeling is often included in the flow assurance specialist's job
scope. The results of thermal-hydraulic modeling are also required for modeling
erosion in the system during the design phase. The impact of flow pattern on cor
rosion and erosion has been studied and continues to be an active area of research.
Once an initial flow assurance strategy has been proposed, chemical interac
tions must be considered. The interactions o f production treating chem icals
(e.g., inhibitors, dispersants, additives) with each other, metals, and coatings,
along with produced fluids, are a newer area of focus for flow assurance (chem
ical com patibility). This goes beyond the simple material compatibility issues
and extends into multi-step reactions. These considerations involve detailed
knowledge of all procedures (including one-time uses o f lines such as acidizing
and unloading), chemicals, materials, uses, and both steady and transient opera
FLOW ASSURANCE 29.25
REFERENCE
1. Atkins, P.W . 1990. P hysical Chemistry. 4th Edition. New York. Freem an.
2. Berger, R.K. and M cM ullen N.D. 2001. “Lessons Learned from T roika’s Flow
A ssurance C hallenges.” O ffshore Technology Conference. Houston. Paper OTC
13074.
3. De Boer, R.B., Leerlooyer K., Eigner M .R.P, van Bergen A.R.D. February 1995.
“Screening o f C rude O ils for Asphalt Precipitation: Theory, Practice, and the Selec
tion o f Inhibitors.” SP E Production & Facilities. 55-61.
4. Dyer, S.J., G raham G.M ., and Sorbie K.S. 1999. “Factors Affecting the Therm al
Stability o f Conventional Scale Inhibitors for Application in High Pressure/H igh
T em perature R eservoirs.” Proceedings— SP E International Sym posium on O ilfield
Chem istry. R ichardson, TX . SPE. 167-177.
5. D EEPSTA R. 1998. “Flow Assurance Issues— Deepstar III Project Reference
L ibrary.” Internet D ocum ent Library prepared for DEEPSTAR by M ultiphase
Solutions, Inc.
6. Fu, B. April 2000. “ M anaging Fluid Behavior, Solids Deposition to Ensure O pti
mum Flow .” O ffshore M agazine. 85+.
7. H udson, J.D ., D ykhno L.A ., Lorim er S.E., Schoppa W., and W ilkens R.J. 2000.
“Flow A ssurance for Subsea W ells.” Proceedings o f the A nnual O ffshore Technol
ogy Conference. Vol. II. 515-520.
8. Hyne, N.J. 1991. D ictionary o f Petroleum Exploration, Drilling, and Production.
Tulsa, OK. Pennwell.
9. K aczm arski, A.A . and L orim er S.E. 2001. “Em ergence o f Flow A ssurance as a
Technical D iscipline Specific to Deepwater: Technical Challenges and Integration
into Subsea System s E ngineering.” O ffshore Technology Conference. Houston.
Paper OTC 13123.
10. Kleinberg, R.L. and B rew er P.G. M ay-June 2001. “Probing Gas Hydrate D eposits.”
Am erican Scientist. 244-251.
29.26 FLUID FLOW HANDBOOK
11. Leontaritis, K.J. and M ansoori G.A. 1988. “A sphaltene Deposition: A Survey of
Field Experiences and Research A pproaches.” Journal o f Petroleum Science and
Engineering, Vol. 1. 229-239.
12. Lorim er, S.E., and Ellison B.T. M arch 1998. “ Subsea Oil System Design and O per
ation to M anage W ax, A pshaltenes, and H ydrates." Paper 60C presented at the
Internation Conference on Petroleum Phase B ehavior and Fouling, H ouston, TX.
13. M cCain, W .D. 1990. The Properties o f Petroleum Fluids. 2nd Edition. Tulsa,
OK. Pennwell.
14. M ueller, J.B.. W ilkens R.J., W ade K., and Broze G. 2001. “D eveloping a W ax-
D eposition Sim ulator through O perator/V endor C ooperation.” Offshore Technology
C onference. Houston. Paper OTC 13076.
15. Sloan, E.D. 1998. Clathrate H ydrates o f N atural Gases. 2nd Edition. New York.
M arcel Dekker.
16. Speight, J.G. 1980. The Chem istry and Technology o f Petroleum . New York.
M arcel Dekker.
17. W ilkens, R.J. and Flach L. 2001. “An A lgorithm for D eterm ining the Uncertainty of
Com plex M ultiphase Flow M odels.” O ffshore Technology Conference. Houston.
Paper OTC 13151.
18. W u, J., Prausnitz J.M ., and Firoozabadi A. January 2000. “M olecular T herm o
dynam ics o f Asphaltene Precipitation in Reservoir Fluids.” A lC h E Journal. 197-209.
____________CHAPTER 30___________
DRAG REDUCTION BY
POLYMER ADDITIVES TO
TURBULENT FLOW SYSTEMS
3 0 .1 1NTRODUCTION
30.1
30.2 FLUID FLOW HANDBOOK
1. Additives
2. Surface
3. External
Energy conservation implies the origins of the term “drag reduction.” The ini
tial studies (1940s and 1950s) were related to unique turbulence or wall effects in
flow through tubes. The pumping power required for fluid transport through pipe
(tubes) is related to the energy dissipation of flow eddies, or turbulence. Under tur
bulent flow conditions, there is net flow in one direction but along points in all dimen
sions there is chaotic motion of fluid on a molecular (micro) level. This chaotic
motion leads to so-called “drag” losses.
The term “drag” comes from the pulling of an object through water. In the
early 1960s the term “drag reduction” was coined. It has been used ever since as
a descriptor for friction loss decreases in turbulent non-Newtonian fluid flow.
There have been many reviews of polymers for drag reduction [1-13]. Excel
lent monographs were recently published [14-16]. Valuable information is also
found in topical conference proceedings [17-29]. The engineering and scientific
literature in the 1960s and 1970s included many papers on drag reduction, many
of these are cited here. These research studies were most interested in trying to
elucidate the mechanism of drag reduction and the degradation of polym er addi
tives due to turbulent flow. There are three methods to effect drag reduction as
listed above.
An accepted definition of drag reduction is “the reduction of skin friction in tur
bulent flow below that of the solvent along” [1 ]. This definition implies two con
straints. First, the flow regime is turbulent after the polymer is added to the solvent.
Second, the skin friction is not simply lower than that of a Newtonian fluid having
DRAG REDUCTION BY POLYMER ADDITIVES TO TURBULENT FLOW SYSTEMS 30.3
Water Soluble
Polyethyleneoxide Poly ox 25322-68-3 PEO
Polyacrylam ide Cyanamer 9003-05-8 PAAm
Polyacrylam ide (partially hydrolyzed) Separan PAMH
G uar Gum J-2FP 9000-30-0 GGM
Hydroxyethyl Cellulose Cellusize 9004-62-0 HEC
Carboxym ethylcellulose Sodium 9004-32-4 CM C
Sodium Polystyrene Sulfonate 9003-59-2
Potassium Polyphosphate
Xanthane
Polyvinyl Alcohol 9002-89-5 PVA
O il Soluble
Polym ethylm ethacrylate Lucryl 9011-14-7 PMMA
Polyisobutylene Vistanex 9003-27-4 PIB
Polycisisoprene 9006-04-6 PCIP
Polydim ethylsiloxane PDMS
Poly isodecylm ethacry late
Polystyrene 9003-53-6
Polystyrene Sulfonate
Polyethylenim ine PEI
Polyvinylpyrrolidone 9003-39-8 PVP
the same viscosity as the polymer solution at the wall shear in question. In practi
cal terms, drag reduction (DR) is defined by two formula. The percent DR is
AP, - APd
DR = — -----------E X 100 (30.1)
AP,
where
Table 30.2 presents pressure drop and drag reduction calculated results.
Since pressure differences are directly proportional to friction factors, f, drag
reduction can also be quantified by friction factor differences,
DR EE |f , - M i oo
Friction factors differences are often used to quantify the results o f polym er
additives to turbulent flow. The polymer solutions have friction factors always less
than the solvent (fluid). This means that from a quick view of a Moody plot of fric
tion factors versus Reynolds number there can be friction factor values below that
o f a smooth tube.
After a historical perspective is presented, this chapter describes a num ber of
characteristics of polymer solutions for drag reduction. A brief outline of turbulence
is given here. Two areas of great drag reduction interest are included, polymer degra
dation and scaleup. Finally, the chapter concludes with applications.
30.2 History
The earliest reported observations of pressure drop decreases due to additives or con
taminants in turbulent water flow existed prior to 1900 [8]. In 1897 Hele-Shaw
reported the results of experiments on the nature of surface resistance in pipes and
ships. Pipes were partly lined with rough surface section (flannel) and the other side
being smooth (brass). From the rough side, bile was injected in flowing water and
visual observations were made of increased velocity.
In the early pressure drop decrease studies, the contaminants were algae, bile or
slime. In the paper manufacture of, pulp suspensions were reported to have higher
velocities than water [12]. During the closing years o f World War II, two indepen
dent studies of pressure drop decreases in turbulent flow were reported [ 17,18,23,30].
(See Figures 30.1 and 30.2.)
Initial Second
M easuring Measuring
Point Point
Pressure Pressure Difference
Fluid -------------------------- kg/cm 2
P RE S S U R E GRADI ENT
FIGURE 30.1 Schem atic o f experim ental results for flow friction-reducing effect by Mysels
(1945), adapted from [23).
At that time information was needed on the flow characteristics of gasoline thick
ened with soaps. In 1945, experim ents by Mysels et al., were reported to the gov
ernment, and the papers for a patent were initiated; a patent was issued in 1949. In
the report it states, “the thickened gasoline flows faster than the unthickened.
This paradoxical result has not been checked by direct experiments. Its expla
nation may lie in the fact that in the region considered, the flow of gasoline is strong
ly turbulent and would be 70 times faster were it streamlined. If the flow of gel is
less turbulent, it may cause a smaller pressure drop despite a higher viscosity.”
Subsequent experiments conducted prior to October, 1945 confirmed “that the
effect was real, as drag reductions of 66 percent were obtained over a consider
able range of flow rates. (See Figure 30.1.)
In 1948, in one of the earliest experimental reports of friction-reduction effects,
Toms showed that a dilute polymethylmethacrylate solution in monochlorobenzene
30.6 FLUID FLOW HANDBOOK
exhibits unusually low friction coefficients in turbulent pipe flow [18]. These unusual
results were broadly consistent with a model for turbulent flow that included a “wall
effect” [17]. By “wall effect” it means a region near the pipe wall where polymer
molecules would be excluded due to their bulky size. Toms obtained reductions in
friction o f up to 50 percent as reported by Hoyt [6].
Two doctoral studies were published in 1959 that described extensive exper
iments about the behavior of non-Newtonian fluids in turbulent flow through straight
tubes [31,32]. In comparison to Newtonian fluids, one study reported a thicker non-
turbulent layer at the wall and lower friction factors [31]. The other study reported
typical friction factor versus Reynolds numbers plots over both the laminar and tur
bulent flow regimes; the non-Newtonian polymer solutions (an aqueous solution of
0.2% Carbopol 934) had lower friction factors than Newtonian fluid (w'ater) [32].
In the early 1960s three important studies were reported [6]. First, industrial
applications were begun using guar gum solutions to suspend sand in sand-water
mixtures for the fracturing of oil reservoirs. This solution decreased the friction
losses in the downpipe. Use of such polymer solutions continues to the present.
DRAG REDUCTION BY POLYMER ADDITIVES TO TURBULENT FLOW SYSTEMS 30.7
Second, researchers with the U.S. Navy surveyed many polymers for potential fric-
tion-reduction applications and identified polyethyleneoxide (PEO). This linear
polymer with a molecular weight in the millions is effective at concentrations (in
water) of a few ppm [33]. Finally, the term “drag reduction” was coined by
Savins and continues to be used as the descriptor for the turbulent flow of non-
Newtonian fluids that exhibit lower friction loses [34].
Since the early 1960s, the engineering and scientific literature in various dis
ciplines has described a myriad number of academic and industrial studies of drag
reduction, see references. The following subsections 30.3 to 30.10 describe salient
aspects of drag reduction by polymer additives.
30.3 Onset
The onset of drag reduction is measurable in turbulent flow, but its prediction is
semi-empirical [35, 36]. Basically, primary differences between turbulent and lami
nar flow are the existence of fluctuating velocities and the average (steady-state)
flow rate. In laminar flow there is an absence of fluctuations, and the average flow
rate is smaller than in the turbulence flow regime. In terms of the Reynolds num
ber, DVp/p, the upper limit to laminar flow occurs, classically, at Reynolds number
of slightly above 2000. Turbulence begins at Reynolds numbers of less than 10,000.
In turbulent flow with the presence of drag reduction, the friction factors for a
given Reynolds number are less than that for a smooth pipe (tube) or the Prandtl-
Karman law for Newtonian fluids (see Figure 30.3).
In the direction of flow down a pipe with an average flow rate = U; and with
velocity components, u, v, and w and fluctuating velocity components u ', v \ and
w \ the velocity components in 3-dimensions are:
u = U + u'
v = v'
w = w'
The fluctuating velocity components (primed values) have average values of zero.
But their squares are finite; or the time-averaged root mean squares are finite [11,
37, 38]. O f course, for an impermeable pipe wall all the above velocity values are
zero. However, at slight distances away from the boundary wall, certain velocity
com ponents are finite; namely,
The most important of these is u 'v '. Since pressure loss in a pipe depends
strongly on p u 'v '. This term is the turbulent shear stress in the direction of flow.
30.8 FLUID FLOW HANDBOOK
At points very near the wall the values v '2 and u V are zero. The shear stress at
the wall, since fluctuating velocities are absent, becomes Newton’s law
Au
rw = iti —
Ay
where
The wall shear stress for turbulent flow of a dilute polymer solution can be
approximated by an empirical law,
w h ere
Flow friction factors are well established for Newtonian fluids [39]. In general,
these depend on two parameters, the system Reynolds number and the surface
roughness. But for laminar flow in smooth tubes the friction factor depends only
on the Reynolds number:
.1/2
5.75 log + 1.75 (30.2)
The turbulent flow equation is a form of the Prandtl-Karman law, see Figure 30.4.
Friction factors for non-Newtonian systems, dilute polymer solutions can be
estimated based on empirical data. Friction factors for dilute drag reducing poly
mer solutions are presented schematically in Figure 30.5. The coordinates of this
graph are according to Prandtl-Karman law, f ~,/2 versus R ef-1/2.
30.10 FLU ID FLO W HANDBOOK
t
f
Re
FIGURE 30.4 Friction factors versus Reynolds number for pipe flow, adapted from [6|.
Between the Prandtl-Karman law for Newtonian turbulent flow (in smooth
tubes) and the asymptotic regime of maximum possible drag reduction in which
the friction factor is insensitive to the polymer solution employed, a drag-reduction
regime exists with a friction factor dependent on the nature o f the polym er solu
tion [7]. In this regime an approximate friction factor equation
j/2
y ) = (4.0 + <5)log(Refl/2) 0.4A- <X51
pl/2\ - A _ ( 2 1/2DW*)
log (30.3)
where
R e (SO L V E N T ) ----------------
FIGURE 30.5 Polymer concentration effect on drag reduction, adapted from |8].
There is a turbulent flow regime in which the frictional losses can not be decreased
further below the Prandtl-Karman law, Equation 30.2. In other words the friction
30.12 FLUID FLOW HANDBOOK
FIGURE 30.7 Schematic of friction factors for pipe flow, adapted from [7].
DRAG REDUCTION BY POLYMER ADDITIVES TO TURBULENT FLOW SYSTEMS 30.13
Equation
D R _
Ts"
1 f l 'N T R l A r u n
r~s“ j
I I HP ATI . I ) ' | | j
c 11 IN T R F A T E I)
= Ax ) + B (7)
ppm \ DR
factor is independent of pipe diameter and the polymer solution used to affect drag
reduction, see Figure 30.7. This limit corresponds to maximum drag reduction.
The maximum drag reduction correlation for pipes is:
1/2
= (19.0 ± 0.4) lo g (R ef,/2) - (32.4 ± 1.2) (30.4)
This phenomenon was first reported in 1964 by Hoyt and Fabula, confirmed by
others, and the correlation developed by Virk and co-workers in 1970 [6, 41, 42].
It is based on drag reduction studies from nine independent sources that generated
315 data points. The polymer solution solvent, except for one, was water or aqueous
solutions. Data includes results using two polymers, PAM and PEO, of various mol
ecular weights and concentrations; pipe diameters ranged from 0.128 to 1.27 cm.
30.14 FLUID FLOW HANDBOOK
If the system is not a pipe but a flat plate, the maximum drag reduction asymp
tote correlation is [l 1].
where
VL
ReL = —
v
L = length of the plate
In turbulent flow, polymer additives affect the mean velocity profiles such that the
two-layer model for Newtonian fluids is modified to a three-layer model. The New
tonian fluid two-layer model includes a viscous sublayer and a logarithmic layer:
(viscous) u+ = y+ (30.6)
where
Equation 30.8 indicates an increase in velocity due to S+. This increase is due
to the interaction of polymer additive molecules and velocity fluctuations in the
region next to the viscous sublayer. This velocity profile can be generalized to
u+ = A log (y +) + B + AB (30.9)
The constant A does not change in drag reducing flows, see Figure 30.9. The
quantity AB is a complicated function of the polymer solution properties and relax
ation time [7]. The elastic sublayer is a flow regime characteristic of drag reduc
tion. It is useful in the scale-up of bench-scale data.
A number o f velocity profile measurements have been made using LASER-
Doppler techniques and other methods [19,33,43-51 ].
30.16 FLU ID FLO W HANDBOOK
30.7 Scale-Up
The quantification o f pipe diameter effect on the performance of drag reducing poly
mer solutions in commercial pipe based on data from small pipe has only recently
been achieved. The prediction of drag reduction in large pipes based on bench-scale
studies is important for the practical applications of polymer solutions to conserve
energy, pumping power costs [52,53]. In the 1990s a few papers were published
about pipeline scale-up [54-56]. The meager number of prior reports is indicative
of the complexity o f the scale-up problem. The earlier attempts to develop scale-up
methods can be found in the most recent publication [56].
Equation 30.9 can serve as a scale-up basis by transferring of the AB term from
small to larger pipe [54, 55]. For the same polymer solution and at the same shear
velocity (//*). Experiments were conducted in two pipe sizes (1.03 and 6.15 inch
diameter) using the same polymer solution (polyacrylamide in water) at six differ
ent concentrations. In the non-dimensional velocity profile, Equation 30.9, the AB
represents drag reduction and was found to be independent of diameter for the
same polymer solution. On this basis, the results from the smaller pipe were used
to predict the drag reduction of the larger pipe.
DRAG REDUCTION BY POLYMER ADDITIVES TO TURBULENT FLOW SYSTEMS 30.17
The experimental data set can be transferred point by point; see results in Figures
3 0 .10 and 3 0 .11. Figure. 3 0 .10 shows data for the lowest polyacrylamide concen
tration, 10 ppm and data for water without polymer. Figure 30.11 presents results at
the highest polyacrylamide concentration, 135 ppm. Note the increase in drag reduc
tion or lower friction factors.
Re
Reynolds Number
30.8 Polymers
Many polymers have been used as drag reducing agents, or flow enhancers [72,73j.
For a polymer to be effective it needs to have the following general characteristics:
2 «•
*§ 1 - D = 1194 mm
> o
0 10 15 20 25 30 35 *0
2.5 -
1.5 r
1 -
0 5 f D = 52.5 m m
0 10 15 20 25 30 35 40
"3
a D = 26.6 m m
>
10 20 30 40 50 60 70
shear stress, P a
The desired mechanical attributes are to attenuate flow velocity fluctuations (ju\
v', and ju'v'. These attributes include sufficient length to attenuate crosswise fluc
tuations, mass to dampen both axial and crosswise fluctuations, rigidity to absorb
fluctuations, and elasticity to provide a delayed response to fluctuations [10]. These
characteristics are important to polymer degradation, see Section 30.9.
Polymers marketed and used for drag reduction are listed in Table 30.1. In this
table, the Chemical Abstracts registry numbers are included, this permits easier
access to full information about the polymer. These are mostly synthetic materials
and modified natural sources of polysaccharides. Exocellur polysaccharides from
algae and bacteria have caused anomalously low friction and hydrodynamic test
ing using tow tanks and water tunnels [II].
30.9 Degradation
Trace amounts of high-molecular weight straight chain polymers can reduce drag
losses in turbulent flow through pipe and across flat plates. Drag reductions with
3 0 .2 0 FLUID F L O W H A N D B O O K
FIGURE 30.13 Variable mixing length model comparison of predicted and actual velocities [56].
a few ppm polym er in water (or oil) amounts to 25 to 50 percent, in some in
stances up to 70 percent depending on the polymer and concentration. The poten
tial applications and energy savings are great, (see 30.10).
The widespread application of polymers for drag reduction has been limited
because the polymers can loose their effectiveness under conditions of high shear,
see Figure 30.14. This is called degradation, or the decrease in drag reduction with
flow time [8].
Simply, degradation occurs when the polymer (picture a ball of spaghetti) un
winds into individual strands. The strands are also shorter than the original poly
mer chain due to permanent scissions. Polymer disentanglement requires a certain
shear rate to be exceeded. The shear arises from the turbulent flow fluctuating veloc
ity components. Degradation of polymer may also arise from chemical, thermal
and biological effects.
The mechanical degradation of polymer solutions has been discussed in a num
ber of reports [54,73-77]. Mechanical degradation is a process the mechanical
action in the polym er chain (due to flows forces) exceeds the chemical activation
energy in the polym er chain and bond rupture occurs. In the numerous investiga
tions some of the results are in conflict. Apparently, the polymer scissions tend to
D R A G R E D U C T I O N B Y P O L Y M E R ADDITIVES T O T U R B U L E N T F L O W S Y S T E M S 30.21
R E L A T IV E
DRAG
R E D U C T IO N
NUMBER OF P A S S E S
occur around the chain midpoint. This implies that the polymers are fully extended
before their breakage. At this time there is no definitive theoretical model for
polymer degradation in turbulent flow.
The mechanism(s) for degradation are beginning to be understood [12]. There
exists an enormous amount of degradation literature [27,54,74,78-100]. These
studies attempt to relate molecular properties with time-dependent friction losses
of flow systems. Additional theoretical studies are needed to eliminate degradation
effects as a constraint to polymer usage for drag reduction applications.
The enzymatic degradation of natural water-soluble polymers has recently been
investigated using guar galactomannan as a surrogate [101-104]. This study helps
to predict guar gum solution viscosity. A unique correlation was developed between
molecular weight, viscosity and degradation time. Degradation is also related to poly
mer extensional flow behavior [105-107].
30.10 Applications
The addition of polymers to systems with turbulent How for drag reduction pur
poses requires not only the technical benefit of reduced energy consumption, but
also economic and environmental benefits.
30.22 FLUID F L O W H A N D B O O K
Many possible applications of drag reduction with polymer additives have been
suggested over the last 50 years. Most of these suggestions have remained commer
cially unexploited because of two reasons. Economics, material costs for polymer
additions are too high compared to pumping power savings; plus any polymer addi
tions would require injection equipment and controls that need to be purchased.
Performance, the drag reduction effectiveness of polymer additions, is curtailed by
rapid shear degradation. With regard to adverse environmental impact, the poly
mer/additives are relatively innocuous.
In spite of the two limitations, costs and degradation, there are commercial appli
cations for polymeric drag reducing agents. Polymers are being used to transport
crude oil (20 years) and in oil well fracturing operations (40 years). Other sug
gested applications are listed in Table 30.4.
Research of new' applications is ongoing by polymer manufacturers and aca
demicians throughout the world. The only two successful commercial applications
of polymers for drag reduction are described next.
Field Date
Pipeline
Crude Oil
Trans-Alaska 1979-present
Off-shore Platforms present
Refined Petrol Products 1981
Slurry
Fly-Ash 1982/inactive
Coal 1986/inactive
Water
Fire Fighting 1971/inactive
Hydraulic Capsule [121-124] 1995/research
Open Channel
Flood Water
Irrigation
Marine
Ship
Instruments
Hydrofoils
Submarines
Torpedoes
Miscellaneous
Oil-Reservoir Fracturing 1963-present
Cavitation Noise
Medical
Jet Cutting
Hydraulic Machines
Aircraft Refueling
D R A G R E D U C T I O N B Y P O L Y M E R ADDITIVES T O T U R B U L E N T F L O W S Y S T E M S 30.23
Crude Oil Pipelines [16, I OH-112]. For twenty years, polyalphaolefins have been
used to save energy in the pumping of crude oil in the Trans-Alaska Pipeline.
Now, with the polym er concentration at about 1 ppm the oil flow rate is increased
by 33%. Initially, the polymer concentration was a factor of ten greater. At the
higher concentration, the initial drag reduction tests were sufficiently favorable that
two additional pumping stations were not constructed. Over the last 20 years the
straight-chain alpha-olefin has been improved. For example, the molecular weight
has been increased to more than 2 0 x 1 0 6 .
In the Trans-A laska Pipeline there are 25 injection sites, each located dow n
stream o f pumping stations, along the 1300 km pipeline. The pipeline is 1.3 meters
in diameter. “Polymer additives are being used in pipeline segments from a few
hundred meters up to over 400 km in length” [ 1111.
The application of polyalphaolefins in the Trans-Alaska Pipeline results in no
adverse environmental impacts. Drag reduction increases oil flow rate of an exist
ing pipeline. But, drag reduction reduces heat transfer rates between the oil and the
ambient air (or seawater, etc.). Thus, in the transportation of oil its cooling is reduced.
The crude oil remains warmer in the pipeline. This fact lowers oil viscosity which
results in greater savings in pumping power.
Today, the application of polymer additives for crude oil transportation also
includes many off-shore pipelines in different parts of the world. For this field, a
series of practical articles on drag reduction was published by Lester in 1985
[40,113-1 15]. Table 30.3 lists practical formula to calculate drag reduction per
centage for given polymer concentration.
Oil-Well Fracturing [8, 16]. Since the 1960s, polymers were added to drilling
fluids used to open oil reservoirs. The polymer additives reduced the friction losses
in the downpipe. This allowed the recovery of oils from greater depths using the
same size pipe. This application is ongoing.
Ten previously suggested and tested applications of polymer additives for
drag reduction are mentioned (in alphabetical order) below [16].
1. Biomedical [8]. The use of polymers for drag reduction has been tested in
human blood to improve flow during transfusions and in occluded tubes
(arteries). Such tests have been conducted since the late 1960s. In the early
1970s the University of Akron had a 5-year program in place on drag
reduction in blood flow. One polymer, polyethylene oxide has been tried at
low concentrations, between 5 and 100 ppm.
2. Firefighting [8]. One of the first tested concepts for the application of poly
ethylene oxide for drag reduction was in pumpers of the New York Fire
Department. The use of polyethylene oxide decreased pumping power,
increased throw, allowed longer hose lay, and higher delivery rates. The
polyethylene oxide also enhanced the coherence of the water jets.
3. Hydraulic machines [8J. In the early 1970s, a few tests o f drag reducing
polymers were conducted that indicated that the performance of centrifugal
30.24 FLUID F L O W H A N D B O O K
attenuated by using the drag reducing agent in a slurry form [ I 16-120]. The poly
mer in powder would dissolve, replenishing the dissolved polymer that degraded
due to mechanical shear, chemical or temperature effects.
Greek
A difference
8 slope increment, flow
£ density
n viscosity, absolute
V viscosity, kinematic
T shear stress
Q onset constant
Subscripts
1 liquid, untreated
p polymer solution
30.26 FLUID F L O W H A N D B O O K
w wall
L length, plate
g gyration
T turbulence
Superscripts
' fluctuation, velocity
30.11 REFERENCES
1. Lumley, J.L. 1969. “ Drag Reduction by Additives.” A nnual Rev. F luid M echanics
367-384.
2. Patterson, G.K., Zakin, J.L. and Rodriguez, J. M. 1969. “ Drag Reduction: Polymer
Solutions, Soap Solutions and Solid Particle Suspensions in Pipe Flow.” Ind. Eng.
Chem. 61: 22-30.
3. Gadd, G.E. 1971. “ Friction Reduction,” E ncyclopedia o f P olym er Science and
Technology , Vol. 15: 225-253. Wiley, New York.
4. Liaw, G.C., Zakin, J.L. and Patterson, G.K. 1971. “ Effects of Molecular Character
istics of Polymers on Drag Reduction.” A lC h E J. 17 (2): 391-397.
5. Govier, G.W. and Aziz, K. 1972. The F low o f C om plex M ixtures in Pipes. Van
Nostrand Reinhold, New York.
6. Hoyt, J.W. 1972. “ The Effect of Additives on Fluid Friction.” Trans. ASM E, J.
Basic Engin. 94: 258-285.
7. Virk, P.S. 1975. “ Drag Reduction Fundamentals A lC h E J. 21 (4): 625-656.
8. White, A. and Hemmings, J.A.G. 1976. “ Drag Reduction by Additives— Review
and Bibliography.” BHRA Fluid Engineering: Bedford, England.
9. Shenoy, A.V. 1984. “ A Review on Drag Reduction with Special Reference to
Micellar Systems.” Colloid & P olym er Science 262: 319-337.
10. Hoyt, J.W. 1985. “ Turbulent-Flow Interactions and Drag Reduction,” Proc. AIP
Conf., Number 137, Rabin, Y. (Ed.) New York: American Institute of Physics.
11. Hoyt, J.W. 1986. “ Drag Reduction,” E ncyclopedia o f Polym er Science a n d E ngi
neering , Vol. 5: 129-151, 2nd Ed. Wiley, New York.
12. Kulicke, W.M., Kotter, M. and Grager, H. 1989. “ Drag Reduction Phenomenon
with Special Emphasis on Homogenous Polymer Solutions.” A dvances in Polym er
Science 89: 1-68. Springer-Verlag, Berlin.
13. Matthys, E.F. 1991. “ Heat Transfer, Drag Reduction, and Fluid Characterization for
Turbulent Flow of Polymer Solutions: Recent Results and Research Needs.” J. Non-
Newtonian Fluid M echanics 38: 313-342.
14. Patterson, G.K. 1985. “ Model for Effects of Degradation on Polymer Drag Reduc
tion,” The Influence o f P olym er A dditives on Velocity a n d Tem perature F ields,
Gampert, B. Ed. Springer-Verlag, Berlin, pp. 173-179.
15. Wilson, K.C. 1989. “Two Mechanisms for Drag Reduction,” D rag R eduction in
Fluid Flows: Techniques fo r Friction C ontrol , Sellin, R.H J. and Moses, R.T. (Eds.)
Chapter 1.1. Ellis Horwood. (Preprint), Chichester.
D R A G R E D U C T I O N B Y P O L Y M E R ADDITIVES T O T U R B U L E N T F L O W S Y S T E M S 30.27
16. Gyr, A. and Bewersdorff, H.W. 1995 “ Drag Reduction of Turbulent Flows by Addi
tives.” Dordrecht: Kluwer Academic.
17. Oldroyd, J.G. 1948. “ A Suggested Method of Detecting Wall-Effects in Turbulent
Flow Through Tubes." Proc. International Rheological Conf., Vol. II: 130-134.
18. Toms, B.A. 1948. “ Some Observations of the Flow of Linear Polymer Solutions
Through Straight Tubes at Large Reynolds Numbers." Proc. International Rheologi
cal Congress, Vol. II: 135-141.
19. Hoyt, J.W. 1965. “ A Turbulent-Flow Rheometer,” Symposium on Rheology. Mar-
ris, A.W. and Wang, J.T.S. (Eds.) New York: ASME, pp. 71-82.
20. Fortuna, G. and Hanratty, T. J. 1971. “ Use of Electrochemical Techniques to Study
the Effect of Drag-Reducing Polymers on Flow in Viscous Sublayer.” Drag Reduc
tion, AIChE Symposium Series III, Vol. 67, pp. 90-92.
21. Hand, J.H. and Williams, M.C. 1971. “The Role of Polymer Conformation in Drag
Reduction.” Drag Reduction, AIChE Chemical Engineering Progress Symposium
Series III, Vol. 67: 6- 8.
22. Hansen, R.J. and Little, R.C. 1971. “ Pipe Diameter, Molecular Weight, and Concen
tration Effects on the Onset of Drag Reduction.” Drag Reduction, AIChE Chemical
Engineering Progress Symposium Series III, Vol. 67: 93-97.
23. Mysels, K.J. 1971. “ Early Experiences with Viscous Drag Reduction.” Drag Reduc
tion, AIChE Chemical Engineering Progress Symposium Series III, Vol. 67: pp.
45-49.
24. Savins, J.G. and Virk, P.S., Eds. 1971 “ Drag Reduction.” Chemical Engineering
Progress Symposium Series III. Vol. 67, New York: American Institute of Chemical
Engineers.
25. Gordon, R.J., Balakrishnan, C. and Pahwa, S. 1973. “ Importance of Filament Forma
tion in Turbulent Drag Reduction.” AIChE Symposium Series 130, Vol. 69: 33-37.
26. Kumor, S.M. and Sylvester, N.D. 1973. “ Effects of a Drag-Reducing Polymer on
the Turbulent Boundary Layer.” AIChE Symposium Series 130, Vol. 69: 1-12.
27. Sylvester, N.D. and Kumor, S.M. 1973. “ Degradation of Dilute Polymer Solutions
in Turbulent Tube Flow.” AIChE Symposium Series 130, Vol. 69: 69-80.
28. Meng, J.C.S. 1998 “ Proceedings of the International Symposium on Seawater Drag
Reduction.” Washington, DC: U.S. Office of Naval Research.
29. Kajishima, T. and Miyake, Y. 1999. “ DNS of Drag Reduction by Polymer Addi
tive.” Proc. 3rd ASME/JSME Joint Fluids Engineering Conf. July 18-23, 1999, San
Francisco, CA, FEDSM99-7796.
30. Toms, B.A. 1977. “ On the Early Experiments on Drag Reduction by Polymers.”
Phys. Fluids 20 (10, Pt. 11) S3-S5.
31. Dodge, D.W. and Metzner. A.B. 1959 “Turbulent Flow of Non-Newtonian System.”
A IC hE J. 5: 189-204.
32. Shaver, R.G. and Merril, E.W. 1959 “ Turbulent Flow of Pseudoplastic Polymer
Solutions in Straight Cylindrical Tubes.” A IC h E J. 5: 181-188.
33. Oliver, D.R. and Bakhtiyarov, S.I. 1983 “ Drag Reduction in Exceptionally Dilute
Polymer Solutions.” J. N on-N ew tonian Fluid M ech. 12: 113-118.
34. Savins, J.G. 1964 “ Drag Reduction Characteristics of Solutions of Macromolecules
in Turbulent Pipe Flow.” Soc. Petrol. Eng. J. 4: 203-214.
35. Virk, P.S. 1971. “ Drag Reduction in Rough Pipes.” J. Fluid M echanics 45 (Pt. 2):
225-246.
36. Naudascher, E. and Killen, J.M. 1977 “ Onset and Saturation Limit of Polymer
Effects in Porous Media Flows.” Phys. Fluids 20 (10, Pt. II): S280-S283.
30.28 FLUID F L O W H A N D B O O K
37. Seyer, F.A. and Metzner, A.B. 1969. “Turbulence Phenomena in Drag Reducing
Systems ." A I C h E J . 15:426-^434.
38. Gordon, R. J. and Balakrishnan, C. 1972. “ Vortex Inhibition: A New Viscoelastic
Effect with Importance in Drag Reduction and Polymer Characterization." J.
A pplied P olym er Sci. 16: 1629-1639.
39. Moody, L. F. 1944. “ Friction Factors for Pipe Flow.” Trans A SM E 66: 671-684.
40. Lester, C.B. 1985 “The Basics of Drag Reduction ” Oil G as J. 83 (5): 51-56.
41. Virk, P.S., Merrill, E.W., Mickley, H.S., Smith, K.A. and Mollo-Christensen, E.L.
1967. “ The Toms Phenomenon: Turbulent Pipe Flow of Dilute Polymer Solutions.”
J. Fluid M ech. 30 (part 2) 305-328.
42. Virk, P.S., Mickley, H.S. and Smith, K.A. 1970. “ The Ultimate Asymptote and
Mean Flow Structure in Toms’ Phenomenon.” Trans. ASM E, J. Appld. M echanics
92 (2): 488-493.
43. White, W.D. 1969. “ Drag-Reduction Measurements for Three Polymers at 4°C,”
Viscous D rag R eduction , C.S. Wells (Ed.) Plenum, New York. pp. 173-181.
44. Smith, R., Edwards, M.F. and Wang, HZ. 1982. “ Pressure Drop and Mass Transfer
in Dilute Polymer Solutions in Turbulent Drag-Reducing Pipe Flow.” Int. J. Heat
M ass Transfer 25 (12): 1869-1878.
45. Mizunuma, H. and Kato, H. 1983. “ Frictional Resistance in Fiber Suspensions.”
Bulletin J S M E 26 (219): 1567-1574.
46. Usui, H., Kodama, M. and Sano, Y. 1988. “ Laser-Doppler Measurements of Turbu
lence Structure in a Drag-Reducing Pipe Flow with Polymer Injection.” J. Chem
Eng. Japan 21 (2): 134-140.
47. Lodesova, D. and Lodes, A. 1989. “ Thickness of the Elastic Sublayer in the Veloc
ity Field of Diluted Polymeric Solutions.” Exp.Fluids 1: 379-382.
48. Tam, K. C., Tiu, C. and Keller, R.J. 1992. “ A General Correlation for Turbulent
Velocity Profiles of Dilute Polymer solutions.” J. H ydraulic Research 30 (1):
117-142.
49. Vlassopoulos, D. and Schowalter, W.R., Eds. 1993 “ Characterization of the Non-
Newtonian Flow Behavior of Drag-Reducing Fluids.” J. Non-N ew tonian Fluid
M echanics 49: 205-250.
50. Vlassopoulos, D. and Schowalter, W.R. 1994. “ Steady Viscometric Properties and
Characterization of Dilute Drag-Reducing Polymer Solutions.” ./. R heology 38 (5):
1427-1446.
51. Choi, H.J. and John, M.S. 1996 “ Polymer-Induced Turbulent Drag Reduction.” Ind.
Eng. Chem. Res. 35: 2293-2998.
52. Interthal, W. and Wilski, H. 1985. “ Drag Reduction Experiments with Very Large
Pipes.” C olloid and Polym er Science 263: 217-229.
53. De Gennes, P.G. 1986. “Towards a Scaling Theory of Drag Reduction.” Physica
140A: 9-25.
54. Nguyen, T.Q. and Kausch, H.H. 1992 “ Mechanochemical Degradation in Transient
Elongational Flow.” Adv. Polym. Sci. 100: 73-182.
55. Anderson, G.W., Rohr, J.J. and Hoyt, J.W. 1996. “ An Experimental Investigation of
Polymer Drag-Reduction Scale-Up.” ASME, Fluids Engineering Division Conf.,
Vol. 237, Vol. 2, pp. 19-24.
56. Sood, A. and Rhodes, E. 1998 “ Pipeline Scale-Up in Drag Reducing Turbulent
Flow.” Canadian J. Chem. Eng. 76: 11-18.
57. Meyer, W.A. 1966. “ A Correlation of the Frictional Characteristics for Turbulent
Flow of Dilute Viscoelastic Non-Newtonian Fluids in Pipes.” AIC hE J. 12: 522-525.
D R A G R E D U C T I O N B Y P O L Y M E R ADDITIVES T O T U R B U L E N T F L O W S Y S T E M S 30.29
58. Everage, A.E. and Gordon. R.J. 1972. “ Modified Bead-Spring Theory of Dilute
Polymer Solutions II. Effects of Polydisperity and Comparison with Experiment."
J. A p p lied Polym er Sci. 16: 1967-1982.
59. Kelkar, J.V. and Mashelkar, R.A. 1972. “ Drag Reduction in Dilute Polymer Solu
tions” J. Appl. P olym er Sci. 16: 3047-3062.
60. Parker, C.A. and Hedley, H. 1974. “ A Structural Basis for Drag-Reducing Agents.”
J. Appl. P olym er Sci. 18: 3403-3421.
61. Little, R.C., Hansen, R.J., Hunston, D.L., Kim, O.K., Patterson, R.L. and Ting, R.Y.
1975. “ The Drag Reduction Phenomenon. Observed Characteristics, Improved
Agents, and Proposed Mechanisms.” Ind. Eng. Chem., Fundam. 14 (4): 283-296.
62. Berman, N.S. 1977. “ Flow Time Scales and Drag Reduction.” Phys. Fluids 20 (10,
Pt.II): S168-S174.
63. Dunlop, E.H. and Cox, L.R. 1977. “ Influence of Molecular Aggregates on Drag
Reduction.” Phys. Fluids 20 (10, Pt. II): S203-S213.
64. Hinch, E.J. 1977. “ Mechanical Models of Dilute Polymer Solutions in Strong
Flows.” Phys. Fluids 20 (10): S22-S30.
65. Tiu, C. and Chee, N.O. 1979. “Turbulent Flow Behaviour of Dilute Polymer Solu
tions in an Annulus.” Canadian J. Chem. Eng. 57: 572-577.
66. Darby, R. 1985. “ An Engineering Approach to Modeling Complex Flow Behavior
in Polymer Solutions,” The Influence o f P olym er Additives on Velocity a n d Temper-
ative Fields , Gampert, B. (Ed.) pp. 325-334. Springer-Verlag, Berlin.
67. Shenoy, A.V. and Shintre, S.N. 1986. “ Developing and Fully Developed Turbulent
Flow of Drag Reducing Fluids in an Annular Duct.” Canadian J. Chem. Eng. 64:
190-195.
68. Ryskin, G. 1987. “ Turbulent Drag Reduction by Polymers: A Quantitative Theory.”
Phys. Rev. U tte r s 59 (18): 2059-2062.
69. De Gennes, P.G. 1990 “ An Elastic Theory of Degradation,” Introduction to P oly
m er D ynam ics , Chapter 4, Cambridge University, Cambridge.
70. McCormick, C.L., Hester, R.D., Morgan, S.E. and Safieddine, A.M. 1990. “ Water-
Soluble Copolymers. 30. Effects of Molecular Structure on Drag Reduction Effi
ciency.” M acrom olecules 23: 2124-2131.
71. Malik, S. and Mashelkar, R.A. 1995 “ Hydrogen Bonding Mediated Shear Stable
Clusters as Drag Reducers.” Chem. Eng. Sci., 50: 105-116.
72. Anonymous. 1971. “ Using Polyox® Water-Soluble Resins to Reduce Hydrody
namic Drag.” Union Carbide Corp., Bulletin F43273.
73. Anonymous. 1996. “ Polyox® Water-Soluble Resins: Dissolving Techniques.”
Union Carbide Corp.
74. Brostow, W. 1983. “ Drag Reduction and Mechanical Degradation in Polymer Solu
tions in Flow.” Polym er 24: 631-638.
75. Moussa, T. and Tiu, C. 1994 “ Factors Affecting Polymer Degradation in Turbulent
Pipe Flow.” Chem. Eng. Sci., 49: 1681-1692.
76. Den Toonder, J.M.J., Draad, A.A., Kuiken, G.D.C. and Nieuwstadt, F.T.M. 1995.
“ Degradation Effects of Dilute Polymer Solutions on Turbulent Drag Reduction in
Pipe Flows.” A pplied Scientific Research 55: 63-82.
77. McCoy, B.J. and Madras, G. 1997. “ Degradation Kinetics of Polymers in Solution:
Dynamics of Molecular Weight Distributions.” A IC hE J. 43: 802-810.
78. Kim, C.A., Kim, J.T., Lee, K., Choi, H.J. and Jhon, M. S. 2000. “ Mechanical
Degradation of Dilute Polymer Solutions Under Turbulent Flow.” P olym er 41:
7611-7615.
30.30 FLUID F L O W H A N D B O O K
79. Gadd, G. E. 1965. “ Turbulence Damping and Drag Reduction Produced by Certain
Additives in Water.” N ature Vol. 206 (4983): 463^67.
80. Paterson, R.W. and Abernathy, F.H. 1970. “Turbulent Flow Drag Reduction and
Degradation with Dilute Polymer Solutions.” J. F luid Mech. 43 (Pt. 4): 639-710.
81. Fisher, D.H. and Rodriguez, F. 1971. “ Degradation of Drag-Reducing Polymers.”
J . A pplied Polym er Sci. 15: 2975-2985.
82. Hoyt, J.W. 1972. “Turbulent Flow of Drag-Reducing Suspensions.” Naval Undersea
Center, NUC TP 299, AD 746485.
83. Ting, R.Y. and Little, R.C. 1973. “ Characterization of Drag Reduction and Degra
dation Effects in the Turbulent Pipe Flow of Dilute Polymer Solutions.” J. Appl.
P olym er Sci. 17: 3345-3356.
84. De Gennes, P.G. 1974. “ Coil-Stretch Transition of Dilute Flexible Polymers Under
Ultrahigh Velocity Gradients.” J. Chem Phys. 60 (12): 5030-5042.
85. Durst, F. and Rastogi, A.K. 1977. “ Calculations of Turbulent Boundary Layer
Flows with Drag Reducing Polymer Additives.” Phys. Fluids 20 (12): 1975-1984.
86. Sedov, L.I., Ioselevich, V.A., Pilipenko, V.N. and Vasetskaya, N.G. 1979. “Turbu
lent Diffusion and Degradation of Polymer Molecules in a Pipe and Boundary
Layer.” J. F luid Mech. 94 (3): 561-576.
87. Vasetskaya, N.G., Ioselovich, V.A. and Pilipenko, V.N. 1979. “ Mechanical Degra
dation of Polymer Molecules in Turbulent Flow.” Fluid M echanics —Soviet
Research 8 (6): 58-74.
88. Lee, H.S., Irvine, T.F. and Kwack, E.Y. 1980. “ Anomalous Effects in Drag Reduc
tion Measurements of Polymer Solutions.” Letters in H eat M ass Transfer 7: 7-13.
89. Gampert, B. and Wagner, P. 1982. “ Turbulent Flow with Polymer Additives.”
Archives o f M echanics (W arszawa) 34 (4): 493-502.
90. Lee, H.S. and Irvine, T.F. 1984. “ Degradation Effects and Entrance Length Studies
for a Drag Reducing Fluid in a Square Duct.” Chem. Eng. Commun. 25: 79-91.
91. Cho, Y.I. and Hartnett, J.P. 1985. “ Non-Newtonian Fluids,” Handbook of Heat
Transfer Applications, 2nd Ed. Rohsenow, W.M., Hartnett, J.P. and Ganic, E.N.
(Eds.) Chapter 2. McGraw-Hill, New York.
92. Gampert, B. and Wagner, P. 1985. “The Influence of Molecular Weight and Molec
ular Weight Distribution on Drag Reduction and Mechanical Degradation in Turbu
lent Flow of Highly Dilute Polymer Solutions,” The Influence o f P olym er Additives
on Velocity and Tem perative F ields , Gampert, B. (Ed.) pp. 73-84. Springer-Verlag,
Berlin.
93. Kulicke, W.M. 1985. “ Aging of Aqueous Polymer Solutions,” The Influence o f
Polym er Additives on Velocity and Tem perature F ields , Gampert, B. (Ed.)
pp. 163-171. Springer-Verlag, Berlin.
94. Nguyen, T.Q. and Kausch, H.H. 1986. “ Degradation of a Polymer Solution in Tran
sient Elongational Flow: Effect of Temperature.” C olloid & P olym er Sci. 264:
764-772.
95. Donbrow, M. 1987. “ Stability of the Polyoxyethylene Chain,” Nonionic Surfactants:
Physical C hem istry , Schick, M. J., Ed, pp. 1101-1072. Marcel Dekker, New York.
96. Brostow, W., Ertepinar, H., and Singh, R.P. 1990 “ Flow of Dilute Polymer Solu
tions: Chain Conformations and Degradation of Drag Reducers.” M acrom olecules
23:5109-5118.
97. Jou, D., Casas-Vazquez, J. and Criado-Sancho, M. 1995. “ Thermodynamics of
Polymer Solutions Under Flow: Phase Separation and Polymer Degradation.”
A dvances in P olym er Science 120: 207-266.
D R A G R E D U C T I O N B Y P O L Y M E R ADDITIVES T O T U R B U L E N T F L O W S Y S T E M S 30.31
98. McCormick, C.L., Morgan. S.E. and Hester, R.D. 1991. “ Roles of Molecular Struc
ture and Solvation on Drag Reduction in Aqueous Solutions,” W ater-Soluble Poly
mers, Shalaby, S.W., McCormick, C.L. and Butler. G.B. (Eds.) ACS Symposium
Series, No. 467, pp. 320-337.
99. Wang, M., Smith, J.M. and McCoy, B.J. 1995. “ Continuous Kinetics for Thermal
Degradation of Polymer in Solution.” A IC hE J. 41: 1521-1533.
100. Clay, J.D. and Koelling, K.W. 1997. “ Molecular Degradation of Concentrated Polysty
rene Solutions in a Fast Transient Extensional Flow.” Polym er Eng. Sci. 37: 789-800.
101.Kenis, P.R. 1971. “Turbulent Flow Friction Reduction Effectiveness and Hydrody
namic Degradation of Polysaccharides and Synthetic Polymers.” J. A pplied Polym er
Science 15: 607-618.
102.0zari, Y. 1981. “ Friction Reduction Process.” Am er. Chem. Soc., Div. Polym. Chem.
22(1): 145-6.
103. Deshmukh, S.R. Sudhakar, K. and Singh, R.P. 1991. “ Drag-Reduction Efficiency,
Shear Stability, and Biodegradation Resistance of Carboxymethyl Cellulose-Based
and Starch-Based Graft Copolymers.” J. Appld. Polym. Sci. 43: 1091-1101.
104.Tayal, A., Kelly, R.M. and Khan, S.A. 1999. “ Rheology and Molecular Weight
Changes During Enzymatic Degradation of a Water-Soluble Polymer.” M acrom ole
cules 32: 294-300.
105.Metzner, A.B. and Metzner, A.P. 1970. “ Stress Levels in Rapid Extensional Flows
of Polymeric Fluids.” Rheologia A cta Vol. 9 (2): 174-181.
106. James, D.F., McLean, B.D., and Saringer, J.H. 1987. “ Presheared Extensional Flow
of Dilute Polymer Solutions.” J. R heology 31 (6): 453-481.
107. Lee, E.C. and Muller, S.J. 1999. “ Flow Light Scattering Studies of Polymer Coil
Conformation in Solutions in Extensional Flow.” M acrom olecules 32: 3295-3305.
108. Ram, A., Finkelstein, E. and Elata, C. 1967 “ Reduction of Friction in Oil Pipelines
by Polymer Additives.” Ind. Eng. Chem. Proc. Des. Dev. 6(3): 309-313.
109. Brod, M., Deane, B.C., and Rossi, F. 1971 “ Field Experience with the use of Additives
in the Pipeline Transportation of Waxy Crudes.” J. Inst. Petrol. 57 (554): 110-116.
110. Bzailevich, V.A. and Shabrin, A.N. 1971 “ Reducing the Hydraulic Drag in Pipe
lines with Polymer Additives.” F luid M echanics Sov. Res. 1(5): 59-67.
111. Lescarboura, J.A., Culter, J.D., and Wahl, H.A. 1971 “ Drag Reduction with a Poly
meric Additive in Crude Oil Pipelines.” Soc. Petroleum Engrs. J. 11(3): 229-235.
I 12. Nijs, L. 1995. “ New Generation Drag Reducer.” Pipeline Technology , Denys, R.
(Ed.) Vol. II, Amsterdam: Elsevier Science pp. 143-149.
113. Lester, C.B. 1985 “ Here’s How Drag-Reducing Agents Outweigh Looping or
Boosting.” O il Gas J. 83(7): 76-80.
114. Lester, C.B. 1985 “ How Active, Passive Drag Affect DRA Injections.” Oil G as J.
83(9): 107-110.
115. Lester, C.B. 1985 “ What to Expect From and How to Handle Commercially Avail
able Drag-Reducing Agents.” O il Gas J. 83(10): 116-122.
116. Lumley, J.L. 1977. “ Drag Reduction in Two Phase and Polymer Flows.” Phys. F lu
ids 20 (\0, Pt. 11): S64-S71.
117. Little, R., Smidt, S. Huang, P., Romans, J., Dedrick, J., and Matuszko, J. S. 1991.
“ Improved Drag Reduction by Control of Polymer Particle Size.” Ind. Eng. Chem.
Res. 30: 403-407.
118. Bewersdorff, H.W., Gyr, A., Hoyer, K., and Tsinober, A. 1993. “ An Investigation
of Possible Mechanisms of Heterogeneous Drag Reduction in Pipe and Channel
Flows.” Rheologica A cta 32: 140-149.
30.32 FLUID F L O W H A N D B O O K
119. Mamonov, V.N., Mironov, B.P., Mustafaev, R.F. and Ginzburg, D.I. 1995. “ Reduc
tion of Hydrodynamic Friction Drag Using Quick-Prepared ‘Solutions’ of Polyeth
ylene Oxide.” P hysics-D oklady 40 (3): 138-141.
120. Hoyer, K.W. and Gyr, A. 1996. “ Heterogenous Drag Reduction Concepts and
Consequences.” ASME, Fluids Engineering Division Conf., Vol. 237, Vol. 2,
pp. 151-158.
121. Vlasak, P. 1995. “ The Toms Effect in Capsule-Liquid Flow." Proc. 8th Intl. Freight
Pipeline Symposium, Pittsburgh, PA.
122. Huang, X., Liu H., and Marrero, T.R. 1997. “ Polymer Drag Reduction in Hydraulic
Capsule Pipeline.” A IC h E J. 43: 118-1121.
123. Wu, G., Xu, J., and Miles, J. 1998. “ Polymer Drag Reduction in Large Diameter
Coal Log Pipeline.” Proc. 23rd International Techn. Conf. Coal Utilization Fuel
Sys., Clearwater, FL, pp. 889-900.
124. Marrero, T.R. and Kuhlman, G.S. 2000. “ Drag Reduction in Hydraulic Capsule
Pipelines.” 10th International Freight Pipeline Symposium, May 29-June 1, 2000,
The Dead Sea, Israel. See also: 2001. Handbook o f Conveying and H andling o f
P articulate Solids , Levy, A. and Kalman, H. Eds. pp. 513-520. Elsevier Science,
Amsterdam.
CHAPTER 31
TURBULENT FLOW
INTRODUCTION
31.1
3 1 .2 FLUID F L O W H A N D B O O K
31.1 CHARACTERISTICS OF
TURBULENT FLOWS
over the members of the ensemble, at a fixed time measured from the corre
sponding time origin, is called ensemble averaging, to be denoted by brackets
(e.g. < U i > ) .
A random process is called stationary, if its statistical properties are not affected
by a shift in the time origin. Similarly, two random processes are called jointly sta
tionary, if their joint statistical properties are not affected by a shift in the time ori
gin. For a stationary process, a time average, to be denoted by an overline (e.g. U{),
can be determined from each time series. In general, each time series of a station
ary random process would give a different time average, such that this average would
be a random variable. A stationary random process is called ergodic if the ensem
ble average of any of its time series is equal to the corresponding time average.
A turbulent flow is called homogeneous if its statistical properties are inde
pendent of translations of the coordinate system. It is called isotropic if its statis
tical properties are independent of rotations and/or reflections of the coordinate
system. A homogeneous and isotropic flow cannot be stationary, but when non-
stationarity is sufficiently weak, time averages may be defined over a properly
selected, finite time interval.
^ = 0
OXj
The above equations are subject to Einstein’s summation convention, by which each
term that contains a repeated index is summed over that index (e.g. w,w, - u xu\ +
u2u2 + w3w3 = u\ + u \ + u 3 ); Kronecker’s delta is defined as Sy = 1, if i = j, and
31.4 FLUID F L O W H A N D B O O K
d 2U; d 2Uj
+ vl oxk oxk > + <UJ dxk
T ~ oxk
*7>
dk dk d<Uj> d p d
— + <Uk> - — = - <UjUj> — ------------ — <uj - > - —— <u.k>
dt dxk dxj dxj p dxj
d 2k d 2<UiU;> l
+ v —— - — + v — — ----------- e
dxjdxj dxjdxj 2
where the turbulent kinetic energy dissipation rate per unit volume is defined as
f t
* \
a m *
• • •
4
<
+ «2
* L
J
54 *
& V*
* 4
eddies, until eddies small enough to be dominated by viscous actions are produced.
The latter eddies are hydrodynamically stable and their kinetic energy is dissi
pated by friction with other eddies and converted irreversibly to random molecular
motions, that is to heat. As the Reynolds number increases, the number of stages
in the energy cascade becomes sufficiently large for the motion of the small eddies
to be plausibly decoupled from the macroscopic structure of the flow. Thus, the
statistical properties of the fine structure of large Reynolds number turbulence
would likely be independent of axis orientation, which justifies the designation of
such flows as locally isotropic.
the inertial range of the three-dimensional spectrum (see Section 3 1.4.4) is not
exactly -5 /3 , but has a smaller magnitude that depends on the Reynolds number
value [7].
A main criticism of Kolm ogorov’s scaling laws is that they assume a constant
dissipation rate, averaged over a volume comparable to the size of the energy con
taining eddies (see Section 3 1.4.4). However, the fine structure which contributes
most to the energy dissipation is not uniformly distributed in space, but it is con
centrated on thin regions within the turbulent fluid. Across these regions, the local
velocity gradient is very steep, causing excessive frictional forces and, thus, large
energy dissipation. The phenomenon of spotty spatial distribution of vorticity and
dissipation rate (Figure 31.2) is known as internal intermittency. An indication of
internal intermittency is the strong non-Gaussian nature of the pdf of velocity deriv
atives, in sharp contrast to the pdf of the velocity itself, which is often essentially
Gaussian. Instantaneous velocity profiles across a turbulent flow display ramp-like
appearance with sharp drops followed by gradual increases (Figure 31.2). Various
u ->
X1
duj/dx ->
Xj
>
x
FIGURE 31.2 Illustration of the ramp-like structure of the velocity fluctuations, the variation of the
streamwise velocity derivative and the spotty character of the turbulent kinetic energy dissipation rate.
3 1 .8 FLUID F L O W H A N D B O O K
corrections to the Kolmogorov scaling laws and alternative approaches have been
suggested, but this issue remains a subject of intense research.
R"<ri = <„?«>
• temporal cross-correlation coefficient
<Uj(t)U j(t + T)>
,00
= 2 < U i(t)U i(t + t)> c o s (o > t )< / t
'0
• frequency cross-spectrum (complex function)
Tij = [ R ij( z ) d t
Jo
L ij,k = \ K ,M k )d ?
Jo
• three-dimensional spectrum tensor (ic: w a v e n u m b e r v e c to r)
E( k ) = £ < & „ (£ ) do
• property
f 00 1
E(K)dK = - <UjUj> = k
Jo 2
• one-dimensional (streamwise) spectrum tensor
• property
1 3 £ , i ( k)
£ ( 0 = —k ^ ~ -
2 dk
1.00
Note that most references dealing with shear flows use different definitions
for the Taylor microscale and the turbulence Reynolds number:
(duy/du\)2
• streamwise turbulence Reynolds number
*A1 =
V
The turbulence Reynolds number is the best parameter by which one may
compare the level of turbulence activity in different flows. The turbulence in
flows with R m < 100 is considered relatively weak, and it is not expected to have
a measurable inertial subrange. For a flow to be considered strongly turbulent,
Rx\ must be greater than at least 1000. Most laboratory shear flows have Rx\ in
the range 100 to 1000, while in atmospheric and oceanic turbulence R ^ usually
exceeds 1000.
TURBULENT FLOW 31.11
Lf =
Jo
f f{r)dr
ii 3
£ -------
/
y ~ R, 1/2 ~ R ?
5 ~ R f 3'4 ~ R-,™
R - r^ - r -^
— ~ R ~ ]/4 ~ R ~ U2
u
TprU , ,
^ ~ R r l/2 ~ r ~,]
dE(K,t)
= T(K,t) - I v i f i K t )
dt
where the spectral energy transfer function T(K,t) is related to the Fourier trans
form of third-order two-point correlations.
Physical significance of terms:
• lhs: net rate of change of the turbulent kinetic energy at wave number k (cor
responding roughly to eddies of size l I k )
3 1.12 FLUID F L O W H A N D B O O K
• 1st term, rhs: net rate of energy transfer from all wave-numbers smaller than
k to all wave-numbers larger than k\ as a result of non-linear interactions
• 2nd term, rhs: rate of turbulent kinetic energy dissipation by viscous forces.
F IG U R E 31.4 Typical shapes of the energy spectrum, the spectral energy transfer func
tion and the dissipation spectrum in high Reynolds number isotropic turbulence.
TURBULENT FLOW 31.13
FIGURE 31.5 Typical shapes of the energy spectrum and the dissipation spectrum in relatively
low Reynolds number isotropic turbulence (left); effect of decay time on the shape of the energy
spectrum (right).
• The range of largest eddies (permanent eddies). These decay the slowest but
contain a very small amount of the total energy and are not dynamically
important. They follow the law
E( k) = I k 4, / = const., k « Ki ~ ~ j
FIGURE 31.6 A schematic representation of the different subranges of the energy spectrum
in large Reynolds number turbulence.
31.14 FLUID F L O W H A N D B O O K
These are flows that, ideally, would have no solid boundaries, while, in practice,
may have solid walls, but sufficiently far away not to affect the flow. Sketches of
the geometrical appearance of these flows and the corresponding mean velocity
profiles have been provided in Figure 31.7.
As there is no energy production, k will decay. One may use the following
decay laws:
!*■>£ L kL
r\*^> o
•» $
AU'
xc r thermal plume
x,
x 2 or 1
u .max
t jz-mrflinilTTTT 4 max
T
- T >
J
-yU n
2-D mixing layer
71U heat
- dk __________ d U x I
U i-----= - U \ U 2— ------ £
dxx dx2 2
~2 ~2 ~2
U\ U~> U3 U \U 2 n n
— = 1.02, — = 0.40, — = 0.58, —— - = - 0 .3 0
k k k k
x2 or
where / is the transverse width of the flow (see, for example, the sketch for a wake,
in Figure 3 1.7). For wakes and mixing layers, the mean velocity is normalized by
the maximum mean velocity difference AUmax at a certain streamwise location; for
jets in still fluid and thermal plumes, by the local maximum velocity Umax. Finally,
the effective turbulence Reynolds num ber is defined as
o ^ U maxl D Umaxl
A y = --------- or A y = -------
Vf Vf
• Two-dimensional wakes
^ = 1. 5 8 / - ^ - , - L ~ 0 .2 5 2 I - * - ,
t/o o V X, oM \ 8m 2
Heated, two-dim ensional wakes (H is the heat flux per unit span, p is the
fluid density and cP is the specific heat under constant pressure)
T - T ~ = e- m ? A Tmax = 1.58 H
A T m ax ’ Too pc„Ux Tx 8M V AT,
31.18 FLUID F L O W H A N D B O O K
Two-dimensional jets in still fluid (Uj is the jet speed at the exit of the nozzle)
^ = s e c l ,2( w ) NM:
11 / y- \ —1/2 /
_ 2-7(-£]_J ? _L = 0.078, R T =* 14.1
Uj \ d ) x,
Axisymmetric jets in still fluid
= = — = 0.067, /? r ~ 32
^m ax \ 4 / \ d ) Aj
yvnT^pCp) g
u
'-'m ax =1I gRlH
rr* Y
I
1n T „ p cp ]
V - ’ ArmoA" I
\ g n p cp
9 ? 1
tinct length scales, the viscous length v/uz and the much larger physical thickness
6 of the boundary layer. One may distinguish an inner region (typically, but not
universally, for y/S < 0 .1 , where y is the distance from the wall) and an outer
region (typically, but not universally, for y/5 > 0.1). In the inner region, viscous
stresses are significant by comparison to the Reynolds shear stresses and it is appro
priate to non-dimensionalize y by the viscous length, as y + = u Ty/v. Then one
may distinguish the following distinct regions near the wall, in which the dimen
sionless velocity = U/uT varies as described below.
• 5 < y + < 30: Buffer sublayer. This is an intermediate region for which there
is no theoretical expression. A num ber o f semi-empirical relationships have
been proposed for this sublayer, also valid in the adjacent sublayers, thus rep
resenting the entire inner boundary layer and referred to as the law of the wall.
Among the most popular laws of the wall, particularly in computational stud
ies, is based on van Driest’s expression for the dim ensionless mixing length
/+ = Ky+[ \ - e ~ r /A+]
where k is discussed in the next paragraph and A+ = 26. This provides the
following expression for the dim ensionless velocity
V i + 4 c 2oo
• 30 < y + : Logarithmic layer. The upper limit of this layer increases with
increasing Reynolds num ber from values below y + = 100 to values approach
ing 105 [32].
u+ = —In y* + B
K
where typical values (within 5% [2]) of the constants are k = 0.41
(von Karman constant) and B = 5.2.
d x #
wall
It is generally accepted that the above laws are independent of the conditions
away form the wall. In the outer region, the velocity variation depends on many
parameters, such as the shape of the wall and the pressure gradient. It is com
monly expressed as the sum of the law of the wall and an empirical function,
called the wake function. A more detailed discussion of turbulent boundary lay
ers has been included in Chapter 20.
In the case of a rough wall, the roughness height e (i.e. an average value of
the height of the roughness elements) is an additional length scale. The ratio of
the roughness height and the viscous length may be also viewed as a roughness
Reynolds number Re = u Te/v. When R(, < 5, the roughness elements are entirely
submerged in the viscous sublayer, the wall is called hydraulically smooth and
roughness plays no appreciable role in the flow development. On the other extreme,
when Re » 1, the roughness elements generate wakes with additional drag, the
wall is called fully rough and one may use a logarithmic law
u+ = L\nl + B
K e
where B' — 8.5 for u z/ v > 70. For intermediate values (5 < Re < 70) one may use
a similar expression, but with the value of the constant B' depending on Re. Addi
tional information on turbulent boundary layers may be found in section 20.2.4.
where AP is the frictional pressure loss in fully developed, turbulent flow in a pipe
with a diameter D and a length L\ Uh is the bulk velocity, namely the cross-sectional
average of the mean velocity in the pipe. A number of expressions are available for
the friction factor, taking into account various effects. For fully developed, incom
pressible flow in smooth circular pipes, one may use Prandtl’s friction law
TURBULENT FLOW 31.21
where the bulk Reynolds number is Rh = UhD/v. For rough walls, it is most conve
nient to use M oody’s diagram [33]. For ducts and channels with non-circular cross-
sectional shapes, one may substitute the diameter D by the hydraulic diameter Dh
= 4A/Wp, where A is the cross-section area and Wr is the wetted perimeter, namely
the length o f the part of the wall in contact with the fluid [33].
Despite their apparent randomness, turbulent shear flows contain recurring elements
with distinct features, called coherent structures. These usually appear in the form of
connected, large-scale turbulent fluid masses with a phase-correlated vorticity over
their spatial extent [34]. Coherent structures sometimes dominate the macroscopic
evolution of a flow but coexist with fine-grain, disorganized turbulence. Examples
of dominant coherent structures in turbulent shear flows include the following:
which is a linear relationship between the Reynolds stresses and the mean strain.
An illustration of its physical relevance is given in Figure 31.9. Assuming that a fluid
particle conserves its momentum as it wanders randomly in a flow with a mean
gradient (strain rate) d<U\>/dx2 > 0, it is easy to see that it will create a negative
Reynolds stress < u lu2>, because, if it moves towards higher mean speeds (w2 > 0),
it will on the average carry a velocity deficit (u\ < 0) and vice versa.
31.8.1.2 Two-Equation Models. These require the solution of two additional dif
ferential equations. The most common model is the k-e model that requires the
ju ^ 0
^ ' Uj< 0
UjU2<0
d \ j 1/ d x 2> 0
u2< 0
Uj>0
UjU2< 0
solution of modeled equations for the turbulent kinetic energy k and its dissipa
tion rate £. These are as follows.
• Turbulent viscosity
k2
VT ~
£
Very close to a solid wall, the k-£ model requires a correction, for example
the specification of a damping function j ^ such that
f M = 1 - exp(-0.0002y+ - 0.00065v+2)
31.8.1.3 Second M om ent Closures. In addition to the equations for k and £, these
closure schemes also solve algebraic or differential equations for the Reynolds
stresses. An example of an algebraic stress model is the following
where
3<Uj> d<Uj>
n j = ~ < U i U m > — ----------- < U j U m> — —
o x o x ...
3 1 .2 4 FLUID F L O W H A N D B O O K
where the turbulent diffusion tensor, d ijm, the pressure-strain-rate covariance ten
sor, (pjj, and the dissipation tensor, £,y, are modelled in terms of parameters already
introduced in the previous equations and adjustable constants.
energy of motions which have scales finer than the mesh size. Aliasing may affect
significantly phenomena associated with the fine structure, such as internal intermit-
tency and energy dissipation, and, to a lesser degree, phenomena associated with the
large-scale motions, such as the total kinetic energy.
The specification o f boundary conditions is one of the difficult issues in DNS.
If there is a direction o f hom ogeneity (for isotropic turbulence, all directions; for
two-dimensional flows, the spanwise direction; for fully developed channel flows,
the streamwise direction), then the condition imposed is that of periodicity, namely
the requirement that the solution must be identical on two parallel planes that bound
the computational domain and are normal to the homogeneity direction. For the
inflow condition, a sensible approach is to compute first a three-dimensional, homo
geneous turbulent flow within a box, matching certain specifications in terms of
energy and scale, and then to “convect” this box through the inlet plane by an aver
age speed. This box turbulence will evolve, under the influence of the additional
imposed conditions in the main computational domain, and after a certain time or
distance from the inlet plane, it will attain the structure of the configuration under
study. Because spatially evolving flows are more difficult to simulate, many DNS
have been performed on flows (e.g. mixing layers) with streamwise homogene
ity, but temporally evolving, rather than inhomogeneous and stationary, as is the
usual case with experiments. The correspondence of temporally and spatially evolv
ing flows is not exact, but permits certain comparisons when the latter are viewed
as evolving in a frame convected with a constant speed.
where the filter function g(xj,t) may assume one o f several different forms, for
example the box shape
Following standard LES notation, the tilde over a symbol will indicate a filtered
property. Filtering and conventional averaging are two distinct processes and pro
duce different results. Then one may decompose the instantaneous velocity Uj
into a filtered component and a residual component, as
Uj = Uj + u[
Again, this process is distinct from the Reynolds decomposition. Next, one may
introduce the above components into the time-dependent Navier-Stokes equations
and derive equations for the filtered components, which, however, would be sub
ject to the closure problem, as they contain the residual stresses
With the use of a model for r,y, the system o f equations for the filtered field
may be solved. The most common model is the Smagorinsky model, by which
= — T k k 8jj - 2 Vr p Sjj
In this expression, the rate of strain tensor of the resolved field is defined as
~ i/a a a tn
2 ^ dxj dx, j
and the effective eddy viscosity of the residual field is determined from the fil
tered width as
vT = ( Q A ) 2V 2 S , JSlJ
The most reliable information for turbulent flows of technological interest is still
based on experiment. Even so, one must bear in mind that the design and realiza
tion of meaningful experim ents in turbulence is not an easy task. For this reason,
much of experim ental turbulence research has been conducted in simplified flow
set-ups, which are relatively easily controlled in the laboratory. The spatial, tem
poral and amplitude resolution requirements imposed on a turbulence measuring
method are extremely stringent; ideally, such methods should resolve sim ultane
ously all motions of eddies spanning a very wide range from the full flow extent
to the Kolmogorov scale [ l l , 55]. Instruments that provide single-point measure
ments with very high resolution (e.g. hot-wires) are available. In principle, such in
struments could be arrayed to provide multi-point measurements; this approach,
although cumbersome and costly, has been applied successfully in a number of spe
cialized laboratories. On the other hand, global methods, which provide turbulence
characteristics over an area or volume, are also available, although with a limited
resolution, usually unable to provide accurate measurements of the fine structure.
Measured properties include statistical moments, correlations, spectra, probability
density functions and scales o f a flow parameter and its derivatives. The acquisi
tion of turbulence m easurem ents and the interpretation of the results require an
in-depth understanding of the techniques involved and of the physical and chem
ical characteristics of the flow under study.
A sensible initial step in any experiment in fluid mechanics, and particularly
in those dealing with turbulent flows, is to perform flow visualization. A great vari
ety of flow visualization methods are available, including both marker methods (e.g.
dye and smoke injection) and optical methods (e.g. shadowgraph and Schlieren
methods, interferometry and holography) [55-58J. Such methods provide a gen
eral view of the entire flow structure and could be refined to the point that they
could measure statistical characteristics of the energy containing eddies.
The most popular quantitative methods for measuring turbulent velocity are
the following:
1. Hot-wire anemometry (HW A) [11, 55, 56, 59-61]. Also known as thermal
anemometry, this method measures velocity by its relationship to convec
tive heat transfer from fine heated elements. Sensors include hot-wires
(with typical lengths o f the order of 1 mm and diameters of the order of
5 /J.m), used exclusively in gas flows, and hot films, which may be used in
gas or liquid flows. Heating o f the sensor is achieved by passing an electric
current through it, supplied by a bridge with active feedback that maintains
a constant resistance (i.e. temperature) of the sensor. A combination o f two
sensors inclined with respect to each other and to the mean flow can be
used to measure sim ultaneously two velocity components and a turbulent
shear stress (cross-wire anemometry). Three or more sensors in proper
arrangements have been used to resolve the velocity vector in both m ag
nitude and direction, and even components of the fluctuating vorticity.
3 1 .2 8 FLUID F L O W H A N D B O O K
NOMENCLATURE
Greek Symbols
• • • 0 0
y thermal diffusivity or molecular diffusivity, ft /s (m /s)
A LES filter width, ft (m)
AP frictional pressure loss, psi (Pa)
hntix maximum velocity difference, ft/s (m/s)
8 boundary layer (disturbance) thickness, ft (m)
Su Kronecker’s delta
8m wake momentum thickness, ft (m)
£turbulent kinetic energy dissipation rate, ft2/s3 (m2/s3)
o 'x "X
£U turbulent kinetic energy dissipation rate tensor, ft / s ' (m /s )
n Kolmogorov microscale, ft (m)
e temperature fluctuation, °R (K); also concentration fluctuation
K wave number, f t' ( rr f ’); also von Karman constant
K wave number vector, ft" (m-1)
*7 wave number at the peak of the energy spectrum, ft-1 ( r r f ')
A Taylor microscale, ft (m)
Xf stream wise Taylor microscale, ft (m)
Other Notation
REFERENCES
1. Van Dyke, M. 1982. A n A lbum o f F lu id M o tio n , Parabolic Press, Stanford.
2. Pope, S.B. 2000. Turbulent Flows. Cambridge University Press, Cambridge (UK).
3. Hinze, J.O. 1975. Turbulence. 2nd Edition. McGraw-Hill, New York.
4. Tennekes, H. and Lumley, J.L. 1972. A F irs t Course in Turbulence. MIT Press,
Cambridge, MA.
5. Mathieu, J. and Scott, J. 2000. An Introduction to Turbulent F lo w . Cambridge Uni
versity Press, Cambridge (UK).
6. Libby, P.A. 1996. Intro ductio n to Turbulence , Taylor & Francis, Washington.
7. Frisch, U. 1995. Turbulence. Cambridge University Press, Cambridge (UK).
8. Monin, A.S. and Yaglom, A.M. 1971 (Vol. 1) and 1975 (Vol. 2). Statistical F lu id
M echanics. MIT Press, Cambridge, MA.
9. Townsend, A. A. 1976. The Structure o f Turbulent S h ear F lo w . 2nd Edition. Cam
bridge University Press, Cambridge (UK).
10. Batchelor, G.K. 1956. The Theory o f Hom ogeneous Turbulence. Cambridge Univer
sity Press, Cambridge (UK).
11. Bradshaw, P. 1971. A n Introduction to Turbulence and its M easurem ent. Pergamon
Press, Oxford.
12. Landahl, M.T. and Mollo-Christensen, E. 1986. Turbulence a n d Random Processes
in F lu id M echanics. Cambridge University Press, Cambridge (UK).
13. Sreenivasan, K.R. 1999. “ Fluid Turbulence.” Rev. M od ern Phys.. 71 (2): S383-S395.
14. Papoulis, A. 1991. P rob a b ility, Random V ariables and Stochastic Processes. 3rd
Edition. McGraw-Hill, New York.
15. Batchelor, G.K. 1970. A n Introduction to F lu id D ynam ics. Cambridge University
Press, Cambridge (UK).
16. Mankbadi, R.R. 1994. Transition, Turbulence a n d Noise. Kluwer Academic Pub
lishers, Boston.
17. Narasimha, R. and Sreenivasan, K.R. 1979. “ Relaminarization of Fluid Flows.”
Adv. A ppl. M ech.. 19: 222-309.
18. Schlichting, H. 1979. B o un dary-L ayer Theory. 7th Edition. McGraw-Hill, New York.
3 1 .3 2 FLUID F L O W H A N D B O O K
19. Fung, Y.C. 1969. A F irst Course in Continuum M echanics. Englewood Cliffs:
Prentice Hall.
20. Hunt, J.C.R. et al. (Editors). 1991. Turbulence a n d Stochastic Processes:
K o lm o g o ro v's Ideas 5 0 Years On. The Royal Society, London.
21. Sreenivasan, K.R. and Antonia, R.A. 1997. “The Phenomenology of Small-Scale
Turbulence.” Ann. Rev. F lu id M ech.. 29: 435-472.
22. Comte-Bellot, G. and Corrsin, S. 1966. “ The Use of a Contraction to Improve the
Isotropy of Grid-Generated Turbulence.” J. F lu id M ech. 25: 657-682.
23. AGARD Advisory Report 345. 1998. A Selection o f Test Cases f o r the V alidation
o f L a rg e -E d d y Sim ulations o f Turbulent Flow s. AGARD, Neuilly-sur Seine
(France).
24. Mansour, N.N. and Wray, A.A. “ Decay of Isotropic Turbulence at Low Reynolds
Numbers.” Phys. Fluids 6: 808-814.
25. Rogallo, R.S. and Moin, P. 1984. “ Numerical Simulations of Turbulent Flows.”
Ann. Rev. F lu id M ech. i 6: 99-137.
26. Tavoularis, S. and Corrsin, S. 1981. “ Experiments in a Nearly Homogeneous Shear
Flow with a Uniform Mean Temperature Gradient.” Part 1. J. F lu id M ech . 104:
311-347.
27. Tavoularis, S. and Karnik, U. 1989. “ Further Experiments on the Evolution of Tur
bulent Stresses and Scales in Uniformly Sheared turbulence.” J. F lu id M ech . 204:
457-478.
28. Holloway, A.G.L. and Tavoularis, S. 1992. “The Effects of Curvature on Sheared
Turbulence.” J. F lu id M ech. 237: 569-603.
29. Rogers, M.M. and Moin, P. 1987. “ The Structure of the Vorticity Field in Homoge
neous Turbulent Flows.” J. F lu id M ech. 176: 33-66.
30. Lee, M.J., Kim, J., and Moin, P. 1990. “ Structure of Turbulence at High Shear
Rates.” J. F lu id M ech. 216: 561-583.
31. White, F.M. 1991. Viscous F lu id F lo w . 2nd Edition. McGraw-Hill, New York.
32. Zagarola, M.V. and Smits, A.J. 1997. “ Scaling of the mean velocity profile for tur
bulent pipe flow.” Phys. Rev. Lett. 78: 239-242.
33. Fox, R.W. and McDonald, A.T. 1998. Introduction to F lu id M echanics. 5th Edition.
John Wiley & Sons, New York.
34. Hussein, A.K.M.F. 1983. “ Coherent Structures— Reality and Myth.” Phys. Fluids
26:2816-2850.
35. Cantwell, B.J. 1981. ‘‘Organized Motion in Turbulent Flow.” Ann. Rev. F lu id
M e c h .:\3 y 457-515.
36. Robinson, S.K. 1991. “ Coherent Motions in the Turbulent Boundary Layer.”
Ann. Rev. F lu id M ech. : 23, 601-639.
37. Falco, R.E. 1997. “ Coherent Motions in the Outer Region of Turbulent Boundary
Layers.” Phys. F luids 20: S124—S132.
38. Head, M.R. and Bandyopadhyay, P. 1981. “ New Aspects of Turbulent Boundary-
Layer Structure.” J. F lu id M ech. 107: 297-338.
39. Antonia, R.A. 1981. “ Conditional Sampling in Turbulence Measurement.”
Ann. Rev. F lu id M e c h .:\3 , 131-156.
40. Blackwelder, R.F. and Kaplan, R.E. 1976. “ On the Wall Structure of Turbulent
Boundary Layer.” J. F lu id Mech. 76: 89-112.
41. Berkooz, G., Holmes, P., and Lumley, J.L. 1993. “ The Proper Orthogonal Decom
position in the Analysis of Turbulent Flows.” Ann. Rev. F lu id M ech. 25: 539-575.
TURBULENT FLOW 31.33
42. Holmes. P., Lumley. J.L. and Berkooz, G. 1996. Turbulence, C oherent Structures,
D y n a m ic a l Systems and Symmetry. Cambridge University Press, Cambridge (UK).
43. Launder, B.E. and Spalding, D.B. 1972. M ath em a tica l M odels o f Turbulence. Acad
emic Press, London.
44. Rodi, W. 1980. Turbulence M odels a n d T h eir A p plicatio n in H y d ra u lic s —A State
o f the A rt R eview . International Association for Hydraulics Research, Delft
(The Netherlands).
45. Chen. C.J. and Jaw, S.Y. 1998. F undam entals o f Turbulence M o d elin g . Taylor and
Francis, Washington D.C.
46. Wilcox, D.C. 1993. Turbulence M o d e lin g f o r C F D . DCW Industries Inc., La
Cagton D.C.
47. Gatski, T.B., Hussaini. M.Y., and Lumley, J.L. 1996. Sim ulation and M o d e lin g o f
T urbulent Flow s. Oxford University Press, Oxford.
48. Bradshaw, P.. Launder, B.E., and Lumley, J.L. 1996. “ Collaborative Testing of Tur
bulence Models.” Jour. Fluids Eng. 118: 243-247.
49. Speziale, C.G. 1998. “ Turbulence Modeling for Time-Dependent RANS and VLES:
A Review.” A IA A Jou rn al. 36 (2): 173-184.
50. Rodi, W. and Mansour, N.N. 1993. “ Low Reynolds Number k-e Modelling with the
Aid of Direct Numerical Simulation Data.” J. F lu id M ech. 250: 509-529.
51. Moin, P. and Mahesh, K. 1998. “ Direct Numerical Simulation: A Tool in Turbu
lence Research.” Ann. Rev. F lu id M ech. 30:539-578.
52. Reynolds, W.C. 1989. “The Potential and Limitations of Direct and Large Eddy
Simulations.” In W hither Turbulence? Turbulence at the Crossroads. Lumley, J.L.
(Editor). Springer, New York.
53. Lesieur, M. and Metais, O. 1996. “ New Trends in Large-Eddy Simulations of
Turbulence.” Ann. Rev. F lu id M ech. 28: 45-82.
54. Meneveau, C. and Katz, J. 2000. “ Scale-Invariance and Turbulence Models for
Large-Eddy Simulation.” Ann. Rev. F lu id M ech. 32: 1-32.
55. Tavoularis, S. 1986. “ Techniques for Turbulence Measurement.” In E ncyclopedia o f
F lu id M echanics. Cheremisinoff, N.P. (Editor). Vol. 1, Ch. 36, pp. 1207-1255. Gulf
Publishing Co, Houston.
56. Goldstein, R.J. (Editor) 1996. F lu id M echanics M easurem ents. 2nd Edition. Taylor
and Francis, Washington DC.
57. Yang, W.J. (Editor) 1989. H andbook o f F lo w Visualization. Taylor and Francis,
Washington DC.
58. Merzkirch, W. 1987. F lo w Visualization. (2nd Edition). Academic Press, New York.
59. Bruun, H.H. 1995. H o t-W ire Anem om etry. Oxford University Press, Oxford.
60. Perry, A.E. 1982. Hot-Wire Anemometry. Clarendon Press, Oxford.
61. Lomas, C.G. 1986. Fundam entals o f H o t W ire Anem om etry. Cambridge University
Press, Cambridge.
62. Durst, F., Melling, A. and Whitelaw, J.H. 1976. P rinciples and P rac tic e o f L as er
D o p p le r Anem om etry. Academic Press, New York.
63. Raffel, M., Willard, C. and Kompenhans, J. 1998. P a rtic le Im age Velocim etry —
A P ra c tic a l G uide. Springer-Verlag, Berlin.
INDEX
1.1
1.2 INDEX
M p
MacCormuck method, 24.24
Packed reactors, 23.2
Mach number, 3.19, 5.19, 9.17, 20.27,
Panhandle equation, 9.3
22.5, 28.5
Parshall flume, 19.34
Macroscopic flow, 14.4
Permeability, 7.13, 21.7, 21.20, 21.22
Magnus effects, 20.38
Peng Robinson EOS, 3.8
Manning’ s roughness coefficient 19.7
Photophoresis, 14.19
Manometers, 4.9-4.10
Piezometers, 4.7, 4.9
Maxwellian distribution, 14.7
Pigging, 29.12, 29.23
Maxwellian velocity, 14.14
Pipe
Mean distance, 14.1
contraction, 10.14
Mean free path, 14.2, 14.32
exit, 10.12
Mean temperature, 27.8
inlet, 10.12
Mechanical equivalent, 3.2
Plug flow, 23.1,23.2
Method of characteristics, 22.9
Polymers, 30.2, 30.8
Micro pump, 28.5
Polymers additives, 30.1
Minimum bed fluidization, 23.3
Porous medium, 7.10, 7.17, 21.1, 21.31,
Mixers, 23.27
21.37
Mixer effenicy, 23.33
Porosity, 7.13, 21.24, 21.39
Molar volume, 2.6, 2.8
Positive displacement, 17.3, 28.5
Molecular dynamic method, 14.32
Pour point, 29.10
Molecular effusion, 14.15
Power consumption, 23.17-23.18, 23.32
INDEX 1.5
T W
Taper ratio, 20.36 Wall-bounded flow, 31.18
Thermal process Water hammer, 22.7, 22.39
adiabatic, 3.13, 3.21 Wave drag, 20.5
isenthalpic, 3.15, 11.44 Wax, 29.9
isentropic, 3.16, 14.14 Wax appearance temperature 29.11
isothermal, 3.12, 3.21 Wax deposition, 29.11
non-adiabatic, 3.15 Wax inhibition, 29.11
Thermal velocity, 14.5 Weber number, 5.18
Thermophoresis, 14.19 Weeping, 23.24
Three-K method, 12.23 Wetted perimeter, 19.3-19.4
Three-layer model, 30.15, 30.17 Weymouth equation, 9.5
Tip speed, 23.29 Windkessel model, 26.3
Toroidal path, 17.19 Womersley number, 26.9
Trickle-bed, 23.4
Two-K method, 10.4-10.5
Two-point correlation, 31.8
Taylor’ s frozen flow, 31.9
Transpiration, 14.15
v
Valve characteristics, 16.7-16.8
Valves, 8.7, 12.19, 10.1, 10.20, 10.24,
22.14
Van der Walls, 28.27
Variable soft-sphere model, 28.23
Velocity, 12.3, 12.4
mixture, 11.2
superficial, 11.2, 11.27, 23.14
true, 11.3
Velocity numbers, 11.13
Ventilation drag, 20.4
Viscose stress, 24.2
Viscosity, 2.22,7.11,30.8
Volume fraction, 13.1, 13.4
Von-Karman constant, 31.19
Von Neuman stability, 24.38
Vortex, 31.2
NOTES
NOTES