You are on page 1of 303

Foreword

Momentum, heat and mass transport phenomena can be found nearly

everywhere in nature. Even an early morning activity such as boiling an egg*
or making tea is governed by laws which will be treated here. A solid under-
standing of the principles of these transport processes is essential for those who
apply this science in e.g. chemical and process engineers.
The history of teaching transport phenomena went from a practical but less
fundamental approach, via a short period of a practical and academic approach,
to the present sophisticated approach. Our experience in education and in
industry is that today's abstraction does not appeal to all students and engineers,
who feel themselves easily lost in vast literature and difficult mathematics.
Hence, our objective in writing this book was to digest the enormous amount
of new knowledge and present it in a form useful for those who work as pro-
fessional engineers or who study engineering.
The present book incorporates much fundamental knowledge, but we have
also tried always to illustrate the practical application of the theory. On the
other hand, we have also included practical information and have not shied
away from giving one or two useful empirical correlations, where theory
would have been too difficult. The book is based on the text for a course in
transport phenomena given by W. J. Beek at Delft University from 1962 to
1968. Parts of the last draft have been used, together with most problems
Each chapter ends with a number of problems which form an integral part
of the book. We would like to ask the student to solve as many of these problems
as possible-this is the best way to absorb and digest the theory. We have given
the answers of all problems so that the reader can check his results. Where we
expected difficulties to arise, we have explained some problems in greater detail.
Furthermore, we invented John, our scientific sleuth, and we hope the reader
likes his way of solving problems. The reason for his ability to do so is evident :
he has read the present book and has worked through the problems!
*For identical temperature equalization in the egg the Fourier number must be Fo = atf d
2
=
constant, i.e. for all eggs:
boiling time x (mass)-
213
= constant
is valid. Knowing the optimal bojling time for one species of egg, we can thus predict the boiling
time for other eggs, even ostrich eggs.
Contents
CHAPTER I INTRODUCTION TO PHYSICAL TRANSPORT
PHENOMENA 1
I.l. Conservation laws 2
1.2. Rate of molecular transport processes 11
I.3. Microbalances 15
1.4. SI units 20
1.5. Dimensional analysis . 21
1.6. Problems . 28
CHAPTER II FLOW PHENOMENA
ILL Laminar flow 37
1. Stationary laminar flow between two flat horizontal plates 38
2. Flow through a horizontal circular tube 40
3. Flow through a horizontal annulus 42
4. Flow caused by moving &urfaces 45
5. Flow through pipes with other cross-sections 48
6. Non-stationary flow 49
7. Problems 52
II.2. Turbulent flow . 55
1. Turbulent flow in pipes 55
2. Pressure drop in straight channels 60
3. Pressure drop in pipe systems . 64
4. Problems 74
II.3. Flow with negligible energy dissipation 79
1. Flow of a liquid from an orifice 80
2. Flow of g a ~ e s through orifices 82
3. Flow through,weirs 85
4. Problems 88
II.4. Flow meters , . _ 91
1. Venturi tube 92
2. Orifice plate 93
3. Rotameter 94
4. Problems 95
II.5. Flow around obstacles. 98
1. General approach . 98
2. Spherical particles . 101
3. Free fall of droplets 104
4. Particles in non-stationary flow 105
Vlll
5. Rate of sedimentation of a swarm of particles
6. Cylinders perpendicular to the direction of flow
7. Problems
Il.6. Flow through beds of particles
I . Fixed bed
106
107
108
112
112
115
116
117
120
120
122
124
124
125
125
127
129
131
137
140
2. Filtration through a bed of particles
3. Fluidized bed
4. Problems
II. 7. Stirring and mixing
1. Types of stirrer and flow patterns
2. Power consumption.
3. Pumping capacity and mixing time
4. Problems
II.8. Residence time distribution
1. The F function
2. The E function
3. Simple applications ofF and E functions
4. Continuous flow models .
5. Dispersion in fl ow systems
6. Problems
CHAPTER III HEAT TRANSPORT
III. I. Stationary heat conduction . 145
I. Heat conduction through a wall 146
2. Heat conduction through cylindrical walls 147
3. Heat conduction around a sphere 148
4. General approach for the calculation of temperature distribu-
tiom 1 ~
5. Temperature distribution in a cylinder with uniform heat
production 150
6. Problems 152
III.2. Non-stationary heat conduction 156
1. Heat penetration into a semi-infinite medium 157
2. Heat penetration into a finite medium 161
3. Influence of an outside heat transfer coefficient 164
4. Problems 167
III. 3. Heat transfer by forced convection in pipes 171
I . Heat transfer during laminar flow in pipes 171
2. Heat transfer during turbulent flow . 174
3. Partial and total heat transfer coefficients 176
4. Problems 178
III.4. Heat exchangers . 182
1. Determination of mean temperature difference 182
2. Height of a transfer unit . 186
3. Design of heat exchangers 187
4. rroblems 189
III.5. Heat transfer by forced convection around obstacles
1. Flow along a fiat plate
2. Heat transfer to falling films
3. Flow around spheres and cylinders
4. Heat transfer in packed beds
5. Heat transfer in fluidized beds
6. Problems
III.6. Heat transfer during natural convection .
1. Heat transfer during natural convection
2. Problems
III. 7. Heat transfer during condensation and boiling
1. Film condensation .
2. Dropwise condensation
3. Boiling .
4. Heat transfer in evaporators
5. Problems
III.8. Heat transfer in stirred vessels
1. Problem
III.9. Heat transport by radiation .
1. Problems
CHAPTER IV MASS TRANSPORT
IX
194
194
194
195
197
198
199
201
201
204
208
209
211
212
213
214
215
219
220
223
IV . 1. Stationary diffusion and mass transfer 227
1. Stationary diffusion 227
2. Mass transfer coefficients 231
3. General approach for the calculation of concentration dis-
tributions 233
4. Film theory 234
5. Problems 236
IV.2. Non-stationary diffusion 240
1. Problems 242
IV.3. Mass transfer with forced convection 246
1. Analogy with heat transfer 246
2. Mass transfer 'during laminar flow 248
3. Mass transfer during turbulent flow 253
4. Problems .- 254
IV.4. Mass exchangers 257
l. Thermodynamic equilibrium 258
2. Choice of the apparatus 259
3. Size determination of the mass exchanger 260
4. The concept of theoretical plates 263
5. Problems 265
IV.5. Mass transfer with chemical reaction 270
1. Slow homogeneous first-order reactions 271
2. Fast homogeneous first-order reactions 273
3. Homogeneous nth-order reactions 275
X
4. Homogeneous second-order reactions
5. Mass transfer with heterogeneous chemical reaction
6. Problems
IV.6. Combined heat and mass transport
1. Drying .
2. Problems
INDEX
276
280
283
292
292
294
296
CHAPTER I
Introduction to Physical Transport
Phenomena
During the designing of industrial process plant qualitative and quantitative
considerations play a role.
On the basis of qualitative (sometimes semi-quantitative) considerations a
preselection of feasible concepts of processes suitable for carrying out the desired
production in an economical way is made. The type of operation, e.g. distillation
against extraction or the choice of a solvent, will also be fixed by this type of
reasoning, in which experience and a sound economic feeling play an important
role.
As soon as one or two rough concepts of a production unit are selected, the
different process steps will be analysed in more detail. This asks for a quantitative
appro3.:ch with the aid of a mathematical model .of the unit operation. The
experience that mass, energy and momentum cannot be lost provides the three
conservation laws, on which the quantitative analysis of physical and chemical
processes wholly relies and on which the process design of a plant is based.
This kind of design, which aims at fixing the main dimensions of a reactor or an
apparatus for the exchange of mass, momentum and energy or heat, is the
purpose of the disciplines known as 'chemical engineering' and 'chemical
reaction engineering'. The basic ideas behind these disciplines are found under
the headings 'transport phenomena' and 'chemical (reaction) engineering
science', which rely. on deductive science and, hence, have the advantage of
analytical thought but which, because of that, lack the benefit of induction
based on experience when aiming at a synthesis.
Qualitative and quant-itative reasoning cannot be separated when setting up
a plant, or to put it in another way:
no apparatus, however good its process design might be, can compete with
a well-designed apparatus of a better conception,
which can be the device for a process designer, or
no research, however brilliant in conception it might be, can result in a
competitive production plant without having a quantitative basis,
which can be a motto for a research fellow.
2
Examples of questions, in which feeling and reasoning have to match well
before science is used to some profit, are to be found in the following areas : the
potential possibilities of raw materials, intermediate and end products, the
choice of materials and especially materials of construction, the influence of
side reactions on the performance of subsequent process steps and the con-
siderations on quality and end-use properties of a product. This type of question,
although of importance for the integral approach of a design engineer, will not
be dealt with in this book, which will find its limitations just there. This book
treats the practical consequences of the conservation laws for the chemical
engineer in an analytical way, trying not to exaggerate scientific nicety where
so many other important questions have to be raised and answered, but also
pretending that a solid understanding of the heart of the matter at least solves
a part of all questions satisfactorily.
The laws of conservation of mass, energy and momentum are introduced in
paragraph 1.1. They are extended to phenomena on .a molecular scale in para-
graphs 1.2 and 1.3. Paragraph 1.4 is concerned with .dimensions of physical
quantities, especially SI units, whereas paragraph 1.5 discusses the technique
of dimensional analysis. We will end this chapter (and most paragraphs of the
following chapters) with some proposals for exercising and comments on the
solution of some of the problems given.
After this, three main chapters follow, each of which concentrates on one of
the conserved physical quantities: hydrodynamics (mainly momentum transfer)
in chapter II, energy transfer (mainly heat transfer) in chapter III and mass
transfer in chapter IV. These chapters elaborate the ideas and concepts which
are the subject of the following introduction.
1.1. Conservation laws
John looked at the still smoking ashes of what had once been
the glue and gelatine factory. The fire had started with an
explosion in the building where bones were defatted by
extraction with hexane. John remembered that the extrac-
tion building had a volume of 6000 m
3
and that the tempera-
ture in the building was always 3rfC higher than outside.
He knew that per 24 h, 70 ton steam were lost as well as 9 ton
hexane. He made a quick calculation and concluded that the
steady-state hexane vapour concentration in the plant was
well below the explosion limit of 1 2 per cent by volume and
chac some accident muse haue happened which subsequently
led to the explosion.
Physical technology is based on three empirical laws: matter, energy and
momentum cannot be lost.
The law of conservation of matter is based among other things on the work
of Lavoisier, who proved that during chemical reactions no matter, i.e. no
mass (mass being the most important property of matter), is lost. The law of
conservation of matter is not always valid: in nuclear technology matter is
transformed into energy but for chemical or physical technology this exception
3
is of no importance. It is, of course, possible that matter is transferred from a
desired form into an undesired one (e.g. the degradation of a polymer, which
2
, H
2
0, etc.).
The law of conservation of energy is based among other things on the work
of Joule, who proved that mechanical energy and heat energy are equivalent.
His work finally led to the first law of thermodynamics, which, when formulated
for a flowing system, is the law of energy conservation we are looking for. It
is historically remarkable that it took more than two centuries before this
law, formulated initially for a closed system, was translated into a form in which
it could be applied to flowing systems.
The law of conservation of momentum was finally formulated in its simplest
form for a solid body by Newton : if the sum of the forces acting on a body is
different from zero, this difference is (in size and direction) equal to the acceler-
ation of that body. Together with his second law, action equals reaction, this
formed the basis for dynamics and hydrodynamics. This time it did not take
much more than one century to transpose the concept, originally formulated
for a rigid body, to the more general case of flow in fluids.
These conservation laws play in daily life the same role as the experience that
a pound cannot be spent twice and that the difference exists between the
pound you owe somebody and the one somebody owes you. The economic
rules and the conservation rules of our study are used in the same manner:
balance sheets are set up which account for inflow and outflow and for the
accumulation of the quantity under consideration.
Let us denote by X a certain amount of money, mass, energy or momentum.
Then the general law of on which all phenomenological descrip-
tions of change in the physical world are based; reads as follows:
accumulation of X in system
unit time
flow of X into system flow of X out of system production of X in system
. . + ..
umt time umt time unit t ime
(1.1)
The system may be a country, a concern, a factory, an apparatus, a part of an
apparatus (e.g. a tray.), a . ..pipe or an infinitely small element of volume, etc. This
sounds very general and easy, but daily practice proves that we have to develop
the qualitative judgement for defining the system such that the analysis stays
as easy as possible. To this end, to develop a feeling for the qualitative aspects
of analytical science, we have to go through many a quantitative exercise.
Introductions into a discipline, such as this one, may easily confuse the reader
if these points are not made clear in the beginning; the subject of our study is
the three laws of conservation, which will appear in many forms because we
will study many different systems and not because the basic rules are many.
If we are only interested in macroscopic properties like mean concentration
in the chosen control volume or the rate of change of the mean temperature in
4
this volume we can choose a macroscopic control volume, e.g. a complete
reactor, a complete catalyst particle, etc., for setting up a balance of the desired
physical property. If, on the other hand, we are interested in temperature or
concentration distributions we have to start by setting up a microbalance, Le. a
balance over an infinitesimally small volume element, and to integrate the
differential equations obtained over the total (macroscopic) volume. Both
balances will be treated in d e t a i ~ in this and in the following paragraph.
Let us now try to formulate the law of conservation which we have just
defined in words (equation I. l) in a more precise way. In order to do so, we
need some symbols :
V for the volume of the system (space volume, number of inhabitants, etc.),
<l>v, in and cf>v.out for the ingoing and, outgoing volumetric flow rates (see
Figure IJ),
r for the velwnetric production of X per unit of time and
X for the volumetric concentration of the physical quantity in study.
Cont rol vol ume
r--1--------,
I I
I
<l>v, out
1
V, r
I I
L-----------J
Figure J.l Macrobalance
The accumulation per units of volume and time can now be denoted by dX /dt ;
hence, the accumulation in the system per unit of time is given by VdX/ dt.
A flow at a rate cf>v, containing a volumetric concentration X of the considered
quantity, represents a flow rate <l>vX of this quantity. Hence, our formulation of
the conservation law reads as follows:
(1.2)
We will now apply this law to the following quantities: money, mass, energy
and momentum.
The money balance
Let us start with the most common daily practice, the conservation law for
pocket money (the pocket being defined as the system, although some operate
is very unsystematically). The number of pence in the system is given by X
( J = 1 ), and equation (1.2) can be read as follows: the accumulation of pence
*In order to be completely exact we have to write this equation as :
- u
dX - 1 -1
vdt = cl>u, inxin- rPu,oulxOUI + Vr
""'here -. ndic.ltes a vol ume average and -!a fl ow average.
5
in my pocket pt;r week (which may prove to be negative!) equals the difference
between the number of pence I have taken in during this week and the number
I have spent in this time, increased by the number of pence I produced in the
meantime. The last contribution sounds somewhat cryptic or even illegal, but
an honest production of pence would be to change other coins into pence.
Similar laws can be expressed for the other coins, as well as for the overall
contents of the pocket (moneywise). From all these statements it follows that
the sum of all the production terms r must equal zero (expressed as an intrinsic
value and not as numbers) or, to put it in another way, changing in its own
currency never results in a positive gain. To find out how simple these statements
might be, try and see what happens to your thinking when X no longer stands
for money, but for mass, energy or momentum!
The mass balance
The mass balance still looks familiar and comes close to the money balance.
Here X = cA, the volumetric concentration of component A in a mixture; the
units in which c A is measured are kg/m
3
*
Hence, in a rayon factory, for instance, where NaOH (in the form of viscose)
and HCl (in the form of the spinning bath) are used, the conservation law for
NaOH {A) reads as follows: the accumulation of NaOH on the site in a month
(or any other chosen time unit) is equal to the delivery ofNaOH to the factory
in that time, subtracting the amount of NaOH distributed from the site in that
unit of time as NaOH and adding the amount ofNaOH produced in the factory
during that time (a production which is negative where NaOH is used as a
reactant).
A similar mass balance can be set up for the other reactant, HO (B), as well
as for the products NaCl (C) and H
2
0 (D). Again, the sum of all individual
production rates must be zero: r A + r
8
+ rc + r
0
= 0 (Lavoisier:t. More olten
than not, the accountants of a factory are more aware of the implications of the
conservation laws and the dynamic consequences of varying inflows and out-
flows of a factory than the engineers in charge of production.
' f#>v, i n
l#>v, out
v c
Figore 1.2 Mass balance of a stirred
vessel
The wellstirred continuous flow tank reactor of Figure I.2 gives an example
of the applications of the mass balance. By 'wellstirred' we mean that the
* From the onset, we will accept kg, m, s, oc as the basic units for comparing physical phenomena
(paragraph 1.4).
6
concentration c of a certain compound (e.g. salt) is the same at all. places in the
vessel so that the salt concentration in the effluent stream equals the con-
centration in the vessel. Let us assume that the salt concentration in the vessel
at time t = 0 is c
0
and that, from t = 0 on, a continuous stream of pure water
(salt concentration c = 0) is passed through the vessel. The question is then
how does the salt concentration in the vessel change with time?
We can now set up two mass balances, one for the water and one for the salt.
The first balance says that, if the liquid volume in the reactor is constant, the
flow rate of liquid out of the reactor must equal the flow rate into the .reactor,
<l>v,out = <l>v.in . The second balance says that the decrease in the amount of salt
in the vessel (Vc) per unit of time must equal the mass flow rate of salt from the
vessel (</>vc). Thus :
d(Vc) =
0
_ c
dt v
or:
de c
- =
dt r
where r = Vf4>v is the mean residence time of the fluid in the vessel. Integration
between t = 0, c
0
and t , c yields:
.:_ = exp (!)
Co T
f / T -
Figure 1.3 Change of salt concentration in
stirred vessel
This function is shown in Figure 1.3, which shows that at the time t = 't still
3"" per cent of the initial salt is present in the reactor. For t = i r and ~ r these
values are 65 and 22 per cent respectively. Apparently, the liquid in the reactor
has a large distribution of residence time.
7
The energy balance
Before we can attempt to interpret equation (1.2) in terms of an energy
balance, we must define the meaning of X as ' the volumetric concentration of
energy' Er. This quantity will contain internal energy and potential energy.
Internal energy again comprises perceptible and latent heat and pressure
energy. Pressure energy per unit of volume is nothing but an elaborate expres-
sion for merely the pressure inside the volume. Perceptible and latent heat per
unit of volume (U) may be expressed as :
(1.3)
in which p is the specific gravity, cP the heat capacity per unit of mass, T the
temperature, T, a reference temperature for calculating U, !J.H the latent heat
per firstorder phase transition (melting, evaporation) per unit mass and La
symbol to indicate that the contributions of all phase transitions between T,.
and T have to be taken into account. If the process studied does not show phase
transition, the reference temperature T,. can be chosen high enough to keep
latent heats out of the analysis. If, furthermore, the heat capacity per unit of
volume (pc p) is independent of temperature, U = pc P (T - T,).
The kinetic energy per unit of volume is fpv
2
, when v is the velocity of this
volume. If different elements of the volume flow have different velocities, then
the flow average of the squared velocity has to be used for calculating the flow
of kinetic energy, <Pv !p(v
2
) , with
1
2
1 f pv
2
v dA
2 p < v ) = 2 f v dA
(1.4)
where A is the cross-sectional area of the flow channel.
The potential energy per unit of volume is pg dh, in which h is the covered
distance, h, a point of reference along the path of movement and g a field force
density in the direction of movement (e.g. the gravitational acceleration of earth).
If both p and g are independent of h, the volumetric potential energy is expressed
by pg(h - hr) = pgh (hr often being put arbitrarily equal to zero as the definition
of the -
Hence. X as the volumetric concentration of energy can be evaluated from
if the conditions mentioned apply.
With regard to the production term in equation (!.2), we can distinguish
between mechanical and thermal energy and denote the amount of added
mechanical energy by A (e.g. supplied by a pump or used by driving a turbine)
and the amount of heat flow from the control volume by <PH.
8
Thus, the general balance of equation (1.2) can be written as an energy balance
as follows:
dE,
V dt = cPu. inEr,in - <Po,outEr,out + cPA - f/JH
(1.6)
where E, is defined by equation (I.5).
If heat is transferred between the control volume and its surroundings it is
accepted practice to put the heat transfer rate per unit of boundary surface area
proportional to the temperature difference between the surroundings and the
system. The proportionality factor thus defined has been named the heat
transfer coefficient and is given the symbol a; the practical significance of this
coefficient will be treated in more detail in chapter III. In this chapter we shall
indicate how the heat transfer coefficient can be calculated from first principles
in a number of heat transfer situations. In complicated cases, however, this
coefficient is nothing more than a numerical factor which has to be obtained
from formerly reported, consistent experience. It is one .of the goals of the
discipline to keep scientificalJy well-organized records of this experience.
The momentum balance
The amount of momentum per unit mass is mv/ m = v, whereas the amount
of momentum per unit of volume is mv/V = pv. Thus the momentum flux per
unit volume is pvv. Because momentum has direction as well as size we must
use the three components of momentum in the x-, y- and z-directions, namely
pvx , pvY and pvz . For each direction the law of conservation of momentum is
valid and thus we obtain three separate equations. The momentum-producing
terms must be interpreted as forces {Newton). The forces which can occur in a
flow system are pressure forces, friction forces (caused by shear stresses) and
potential forces (weight forces due to gravitational acceleration). Thus the
momentum balance in the x-direction is found by replacing X = pv in
equation (1.2):
(1.7)
where I Fx indicates the sum of all forces acting in the x-direction. The balances
for the y- and z-directions can be written analogously. The dimension of all
terms in the above balance is the Newton (N).
Application of balances to a pipe corner
We will now, at the end of this paragraph, illustrate the application of the
conservation laws with a practical example. Let us consider a horizontal pipe
comer of cross-section A which is well insulated against heat loss. The mass
balance then reads for an incompressible liquid flowing through the pipe
(equation 1.2 with X = p, the total concentration, and r = 0) under steady-
state conditions :
dp
V- = 0 = . p. - p
dt <f' v,1n 1n "t' v,out out
and, with <l>v. l = v
1
A
1
and <l>u,
2
= v
2
A
2
and constant specific gravity:
vt A 1 = v2A2
9
indicating that at constant cross-section of the pipe the entrance and exit
velocities of the liquid must be equal.
The energy balance reads in the stationary state :
X
L ,
Figure 1.4 Pipe corner
dE,
V dt = Q = <l>v.inEr,in - v,outr ,out + </>A - </> H
(1.8)
and since A = <I>H = 0 and </>r;, in = v.out:
E r,in = E r,out
or, with the definition of E, (equation 1.5), if A
1
= A
2
and thus v
1
= v
2
and if
there is no phase transition (h
1
= h
2
because the pipe is horizontal) :
(1.9)
Equation I.9 indicates that the pressure drop can only be predicted if the
amount of frictional heat liberated per unit of volume is known. This is a prin-
cipal problem of hydrodynamics that can only be solved theoretically
in a few cases. In all other cases we have to rely on empirical correlations for
estimating the pressure drop in flow systems (see paragraph !1.2.2 and II.2.3).
The momentum balance over the pipe corner in the x-direction yields for the
stationary state:
dpvx "
V dt = 0 = pv
1
A
1
v
1
- pv
1
A
2
0 + '- Fx (1.10)
The second term on the right-hand side of this equation expresses the fact that
no momentum in the x-direction is taken away with the fluid. The forces acting
in the x-direction on the liquid are the pressure force p
1
A
1
and the reaction
10
force the wall exerts on the liquid, Fx,w-
1
, thus:
L Fx = P1A1 + Fx,w- f
(1.11)
and combining equations (1.10) and (I.ll) we find for the momentum balance
in the x-direction :
dpvx 2
vdt = 0 = pv1A1 +PIAl+ Fx,w- J
Thus the wall has to produce a force in the negative x-direction and withstand
a force Fx.J- w ( = - Fx,w-
1
) in the positive x-direction (action equals reaction-
first law of Newton). This force equals the sum of the pressure force and the
force which is caused if all the momentum of the flowing liquid is taken up by
the wall*:
Fx,w- f = - Fx,J-w = - pvtAt - P1A1 (I.12a)
Analogously to the above, we can develop a momentum balance for the y-
direction and find that the force the liquid exerts on the wall in the y-direction
is given by :
(I.l2b)
- - -- -
i /
L_ y -I
2
Figure 1.5 Forces acting op pipe corner
With this information the resulting force of the fluid on the wall can be con-
structed as shown in Figure 1.5. The pipe corner must be fixed in such a way
that these forces can be taken up by the holding structure.
In the above example we have assumed that the velocity distributions of
the liquid entering and leaving the pipe are uniform and that there are no
velocity differences over the cross-section. In that case, the momentum of the
inf.owing liquid is indeed pvi A
1
. If the above assumption of uniform flow
technical flow velocities in pipes are 1 m/s for liquids and 15 m/s for gases. The reader
check that the momentum force on a pipe comer is mostly negligible as compared to pressure
forcu
11
velocities is not allowed, we have to write p(vT)A
1
, indicating that vr must be
averaged over the cross-section. Since pA
1
(vr) =

we say that the

momentum per unit mass is given by (vi)/(v
1
). For the same reason we write
(v
3
)/2(v) for the kinetic energy per unit mass. Such expressions for physical
.quantities averaged over the flow will be found regularly in our discipline.
1.2. Rate of molecular transport processes
In the foregoing paragraph we have defined the various terms in the balances
based on the three laws of conservation. In so doing we-have concentrated on
transport with the flowing fluid, the so-called convective transport caused by
movement of the molecules, and neglected the statistical transport. We have
mentioned briefly the conduction of heat through the boundaries of a system
as a means of heat transport, viscous friction as a means of momentum transport
and viscous dissipation as a possible contribution to heat transport.
The opportunity of introducing diffusion of matter through the system
boundaries as a means of transporting a component has been missed until now.
Together with heat transport by conduction and momentum transport due to
viscosity, diffusion of matter is the third molecular process which is of interest
to us.
It might be imagined that a difference in the concentration of a component
on both sides of a boundary enhances a net mass flux through that boundary
because of the Brownian (heat) movement of the molecules of this component:
the bruto flux from the side of the higher concentration will override the bruto
flux from the other side, simply because more of the specific molecules are
present on that side while each has an equal chance to pass the boundary at a
given uniform temperature. Brownian motion also explains the conduction of
heat: a temperature gradient over the system boundary produces a net flux of
energy, because the molecules on the hotter side move faster and carry a
higher energy.
Molecular movement again is a reason for stress in a flow field with a velocity
gradient. Due to Br:ownian movement molecules are exchanged in all directions
and hence also in the direction of the gradient. Those coming from the faster
moving side have a higher velocity in the direction of flow and, hence, carry a
bigger momentum" those coming from the slower moving side. This
results in a net flux of momentum from the faster flowing area to the slower
moving parts or, in other words, slow neighbours try to decelerate fast nt:igh-
bours and vice versa. So, a net transversal transport of momentum-in-the-
direction-of-flow can be interpreted as stress. It is good to notice again the
vectorial character of momentum flow: both the direction of transport
(transversal) and the direction of the quantity transferred (longitudinal) are
important. This means that in an arbitrary plane, stresses in all three directions
can be found when velocity gradients in all these directions are present.
Transport of heat, mass or momentum by molecular processes is sometimes
called statistical transport because it is caused by the random Brownian
12
movement of the molecules.* Statistical transport is usually much smaller than
convective transport, unless very low flow velocities occur. For this reason
statistical transport in the direction of flow can mostly be neglected.
Our elaboration of the conservation concept has to account for mass, energy
and momentum fluxes due to processes on a molecular scale. Generally, these
fluxes are found to be proportional to the gradient dX/ dn, where n denotes the
coordinate in the direction perpendicular to the considered plane. The pro-
portionality constant between the flux and the driving force dX jdn is known as :
the mass diffusivity, [},in the case of mass transfer (Fick's law),
the thermal diffusivity, a = 2/ pc
9
, with 2 being the heat conductivity, in the
case of heat transfer (Fourier's law)t and
the kinematic viscosity, v = 11/ p, with 17 being the dynamic viscosity (Newton's
law).
The dimension of each of these constants is m
2
/s.
When " denotes the flux (the flow rate per m
2
) of the quantity X due to
molecular movement in a gradient of X, we thus writet:
dX
</>" = - (0 or a or v)-
dn
(I.l3)
The minus sign denotes that the net flux is positive in the direction of diminishing
concentrations of X. Thus the mass, beat and momentum fluxes in the positive
direction of n are given by :
mass flux:
cf>':n =
_[}de
dn
(kg/m
2
s) (Fick)
heat flux :
4>" -
d(pcPT)
CW/m
2
)
(Fourier)
H-
-a
dn
momentum flux: F'' = r =
d(pv)
(N/ m
2
) (Newton) -v--
dn
Figure 1.6 illustrates the above relations for mass and heat transport caused by
* Besides statistical molecular transport there can also occur statistical transport by eddies (para-
graphs 11.2, 111.3.2 and IV.3.3). We speak of laminar flow if there is only statistical molecular
transport.
t Here, heat transfer has been used deli berately as a description which is not completely s1rnilar
to energy transfer. Heat transfer accounts only for the transport of perceptible and latent heat,
not fo: the transport of other forms of energy.
! The :-elation between momentum flux and velocity gradient discussed here is restricted to New-
tamar. llquids. For non-Newtonian liquids more complicated relations exist which will be
discussed i:n the next chapter.
...
0
Q
. ~ . . . I = _ ( "' ) d ( c or pc p T)l
't"m orH" " ora
, "o dn no
"o
___ ,
Figure 1.6 Diffusion and heat con-
duction
13
In the case of momentum transport there is the extra complication that we
have not only to specify the plane through which the transport occurs (e.g.
x = x
0
) but also the direction of the momentum considered (e.g. y). The cor-
responding momentum flux is then written as F;
1
, indicating that there is
transport of y-momentum in the x-direction at the plane x = x
0
caused by a
velocity gradient dvy/dx. In other words, viscous shear is transversal transport
oflongitudinal momentum, as indicated in Figure I. 7. This momentum transport
represents itself as shear which, per unit of surface, is called the shear force -r.
Thus the momentum flux F ~ Y corresponds with a shear force txy in they-direction
j
I
r; I = - "dIP vy l I
Y x dx .t
0 0
i
Figure 1.7 Shear in terms of
momentum flux
r xy I =- " d ( p Vy ) I
xo dx X
0
- - x
Figure 1.8 Shear in terms of shear forces
14
at x = constant. This is illustrated in Figure I.8. The exact definition of the shear
force needs, therefore, a statement of the plane and direction considered, e.g. :
d(pv)
-rxy = - v dx
The shear force is defined in such a way that the force the left-band fluid exerts
on the right-hand fluid is positive if the gradient {from left to right) is negative.
We have seen in the foregoing that the statistical transport through
a plane can be described quantitatively with equation (1.13) if the proportionality
constant([), a or v) and the gradient at that plane is known. In practice we are
interested in transport through a plane fonning the boundary of a system
(e.g. a wall, an interface, etc.), but, alas, often the concentration gradient at the
boundary is not known beforehand and a practical approach has to be followed
when qualifying the flux. In these cases, empirical phenomenological transfer
coefficients are used, which are defined as follows :
=
_[) de
= k(coo - cw)
(1.14a)
dn
w
</>" -
d(pcPT )
T..,)
(1.14b)
n-
- a
dn
w
d{pv)
w = (;pv)(v- Vw)
{l.14c)
r =
-v--
w
dn
in which w stands for wall, oo for system (indicating a characteristic position in
the system, such as, for instance, the axis), k for mass transfer coefficient, a. for
heat transfer coefficient and fpvcof2 for momentum transfer coefficient (which
looks awkward for historical reasons) ; f itself is better known as the Fanning
friction factor.
These empirical relationships, to which much background can be added, as
will be seen in the following chapters, are of practical use for establishing
transport rates through boundaries only if we can resort to a physically com-
pletely similar situation on which quantitative data are available. This has
already been brought out when we discussed equation (1.2) in terms of heat,
introducing the concept of the heat transfer coefficient at an earlier stage.
Physical similarity, however, does not mean being identical, as wiJl become
clear when the concepts are worked out in the subsequent chapters.
Since order of magnitude estimations are of great importance in our discipline
it is useful to realize the order of magnitude of ITJ, a and v (or 11 = vp) for gases,
liquids and solids. According to the kinetic theory for gases:
The velocity at a solid wall is zero, which is one of the hypotheses on which hydrodynamics rests;
vw nearly always equals zero.
15
where ii is the mean velocity of the molecules ( -)2RfiM) and I the mean free
path of the molecules (--Mf p = RTj p). At normal pressure and temperature
for gases. D. a and v are of the order 05-2 x lo-s m
2
f s. The kinetic theory
further predicts that 11 and .A (the thermal conductivity = a . pc ) for ideal gases
p
are mdependent of pressure and that they are approximately dependent on the
square root of absolute temperature.
For the diffusion coefficients of liquids at room temperature we find values
of the order 0) = 10-
8
-10-
9
m
2
js ; for other temperatures diffusivities can be
estimated by means of the Einstein- Nemst- Eyring relation D17/ T = constant,
where 11 is the viscosity of the solvent. The viscosity 11 of liquids varies widely
(e.g. at 20C for water '1 = 10-
3
Ns/m
2
, for glycerol 11 = 15 Ns/ m
2
) and is
strongly dependent on temperature ; in the first approximation t7 - exp {1/ T).
The termal diffusivity of liquids is about a 10-
7
m
2
js with the exception of
molten salts and liquid metals, which show considerabTy higher thermal
diffusivities.
The diffusion coefficients of solid materials at room temperature are of the
order 10-
11
- 10-
13
m
2
j s. The thermal diffusivity of non-metallic solids is
roughly a 10-
7
m
2
js, while metals belong (because of the effective heat
transport by the free electrons) to the best heat conductors (a 5-100 x 10-
6
m
2
f s). Also, solid materials have a very high viscosity. The ratio of the viscosity
and the elasticity modulus (having the dimension of time) gives an impression
of the time necessary before an initially elastic deformation can be observed as
flow. On the other hand, many liquids (e.g. molten polymers) show elastic
behaviour. For these liquids the time necessary before flow is observed is
generally much greater than 01 sand it is therefore not possible to describe the
flow of these materials by means of a simple shear stress- velocity gradient
relationship, because the previous history of each particle has to be included in
the description. These problems belong to the area of rheology and we will not
go into more detail here.
1.3. Microbalances
In some cases concentration, temperature and flow velocity distributions
in a system can be by starting with the principle of conservation and
applying this principle to every small volume element of the system. This leads
to the so-called micrQ.balances, as compared to the macrobalances of the
foregoing section (I.l .) which give no insight into the distribution of the quantity
over the system. These microbalances can be formulated in a general way, as
will be shown below. However, the partial differential equations which will
emerge from this microconcept can be solved only for relatively simple situ-
ations. Therefore, dependence on equations (I.14) is daily practice, but the idea
behind a practical correlation for the transfer coeffici ents is often based on a
fundamental analysis of a somewhat simplified and idealized situation which
can still be treated analytically. Let us, therefore, see how we arrive at a micro-
balance.
16
Let us consider a Cartesian element (Figure 1.9) of volume dx dy dz, where
x, y and z are the three coordinates in space. The accumulation of the quantity
X considered per time interval dt in the system is now given by :
dX
dt dx dy d2
The net inflow of X in the x-direction is :
z
Figure 1.9 Volume element in rect-
angular coordinates
(115)
(1.16)
where :is the flux of the quantity X in the x-direction. Similar expressions hold
for the net flow into the system in the y- and z-directions. We can therefore
write equation (1.1) for an infinitesimal volume element as :
dX d" dq/' de/>"
-dxdydz = --" dxdydz- _Y dx dydz- _z dxdydz + r dxdydz
dt dx dy dz "
or
dX d; d; dtp;
- -----+r
dx dy dz
(1.17) - =
dt
Equation (117) is the basic microbalance which can be further evaluated When
doing so, we have to realize that the flux of X on this microscale consists of a
convective transport term and a statistical transport term. These fluxes in the
n-direction are therefore given by* :
dX
4 > ~ = -([}or a or v) dn + v,X (1.18)
Thus, if we take II) as a representative of I!J, a or v, we obtain from equation (1.17)
with equation (1.18):
dX d(vxX) d(vyX)
- = - ___;_;.__...
dt dx dy
(1.19)
Inter?reted as the momentum balance, equation 1.18 is only valid for Newtonian liquids. We will
d;scuss this problem in more detail immediately.
17
If we interpret equation (1.19) as a mass balance for a component A in a
mixture of substances A, B, C, etc. (thus X = CA. C
8
, etc.), we know that :
C A + Ca + Cc + ... = p (the specific gravity of that mixture)
and
rA+ r
8
+rc + ... =0 (Lavoisier)
Summing up the mass balances for all components we find:
dp d(t.'xP)
-=---
dt dx
(1.20)
which is known as the equation of continuity. For a stationary situation, dpjdt
equals zero. When pis constant over the flow field, we find furthermore:
dvx dv.v dv,.
0 -+-+-=
dx dy dz
(1.21)
which is a very familiar expression for the conservation of total mass.
If equation (1.19) is to be used as a microbalance for the conservation of heat
(or of energy if the transport of heat exceeds the transport of other forms of
energy, as given in equation 1.5), X stands for pcPT This microbalance will be
further evaluated in paragraph III.l.4, whereas the micromass balance for one
component will be discussed in more detail in paragraph IV.1.3.
We will now concentrate on the further development of equation (1.17) as a
micromomentum balance. In order to obtain a relation which is valid for
liquids of different rheological behaviour (thus Newtonian and non-Newtonian
fluids) we will
1
instead of equation (1.18) use the more general expression for
the momentum flux in the n-direction :
(1.22)
which is valid for all rheologies. Here X stands for pv. The momentum pro-
duction term in the n-direction is the sum of the pressure and gravity forces
acting in that direction-on the control volume:
dp
rn = - - + pg
dn n
(1.23)
With the above expressions we obtain from equation (I.17) for the micro-
momentum balance in the x-direction :
dt
(1.24)
18
With the aid of the continuity equation (1.20) this expression can be simplified
to :
dvx dvx dv.x dvx dT.x.x dTy.x dT:x dp )
P dt = - pv - - pv - - pv - - - - - - - - - + pg {1.25
x dx , dy z dz dx dy dz dx x
Analogous expressions can be written for the micromomentum balance in the
y- and z-directions as follows :
dvv dv>' dv, dvy d-r"Y d-r,,. dtz,. dp
1
,
6
)
p dt = - PVx dx - pv, dy - PVz dz - dx - dy - dz - dy + pgY ( -
and
dvz dvz dvz dvz d-rxz d-ryz dtzr dp (l.
27
)
p-- = - pv - - pv - - pv - - - - -- - - - - + pg
d t x dx Y dy z dz dx dy dz dz z
If, instead of a Cartesian volume element in rectangular coordinates, we had
considered a volume element in cylindrical coordinates-, as indicated in Figure
1.10, we would have obtained the following micromomentum balances in the
r-direction :
'
z
Figure 1.10 Volume element m
cylindrical coordinates
p dv, = _ p {v dv, + v6 dv, _ vi + v dv,} _ ~ d(rt,) _ ~ d-r,
6
+ -r
80
dt ' dr r d8 r z dz r dr r d8 r
d-r,% dp
---- +pg
dz dr r
in the 8-direction :
p dv
8
= _ p {v dv8 + v8 dv8 + v,v6 + v dv
0
} _ ~ d(r
2
-r,
9
) _ ~ dt
89
dt 'dr r d8 r z dz r
2
dr r d9
(1.28)
d - r
8
~ 1 dp
- dz. - ~ d8 + pge (1.29)
19
and in the z-direction:
dv: { dv: v
8
dv: dv:} 1 d(r't"rz) 1 dr
8
% d-rn dp
p dt = - p Vr dr +-;: d9 + Vz dz -; dr +; d9 + dz - dz + pg.
(1.30)
These micromomentum balances can naturally also be formulated in spherical
coordinates. The relations obtained can be found in many handbooks.
The above momentum balances in terms of shear stresses are valid for all
fluids, because the shear stresses are independent of the rheological behaviour
of the liquid. In the special case of Newtonian liquids -r is proportional to the
d(pvy)
rxy = - v d.x
Introducing this expression into the above microbalances we obtain, for rect-
angular coordinates, in the x-direction :
dvx _ { dvx dvx dvx} {d
2
(pvx) d
2
(pv,) d
2
(pvx) }
p dt - - P Vx dx - l' y dy - v. dz + v dx2 + dy2 + dz2
dp
-- + pg (1.31)
dx x
and in the z-direction:
dvz _ _ { dv. dvz dvz} {d
2
(pvz) d
2
(pvz) d
2
(pvJ}
p dt - p v x dx .+ v >' d y + v z dz + v dx
2
+ d y
2
+ dz
2
- ~ ~ + pgz (1.33)
In spherical coordinates Newton' s law (equation 1.13) reads in the z- or r-
directions :
d(pv
6
) d(pv )
't"gz = - v dz and rzr = - v dr %. ' respectively
and in the 9-direction:
(1.34)
20
Introducing this into the general microbalance in terms of shear stresses we
obtain in the r-direction :
dv,. _ { dv,. v8 dv,. v ~ dv,.} { d ( 1 d(prv,.))
p-- - p v- +- --- + v- + v - - ___:.____:_
dt ,. dr r dO r z dz dr r dr
+ _!_ d
2
(pv,) _ ~ d(pv
8
) d
2
(p . v,.)} _ dp
r
2
d8
2
~ d8 + dz
2
dr + pg,.
in the 0-direction :
dv
6
p-=
d t
and in the z-direction:
dvz _ { dvz v8 dvz dvz} { 1 d ( d(pvz)) 1 d
2
(pvz)
p dt - - p v,. dr + 7 dO + Vz dz + v ; dr r dr + r
2
d0
2
d
2
(pvz)} dp
+ dz2 - dz + pgz
(1.35)
(1.36)
(1.37)
The micromomentum balances given above form the basis for the calculation
of velocity distributions and flow rate- pressure drop relations during laminar
fl ow; we will apply them many times in the next chapter.
1.4. SI units
Different systems of units are applied at the moment in different countries in
industry, research and development. The most important systems are :
the c.g.s. system, based on the centimeter (em), gramme mass (g), second (s)
and degree Celcius (C). For the amount of heat the calory (cal) is used
(derived unit);*
the SI system, based on the metre (m), kilo gramme (kg), second (s) and degree
Kelvin (K). The amount of heat (derived unit) is expressed as Joule (J),
which is identical to the unit for mechanical or electrical energy (1 J =
1 kgm
2
js
2
= 1 Nm = 1 Ws, where N =Newton, W = Watt) ;
the metric system, based on the metre (m), kilo gramme force (kgf), second (s)
and degree Celsius (C). The amount of heat is expressed in kilocalories (kcal) ;
the foot-pound-second system, based on the foot (ft), pound mass (lb),
second (s) and degree Fahrenheit (F). The derived unit for the amount of
heat is the British Thermal Unit (BTU) ;
The units for electric current and for luminous intensity will not be discussed here because they
play no role in this book.
21
the British engineering system, based on the foot (ft), pound force (lbf),
second (s) and degree Fahrenheit (F).
For scientific research the c.g.s. system is used all over the world. Engineers in
Anglo-Saxon countries generally use the British engineering system; in other
countries the metric system is generally used.
In the last few years the SI system (Systeme International d 'Unites) has been
more and more applied. This system has been adopted by the International
Organization for Standardization and is recommended by a large number of
national standard organizations. For that reason we will use SI units throughout
this book. In Table I.l a survey of the basic and derived SI units important for
us is given. The reader is advised to use one consistent system of units, preferably
the SI units.
Table 1.1 Basic and derived SI units
Length
Mass
Time
Temperature
Force
Quantity
Work, energy, quantity of heat
Power
Pressure
Dynamic viscosity
Kinematic viscosity
Surface energy or tension
Enthalpy
Heat capacity
Heat transfer coefficient
Mass transfer coefficient
Thermal conductivity
SI unit
metre
kilogram me
second
degree Kelvin
4
Newton
Joule
Watt
m
kg
s
OK
Unit symbol
N = kgm/s
2
J =Nm
W = Jj s
Nj m
2
Nsf m
2
m
2
/s
Jj m
2
or N/m
Jjkg
Jj kg C
Wjm
2
C
m/s
W/ m C
4
Temperature difference is commonly expressed in degrees Celsius instead of degrees Kelvin.
from one system to the other. In order to facilitate this, Table 1.2 gives a collection
of conversion factor\$. Table !.3 provides some information giving orientation
about the approximate values of some common properties of gases, liquids and
solids, which can be useful for rough calculations.
1.5. Dimensional analysis
In chemical engineering, relations between parameters are often expressed by
means of dimensionless combinations of physical variables. The advantages of
this technique are the easy control of dimensional homogeneity of the relation,
the constants of which are then independent of the system of physical units
applied, and the fact that the number of variables is reduced, which simplifies
22
Table 1.2 Conversion factors
Multiply by
Magnitude Expressed in Divide by In SJ units
Length inch (in) 0.()254 m
foot (ft) 0305
yard (yd)
().914
rrule 1609
Angstrom (A) 10
- 10
Area in
2
645 X 10-
4
m2
ft
2
00929
yd2
0836
acre
4047
mile
2
259 X _10
6
,
Volume in
3
}64 X 10- S
m3
ft
3
00283
yd3
0765
UK gallon
455 X 10-
3
'
US gallon
3785 x w-
3
Time
minute (min) 60 s
hour(h) 3600
day
864 X 10
4
year
316 X 10
7
Mass
gram
648 X w-s
kg
ounce (oz)
284 x w-
2
pound (lb) 0454
(cwt) 508
ton
1016
Force
poundal (pdl) 0138 N
pound force (Jbf)
4-45
dyn
w- s
kg force (kgf) 981
Volumetric How ft
3
/min
472 x w-
4
m
3
/ s
UK galj min
7-58 x w-s
US gal/min
631 x w-s
Mass flow
lb/ min
210 x w-
6
kgls
tonjh 0282
Densit y lb/ in
3
277 X 10
4
kglm
3
lbjft
3
160
Pressure
lbfj in
2
6-89 X 10
3
N/ m
2
lbf/ ft
2
479
dynicrn
2
01
kgf/ cm
2
(=at)
981 X 10
4
atm (standard)
1013 X 10
5
bar
105
in water
2-49 X 10
2
ft water
299 X 10
3
in Hg
339 X 10
3
mm Hg (torr) 133 X 10
2
23
Table 1.2 (com.) Conversion factors
Multiply by
Magnitude Expressed in Divide by In SI units
Dynamic viscosity lb/ft h
413 x w-
4
Ns/ m
2
lb/ft s 149
Poise (P = gjcm s) 01
Centipoise (cP)
w-3
Kinematic viscosity ft
2
/h
258 X JO-S
m
2
/s
Stokes (S = cm
2
j s)
w-4
Centistokes (cS)
w-6
Surface tension dynjcm ( = erg/em
2
)
w-3
Nj m
Temperature difference degree F (or R) 5/9
ocrK)
Energy (work, heat) ft lbl 0-0421 1 = Ws = Nm
ft lbf 136
BTU 106 X 10
3
CHU 1899
hph 268 X 10
6
erg
w-7
kgfm 981
kcal 419 X 10
3
kWh 360 X 10
6
Power (energy flow) BTU/h
0293
'fl\T
CHUjh 0528
ft lbf/s 136
hp (British) 746
hp (metric) 736
erg/s
w-7
kcal/h 1163
calfs
419
Heat flux BTU/ft
2
h
315 W/ m
2
caljcm
2
s 419 X 10
4
kcal/m
2
h
1163
Specific heat BTU/IbF
419 X 10
3
JjkgC
kcal/kgcC 419 X 10
3
"
.-
Latent heat BTU/ lb
233 X 10
3
J/ kg
kcaljkg 419 X 10
3
Heat conductivity
BTU/ ft h OF 173
W/ rn oc
caljcrn soc 419 X 10
2
kcaljm h oc 1163
Heat transfer coefficient BTU/ft
2
h F 568
caljcm
2
s oc 419 X 10
4
kcaljm
2
hoc 1163
24
Table 1.3 Physical properties of some materials
Air Water
20C 20C
p [kg;m
3
] 120 998
11 [Ns/m
2
]
17 X J0-
6
w-3
v = ('1'// P) [m
2
/s]
142 x w-
6
w-6
cp [J/ kg oq 103 X 10
3
4-19 X 10
3
A [W/ m
0
C) 0025 06
u[N/ m]
1 x w-
2
ll)Hz [m2/s]
-2 X 10-S 5 x w-
9
a = A./ p cP [m
2
/s]
20 X J0-
6
0143 X 10-
6
Pr = vf a = C,!1/A.
071 70
Sc = vf O 071 200
~ H v [J/kg]
245 X 10
6
~ H m [Jjkg)
335 X JOS
Gas Law Constant R = 8310 1/ kmol K
Avogadro's number N = 6023 x 10
26
kmol-
1
Gravitational acceleration g = 981 m/ s
2
Stefan- Boltzmann constant <J = 561 x 10-
8
Wfm
2
K
4
Volume of ideal gas at STP = 2241 m
3
/ kmol
Bean oil
lOOOC
870
1 x w-
3
s x w-
6
212 X 10
3
0156
29 X 10-
2
24 X 10-
9
0084 x w-
6
95
300
-
Standard Pressure and Temperature (STP) = 1013 Nfm
2
and 2731SOK
Stainless steel
200C
7750
045 X 10
3
26
7-45 x w-
6
the planning of experiments. Furthermore, dimensionless numbers form the
basis for experimental work with models and model fluids.
We will illustrate the use of dimensionless numbers in a few examples.
Figure 1.11 shows the end of a pipe through which gas flows at a rate v. Another
gas A present at a concentration C
00
outside the pipe will diffuse into the pipe.
====*==> Diffusion
c(l)
Convection c<::=::=t=:= V
1-----x
x =constant
Figure 1.11 Transport by diffusion
and convection in a pipe
Under stationary conditions the rate of diffusion into the pipe equals the rate
of discharge of A with the flow. This situation occurs, for example, in pneumatic
control equipment where a small gas leak is applied (bleeding) in order to
prevent dust or water vapour entering the apparatus. The concentration
distribution c of the gas A in the pipe as a function of the distance x from the
opening is easily calculated if we realize that in the stationary state the amount
diffusing through each cross-section x = constant into the pipe equals the
amount of A removed with the flow, thus:
de
</>" = 0 = - []l- - vc
m dx
(1.38)
25
with c = c:r:: at x = 0. The solution is:
~ = exp (-vx)
c ~ U)
(1.39)
which is a dimensionally homogeneous equation. The relative change in con-
centration appears to be a function of the dimensionless number vxf[ll {the
so-called Peclet number, Pe), which is a measure of the ratio of mass transport
by convection and mass transport by diffusion. If this ratio is big, the relative
concentration of A is smalL Each dimensionless number can be interpreted as
the ratio of the effects of two physical mechanisms.
If we had not been able to produce a complete analysis of the above problem
we could have written formally on the basis of physical intuition:
C = f(x, C
00
, V, 0) (1.40)
indicating that the concentration of A must be dependent on the distance from
the opening x, the outside concentration of A, cro, the rate of gas flow v and the
diffusivity of A, [). Since the dimension 'mass' occurs only inc and C
00
dimen-
sional homogeneity is only possible if the ratio cf cro of these two variables
occurs in the solut ion. Analogously, the dimension 'time' occurs only in v and
D, indicating that these two variables must occur as the ratio v/ D in the solution.
This ratio has the dimension of 'length' and this dimension can be eliminated
by introducing distance into the ratio. Thus the solution we find from dimen-
sional considerations is :
!__ = !(vx)
Co [ll
(1.41)
This is in agreenlent with the result obtained earlier. That the function f is an
exponential function cannot be found from dimensional considerations, this
result can only be obtained from a complete analysis. Dimensional analysis
has resulted in reducing the problem with initially five variables to a problem
with only two paiameters. Furthermore, comparisons between analogous
situations are possible without the complete solution. If we consider, for ex-
ample, two situations where the gas velocities differ by a factor 2, we find tire
same relative concenlratlons at distances which differ by a f a c t o r ~ (i.e. the term
vx stays constant).
As a second illustration of the procedure followed during dimensional
analysis, we will consider the force F acting on a sphere which is placed into a
flowing fluid with velocity v, relative to the sphere (see also paragraph 11.5.2).
Based on physical considerations we can state that tbe force on the sphere will
be given by:
F = f(D, v, , p, 17) (1.42)
where D is the diameter of the sphere, p the specific gravity and 11 the viscosity
of the fluid. This relation must be dimensionally homogeneous, i.t:. the dimen
26
sions occurring on the left-hand side must equal the dimensions on the right-
hand side of the equation. This must also be the case if we write :
F = D"v:pc1J
4
constant
The dimensions of these quantities can be expressed in length (L), time (T) and
mass (M) and we obtain:
LT-
2
M = L
4
(L1
1
)b(L -
3
M)'(L - l T -
1
M)d constant
The condition of homogenity leads thus to the following three expressions (one
for each dimension) between the exponents a, b, c and d:
L= L
0
L"L -
3
cL - d or 1 =a+ b - 3c- d
r -
2
= T
0
T -"T
0
T-d or -2 = - b- d
M = M
0
M
0
j"A' Md or 1 = _c + d
Solving these equations for a, b and c we find :
a=2-d
b=2-d
c =l - d
The following expression for the force F is thus dimensionally homogeneous:
F = D<
2
-

4
> p
0
-
4
>tr
4
constant
or, also :

2
= (-
11
-)
4
constant
pv,.D pv,.D
(I.43)
Since both F 1 pv; D
2
and 11/ pv,.D are dimensionless of the variables,
any relation between these two groups is also dimension;llfy homogeneous. We
can therefore conclude that : /
F ( 17. ) ( 1 )
pv;D
2
= f pv,.D = f Re
(1.44)
Here Re = pv,.D/ 11 is the Reynolds number, which can be interpreted as the
ratio of momentum transport by convection ("'"' pv,v,.) and by internal friction
("' 1]V,./ D).
Again, the number of variables in this problem of initiall y five has been
reduced to two. In most cases the number of dimensionless groups obtained by
dimensional analysis equals the number of variables occurring in the problem
minus the number of independent dimensional equations which can be set up.
In flow and mass transport problems the number of dimensional equations is
three (L, T and M) and in heat transport problems four {L, T, M and tempera-
ture).
27
The function fin the result obtained is again unknown and can only be found
from a complete analysis of the problem. Sometimes the .result obtained from
dimensional analysis can be simplified further by making use of former physical
experience. In the case of the sphere, for example, experience tells us that at very
low flow velocities the force F on the sphere is independent of the specific
gravity of the fluid. Thus the relation between the two dimensionless groups
must be :
or
F '1
2
D
2
= -D constant
pv,. pv,
F
-- = constant
17v,D
(1.45)
This equation represents Stokes's law. That the numerical value of the constant
is 3n can only be found by solving the micromomentum balances for this
problem.
As a last example we will discuss the droplets formed, for example, during
condensation of water vapour at the underside of a horizontal surface. The
hanging droplets grow until they reach a critical volume ~ r and fall down. A
dimensionally homogeneous relation for ~ r follows from the consideration
that ~ - r depends on the density of the liquid p, the gravitational acceleration g,
the surface tension (J of the liquid and the degree to which the surface is wetted
by the liquid. For the great number of cases where the surface is well wetted by
the liquid we can write : . .
~ r = f(p, g, a) (1.46)
The reader may check that the result is given by:
( )
3/2
V:: r ~ = constant
(1.47)
The numerical value of the constant follows only from an exact analysis or
from experimental results. Analysis yields here constant = 666. The problem
with initially four variables has been reduced to one dimensionless group, the
Laplace number which represents the ratio oft he weight of the droplet (-pg ~ r )
and the capillary force (-a VtJ
If dimensional analysis is applied all relevant variables have to be included
in the considerations. If we had, for example, forgotten the gravitational
acceleration in the analysis of the hanging droplet, we would not have found a
solution. If, on the other hand, we include too many variables, the result becomes
unnecessarily complicated. Most profit is obtained if besides dimensional
analysis a fragmentary piece of physical experience is applied, as illustrated in
the above examples.
28
1.6. Problems
In the following the reader will find a number of exercises. These problems
form an important part of the book a .. 1d the reader is asked to solve as many of
them as possible. They provide a check on the understanding of the principles
discussed, they illustrate the practical application of these principles and
sometimes they extend the matter presented. In the three chapters which
follow each paragraph will end with a collection of problems. The answers to
these problems are given in order to enable the reader to check his own result.
On a few problems (marked with an asterisk) we have also included some
comments in order to illustrate an important point or to show that the solution
is really not very difficult if the basic principles are applied in the right way.
When solving a problem, the most important thing is to establish the physical
mechanism that governs the situation studied (e.g. is convection, conduction
and radiation of heat important or can one or two transport mechanisms
be neglected in this special case?}. Secondly, we have .to decide over which
control volume the balance is made (e.g. over a microscopic volume if informa-
tion about the temperature distribution is required; over a pipe or heat exchanger
or over a total reactor with overall transfer coefficients if only information
about the mean temperature is required). Next, we have to determine whether the
state by assuming the observer to move with the system, e.g. the flowing fluid).
Having come so far we are now in a position to solve the problem with the help
of the basic information provided in this book. This all sounds very difficult-
and it is- but the best way to solve any difficulties is to try very hard.
1. A tank is filled with a liquid having a specific gravity of p = 8 lb/ US gal.
What is the pressure difference between the top of the liquid and a point
4ft below the liquid level? Answer in psi, bar, Njm
2
, atm and kgfjcm
2

Answer : 167psi = 0115bar = 115 x 10
4
N/m
2
= 0113atm = 0117
kgf(cm
2
2. The temperature increase into earth is approx. 0025C/m. If the thermal
conductivity of earth is i.. = 186 W j m C, what is the heat loss per unit
surface area?
Answer : = 00465 W / m
2
*3. 11/s water is pressed through a horizontal pipe. The pressure difference
over the length of the pipe is 2. 10
5
N/m
2
.
(a) What is the amount of power A necessary and what is the temperature
increase of the water ll. T if there is no beat exchange with the surround-
ings?
(b) What are 4> A and ll T if the water flow is doubled and if the pressure
drop is determined by momentum forces (and not by friction forces)?
Answer: (a) cPA= 200 W, /:iT= 0048C
(b) A = 1600 w, !J. T = 019 c
29
4. Two spheres of equal weight fall through air. What is the ratio of their
stationary rates of free fall if the ratio of their diameters Dd D
2
equals 3?
2
= 1
5. Show that the power cpA necessary to rotate a stirrer (diameter d) with a
speed n (s -
1
) in a fluid of density p and viscosity '1 is given by :
cpA = r(pnd2' n2d)
pn3ds r, g
*6. Write down.the conservation laws for a kettle of water on the fire:
(a) during heating up and
(b) during boiling.
*7. At the end of a glass capillary (outer diameter 2 mm) single water droplets
are formed very slowly. What is the diameter of the droplets which fall
from the capillary?
Answer : d = 44 mm
8. Through a well-stirred 15m
3
tank flows 0.()1 m
3
js of coconut oil. From a
certain time on palm kernel oil is passed through the vessel at the same
speed. After what time does the effluent oil contain less than 1 per cent
coconut oil?
Answer : l = 6900 s
9. Water flows at a rate of2 mj s through a horizontal pipe of 10 em diameter.
A large flat plate is fixed near to the end of the pipe and at right angles to it
Find the force on this plate due to the jet of water, assuming that on
reaching the plate the water flows away along the surface of the plate.
10. A pulp containing 40 per cent by weight of moisture is fed into a counter-
current drier at.a rate of 1000 kg/ h. The pulp enters at a temperature of
20C and leaves at. 70C, the moisture content at the exit being 8 per cent.
Air of humidity of 0007 kg water per kg dry air enters the drier at 120C
and the temperature of the air leaving the drier is 80C. What is the rate
of air flow through the drier and the humidity of the air leaving the drier?
Data : heat losses from drier = 5 x 10
6
J, specific heat of water vapour
cP = 2 x 10
3
J{kg C, specific heat of dry pul p cP = 10
3
J/kg C,
tlH v at sooc = 23 x 10
6
Jjkg oc
Answer: </Jg = 230 x 10
4
kgjh, water content = 0021 kg water per kg
dry air
*11. A 100 kmol batch ofaqueous methanol containing45 mol percent CH
3
0H
is to be subjected to simple atmospheric distillation until the residue in
the kettle contains only 5 mol per cent CH
3
0H. What is the bulk compo-
sition of the distillate?
30
Vapour- liquid equilibria at 1 atm (X and Y = mole
fraction CH
3
0H in liquid and vapour phases respectively)
X
y
X
y
002 013
().40
073
005
().27
050 078
010 042 060 083
020 058 080 091
030 067 090 096
Answer : 63 mol per cent
* 12. Show that John reached the right C\")nclusion in the case of the burnt-down
factory discussed at the beginning of this chapter.
Problem 3
A macroscopic energy balance over the entire pipe according to equations
(1.5) and (1.6) yields for the stationary state and with Jlii = 0 (no phase transi-
tion), v
1
= v
2
(constant cross-section of pipe}, h
1
= h
2
(horizontal pipe),
cPv, in = </Jv,out and cPH = 0 (no heat losses) :
Now, the power necessary to bring the liquid to the initial pressure of
2 x 10
5
N/ m
2
is (no temperature change, d T = 0) :
</>A = - </> b.p = - </>m ~ p = 200 W
v p
This power is converted into heat in the pipe (where <P A = 0), thus:
If the water flow is doubled, the pressure drop which is proportional to the
momentum (dp .- .v ,..., v
2
) is increased by a factor 4. Thus 4>,. ("' </J,/lp - v
3
)
is increased by a factor 8 and ~ T (,...., Ap ,..., v
2
) by a factor 4.
Problem 6
The sketch shows the situation of the kettle on the fire. We will set up our
balances over the entir:e kettle. During heating up no mass leaves the kettle,
thus the mass and momentum balances reduce to :
vdPL = 0 and
dt
d(pv) .
V dt = 0 respecttvely
whereas the energy balance becomes:
d(pLcpT)
V dt = tPn
1
- tPn
2
= T)A
1
- (XiT- TaJA:z
During boiling the balances become:
mass: Vd:L = -p,vA3 = -</Jm
d(pv)
momentum: Vdt = 0 =IF
31
where the force in the y-direction is given by FY = Mg + <f>mvy and the force in
the x-direction by Fx = 4>mvx; and:
(where, of course, </>H
1
- 4>H
2
- 4>mllHv = 0) if we neglect all other forms of
energy because theyare so small.
Problem 7
A macromomentum balance over the entire droplet will provide the answer
we are looking for. Because the flow velocities are very low we can neglect the
momentum of the fluid and our momentum balance becomes a force balance:
the forces acting are the weight force of the droplet and the surface
tension of the liquid in contact with the capillary as shown in the drawing.
Thus the force balance becomes:
' n 3
L F = 0 = -d pg - 2nra
6
32
and thus :
Problem 11
t F=21TfCT
I
ru;:;
d = \}pg = 44mm
This problem is easily solved by application of the law of conservation of
matter. The complication here is that the composition (Y) of the vapour phase
changes as the distillation proceeds, because of the changing composition (X)
of the residue. Apparently, the right thing to do therefore is to set up an unsteady-
state material balance for a short time interval dt and to integrate the result
found over the total process time te . The mass balance (equation 1.1) for this
case becomes (no flow into the system, no production of methanol but variable
contents of the evaporator):
d(NX) dN
dt = -4>molY = dt Y
where the symbols have the meaning given in the figure. After partial differentia-
tion we can write:
d(NX) = NdX + XdN
dt dt dt
.L
1
= - - = constant
fi
dN
't'mo d/
Mole fraction CH30H
in vapour= Y
N kmole
mole fraction
CH
3
0H =X
ott =0, x
0
, No
33
and these two equations we obtain:
dX dN dN
N- =-Y- -X = - </> (Y- X)
dt dt dt mol
or:
dX
lPmol dt dN
- = -
Y - X . N N
which yields after integration between t = 0, x
0
and t, x:
N = No exp ( s:, y d: X)
This equation gives the relation between the total number of moles in the
evaporator and the methanol content of this residue. We are, however, interested
in the average composition of the distillate < Y) which is given by :
f
t. Y dt -ft .. d(NX) - (NX)I'x ..
( Y) = 0 mol = 0 = O.Xo
mol te No - N No - N
= -No {xex (Jx dX )} = X0 - X)
N
0
-N P x
0
Y-X . Xo
This is the desired result ; all that remains is for us to evaluate graphically the
integral, as shown in the figure. We find for the area under the curve :
and thus :

dX = 117
Y- X
X a
x-
045 - 005 exp (- 117) 0435
( Y) = = -- = 631 per cent
1 - ex p (- 117) 069
34
The amount of distillate is found as:
N
0
- N = N
0
{1 - exp (J:., yd: x)} = 69 kmol
We see that sometimes the solution of a simple mass balance can be quite
involved. There is, however, a much simpler way to solve this problem approxi-
mately by carrying out the distillation stepwise. We start with 100 kmol of a
mixture containing a mole fraction of X
0
= 0-45 methanol. Let us distil off so
much distillate that X drops t ~ X
1
= 0-35. Apparently,in this range the average
mole fraction of methanol in the vapour phase will be Y
0
= 073 and thus we
can find the total number of moles distilled off from a simple mass balance as
follows:
xl = NoXo- (No- N)Yo = 0-35 = 100 X 045- 073(No- N)
N
0
- (N
0
- N) 100 - (N
0
- N)
We find N
0
- N = 263 kmol, containing 192 kmol of methanol. Now we
repeat this exercise as stated in the following table until we reach the desired
methanol fraction of X = 005 in the liquid phase.
Total number Mole fraction of Mole fraction of Number ofkmol Numberofkmol
of kmol in pot methanol in pot methanol in vapour distilled off of methanol
N X Y N
0
- N distilled off
Total
100
73-7
56-1
431
315
045- 035
035- 0-25
025 - 0-15
015- 0-05
073
067
0-58
042
263 192
17-6 118
130 76
116
4-9
685 435
Thus we find that 685 kmol were distilled off, containing 435 kmol of
methanol , i.e. 635 per cent. There is a residue of 315 kmol, containing
45 - 435 = 15 kmol of methanol, which is near enough to the desired 5 mol
per cent.
The reader will realize that the principal considerations in this approximate
solution are the same as for the exact solution given before. Instead of infinitesi-
mal steps dX or dt which made integration necessary, we just used bigger steps
which enabled us to obtain an approximate analytical solution after a reasonable
Problem 12 John and the burnt-down factory
The flow rate of air through the plant v can be found by means of an energy
balance (equation !.6). Under steady-state conditions dEtfdt = 0 and realizing
that the change in heat content of the air pc P L\ Tis much bigger than the changes
in the other forms of energy (equation 1.5), we can write :
dEt
V dt = 4>vPCp ll.T +A- H = 0
35
Now, no mechanjcaJ energy is added, A = 0 and the amount of heat supplied
equals <PH = tPm 6H" , where </>,. is the rate of steam loss. Thus :
A. _ ,. /l.H., _ 70 X 10
3
23 X 10
6
_
52
3
/
t.y - -
3
- ms
" pcPAT 24. 3600.12 x 10 . 30
If we assume that the air in the building is well mixed, the hexane concentration
in the building will equal the concentration in the exit stream. A material
balance then yields with cin = 0 for the steady state:
de
V- =- c + r = 0
dt v
where r equals the rate of hexane loss. Thus we find:
r 9000
c = - = = 0002 kgjm
3
v 24 . 3600 . 52
and, regarding the hexane vapour as an ideal gas, we find :
86
p =- = 3-83 kg/m
3
224
and thus c = 0052 per cent by volume, which is indeed well below the explosion
limit.
CHAPTER II
Flow Phenomena
In this chapter the transport of momentum is studied and we will discuss the
prediction of flow resistance in pipe systems and of the resistance due to
obstacles placed in a flow field.
The analysis of pipe flow is divided into two sections, laminar and turbulent
flow, the difference being that in the first case shear stresses in the fluid are pre-
dictable from velocity gradients (deformation rates) but in the latter they are
not. This means that the 'theory' of turbulent pipe flow relies strongly on
empirical knowledge; however, this knowledge can be made generally applicable
by scientific reasoning.
In stationary flow through pipes of a uniform cross-section the momentum
balance degenerates to a balance between shear forces and pressure forces, but
this is no longer the case in flow systems in which the cross-section available
for flow changes. Hence, in that case a more general and practical approach
will be needed.
A class of flow problems, in which the enery dissipation due to shearing is
small in comparison to the amount of mechanical energy which is transformed,
e.g. from kinetic energy into potential energy, will be dealt with separately.
This will lead us to the design and use of flow meters.
Flow around obstacles will not only be extended to flow through beds of
particles but also to stirrers. This brings us to the problem of mixing and of the
non-uniform distribution of residence time of the fluid elements which pass
through a continuous flow system.
In all hydrodynamic theory we will make use of the three fundamental
conservation laws which were introduced in chapter I. The idea of conservation
of mass, momentum and energy provides us with five relationships for a three-
dimensional flow field, from which the velocity distribution (v,.., v,. and vz) and
the temperature and pressure distributions can be obtained in principle, as
soon as :
these quantities are defined at the boundaries of the system and
relationships between shear stresses and deformation rates are known.
The whole problem in using the concepts of conservation of mass, energy and
momentum in predicting flow fields and flow resistances is to define the control
,olume in such a way that it facilitates the analysis and to make clever guesses
about the order of magnitude of the several terms in the balances in order to
37
make them amendable for an algebraic treatment. Hence, although the whole
theory on transport phenomena does not go beyond the concepts which are
already introduced in chapter I, the professional engineer needs considerable
experience before he is able to tum these concepts into practical use. T h e r e f o r e ~
this chapter has been written. It treats a number of problems in an elementary
way in order to illustrate the practical application of the basic principles and
it provides a number of elementary solutions which can be the starting point
to the solution of more complex problems.
ll.I. Laminar Oow
John looked at the spoon in his hand. Syrup dripped from it
into his porridge at a decreasing rate. He thought about
writing a criminal story in which someone was imprisoned
until the very last drop of syrup would leave the spoon. John
had a feeling that this would last very long, months, perhaps,.
or years. Being a scientific story writer and working effici-
ent ly, he read the following introduction and found out that
the thickness of the syrup layer on his spoon decreased about
100 times in 25 minutes. Consequently, he thought, t his layer
would diminish to molecular dimensions in about 3 x 10
5
years-a long sentence indeed.
Dtlring laminar flow, fluid elements move along parallel streamlines. Con-
sider, for example, the flow of a liquid film (e.g. syrup) on a vertical plane surface.
Here the weight of the fluid constitutes the driving force, resulting (if the vis-
cosity of the fluid is not too low) in laminar flow. At the wall, fl ow velocity will
be zero and there will be a velocity gradient over the thickness of the film. Thus,
adjacent fluid elements will have different flow velocities, causing shear forces
between the fluid elements.
A great number of laminar flow problems can be solved with the aid of the
balances discussed in the foregoing chapter if, for the fluids used, the relation
between shear forces and the velocity gradient is known. This relation reads
for unidirectional fl o.w of Newtonian fluids :
d(pvx)
r = -v--
yx dy
-
and if the fluids are non-compressible:
dvx
"[ = -11-
y x dy
(II.la)
(I l.Ib)
ryx is the shear force in the x-direction in a plane xz normal to they-axis. This
relationship is shown, together with two other possible relations between ryx
and the velocity gradient, in Figure 11. 1.
To solve a flow problem we start with a momentum bal ance which, for
channels with constant cross-section, takes the form of a force balance, from
which we find a relation between the shear force at any place and the pressure
drop. No assumption about the type of liquid or flow has been made until this
38
-dvx l dy -dv
1
1dy
Newton Bing hom Ostwald
(
dv )"-I dv
T yx : - K _ .::..:.!.
dy dy
Figure D.l Relation between shear force and velocity gradient
point. This relation is always valid for any kind of flow and for any type of
rheology. Using the shear force-velocity gradient relation for the fluid con-
cerned (e.g. equation ILl a for Newtonian liquids) we obtain after integration an
expression for the velocity distribution in laminar flow. From this velocity
profile the mean velocity and the volumetric flow rate can be determined as a
function of the driving force. We will illustrate thi s procedure in the following
paragraphs.
II.l .l . Stationary laminar flow between two fiat horizontal plates
Consider a liquid flowing between two parallel plates, as shown in Figure II.2.
As the control volume we choose the volume between the plane y = 0 (centre
of flow field) and the plane y = constant between the two cross-sections '1' and
'2'. Under steady-state conditions, the momentum balance over this control
volume is given by equation (I. 7) as:
d(pvx)
V dt = 0 = tPv,inPV::c.in - 1Jv.outPVx.out + L Fx
Now, since the cross-section available for flow is constant and the liquid is
incompressible, the amount of momentum transported by the liquid into and
out of the control volume is the same :
-
-
-
-
-
i'l'
Figure 11.2 Flow between two flat plates
39
and we find that the momentum balance reduces to a force balance I Fx = 0.
This is the case in all situations where compressibility of the liquid can be
neglected and where the flow channel has a constant cross-section.
The forces in the x-direction per unit of width acting on the control volume
are the following if, at the two cross-sections, the pressures are p
1
and p
2
respectively:
F
pressure force at ' 1': 'd h = p
1
y
Wl t
F
at '2'. x - - .PzY
width
shear force
(positive, because it acts in the
positive x-direction)
(acts in the negative x-direction)
(negative, because it acts in the
negative x-direction)
We can now set up our force balance and we find :
Fx = 0 = Pt Y - PzY - T xL
Width
1
or
(II.2)
Thus we see that the shear force is maximal at the wall (y = + d/2) and zero
in the centre (y = 0). Applying equation II.lb we obtain for a Newtonian non-
compressible fluid flowing in a two-dimensional slit of constant width :
dvx Pt - Pz
-=- -y
dy 11L
Integrating with the boundary conditions vx = 0 at y = + d/2 (flow velocity is
zero at the wall) we the velocity distribution :
v, = y2)
(II.3)
We notice that the velocity profile is parabolic, the maximum velocity being
(at y = 0):
. _ Pt - P2d2
- 8rfL
The flow rate per unit width is found as:
+ df 2
lf>v = f v dy = Pl - P2 d3 = d( v)
width -d;z x 1211L
(II.4)
------------------------------------------------ -----
40
and we see that the maximum velocity is given by
timax = f(v)
In Figure Il.3 the shear force and velocity distribution found for laminar flow
in a narrow slit are shown. The reader is asked to work out on his own the
X
Pt-P2
L
I
I
I
vmox = 3/2 <v>
Figure 0.3 Shear force and velocity distribution for laminar flow
in a narrow slit
solution to a similar problem of a film falling on a vertical plane surface {prob-
lem 2 on page 52).
We could also have solved the flow problem just discussed by means of the
general micromomentum balance developed in paragraph 1.3. Using equation
(!.25) and realizing that there is no flow velocity in the y- and z-directions
(v
1
= 0, vz = 0), no velocity gradient in the x- and z-directions (dvxf dx = 0,
dvjdz = 0), no shear force in the x- or z-directions (d-rxxfdx = 0, dtzjdz = 0)
and that the plates are horizontal (thus pgx = 0), we find for steady-state
conditions (dv:.:f dt = 0):
or, after integration :
r,x = - ~ ~ y + constant
which is identical with equation (11.2) if we realize that there is no shear force
at the centre of the channel (at y = 0) and hence that the integration constant
must be zero.
II.l.2. Stationary laminar flow through a horizontal circular tube
To solve this problem we can use the same approach as in paragraph II.l.l.
As a control volume we choose a cylindrical plug of length L and radius r as
shown in Figure 11.4. The forces acting on the control volume are again pressure
forces acting on the cross-sections and a shear force acting at the cylindrical
surface in the negative x-direction.
41
X
L
{}+. '
I I
I - 1
P xz -x. P2
Figure 11.4 Flow through a circular tube
W'riting down the force balance for steady-state conditions we obtain :
I Fx = 0 = nr
2
p
1
- nr
2
p
2
-

- x
1
)
or:
r = !:_ P1 - P2 = _ dp) (II.S)
rx 2 x2 - xl 2 dx
This relation between shear stress and pressure drop is generally valid (there
has been no assumption about the type of flow), as was the case with equation
(II.2). The shear stress distribution is linear : r,.x = 0 at the centre (where thus
the velocity gradient is also zero) and maximum at the wall.
The velocity distribution for laminar flow of a Newtonian liquid is found by
eliminating tr.x in equation (11.5) by equation (II.lb) and by integrating the
differential equation obtained with vx = 0 at the wall (r = R). The result found
IS :
1 ( dp) 2 2
vx = - (R - r )
dx
(11.6)
Thus, again, the velocity distribution is parabolic, with maximum flow velocity
at r = 0:
1 ( dp) . 2
Vmax = 4'7 - dx R
The flow rate at a given' pressure gradient is found from the velocity distribution
as:
<P = 2nrv dr = - -- = nR
2
(v)
f
R nR
4
( dp)
,. o x 8rr dx
(II. 7)
Combining equations (Il.6) and (11. 7) we find for the velocity distribution :
(
r2) R
2
( dp) ( r
2
)
v = 2(v) 1 - - = - -- 1 --
x R
2
4'1 dx R
2
(11.8)
and we see that in the case of flow through a circular tube :
Vmax = 2(v)
42
Equation (11.7) is the Hagen-Poiseuille relation which (with some corrections)
is used by various methods for the measurement of viscosities of Newtonian
liquids. In Figure II.5 the shear stress and the velocity distribution found are
illustrated.
I
I
I
r (PJ-P2l
2 -L-
I
I
I
vmax = 2<v>
Figure 11.5 Shear and velocity distribution for laminar flow
in a circular pipe
The velocity gradient at the wall can be found by differentiating equation
(11.8) and by determining the value of the first derivative for r = R. We obtain:
=- = 2: (- = -
(II.9)
where rw is the shear stress at the wall. We will need this result later on, when we
develop equations to predict pressure drop during flow of liquids in pipes. The
reader may check that equation (II.5)" could also have been obtained from the
appropriate microbalance, equation (1.30), and that equation (11.6) can be
derived from equation (1.37):
I 1.1.3. Stationary laminar flow through a horizontal annulus
In many cases it is not possible to define directly a macroscopic momentum
balance as used in the foregoing two examples. In such cases a microscopic
momentum balance must be developed. Consider the annulus shown in Figure
II.6. As a control volume we choose a streamline with radius r, thickness dr and
length L. We can then set up the following force balanc.e over the control volume:
I Fr = 0 = 2nr dr - 2nr drp
2
+ rrx2nrL!, - r,.x2nrLI.+ dr
Dividing by 2nL dr we obtain:
Pt - P2 rr,xlr +dr - rr,xlr
:....:..-...:....;:;; r=
L . dr
The right-hand side of this equation is the mathematical definition of the first
derivative of n rx, so we can write:
d(rr,.x) P 1 - P2
--'---'-= = r
dr L
43
I
---.1
--rr
r !
b
1 I
I I
+--- ------ -:--- --
I I
I I
I :
- - - L--
1 I
...-x 1
P1 P2
Figure 11.6 Laminar flow in an annulus
and integrating this equation we obtain :
r =Pt-P2r + C1
rx 2L r
(11.10)
Applying Newton's law (eq1:1ation Il.lb) and integrating again we obtain:
(11. 11)
In this equation the constants C
1
and C
2
still have to be evaluated. We can do
so by applying the boundary conditions vx = 0 at r = a and at r = b. We then
find for the shear stress distribution (from equB:tio!lll.lO) :
!
/ 1/ - (-ba) z 2
r
Pt - Pz
= r
u 2L b
2ln -
(II.12)
and for t he dis_tribution :
, = Pt - Pz bl 1 _
2
+
1
- (WIn
vx b b
(IL13)
whereas the fl ow rate is given by:
I - -[I - ']'
b lnm
(11.14)
44
In Figure Il.7 the shear stress and the velocity distribution for laminar flow
in an annulus are shown. The reader may check that , indeed:
r,.x = 0 at
and:
vx = vmax at the same value of r
2
j b
2
For a = 0, equations (II.l2), (II.13) and (II.14) reduce to the corresponding
relations for flow through a circular tube. For a very narrow annulus, thus for
a = b - d (d a) and r = a + y (y a), the corre-sponding relations for flow
in a narrow slit between two flat plates are obtained (problem 1).
- vx
+ Trx
Figure ll.7 Shear stress and velocity distribution for laminar flow
in a horizontal annulus
The shear stress and the velocity distribution for flow in a circular pipe can,
of course, also be found from equation (11.11) by applying the proper boundary
conditions Yx = 0 at r = Rand vx = oo at r = 0. The last boundary condition
1
= 0, which we would have also found from equation (II.lO) by
realizing that at r = 0, r,.x = oo. The reader may check that equation (11.10) can
also be derived from the proper micromomentum balance, equation (1.30).
Coming back to laminar flow in an annulus it is interesting to compare the
shear force at the inner and outer wall. From equation (II.l2) we find the force
at the wall as :
at inner wall : Fa = - ~ a L - r w l a = -na(p
1
- P2)
I
45
The sum of the two forces is then (ra.x acts in the negative x-direction !) :
Fa + Fb = -2naLrwla + 2nbLrwlb = n(b
2
- a
2
)(pt - p
2
)
i.e. the total shear force at both walls equals the pressure force on the cross-
section of the annulus, as should be the case.
The ratio of the forces at the inner and outer wall is:
b
2
/a
2
- 1
---- -1
- arax 2ln (bja)
- - - - - - - ~ ~ ~ - -
b2/a2 - 1
b
2
j a
2
- 21n (bi a)
This relation is shown graphically in Figure II.8. We see that for the limiting
case bja ...,. 1 the two forces become equal. The ratio of the force on the inner
cylinder to the total force is given by:
Fa -arax 1
----=--- - ----'---- - ----
Fa + Fb -arax + btbx 2ln (bja)
1
(bj a)
2
- 1
This relation is also shown in Figure II.8. If the ratio of the surfaces of the
outer to the inner cylinder is 5:1, then 27 per cent of the total force still acts on
the smaller cylinder ! Considerations of this kind are of importance during the
extrusion of, for example, plastic pipes.
b l a-
Figure 11.8 Forces acting on the walls of an annulus
11.1.4. Flow caused by moving surfaces
The foregoing sections have dealt with the fl ow of liquids caused by a pressure
gradient over a length of pipe. Flow can also be caused by moving a surface
through a liquid.
46
Let us consider what would happen if, in an infinitely long horizontal annulus
filled with a Newtonian liquid, the inner tubes were moving in the x-direction .
(Figure II.6) at a constant velocity v
0
Since there is no pressure gradient
induced from the outside, the force balance under stationary conditions
becomes for this case:
from which we find :
I Fx = 0 = r,_x1nrLI,. -
d(rr,.x) =
0
dr
Integration, application of Newton's law (equation II.16), second integration
and determination of the two constants of integration via the boundary con-
ditions (r = b, vx = 0 and r = a, vx = v
0
YfVo
1
(11.15) r =-
rx r In (b/ a)
and :
In (bj r)
vx = Vo ln (b/a)
(11.16)
The first relation shows that the forces on both surfaces are equal, as should be
the case:
1tYfVo
Fa = nar:,.aL = Fb = = In (b/a) L
In Figure 11.9 both relations are shown graphically. The reader may check
that for bj a-+- 1 both expressions reduce to the proper relations for two parallel
Figure ll.9 Horizontal annulus with moving inner
cylinder
A different situation occurs if the inner cylinder is not moving in the x-
direction but is rotating at a rotational speed n (s -
1
). This case can easily be
treated by applying cylindrical coordinates (Figure II.lO).
Figure ll.lO Annulus with inner
moving cylinder
A momentum balance leads to :
from which follows:
F = 0 = 2nrLn:,
8
1, - 2nrLn,elr+dr
d(r
2
r..e) =
0
dr
47
Integration of this equation yields, after application of Newton's law, which
reads in cylindrical coordinates for non-compressible fluids:
and second integration:
cl c2
v
8
=-- -r
2w 11
and, applying the boundary conditions v
8
= v
0
at r = a, v
8
= 0 at r = b, we
find (using v
0
= na) : .-
(II.17)
and:
2na
2
b
2
ry
r,e = (bl - a2)r2
(II.18)
As should be the case, the torque at the inner and at the outer wall Fr = 2nr-r,rL
is the same. The shear stress and the velocity distribution are shown in Figure
II. ll.
48
o ~ - - - - - - - - - - - - - - - - - - - - - - ~
a r b
Figure 11.11 Shear stress and velocity distribution
in an annulus with inner moving cylinder .
The above situation is present, for example, in friction bearings. A similar
situation occurs in a rotary viscosimeter. Here the inner cylinder is immobile
and the outer cylinder rotates. The annulus is filled with liquid and the force
on the inner cylinder is measured (e.g. by means of the torque of a wire). The
reader may check that the viscosity of the liquid in the annulus is given by:
F b
2
- a
2
t
'1 = - - -...,..--,--
4rtnL a
2
b
2
if Fr is the torque, L the height of the cylinder and if effects caused by the bottom
of the annulus are neglected.
The reader is asked (problem 4) to solve by himself the analogous simple
problem of two parallel moving plates.
Il.l .5. Flow through pipes with other cross-sections
The calculation of velocity profiles for pipes with non-circular cross-sections
is difficult but possible. Boussinesq developed analytical solutions for the
velocity profiles for Newtonian liquids in pipes with elliptic, triangular and
square cross-section. These results can be summarized by defining a flow
coefficient M
0
for which the volumetric flow rate of flow through a slit with
the same smallest dimension has to be corrected. For flow through two parallel
plates we have developed equation (1I.4), which reads, if B represents the width
of the slit :
tip 3
v == 121JL Bd
Thus, for other cross-sections the flow rate is given by :
llp 3
v = 12'1L Bd M o
Values for the correction factor M
0
can be found from Figure II.12.
(II.19)
Rectangle
!2:{
0 1 8-----1
Isosceles triangle
d/8
Figure 11.12 The flow coefficient lvf
0
for stationary flow of Newtonian
liquids in cross-sectjons of various geometries
II.J.6. Non-stationary flow
49
Until now we have concentrated on stationary (steady-state) situations,
i.e. on situations which were independent of the elapsed time. ln this section
we will consider what happens in the time between starting a flow and reaching
the stationary state (transient behaviour) and how long it takes to obtain a
Consider the situation drawn in Figure II.13. From t = 0 a big fiat horizontal
plate is moved with a constant velocity v
0
. At t = 0 the liquid is stagnant
(va:: = 0). At t > 0 momentum transport in the + y-direction takes place, result-
ing in a velocity distribution which is not only a function of the x-coordinate
but also of time.
50
/aQ < <
Figure D.t3 Horizontal flat plate in infinite viscous
liquid
To analyse this problem we select a control volume parallel to the plate and
of height dy. The micromomentum balance over this' height dy then reads:
d(pvJ d I dr
1
"'
dt y = Tyxy- r,xly+dy = - dy dy
and, after applying Newton's law (equation 11.16) we obtain :
dv d
2
v
X X
-=v--
dt dy
2
The same equation could have been obtained from the general micro-
momentum balance, equation (130). This differential equation is analogous to
the equations describing the penetration of heat and mass:
dT d
2
T de d
2
c .
dt = a dx
2
and dt = []l dx
2
respectively
which will be treated in more detail in chapters III and IV. There, also, the exact
solution of this differential equation will be discussed.
In Figure II.14 the velocity distribution for the system of Figure 11.13 is shown
for two short times. As can be seen, the distribution can be approximated by
straight lines which cut the abscissa at y = foi. This distance foi is called
the penetration depth, i.e. the distance over which momentum has penetrated.
This solution is applicable as long as the penetration depth is smaller than the
depth of the liquid.
We can now estimate the velocity gradient as:
or, since V
00
= 0:
dv"' Vo17
r = -1]-=--
w dy Fvi
(II.20)
j;;t( ;-;;-;-; X
Figure 11.14 Velocity distribution for system of Figure
II.13
51
A somewhat different situation occurs if, instead of a moving plate, liquid
with an initial velocity v
0
flows along a fixed plate. This situation will be dis-
cussed in paragraph II.5. 1 (see Figure II.37).
We can now estimate the distance needed for laminar pipe flow to develop
its velocity distribution completely (Figure 11.15). If we move with the liquid
and assume that the velocity distribution will be developed completely when
the penetration depth at x = xs is equal to the distance R to the centre of the
channel, we can write:
R ~ foj = ~ v Xs
V 1l (v)
Thus, after a distance of approximately:
x ~ (v) Rz ~ 01 cf>v
s rrv v
the velocity profile is fully developed.
Figure II. 15 Development of velocity profiles in tubes
(11.21)
52
JI .l.7. Problems
1. Show that equations (II.12) and (U.13) reduce to equations (II.5) and (11.6)
if af b goes to zero. Show that these equations reduce to the corresponding
expressions for flow in a slit if aj b goes to 1.
2. (a) Develop equations for the shear stress distribution, the velocity profile
and the volumetric flow rate for a liquid film flowing down on a vertical
surface. Use the notation given in Figure II.16 (width of plate = W).
g
Figure 11.16 Laminar film on a vertical plate
Answer : 'txy = - pg(d- x)
v = gp ( dx - 1x
2
)
y 11
cPv = (v ) d = pgd3
w )' 311
(b) On a vertical plane {width B = 1m) 1200 kg/ h water flows downwards.
What is the mean velocity, the film thickness and the maximum velocity?
Answer : < v) = 0 725 m/ s
d = 046 X 10-
3
m
Vma)l = 109 m/s
*3. In a rotary viscosimeter filled with paint (a Bingham liquid) the shear
velocities are measured as a function of shear foree :
Shear force (N/ m
2
)
15
18
21
Shear velocity (s-
1
)
6
12
18
53
If this paint is smeared on a vertical wall, how thick can the layer be before
the paint starts tc drip (p = 1200 kgfm
3
)?
3
m
4. The slit (width 1 mm) between two horizontal flat plates is filled with a
viscous Newtonian liquid ('1 = 100 Nsjm
2
). If the surface of the plates is
10-
2
m
2
, which force has to be applied under steady-state conditions to one
plate in order to move it parallel to the other plate at a velocity of 1 crnfs?
Calculate and draw the shear force and the velocity distribution between
the plates. After what time is the stationary state reached?

5. The 'viscosity' of a non-Newtonian liquid is approximately given by :
'
dvx'
YJ = YJo dr
where '7o is a constant. Find the volumetric fl ow rate of this liquid in a
horizontal pipe (diameter 2R) under laminar and isothermal conditions.
2n , {!;P
Answer : <l>v = 7 Ri
6. Viscous oil ('7 = 01 Nsjm
2
) is pumped through a horizontal pipe (diameter
1 em, length 10m) into an open vessel. The absolute pressure just behind
the pump is 3 x 10
5
Nj m
2
.
{a) Determine the volumetric fl ow rate of t he liquid.
(b) What would be the flow rate if a pipe with rectangular cross-section
(05 x 20 em) was used?
Answer : (a) 4>v = 4-9 x 10-
5
m
3
js
(b) ., = 3-47 x lo-s m
3
/s
*7. Check John's calculations about the syrup dripping from the spoon,
reported at the beginning of this paragraph.
Assume: YJ = p = 2000 kgjm
3
, initial thickness 5 mm, height of
spoon 5 em
Problem 3
For a liquid with Bingham characteristics the following relation between
shear force and velocity gradient is valid if we use the coordinates shown in
the small drawing overleaf:
54
B
The size of the yield value t
0
for this paint can be determined from the experi-
mental results as t
0
= 12 N j m
2

A force balance will lead to the maximum possible layer thickness before
dripping takes place :
and thus:
and we find d < 10-
3
m.
pghdB < -r
0
hB
d
to
<-
gp
Problem 7 John and his spoon
We are dealing with a situation where a liquid film is flowing down from a
vertical surface. An expression for t:f>v was found in problem 2 of this section:
cP = pgd3W = WHdd
v 3 ~ dt
---- dl
d=d dd pg sr =r
L=do d3 - 3 ~ H r=O
and after integration:
1 [ 1 1 ] pgt
- 2 d ~ - d
2
= J17H
or:
d
-=
55
Spoon ( face on)
Spoon ( in profi l e)
l
-w-
H
J
For long times :

_:_:::__..::.. 1
3t7H
\
\
\
'
:-do
I
J
I
I
and we can simplify the above relation to:
d/3riH

For djd
0
= 001 we find t = 1500 s = 25 minutes. The period to diminish the
syrup layer to molecular dimensions is found by substitution of
as 15 x 10
12
s.
11.2. Turbulent flow
d 5 x 10-
10
d
0
= 5 x 10-
3
=
10
-
7
John stood at the edge of a circular sewer channel of 4 m
found floating in the channel 72 km downstream, in a fisher
village, at 11.45 p.m. The suspect had a watertight alibi for
the time before 9.15 p.m. The channel was now half-filled and
a piece of wood passed John at a speed of 1 mjs. One of the
sewage plant operators had informed him that until the time
che body was found the water height had been less than one-
quarter of the diameter of the channel. Thinking about one
of his college courses, of which you will .find a brief outline
below, he made a quick calculation and ordered to sec the
suspected man free.
11.2.1. Turbulent flow in pipes
In a well-known experiment in which he injected a filament of dye into a fluid,
Reynolds demonstrated that above a certain flow velocity in a pipe the flow
changes from laminar into turbulent. In this case velocity fluctuations occur
with components perpendicular to the main direction of the fiow.
56
The turbulence of the flow field makes the effective viscosity (from a macro-
scopic viewpoint) larger than the molecular viscosity 17, because neighbouring
layers of different mean velocity not only transmit momentum to each other by
molecule interaction but also by cross-currents (eddies). These cross-currents
decrease in intensity the nearer they are to the wall. Because of this varying
effective viscosity (which can be supposed to be composed of the t7 and the 'eddy'
viscosity pE) the velocity distribution differs from that in laminar flow. The eddy
viscosity just introduced is, of course, not a physical property of the liquid but
is dependent on the place and on the type of flow. The dimension of the eddy
viscosity pE is kg/ ms ; thus the term E has the same dimension as the kinematic
viscosity v (m
2
js).
Figure Il.17 shows an example of the velocity profile in a pipe with a circular
cross-section for laminar (parabolic) and turbulent flow (more flattened).
X
r Vmox"' 2<v >
I
<v >
I
I
I
I
I
I I
I I
---r--- ----r---
I I
I I
I I
I I
I
I
Vmox = I I to I 2 < v >
1
I
I I
i <V > I
Figure 11.17 Laminar and turbulent velocity profiles in a pipe
If we want to know something about the frictional force exerted on the wall
by a liquid flowing at a velocity (vx), the shear force -rw at the wall must be
known. Since for incompressible Newtonian liquids :
dv.x
!w = r[ dr
r=R
holds (no turbulence at the wall, pEat wall = 0), we must know the velocity
distribution in order to calculate -rw. For turbulent flow this is not possible by
momentum balances, because with turbulence their solution presents insur-
mountable difficulties. However, on the basis of physical intuition it can be
posed that :
-rw = f (p, rJ, (vx), D, geometry)
from which, with dimensional analysis, it follows that the quotient of -rw and
p(vx)
2
for geometrically similar pipes only depends on the Reynolds number
( = p( vx) D/tt). This is usually written as:
(11.22)
57
With this equation the constant f, the so-called friction factor is defined: on
account of the above it is a function of the geometry and of Re. For turbulent
flow this function can be obtained only from experiments on friction losses in
pipes and channels. Conversely, these friction losses are, in practice, calculated
using this friction factor f, the value of which is well known for widely divergent
conditions (see paragraph Il.4.).
Equation (II.22) can be read as follows: the quotient of -rw and p<v,)
2
which
is fully determined by the velocity distribution is for geometrically similar
pipes only a function of Re. This means that in geometrically similar pipes
at the same value of Re the velocity distribution in corresponding cross-
sections must be similar. The two flow fields are then called dynamically
similar. Consequently, the flow pattern is characterized by theRe number. The
change from laminar to turbulent flow is therefore given by a critical Revalue.
The flow through a straight circular pipe is turbulent if Re > Recr 2000.
This also applies to straight, non-circular pipes if, in the Re number, as the
characteristic length the hydraulic diameter is introduced, which is defined as
four times the surface area of the cross-section divided by the wet circumference
{see, further, the next section and check that for a circular pipe the hydraulic
diameter equals D).
TheRe number can be interpreted as the ratio of the momentum transport by
convection (,.., pv
2
) and the momentum transport by diffusion (,.., qvf D).
Apparently, in turbulent flow the transport by convection, i.e. by eddies, prevails.
In a free-falling liquid film these eddies already occur at lower velocities; film
flow becomes turbulent at aRe number 1000. During flow through a curved
pipe (e.g. a spiral; radius of spiral R) the centripetal forces stabilize the flow and
the change to turbulence does not occur until higher Renumbers have been
attained. Empirically it was found for the critical Renumber :
Re" =

ror to'
(11.23)
The fact that turbulence decreases towards the wall has led to the concept
of the laminar boundary layer. No definite thickness can be assigned to such
a layer because the turbulence changes gradually. For the qualitative insight
:t can be useful to scliematize the flow field by a laminar boundary layer, .with
thickness in which the velocity gradient is constant, and an adjacent turbulent
region, with the mean velocity <vx> in which the velocity gradient is
;>ractically zero. The shear force at the wall is then given by:
( vx)
rw='f'/T
(II.24)
and the thickness (J h can be estimated with the help of equation (11.22) as soon
as the friction factor is known.
With the aid of the laminar boundary layer thickness b
1
, we can give a new
mterpretation on the physical meaning of theRe number. The ratio of the tube
58
diameter and thickness of the laminar boundary layer is found to be :
D = Ihw = f Dp(v;;,) = f Re
c5h 1'/(V),) 211 2
(II.25)
This means that for pipe flow at Re = 10s (f = 00045) the distance from the
wall at which turbulent momentum transport overrides the transport by
viscosity is c5
11
4 x 10-
3
D.
A velocity distribution as given above is physically impossible, but it can
serve as a rough model for heat and mass transfers in turbulent flow.
In the case of heat transfer a layer of thickness c5T is assumed over which the
complete temperature drop occurs. Outside this layer the turbulent eddies
disperse the heat so efficiently that a uniform temperature exists in the core of
the flow. Within the layer eddies are assumed to be absent. This layer of thickness
{JT is called the thermal boundary layer analogous to the hydrodynamic bound-
ary layer (thickness oh) treated above. It will be clear 'that both models do not
contradict each other as long as c5T c5
11
.
The question now arises as to whether a relationship exists between the
thicknesses of the hydrodynamic and the thermal boundary layers. Some
insight into this problem can be gained from the fact that the eddy diffusivity
E increases with the third power of the distance from the pipe wall :
E = C
1
y
3
(II.26)
From the definition of o
11
, this means that eddy momentum transfer overrides
viscous transfer if:
pE = pC
1
y
3
11 . or E = C
1
y
3
11/P = v
Hence:
{II.27)
The diffusivity of heat by eddies is also described by equation (ll.26), whereas
the diffusivity of heat by conduction only is given by a= J..jpcP (m
2
/s). Hence,
the distance c5T from the pipe wall at which heat transfer by eddies overtakes
conductive heat transfer is given by a reasoning similar to that applied above
to be :
(II.28)
Hence, the sought relationship between or and oh is obtained from equations
(ll.27) and (11.28) and is found to be:
= t = Prt
Or a
(II.29)
where Pr is named the Prandtl number. For gases Pr 1 and therefore c5
1
, oT,
for viscous liquids Pr 1000 and o
11
lOoT, and for water at room temperature
Pr 7 and c5
11
2c5r.
59
The reader can imagine that mass transfer between a wall and turbulent pipe
How can be depicted by using a concentration boundary layer of thickness be-
The ratio between this thickness and the thickness of the hydrodynamic bound-
{> (v)t
b: = II) = set
(11.30)
where Sc is the Schmidt number. For gases Sc ~ 1 and ~ , ~ l>c and for liquids
Sc ~ 1000 and b" ~ lObe. We see that in these cases the thickness of the hydro-
dynamic boundary layer is greater than those of the respective thermal and
concentration boundary layers, b
11
> bT, oc.
We have .already stated at the beginning of this chapter that it is impossible
to derive theoretically a relation for the velocity distribution during turbulent
Bow. We can, however, construct a relation between velocity and the friction
factor and see whether we can determine the coefficients in this relation by ex-
perimental results. Analogously to the devdopment of equation (II.22), we
assume that the mean velocity is a function of the shear stress at the wall, the
specific gravity and the kinematic viscosity of the fluid and of the pipe diameter ;
so:
Via dimensional analysis we find:
~ = ~ { ~ ~ } = C { ~ ~ r
(II.31)
which can be rearranged to :
rw = f = (C)-2/(l +n>{-v-}2n!<1+n>
p(v)
2
2 (v)D
(11.32)
Equation (11.31) indicates that in the turbulent region also the friction factor
will be a function bf the Reynolds number only. Results of friction factor
measurements in smooth pipes have been collected and correlated by Blasius as:
_ 4f = 0316 Re -
0
.
25
This correlation is discussed in more detail in the next section. At the moment
we can use it to determine the constant n in equation (II.32). Via :
we find n = l
2n
-- =025
1 + n
Repetition of the foregoing procedure for the flow velocity in the x-direction
at any place r yields, if we realize that vx must be zero at 2r = D, instead of
equation (II.32}:
(11.33)
60
Thus, maximum flow velocity occurs in the centre of the tube at r = 0 and we
can write :
_2_ = (l _ 2r)-+
Vmax D
(II.34)
In order to find the ratio between mean and maximum velocity we can calculate
( v) from equation (II.34) as :
( v) = _1_ JD/2 v dr = Vmax JD/2 ( 1 - 2r) -+ dr
D/ 2 0 X D/ 2 0 D
Solving this equation we find:
(v) 7 -
- =- = 0875
Vmax 8
compared with (v)/ vmax =!found for laminar flow in circular tubes.
The velocity distribution (equation II.34) gives a rather good description of
the actual situation, as illustrated in Figure 11.18, where measured values of
(vfvmu? are plotted against (1 - 2r/ D).
I
><
0
E
"'
'
"'
0
i
/
~ 0
5l
6
v
I
0 5
,CY
0
0
~
,..t
c;/
0\
/
0
_.1-2r/D
'o
\
>
~
'o
'
r;\
~
0 5
Figure ll.18 Measured velocity distribution for
turbulent flow
I 1.2.2. Pressure drop in straight channels
At the end of paragraph 11.1.2 we saw that the pressure drop in a pipe system
can only be predicted as a function of the flow rate if the shear force at the wall
is known. For laminar flow this quantity can be predicted directl y because the
relation between -rxy and dvy/dx (e.g. according to Newton, equation 11.1)
characteristic of the fluid offers sufficient additional information. For turbulent
flow we lack this additional information.
In order to be able to predict the flow rate for turbulent flow on the basis
of a given pressure drop (or, inversely, the pressure drop for a given flow rate),
use is made of graphs and tables of earlier measuring data in which numerical
61
values can be found of the dimensionless groups that are important to this
problem. On the basis of dimensional considerations these results can then be
used for solving every new but analogous problem. We shall deal in succession
with the fiow through a straight channel, the fiow through fittings and the
frictionless fiow.
Since in what follows we only speak of a mean velocity and not of velocity
distributions, the microbalances are of no use to us in this chapter. The quanti-
ties which we shall now come across belong to balances over an entire channel
or channel system (i.e. they are macroscopic quantities to be used in macro-
balances).
If the shear stress on the wall of a horizontal straight channel (cross-section A,
circumference S) is known, we find for the stationary state the momentum
L F.x = 0 = p
1
A - p
2
A - twS(x
2
- x
1
)
and, using equation (II.22), we find for the pressure drop:
S(x2 - x 1) Jl < )
2
S(x2 - x
1
)
Pt - P2 = tw A = 2P vx A
(11.35)
From this equation, which is named after Fanning, it is clear that A/S is the
characteristic dimension of the channel cross-section, so that it is obvious to
call A/ S the hydraulic diameter* of the channel. If the numerical factor 4 is
introduced in the above relation, which is done for historical reasons, fS/ A can
be written as 4fS/ 4A and it is still general practice to state the value of 4f
The problem of predicting the pressure drop in straight channels can thus be
reduced to predicting friction factors. We shall now collect data on the friction
factors for Newtonian flow in circular pipes. The Hagen- Poiseuille formula
(II. 7) valid in the laminar field yields, together with equation (11.22), for the
friction factor :
'7 64
4f =
64
p(v)D = -Re
(II.36)
In the turbulent region the friction factor is considerably less dependent on Re,
notably proportional to- .Re-
0

25
-0. As a consequence, the pressure drop is
proportional to (v)
1
.
75
-
2
, so that for a given pipe the pressure drop increases
:nuch more strongly with ( v) than in the laminar field, but the viscosity of the
9ow has less influence.
For pipes with a smooth wall the experimentally found relation proposed by
Blasius:
4f = 0316 Re- o:z.s (4000 < Re < 10
5
)
gives a good description of the dependency off on Re.
(11.37)
The quotient A/S is also called the hydraulic radius, with the consequence that the hydraulic
di ameter is four times the hydraulic radius!!!
'
1
'-
0
u
.E
c:
.Q
u
.:
LL
006

"
"'

"-..

.. ..
Laminar j Turbulent
(circular .
cross- section!
only l Roughness of wall x . ( m l
Drown tube 2xto-6
Steel tube 40xlo-
6
Galvanized iron t50xlo-6
Cost iron 0 3 x 10-3
Wood,concrete 0 2-2ltl0-3
'
........
.........
'

..........
10-
4
..........
:-------=.::::::.....:----t2x 10-
4
...........
to-
4
--.::--._-------15x 10-5

56 s 103 2 3 4 G s 104 2 3 4 6 a to5 2 3 4 6 8 106 2 3 4 6 8 107 2
Re= 4A
II S
Figure 11.19 Friction factor for flow in tubes
---- - -
63
In the case of turbulent flow the roughness of the wall surface also influences
the friction losses. When the mean height x of these irregularities (roughness)
becomes of the same order of magnitude as the thickness of the laminar bound-
ary region (either in the case of great roughness or at high Renumbers), part of
the resistance will be caused by direct momentum transmission between the
turbulent core of the flow and the protuberances of the wall. In extreme cases
(very great roughness or high Re) this causes a pressure loss proportional to the
kinetic energy of the flow, so that the friction factor no longer depends on Re.
The wall roughness is allowed for by the dimensionless quantity i / D = relative
roughness. Consequently, the friction factor is in general a function of Re and
i j D.
In Figure II.l9 the relation between 4f and Re is shown for various values of
the relative roughness. It appears that in the turbulent region two parts can be
distinguished : at high Re numbers, the so-called completely turbulent region
where f only depends on the roughness, and at lower Re, a turbulent transition
region where both Re and roughness play a part. The mean roughness of a
number of materials is also given in Figure 1119.
For pipes with non-circular cross-section the friction factor in the laminar
region is proportional to 1/ Re. The proportionality constant depends on the
form of the cross-section and is known for a number of cases (see paragraph
11.1.5 and figure II.12).
If the flow is turbulent, approximately the same values of the friction factor
apply as in tubes with circular cross-section if, in the Fanning equation (II.35)
and in the Reynolds number, use is made of the hydraulic diameter. Hence the
advantage of using the hydraulic diameter is that (only for turbulent flow) with
one graph for the relation between the friction factor and Re the pressure drop
Flow si tuation Hydraul ic diameter A
o, = 4A ! S
a
Cir;culor pipe D
r. o2
4
8
-0E}r
Concent ri c pipe or sl it 0
2
-01 = 28
.,. 2
0
z >
- ( Dz- I
4
CJe Rect a ngula r p i pe
2 WB
WB
w
W+ B
r? P
4WH
WH
Open channel --
- =-=rH W+2 H

Open channel
2H
J.. Hz
12
2

Half - f i lled 0
.!!:. oz
8

L iquid film 48 a.,. o
Figure Il.20 Hydraulic diameter of various channels
64
can be calculated for all forms of the cross-section and for all Newtonian fluids
(i.e. accurately for circular cross-sections and approximately for other forms).
A number of frequently occurring hydraulic diameters is shown in Figure II.20,
together with the cross sections available for flow.
The pressure drop in spirals is, for the reasons stated in paragraph 11.2.1,
higher than in a straight tube. For a rough estimation of pressure drop of
turbulent flow of waterlike liquids, the empirical relation :
l l p , p h ~ = dp,.n;&b<p;p< X ( 1 + 374 ~ )
(II.38)
can be used (D = tube diameter, 2R = diameter of spiral). In practice, the
chemical engineer often has to determine which flow rate can be attained at a
given pressure drop. Thus he has to solve the relation (11.35) :
L
tlp =4ftp(v)
2
D
for ( v ), which can be done by a trial and error procedure. Rearrangement of
equation (!1.35) yields:
JRe2 = 2llppD3
4Lyt
2
(11.39)
For a given problem the right-hand side of this equation is a known quantity ;
thusfRe
2
is known. In Figure 11.21 the relation betweenfRe
2
andRe is shown
for pipes of differing roughness. With the help of this graph the Re number can
be found and consequently the volumetric flow rate.
Now that withfthe pressure drop in the straight horizontal channel is known,
it is obvious that the energy dissipation in the channel is also fixed. From the
energy equations {1.5) and (I.6) it follows that the frictional beat released for
each transported unit of mass in this case equals (p
1
- p
2
)fp. The energy
necessary for pumping the fluid is then :
A = lPm (pl ~ P2) = lf>v(pl - P2) = 4>m; (v)2 S(x2; X 1)
(11.40)
This means that the dissipated mechanical energy per unit of volume equals the
undergone pressure difference. Thus, for laminar flow 4> AI v is proportional to
(v) and for turbulent flow to (v)
1
'
7 5

I 1.2.3. Pressure drop in pipe systems
The calculations for pressure drop in straight pipes cannot easily be extended
to piping systems with fittings (valves, bends, constrictions, etc.). The reason
is that the forces which the piping system (now not in one but in three directions)
exerts on the flow are position-dependent, which makes the three momentum
balances too complicated for mathematical treatment. It was Bernoulli who
in principle indicated how this problem could be solved By scalar multiplication
65
1
10 6
10
3
10
4
10
5
10
6
10
7
Re--
Figure ll.ll The relation ofJRe
2
andRe for turbulent flow
of the three momentum balances with the corresponding flow velocities he
obtained, after addition of the three equations, a relation between the various
forms of mechanical energy in a flow field Whereas the terms of the momentum
balance have the dimension of a force or momentum flow, multiplication by
the velocity leads to an equation between energy flows.
In the stationary state for part of a pipe between cross-sections 1 and 2
0 = - {f dp + g(h, - hi ) + :l{(v,)
2
- (vi )
2
)} 1/>m + 1/> A - Er,l/>m (11.41)
Err in this equation is the amount of mechanical energy which per unit of mass
is converted into heat owing to internal friction. This constant is always positive
because the dissipation is a result of statistical transport, which according to
thermodynamics is attended by degradation of energy (entropy production).
66
Hence we may conclude from equation (II.41) that owing to friction the total
mechanical potential (the first four terms of the equation) decreases in the direc-
tion of the flow, so that here we can speak of the law of ' non-conservation'
of mechanical energy. The readers should realize that the Bernoulli equation
is not a new balance but rather a summary of the three momentum balances.
If the amount of energy added to the system per unit of mass 4> A = 0 and the
energy loss due to friction is negligible, equation (II.41) can be written for any
point in the flow field as :
1
- dp + g dh + v dv = 0
p
or, also, if at the same time the specific gravity is constant:
!?. + gh + tv
2
= constant along streamline
p .
(11.42)
(11.43)
These are two forms of the well-known equation of Bernoulli, which can also
be derived in a simple way by using the momentum balance for a frictionless
fluid (Newton's first law). The more general equation (11.41), which plays a
very important role in the technical flow theory, is furt her referred to as the
extended Bernoulli equation.
On account of what we know about the pressure drop and the energy dis-
sipation in a straight channel we can relate the constant Err for such a channel
(length L) to the friction factor and the mean velocity (v) in that channel
(equation II.40):
1 2SL
Err= f2.(v) A
(11.44)
Local pressure losses are mainly due to a sudden increase in the cross-section
of the pipe. According to Bernoulli's law in such a diverging flow the velocity
downstream decreases and the pressure increases. This pressure gradient may
cause the fluid near the wall to flow upstream. This results in systematic eddies
(not the same a ~ turbulence!). In these eddies part of the kinetic energy is
destroyed by internal friction. Only if the diameter is carefully widened (e.g.
via a diffuser with a vertical angle of at most 8) do these eddies not occur.
The dissipation per unit of mass in a fitting is described with a friction loss
factor Kw which is defined as :
(II.45)
In practical cases Kw is only a function of local geometry and is for not too
low Renumbers independent of Re. For (v) the mean velocity in the pipe on
the downstream side of the local resistance is usually taken.
To calculate the resistance of a pipe having various pieces of different dia-
meter, bends, valves and other local resistances, both equations (II.44) and
(II.45) should be introduced in the extended Bernoulli equation (11.41), via
67
the lost frictional energy Err. The latter constant integrated over the entire
pipe then becomes :
(ft(v)
2
s:) . + 2;: (K...,t (v)
2
)j
I I J
(11.46)
where the first sum is taken over all pieces of straight pipe and the second over
all local resistances. In air conditioning channels, often short bends have to
be made owing to lack of space. However, the K..., values then become un-
permissibly high so that frequently guide vanes are built in to lower these values
(avoidance of eddies, streamlines).
Empirical correlations between K..., values and the geometry of the pipe
system are known for a great number of situations. A few K..., values are listed
in Table ILL In some cases it is possible to derive K..., values theoretically.
Consider the situation shown in Figure II.22. For the calculation of the energy
dissipation three equations are available, viz.:
I
I
I
1 I
---- :---------------+- ---
1 I
l I
I
I
Figure 11.22 Flow in a channel with changing cross-section
the mass balance :
m = pv1A1 = PVzAz
..
the momentum balance :
0 = P1A1 + mvt - P2A2 - mv2 +F...,
the Bernoulli tquatioQ (II.41) : 0 = - u: ; + f g dh + f v dv} t/>m
+ </>A - Err</>m
The latter equation simplifies for horizontal pipes, no energy input and incom-
;>ressible fl uids to:
Pt - Pz
1 2 2
Err= + 1((vl ) - (vz) )
p
using the above three balances it is possible to estimate the amount of dissipated
!nergy Err if we can find an expression for the force the liquid exerts on the
w We will illustrate the procedure to be applied in four examples.
68
Table II.l Some values of the friction Joss factor
K k{ I - ~ ~ }
2
rY =1 < 1001 10 1201301 40 150 ,600,70,80 190 I
k = 0 0 17 0-41 0 71 0-90 103 112 113 1-10 105
A
2
fr = to l 2o 130 140 !so I so jmo 1800 1
Kw = 0 16 020 024 028_ 031 032 0 34 0 35
..:
Az Kw
__ ,
= ( A2 ) 2
A I
02 0 3 04 0 5 0 6 07 08 0 9
Ao
Ao
A,
A2
A,

0 1 02 03 04 0 5 0 6 0 7 0 8 0 9 1 0
Kw ,. 232 51 20 9 6 5 3 3 1 I 9 12 0 73 048
A
1
> 20A
2
~ ..J- = 20 .. 140 16o lao 19o .. j'oo .. lt2o 1'40 j
~ ~ - - - - : - 1 - - Kw = OD5 0 14 036 0-74 0 98 I 26 186 2-43
{
0 35} ,j-
K = 0131 + 0163 ( - ) -
w R goo
69
Kw (referring to downstream velocity for Re > 10
5
)
, ; 15
1
3()0
1
45
1
60
1
900 I
K,= (}02 0 II 0 26 0 50 120
-- Kw =- 2 --- Kw-075
tPm2
= 0
0 2
OJ6
0 8 1 0
.ml
K.,
= 0 95 088 0 89 095 110 128
.m2 Kw2
Kw2
: - -0 08 -oo 007
0 21 0 35
+ - s - - - ~
~ :
20 30
400 150
600 7QO
80 185
K," 751 118 33 II
4 15 0 52 0 24
Kw = 4 86 206 53 17
55 16 0 29 0 05
Gate valve ' Fraction closed 0 i/8
2/8 3/ 814/815/8
6 / 8 7181
Kw
=005 007 0 26 0 81 2 I 5 5
17 981
.-
Globe 110lve Open K. =
...
4
Sieve plate
:
061 051 0 2
Fraction- free space i Kw
0 7 I 5 20
Cyclone
Kw
10 to 20
Water meter K,., =
6 to 12
70
Example 1. Straight t ube
Here the momentum balance reduces to (A
1
= A
2
, (v
1
) = ( v
2
) ) :
p 1 A I - P2A 2 + F w = 0
Since the liquid exerts on the wall a shear force only:
we find :
F..., = - t w- JSL = -fi pv
2
SL
I
t 2SL
Pt - P2 = 2PV A
The Bernoulli equation reads under these conditio!!.s ((v
1
) = ( v
2
)):
E
_ P1 - P2
fr-:
p
as we found earlier. Substituting for p
1
- p
2
the expression developed from the
momentum balance, we finally find :
t 2SL
Err = f zv A
as should be the case. Extending the d e f i n i ~ i o n of the Kw value to the case of
straight pipes:
it follows that for straight pipes :
K =JSL
w A
Example 2. Sudden expansion
In the case of sudden expansions eddy currents will occur, leading to extra
energy losses. The momentum balance for this case (see Figure 11.23) between
18
'
I
I
I
i 2
I
I
I
Figure 11.23 Sudden expansion of pipe
Pt Al - PzA2 + <f>mvl - <f>mv2 + F..., = 0
If we neglect wall friction, F..., is given by
Fw = (A2 - At)PB
71
where p
8
is the pressure just behind the expansion in plane 'B'. This pressure
will be p
1
< p
8
< p
2
. Only experimental results can show which pressure bas
:o be used.
-\. Let us assume p
8
~ p
2
, thus Fw = p
2
(A
2
- A
1
). Then we find from the
momentum balance :
- c/>m(v1 - Vz) 2 Az(A2 1)
Pt - Pz = = - pv2- - -
A
1
A
1
A
1
We see that the momentum balance predicts correctly that p
2
> p
1
. Sub-
stituting p
1
- p
2
in the Bernoulli equation:
E
_ Pt - P2 + li 2 2)
rr - 2\vl - Vz
p
we find after some sorting and applying of the momentum balance :
This is evidently an impossible solution (Err must be positive, otherwise we
would gain energy), thus our estimate of p
8
must have been wrong.
B. Repeating the exercise assuming that PB ~ p
1
, we find:
and, with the BernouHi equation, finally:
Thus, the resistance factor is given by :
K = _::_1
(
A ~ )
2
..., At
which fit s very well with experimental results.
72
For the case shown in Figure II.24 we can roughly estimate (neglecting
friction losses):
I I
I
I
I
I
F = Pt + P2(A _ A )
w 2 1 2
12
I
'
I
I
I I
I I
--- +---------------+--- --
1 I
Figure 11.24 Gradual expansion of pipe
where (p
1
+ p
2
)/ 2 is the average pressure. With the momentum balance we
find :
and, applying the Bernoulli equation, we get :
E
_ P1
fr -
Now (A
1
+ A
2
)/2 is the mean cross-section, where the mean flow velocity
(v
1
+ v
2
)/2 is present. Thus:
A.. _ v
1
+ v
2
A
1
+ A
2
'+'m - p 2 2
Substituting 4>m in the Bernoulli equation we find Err = 0-there is no energy
loss. Err has been proved experimentally to be negligible indeed if 8/ 2 < 8.
In practical cases of smoothly diverging flow, only small friction losses occur and
a practical value of the friction loss factor Kw is 0-05.
Let us now reverse the reasoning. How great is F w? Knowing that Err = 0
we can derive an expression for Fw. This force acts at a mean cross-section A,
where the flow velocity is given by {v
1
+ v
2
)/2. A mass balance shows:
'lwater = 10-
3
Ns/rn
2
11
= 16 x 10-
6
Nsj m
2
'latr .
Pwater = 103 kgjrn3
Pair = 12 kg(m3
75
2. An air heater consists of a rectangular box in which a bundle of heating
pipes has been fitted perpendicular to the direction of flow. In a scale
model of this heater the pressure drop has been measured as a function
of the air velocity related to the empty cross-section. If in the actual air
heater the diameter of the pipes is twice as large. what is the ratio between
the pressure drops over the two apparatus at the same Re number (related
to the pipe diameter and the said velocity)?
Answer : dp actual : t:.p scale = 1 : 4
3. Through a horizontal smooth tube there is a turbulent flow of liquid.
By what factor does the pressure drop increase if the flow rate is trebled?
What should the diameter of a new tube be to obtain the original pressure
drop at the trebled flow rate?
Answer : pressure increase by a factor of 68; D
2
= 150D
1
4. Develop an expression for the friction factor for laminar flow in a falling
film and for flow between .fiat plates.
6
12
f = Re (plates)
*5. Through a smooth horizontal tube 1 kg/s water is pressed (Re 2 x 10
4
).
The pressure difference over the tube is 2 x 10
5
Nj m
2
What energy Pis
needed and what is the temperature rise .1 T of the water if there is no heat
exchange with th:e environment? What will be the values for P and aT:
(a) if the water fiow_is doubled?
(b) if 1 kg/ s W(\ter is pressed through a tube 15 times as narrow?

(a) 1280 W, Ol6C
(b) i400 W, 034C
6. Calculate the pressure drop over a capillary flow meter with a smooth
wall if the gas velocity is 5 m/ s. 50 mj s and 100 mj s respectively.
capillary length 10 em 1Jair = 2 x 10-
5
Ns/ m2
capillary diameter 1 mm Pair= 12 kg/m
3
2
Nj m
2
65 x 10
3
N/ m
2
21 x 10
4
Nj rn
2
76
*7. Through a steel tube of 60 m length 120m
3
/h of water is pumped against
a height difference of 25 m. \Vhat will be the difference in pressure between
the beginning and end of the tube if:
(a) the tube is of circular cross-section (D = 10 em)?
(b) the tube has a square cross-section (10 x lO em)?
(c) the water flows through an annulus (D
0
= 14 em, Di = 10 em)?
Answer : (a) 34 x 10
5
Nj m
2
(b) 31 x 10
5
N/m
2
(c) 52 x 10
5
Nj m
2
8. In a factory there is an 8 inch pipe from a trough of a rotating filter to a
hollander which is situated 375 m lower. Cellulose slurry which flows
over the brim of the trough is returned to the system through this pipe.
The pipe is 15 m long and has two short 90. bends and one open gate
valve. What is the flow rate ofthe slurry in the pipe.and what is the capacity?
The viscosity of the slurry is 5 x 10-
3
kg/ ms (at 20C, that is five times
as viscous as water) and the density is 1200 kg/m
3

Answer : 45 mj s; 520m
3
fh
9. A vertical pipe (1 inch diameter, 2m long, smooth) contains one rotameter,
three right-angled bends for connecting the pipe to the and two
valves. The Kw value of these when open, is 6. The rotameter, which
has approximately the diameter of the pipe, contains a float of 25 x to-
3
kg
(p = 2600 kg/m
3
). Calculate the pressure drop over this pipe if water
flows through it at an upward velocity of 1 m/s.
Answer : .1p = 28,060 Nj m
2
*10. When combustion gases enter a chimney the pressure is 250 N/ m
2
below
atmospheric pressure. The chimney is made of smooth steel plates which
are bent to an internal radius of 168 m. 15 tjh combustion gases are dis-
charged. The mean temperature of the gases is 260C, the outside tem-
perature is 21 oc and the barometer reading is 1000 mbar. What is the
height H of the chimney? The density of the burned gas at 25C is p = 127
kgfmJ and rt = 156 x 10-
6
kgj ms at 260C .
Answer : H = 47 m
11. Through a pipe with a 5 em internal diameter flows 241/s of water. An
obstacle in the pipe shows a flow resistance of 100 mm water column.
(a) What is the force the flow exerts on the obstacle?
(b) How much energy is lost by friction?
Answer : (a) = 2 N
(b) = 196 w
77
*12. A strong fireman sends up a jet of water of 2 m
3
j min at an angle of 45.
The maximum height he can reach is 20m. What is the force with which
he is pressed against the ground as a result of spouting?
13. A garden hose (Di = 1 ern) of smooth, inelastic material, 20m long, is
connected to a tap in the ground ; the resistance value of the tap related to
the downstream velocity is K .,., = 7-5. The pressure before the tap is
265 x 10
5
N j m
2
A nozzle (Di = 1/ 4 em) can be connected to the spouting
end of the hose, in this case Kw = 01 related to the velocity in the nozzle.
The spouting end is kept 15 m above the ground. Calculate the velocity
vat which water of l8C leaves the garden hose:
(a) without a nozzle.
(b) with a nozzle.
Neglect the water velocity before the tap.
(b) 147 mj s
14. At least how high can a pole vaulter jump if he can run at a speed of
10 mj s?
15. On the Mons St. Nicolas (Luxemburg) lies an artificial lake with a capacity
of 7-6 X 10
6
m
3
. For twelve hours (mainly during the night) the lake is
filled with water from the Our, 500 m lower. During the next twelve hours
turbines on the Our are driven with water from the lake in order to generate
electric power. The water supply from the lake to the turbines takes place
through two parallel circular concrete tubes, 650 m long and 6 m in
diameter.
(a) What power could this station deliver, apart from losses, if it is in
continuous operation for 12 out of 24 hours?
(b) What are the . f r i ~ o n losses expressed as a percentage of the theoreti-
cally possible power? Let the resistance value for the entry and exit
losses, each related to the velocity in the tubes, be 25.
(c) Calculate that, if the pump delivery and the turbine efficiency are
75 per cent, this way of energy storage makes the electric power 18
times as expensive : the engineering works as such may be considered
as being fully written off.
Answer : (a) 550 MW; (b) 0-65 per cent
16. Show that, in the case discussed at the beginning of this paragraph, John
was right in letting the suspected man go.
78
Problem 3
(a) An increase in throughput will affect, amongst others, the Re number and
therefore also the friction factor. (See Figure Il.19.)
(b) Here again you should not forget the change in f due to the changing Re
number.
Problem 5
This problem has been included to show you bow widely Figure 11.19 can
be used, provided that you base your Re number on the proper hydraulic
diameter. Don't forget the hydrostatic head pressure!
Problem 10
A sketch of the chimney with an expression for tlie pressures at the bottom
and top will help you to find an expression for the height of the chimney using
the Bernoulli equation. It turns out that the losses due to friction are negligible.
p , -250
Problem 12
With the substitution of the maximum height h in the Bernoulli equation
you will find the vertical velocity component Vy of the nozzle. The force in the
y-direction can be calculated from the mass flow and vr Do you find the same
answer when you calculate the weight of water the poor fellow has to carry?
Problem 16 John and the sewer murder case
The flow conditions in the sewer are turbulent (Re ~ 4 x 10
6
) and we can
assume f to be practically constant. Because the driving force pg ~ h is constant
we can write :
79
and we find :
v
2
= v
1
) Dh
1
/ D
112
For a half-filled sewer channel Dhl = D, but if the channel is filled to one-
quarter of its diameter :
4
(nR
2
8 _ R
2
sin (J)
360 2
Dh
2
= = 059D
nR8
180
if we use cos 8/2 = (R - tR)/R = f.
Since v
1
= 1 m/ s (half-filled channel) we find v
2
= 077 m/s and the time
necessary for the body to arrive at the fisher village is :
t = ~ = 9360 s = 2 h 36 min
vl
(or longer, if the water level was lower than i D ~ Thus the corpse must have
been dropped into the channel at 909 p.m. or earlier, and John came to the
right conclusion.
ll.J. Flow with negligible energy dissipation
John looked from the corpse to the drum, from which wine
(a poor quality rose, thought John) still spouted through the
two 6 mm holes which two of the bullets had left there. The
this isolated shed with only one other person, who had run
away twenty minutes ago. John thought: the drum of 0-6 m
diameter stands on its fiat bottom, the holes are 1 m from its
top and a wet rim indicates that the wine level has decreased
by 20 em. He remembered what is outlined in the pages to
come and arrested the fat man.
For a great number of flow situations the energy dissipation Err is zero or
negligibly small with resp-ect to the amount of mechanical energy which changes
from one form into the other (see example 3 of the previous paragraph). In those
cases the relation between velocity and driving force is simply given by
Bernoulli's law (in the form of equations 11.42 or II.43). This equation can also
be written in such a way that the terms acquire the dimension of a length :
p ti2
- + h + - = constant along a streamline
pg 2g
(11.47)
respectively ; the sum total of these terms remains constant if there is no energy
80
------y- r----------
-----
- - -
-- -
. -
f=-=-
.
-
- -
1
v2
..
--
p
-
-
- +
-
11 + .1!...
p
-
pg 2g -
-pg --
pg
-
-
--
v
- -
-
-T
-

-
- - -
- - -
-
vzO
-
hI
h
1

h=O
change due to friction or mechanical work. These heads also have physical
significance, as can be seen from Figure II.26. _.
With the left manometer tube only the pressure prevailing in the liquid
measured. The lower opening of the right-hand. tube is opposing the stream.
Right in front of this opening the velocity is negligible so that all kinetic energy
is converted into pressure and causes a thrust equal to !pv
2
This thrust is used
when measuring velocities with a Pitot tube. .
We will treat here two well-known examples of flow with negligible
dissipation, the flow from orifices and the flow over a sharp-edged weir.
ll.3.1. Flow of a liquid from an orifice
Friction hardly plays a role in this type of problem. For calculation of the
outflow velocity use is made of the mass balance and the Bernoulli equation.
Using this mass balance it appears that some distance before the orifice the
velocity is negligible . compared with the velocity in the orifice. Next, with
equation (11.41) it can be derived that for a sharp-edged orifice (Figure II.27)
the volume flow v becomes :
Po
( 0 ) ( b ) ( c )
Figure 11.27 Flow of liquids through holes
<Pv = A'lj2(pl - Po)/p
A'
1
is the surface area of the smallest cross-section of the jet where the pressure
should be p
0
, because no radial accelerations occur. The calculations can
better be based on the actual surface area of the orifice A
1
:
cf>v = CA1j2(pt - Po)/ p
81
C is the flow coefficient and is the product of a coefficient of contraction
Cc = A'
1
/ A
1
and a friction factor C
1
. For the sharp-edged orifice, C
1
is prac-
tically 1, whereas Cc is ca. 062 at sufficiently high velocity.
For a rounded orifice with smallest cross-section A
1
Figure 11.27b applies:
(II.48b)
with c, = 095-099.
If a 'diffusor' (angle not greater than 8 to avoid eddying) is attached to the
rounded orifice then :
(11.49)
where A
0
is the cross-sectional area at the end of the diffusor. (Figure Il.27c).
With the diffusor more liquid flows through the smallest cross-section than
without the diffusor, because in the former case the pressure in the constriction
is lower than p
0
. The diffusor cannot be lengthened infinitely because then the
friction losses become too great On the other hand, if the pressure in the throat
falls below the vapour pressure of the liquid, vapour generation (cavitation)
occurs; the given calculations are then no longer valid.
The following table shows the flow coefficients C for various orifices. These
coefficients must be used with equation (II.48a).
Table ll.2 Flow coefficients for flow
through orifices
c = 0 96
d
c = 0 82
. -
;Iff d/ 2
d-- c ; 0 53
'1
82
11.3.2. Flow of gases through orifices
The approach followed in the last section can also be applied to the flow
of compressible gases through orifices. Friction losses can be neglected, and we
can assume that the change of. state takes place adiabatically. However, due to
the adiabatic expansion, the pressure in the jet will be different from the initial
pressure p
1
and the final pressure p
0
outside the gas jet (see Figure Il.28).
Figure 0.28 Flow of gas through an orifice
In the stationary state the Bernoulli equation (11.41) reads in its differential
form (g dh = 0, </>.,. = 0, Err = 0) for this situation :
0 =- f Hdp + vdv}
whereas the energy balance treated in chapter I (equation 1.6) :
0 = - f {ctu + c t ( ~ ) + vdv + gdh}</>m + q,_.- t/>n
is simplified to (H = 0):
(II. 50)
0 = - f {ctu + c t ( ~ ) + vdv} = - J.' {du + p d ( ~ ) + ~ d p + vdv} (11.51)
From equations (II.50) and (11.51) we thus obtain:
0 = dU + p d ( ~ )
or, realizing that (if there is no change of state) dU = c" dT :
0 = c. dT + p d ( ~ )
(II.52}
Applying the ideal gas law for a number of moles nina volume V, p V = nRT
(R =gas law constant = 831 J/ mol K), we find with M = molecular weight
of the gas (kg/ mol) :
p = nM and p d ( ~ ) = _L d V = R T d V
V . p nM MV
83
R = M(cp- c") and K = cJcr: we now obtain from equation (II.52):
dT dV
0 = T + (K - l)y- (II.53)
tion of this equation finally yields TVIC-
1
= constant, which can be
ormed by means of the ideal gas law to:
p VIC = constant and pp -K = constant
last equations describe the adiabatic compression and expansion of ideal
Returning to the Bernoulli equation (II.50) for the flow of gases through
1
0 = -dp + v dv
p
can now replace p by p = constant p
11
1( and by integration between p
1
, v = 0
-.d p, v we find :
__ 1_ l { p - 1/K+ 1 _ p - 1/K+ 1} + 1v2 =
0
= !v2 + K (!!_ _ P1)
constant - 1
1 2
K - 1 p p
1
-+1
K
V - -l--
2
- 2K p
1
{ ( p ) (K- 1}/IC}
K - 1 Pt Pt
(II.54)
the mass flow rate at any place of the variable cross-section A of the gas jet
.,...,p,p_ the pressure is p is given by:
(
p ) 1/IC
m = pvA = Pt p
1
vA
= A { 2 ~ : ,P.P[(:r _ (:.r >'Jr (11.55)
- \. -
The change of the cross-section of the gas stream with pressure is shown
3Chematically in Figure 11.29, where:
been plotted against pj p
1
_
It can be seen that the cross-section of the jet reaches a minimum at
_!!_ = ( 2 ) K/ (tc- 1)
Pt K + 1
84
p/pl ---
Figure 11.29 Flow of gas through an orifice
At this point the critical conditions Ac, Po vc and Pc are present. For the critical
velocity vc we find with equation (II. 55):
Vc = ( 2K P1)t = (KPc)t = (KRT)t
K + 1 Pt Pc M
(11.56)
which is the velocity of sound at Pc and Pc (for air at 0C, vc = 331 mjs). For gases
consisting of two atoms, K = 14 and consequently pcfp
1
= 0527.
The expression developed for the mass flow rate (equation JI.55) can be sim-
plified for a number of conditions.
(a) p
0
~ p
1
Under these conditions the compressibility of gases can be
neglected and equation (11.55) simplified to:
(II. 57)
where A = cross-section of jet at p = Po, A
1
= cros5-section of orifice and
C = contraction factor which is approximately 06 < C < 07.
(b) Pc < Po < p
1
(for air, Pt = 053pt). Equation (11.55) is valid, A being given
by CA
1
with 06 < C < 07.
(c) p
0
~ Pc ( =053 p
1
for air). When Po is so low that at the cross-section
Ac the speed of sound is reached, a further decrease of p
0
will not result in a lower
pressure at Ac. The reason for this is that pressure waves travel with the velocity
of sound ; hence as soon as the pressure Pc is reached at Ac the pressure difference
p
1
- Pc determines the flow rate completely independently of the value of Po
85
Equation (11.55) becomes:
"'m = p,v.A, = A, {Kp,p,(K! J .. ,y,.-,,r
= C"A,JP:/);
with:
CK = HK ! t)"+l )/(x-1)} t
(II. 58)
and Ac = CcA
1
, where Cc 06 for sharp-edged orifices and Cc 097 for
;ounded orifices. The factor C K can be calculated for each gas from the known
(air: C" = 0684; steam : C K = 0668).
For flow of gases in pipes or pipe systems the same considerations as those
: ast discussed apply. If p
0
p
1
the compressibility can be neglected and the
:elations discussed in paragraph II.2.2 can be applied. If Po < p
1
the calculation
becomes more involved and Figure II.30 can be used for a quick estimation of
pressure drop during turbulent flow. This graph is based on air at l7C. The
irawn lines show the calculation of the pressure drop for a mass flow of 01 kg/ s
air at l7C and 10 atm pressure through a pipe of25 x 10-
3
m diameter and a
total friction loss factor of I Kw = 4. We find 6p = 0.06 atm (for straight
pipes Kw = fSLfA ; for circular pipes at Re > 10
5
, Kw 002 L/ D). We can
further read from the graph that under these conditions v 30 mf s and
o 11 kgjm
3
.
For gases other than air or for other temperatures, the small inserted graph
gives two correction factors, CM and Cr. For these conditions, instead of the
gas pressure p
0
a corrected pressure p
0
CMCr has to be used and instead of
!he flow velocity v the product vCMCr is found, as well as pf CMCr instead of
the specific gravity p. The same applies to the pressure drop determined, which
IS then So, if we had used C0
2
(M = 44) at 100C, CMCr = 13 x 09
= 117, the example drawn in the graph is valid for an initial pressure ofll7 atm.
The flow velocity is then 256 mjs, the specific gravity is 129 kg/m
3
and
the pressure drop would, be Ap = 0052 atm.
I 1.3.3. Flow through
The total liquid in this case, be calculated approximately as a func-
tion of the height h of the liquid surface above the edge (see Figure II.31), with
Bernoulli 's law assuming that mainly potential energy is converted into kinetic
energy.
On the upstream side of the weir at a height h the static pressure is

+ pg(h
0
- h) (p
0
= ambient pressure), whereas the velocity may be con-
sidered negligibly low. The pressure in the overflowing jet is Po in all places,
whereas the velocity of the streamline coming from height h is called vh. Ap-
plication of the Bernoulli equation yields for the velocity distribution :
vh = j2g(h
0
- h)
(II. 59)
86
lnitiaJ tem"'"atwre roc
CorrectiOn factors for other 90ses than
air or ot her temperatures than I 7C
Flow velocity v ( m I s )( cr. eN . v )
.....c........L.......c...- ""--......c...--'......c...-""--......L.--'-'--L-.L.---"-- Density p
0
( kg I m
3
) ( p I C T .c N )
Figure 11.30 Pressure drop for
Reference
line
..
. .
Mass flow rote
(kg/s)
:soo Friction loss
zoo foetor}:; Kw
I
2
.. :bulent flow of gases in circular pipes
Diameter (m)
400 x 10-3
300xlo-3
250x lo-
3
200 xl0-3
150xl0-3
IOOxi0-3
80 X 10- 3
sol( ro-
3
10 X 10-3
8 X 10 - 3
s x ro-3
4xlo-
3
3x lo-3
25)( I0-
3
87
88
Figure ll.31 Flow through a narrow-crested weir
So at h = 0 the flow velocity is maximum and v
11
= 0 at h = h
0
Hence the
volumetric flow rate in the streamline between h and h + dh is :
The total flow rate then becomes theoretically : .
In reality the result must be multiplied by a flow coefficient C = owing
to contraction and some friction loss. Then for L h
0
and h
0
> 001 m the
practical formula becomes :
4>v =
(11.60)
A weir can be used for measuring liquid streams with a free surface area
Besides the rectangular weir there are differently shaped weirs. Figure II.32
gives a survey of the relations applicable to the various weirs.
Occasionally a viscous liquid is passed over a narrow-crested weir. In this
case, the flow rate is given by:
4>v = if 4>v < 02 (11.61)
v Lv
according to Slocum (Can. J. Chern. Eng. , 42, 196, 1964).
I !.3.4. Problems
1. The trough of the rotating cell filter from problem 8 in paragraph II.2.4
has a weir edge which is 285 m broad. The level of the cellulose slurry is,
as a rule, 25 em above the edge. Calculate the amount of slurry flowing
over the edge.
3
jh
2. A free-falling liquid jet contracts owing to acceleration in the field of
gravitation. For a given jet a 20 per cent contraction is attained at a distance
of 8 em from the outlet opening. Calculate at what distance 20 per cent
contraction is found if the outftowing liquid fl ows 15 times as fast.
Rectangul ar wei r ( L > 2h
0
)
Narrow rect angular notch ( ho > L )
Broad- crested wei r ( 0 > 3h
0
)
Tri angular notch
<Pr 0 59 (L-0 2 ho ) J 9/15
( < v > 0 6 m /s; ho > 0 1 m)
4> u : 0 465 (L-02 h
0
) J
( L : width of weir )
0 44 ;-;;g-
tong <P
Figuri ll.32 Flow of water through weirs
89
3. In the bottom of a cylindrical vessel (area of horizontal cross-section
F = 05 m
2
, height h
0
= 3m) there is a sharp-edged round orifice (surface
area F
2
= 5 x 10-
4
m
2
). Calculate the timet necessary for the water-filled
vessel to drain through orifice F
2
The flow C = 0632 is assumed
to be independent of the water height.
4. In the house of Professor T there is a rectangular bath tub. If the tap is
fully opened and the plug is put in the drain the bath-tub is filled in 20
90
minutes. The water height is then 50 em. When the tap is turned off the
bath filled up to this height empties in 32 min through the dr.ain in the bot-
tom. The flow resistance in the hole is then determinative of the velocity.
One day Professor T wants to take a bath and turns the tap wide open,
but in his absent-mindedness he forgets to put the plug in the drain. Cal-
culate the eventual height of the water in the bath-tub.
*5. A gas cylinder has a volume of 40 I and an initial pressure of 200 atm. If we
open the main valve (2 mm diameter), how long would we have to wait
at 20C before the pressure is reduced to 10 atm, if the cylinder is filled
with :
(a) air?
(b) hydrogen?
(b) 82 s
6. A spray column is applied for the extraction of free fatty acids from edible
oils by a solvent, which forms the continuous phase. The oil is supplied under
a cylindrical cap (wall thickness 2 mm, height 10 em), the lower circum-
ference of which is equipped with forty V -notches (equilateral, height
15 mm). Furthermore, the top of the cap has thirty holes (sharp-edged)
of 4 mm diameter. What oil flow rate can be handled by this distribution
device if the oil height in the notches is allowed to be 10 mm? Which per-
centage of oil flows through the holes?
p oil = 800 kg/ m
3
p solvent = 1000 kgjm
3
3
jh; 40 2 per cent through holes
*7. Show that in the case presented at the beginning of this chapter, John took
the right action when he arrested the fat man.
Problem 3
For the volumetric flow rate from the drum, with equation (II.48a) :
"' = A dH = CA J 2(pl -Po)
'+'v 1 dt 2 p
where A
1
= cross-section of drum, A
2
= cross-section of hole, H = height
of liquid above orifice. Further, the liquid pressure is given by:
P1 =Po+ pgH
Substitution and integration between t = 0, H = H
0
, t = te, H = 0 yields the
effluent time.
t-
'1
o.,
1-
1le
er
n-
bt
:d)
:>n
!f-
ok
91
Problem 5
For air, Pc = 053 P
1
and the pressure outside the cylinder p
0
is 1 atm, whereas
lhe lowest pressure in the cylinder is Pe = 10 atm. Thus Pc > Po , i.e. critical
conditions are present and we can use equation (II.58):
A. - d(V Pl) - Vdpi - C r:::-;:-
'Pm - dt - dt- KAcv P1P1
Expressing p
1
in p
1
with the help of the ideal gas law and integrating between
l = Q, p
1
= 200 x 10
5
N/ m
2
and t = te, Pe = 10 x lOs N/ m
2
, we can find
the efftuent time te.
Problem 7 John and the barrel of wine
The flow rate of the wine (irrespective of its quality!) from the barrel at any
height h is given by:
1t 2 dh
> = 2CA = --D -
IJ 4 dt
here C = flow coefficient ( =082 according to Table II.2), A = (n/4) d
2
=
cross-section of one hole, D = diameter of barrel and h = height of wine above
the holes. Note that we have assumed the barrel to be cylindrical.
Integrating between t = 0 and t = t and h = h
0
and h = h, we find t =: 61-
minutes; in other words, the character running away 20 minutes before John
entered could never have fired the fatal shots. For the time being, the behaviour
of the fat man is suspicious.
manufacturer who claimed that the tax inspector in his
factory used a calibration curve based on experiments with
water for the alcohol rotameters. In this way, the manu-
facturer said, he had paid taxes for a production which was
11 per cent higher than his actual production. John asked
this man, who presented to be an experienced chemical
engineer, whether the calibration curve for water of this
;. .- rotameter was dependent on the temperature of the water.
The man answered that chis was nor so and that che tax
inspector used the right float . Hence, said John, you are only
partially right and you can claim no more than to be over-
taxed by 2 per cent. If you are not convinced, I advise you
To measure the flow or the flow rate (volume flow rate l/>,_ or mass flow
e m) of a fluid one can distinguish between:
Displacement meters which divide the gas or liquid stream into known
volume units and give the number of these units which have flown through
during a certain time interval (e.g. wet and dry gas meters, various types of
petrol gauge).
92
B. Flow meters with which the mean velocity of the flow can be
in various ways, e.g. :
(a) dynamic: the indication is based on the flow laws (constriction in
pipe-orifice venturi tube, rotameter and also the weir and
Pitot tube (see the previous paragraph)).
(b) kinematic : the velocity is measured with a mechanical system
(c) thermic : the increase in temperature of the How as a result of a
input of electric energy is determinative of the velocity.
We confine ourselves here to treatment of group B(a) and discuss in
the venturi tube, the orifice plate and the rotameter. These apparatus
designed in such a way that the pressure drop as low as possible (Err
I/.4.1. Venturi tube
This tube consists of a gradual constriction (angle 25- 30) and an even
gradual widening (angle maximal 8) ; Figure II.33. With this
we can be sure that the stream follows the boundary lines and eddy
is excluded. p
1
is measured before (or behind!) the constriction and p
2
in
narrowest cross-section (cross-sectional area A
2
) . For the relation
tPm and (p
1
- p
2
) use is made of the mass balance :
,. = p(v
1
) A
1
= p(v
2
) A
2
and of the Bernoulli equation:
Pt + !p(vt )
2
= P2 + !p(v2)
2
from which follows :
J
Figure 1133 The Venturi meter
The factor (1 - Ai/Ar)- t, if necessary with a correction factor Ji.w for
(slight) energy dissipation, is called the flow coefficient a. of the Venturi tu
93
This factor .Uw is almost 1 (for 0-20 < A
2
/ A
1
< 050 within 10 per cent, provided
the Reynolds' number related to the greatest diameter is higher than 2 x 104).
A neat finish of the Venturi tube is a prerequisite, so that it is more expensive
than the orifice plate to be discussed below. The pressure loss owing to energy
dissipation is, however, small (not more than 20 per cent of the pressure dif-
ference measured), so that Venturi tubes are used if extra pressure losses are
inadmissible or too expensive (very large pipes, water supply). .
Il.4.2. Orifice plate
The principle of this flow meter is that the fluid is allowed through a con-
striction in the pipe (see Figure II.34) and the difference in pressure before and
after the constriction, p
1
and p
2
, is measured. Essentially the same happens as
in the Venturi tube, however :
(a) The smallest cross-section does not coincide with the orifice (cross-section
A
0
) because the liquid jet after the onfice still constricts to the cross-section
A2 (=,UAo)
-
-
-
-

Figure I1.34 ! orifice meter with corner taps
fb) A considerable energy dissipation is given because the flow separates from
the wall and eddying occurs on the side of the orifice.
The relation between the mass flow and the pressure difference measured is
expressed by equation (I1.35), in which A
2
has been replaced by ,uA
0
The
factor ,u(l - ,u
2
times the correction factor for the energy dissipation
is called the flow coefficient rJ. for the orifice plate. If necessary, for the transport
of gases allowance can be made with the compressibility of the medium by
mtroducing, in addition to rJ., another correction factor which lies between 09
and 10 for the usual orifice plates. The flow coefficient <X depends on Re
1
=
94
( v)
1
Dtfv, on the ratio A
0
/ A
1
, on the shape of the orifice and the way in which its
edge has been finished and on the places where p
1
and p
2
are measured.
The finish of the orifice has been normalized in many countries. If tabulated
values of a are used it should be noted that the prescribed dimensioning is
accurately maintained and that the orifice plate is kept at a distance of at least
50 D
1
behind and 10 D
1
before an irregularity (bend, valve, change in diameter).
In the case of careful construction the calibration curve can thus be predicted
to within + 1 per cent. Orifice plates are preferably dimensioned in such a way
that the flow coefficient in the working range is constant (high Re numbers).
(See Table 11.3.)
Table 11.3 Flow coefficient o: for orifices with sharp edges and corner taps
Ao/A1
005 01 02 03 04 05 06 07
Re > 10
4
> 2 X 10
4
>6 X 10
4
>8 X 10
4
> 10
5
>105
>2 X 10
5
>4 X 10
5
(X
0598 0603 0617 0635 0660
().693
0740 0805
Kw
1000 240 52 19 85 4 2 l
A disadvantage of using an orifice plate can be that it causes a considerable
pressure loss in the pipe. The kinetic energy of the flow coming out of the orifice
is only partly converted into pressure. The resistance value Kw of sharp-edged
orifice plates is therefore not zero, as can be seen from Table 11.3.
I I .4.3. Rotameter
A rotameter is a tapered vertical tube (usually made of glass) containing a
float (Figure Il.35). The greater the upward fluid flow, the higher the level where
the float remains. The rotameter can be considered as an orifice plate with.
variable free cross-section A
0
(dependent on the position of the float in the
tube) and a constant pressure drop which is determined by the dimensions and
the weight of the f t o ~ t . The latter can be explained as follows: the pressure on
p,<v>,p
I 1
Figure ll.35 Rotameter
Accurate instructions for the design and use of orifice plates can be found in : V Dl-Durchfluss-
messregeln (DIN, 1952) and R. F. Streams and ot hers, Flow Measurements with Orifice Meters,
Kostrand, New York, 1951.
95
the underside of the float is everywhere almost p
1
and on the upper surface
p
2
, so the force on the float which is in equilibrium with the apparent weight of
the float G
1
= Vg(p
1
- p) is (p
1
-

where Af is the greatest cross-section

of the float. Using this in equation (11.35) leads to the pressure drop/flow rate
characteristic of the rotameter :
4>. = oe(A,- Ar)J 2p(pf p)gV (11.63)
If the Re number in the annulus is sufficiently high the coefficient a has a
constant value which is still dependent on the shape of the float. It appears
from equation (II.63) that{Ab - A
1
) for a certain rotameter fluid combination
is proportional to cPm Since an almost linear dependence of (Ab - A
1
) on the .
height is ensured, the calibration graph of a rotameter (h as a function of <f>m)
is practically linear between 20 and 80 per cent of its full scale.
11.4.4. Problems
1. Calculate the approximate dependence of the loss of mechanical energy
per unit of time on the flow rate in :
(a) an orifice plate ; (b) a rotameter.
Answer : (a) Etrcl>m "' ;
(b) Err4>m "' v
2 (a) Can a rotameter, which has been calibrated for water, be used without
renewed calibration for measuring the mass flow of an alcohol stream?
If so, calculate the conversion factor.
(b) Can a rotameter destined for measuring a H
2
S0
4
flow be calibrated
with a glycerin/ water mixture? What difficulty is encountered?
(palcohol = 760 kgjm
3
, Pnoat = 7000 kg/ m
3
)
A water supply contains a device as sketched in Figure II.36 for
measuring the water consumption. In main I an orifice plate Ma with a
I
-
! R
Db
Figure 11.36 Water meter
96
sharp-edged circular orifice has been incorporated Parallel pipe II con-
tains an orifice plate M b with a sharp-edged circular opening and a
rotameter R. The dimensions have been chosen in such a way that the flow
resistance in the parallel pipe is fully determined by the orifice plate. Cal-
culate the maximum capacity of the rotameter needed at a water flow
of at most 500m
3
(h in the main.
Da = 0105 m
(
D')2
ma = D: = 0345
m, = = 0065
3
/ h
*4. Through a pipe (Di = OlOm) flows an amount of water (T = 20C) from
12 x 10-
3
to 24 x 10-
3
m
3
fs. The flow must be accurately measured with
a sharp-edged orifice plat e with corner taps'. The following requirements
should be met :
(i) The flow coefficient of the orifice plate should be constant over the
entire range.
(ii) The pressure difference to be measured should not exceed 68 em Hg.
Determine the limits between which the opening ratio m should be
chosen.
*5. Show that John gave the greedy alcohol manufacturer the right answer to
the problem described at the beginning of this paragraph (palcohol = 790
kg/m
3
) .
Problem 4
Calculate the Re number for volume flow :
Re = pvl DI = P 08 Vt Df - .
'1 08 17D
1
08 17D
1
= 12 X 10- 3 x . 103 = 15 X 105
Re 08 x 10-
3
x 10-
1
(for minimum flow)
(and for maximum fl ow 3 x 10
5
).
From Table 11. 3 it follows that the upper limit of m is 05. In other words, m
has to be 0 5. The lower limit of m is controlled by the maximum pressure that
n-
a
w
J-
w
is allowed :
<Pm
<l>m =
m <Pm 033

The combination of the two conditions to be fulfilled leads to:
033 m 05
Problem 5 John and the alcohol manufacturer
The flow rate through a rotameter is given by:
t/Jm = <X(A,- AJ)!p(pf p)gV
97
In this story (Ab - A
1
) is. fixed. If we assume for a moment that p
1
p we find:
cPmalcohol cPmwater
-
J JP::::r
c/>malcobol = 089 <f>mwat e r
Here we see the 11 per cent claimed. by the manufacturer!
However, the manufacturer admits that the reading of the flow meter is not
affected by temperature variations. This information can be used to find a
relation between p
1
and p:
d{p(pf - pJ} I = 0
, dp p=pw
He also informs us th.at the float was selected in such a way that the ex value
was independent of the Re number. Now we have:
c/>m alcohol _ <f>mw
137 X 10
3
- 14 X 10
3
c/>malcohol = 098</>mw
and again. John is right.
------------ - - --------
98
n.s. Flow arOlmd obstacles
11.5.1. General approach
Inspector John relaxed for a moment and read a report of an
inventor about a gun which fired almost spherical tear gas
grenades at a muzzle velocity of 45 mfs. The report claimed
that the terminal freefall .velocity of these grenades was
25 mjs and wrote the following text in the margin, before he
turned co work again: ' I am afraid that when fired at a
distance of 200 m from a crowd, they will drop dead just in
between the crowd and the police force.' It took his superior
one night of studying the following text, before he decided
that again John was right.
Flow around obstacles results in a force F of the flowing medium (with
relative velocity v,) on the obstacle. Generally there are two different mecha-
nisms, which together account for the total force, viz. the friction resistance (a)
and the geometric resistance (b).
(a} In the extreme case of flow along a fiat plate (Figure II.37), the force is
only caused by the friction resistance. This figure shows that the disturbance of
v,
Figure 11.37 Flow along a fiat plate
the relative flow velocity vr increases in the y-direction if we move from x = 0
to higher values of x. The velocity distribution in this area can be calculated
with the aid of the boundary layer theory. This theory shows that the thickness
of the hydrodynamic boundary layer is given by:
/Jh = 3R: x
li
r
(II.64)
This result is somewhat different from what we found in paragraph II.1.6, when
we discussed the penetration of momentum for a stagnant liquid and a moving
plate (see page 50). There we had found that:
o =fiW = Rvx
p v
r
99
Apparently, at a given br This is due to the fact that in the case
of the moving liquid there is also a velocity component in the y-direction which
increases the thickness of the boundary layer. This velocity component is zero
in the case of the moving plate.
The presence of a velocity component vy in the flowing liquid is illustrated
by Figure 11.38, where the mean velocities averaged over a constant boundary
v,
x=O
Figure 11.38 Mean velocity of flow along a fiat plate
layer thickness are shown at two values of x. It is evident that (v
1
) > (v
2
)
and therefore some liquid must move upwards at a velocity vr The shear stress
at the wall is now given by :
v,. 2 /v 1 2
!yxly=O = 1Jbh = 3.Y;; 2PV,
(1 1. 65)
and the total force on the plate (both sides, length L, width B) is given by:
1
L ( ,L) -t
F = 2B Jo Tw dx = 266 v v BLtpv;
(II.66)
The dimensionless combination (v,L/v) is here a Reynolds' number related to
the relative velocity vr and the length dimension in the direction of flow.
(b) If a fiat plate with surface area A is placed perpendicular to the flow
{Figure Il.39) an eJqtrem-e case of geometric resistance is attained. Before the
plate a pressure builds up of the order of magnitude of the thrust (!pv;). At the
Figure 11.39 Eddy fonnation behind an obstacle
in flow
100
edge of the plate this thrust has been converted into extra kinetic energy which,
however, owing to eddying in the dead water' behind the plate, is only partly
converted into pressure again. As a result, the order of magnitude of the force
on the plate is given by:
F = A-!pv;
In general, the force on an obstacle in the flow is described with a drag
coefficient Cw which is defined according to:
(II.67)
Table 11.4 Drag coefficients of various objects (10
3
< Re < 10
5
)
\tP

r
Cw .. 14 Cw.,. 0 4
-

-
fJ
-
-
-
-
-
-
Ver tical plote L>D c .... 2 Sphere ell( .. 0 43
-

-
8
-
-
-
-
-
Cylinder L>D Cw 1 2 Cylinder L D Cw 0 62
-

-
c
0
f
I

-
- I
o-:J
-
-
-
Flat circular pl ate
c ... - 1 2
Cor
c ... - 0 4

-
-
- -
- -
0 0
Steam enoine c ...... 1 Bus Cw 06
101
where A.t is the largest cross-sectional area of the obstacle perpendicular to v,.
( = surface area of the projection perpendicular to the flow) and C w is dependent
on Re = vxDfv, with D characteristic dimension of the obst acle. Hence
equation (11.67) has been adapted to the mechanism for the geometric resistance
(b).
At high values of Re this effect predominates the friction resistance so that
Cw is constant. In Table II.4 a survey is given of the C
141
values of a number of
obstacles in the turbulent range. In the extreme case that the total momentum
flux is used for building up the force on the obstacle, this force is given by
F = 4>vPV
Comparing this with equation (1!.67) we see that always Cw ~ 2.
I 1.5.2. Spherical particles
In engineering we often come across situations where liquid or solid particles
are dispersed in a fluid phase. The flow resistance determines the velocity at
which they move under the influence of an external force (gravity, centrifugal
accelerating force, magnetic force or electric force) with respect to the continuous
phase.
Very viscous
f l ow, Re < I
Mainly friction resistance
Release of boundary Ioyer
Turbulent flow , 103 < Re < 105
Mainly geometric resistance
Figure 11.40 Flow pattern around a sphere
The flow field round one single sphere depends on the Reynolds' number,
which is here defined as Re = v,Djv, D being the external diameter (see Figure
Il.40). At highly viscous flow (Re < 1) no inertia effects at all occur and the
streamlines converge again behind the sphere. In this particular case Stokes'
law (which can be derives theoretically) applies:
F = 3nrrDv,
or, with the definition of C.., in (II.67):
c = 24
11
= 24
w pv,D Re
(Re < I)
At higher values ofRe both the friction resistance (a) and the geometric resistance
b) play a role. The influence of the former becomes relatively smaller with
mcreasing Re number.
The geometric resistance is largely determined by the place where the
boundary layer is released from the surface. For 10
3
< Re < 10
5
this release
102
takes place a short distance behind the centre of the sphere where Cw is constant
and approximately 043 (Newton's law of resistance). At Re > 2 x 10
5
the
friction boundary layer becomes as a result of which the point where
the boundary layer is released shifts to the back so that C..., decreases fairly
rapidly. Because of the two mechanisms stated above the drag coefficient Cw is
B lu'tf\.'"rMa eX CJPIR;Ysi're ro me rhcn'on foss ractor trow in
plpeS.
The stationary rate of fall v s of a sphere in a stagnant medium is calculated by
assuming its apparent weight to be equal to the resistance force :
7t 3 1td21 2
(;dP(pp- p)g = Cw4 p"fPVs
The solution for v
5
is found by solving this equation -together with the relation
for C..., as a function of Re. This can only take place analytically in the Stokes
range (Re < 1) : '
2
(pp- p)gdp
vs = 1817
or, in the Newton range (10
3
< Re < 10
5
) :
v = 116j(pP - p)gdP
s p
(Re < 1) (IL68)
(II.69)
In the intermediate range (1 < Re < 10
3
) which is often encountered, only a
numerical solution is possible. The force balance is then written as follows:
Cw Re
2
=

p)g (II.70)
The right-hand term of this equation is a known constant for a given problem.
With the help of Figure II.41 the relevant Renumber can now be found, from
which the stationary rate of fall can.be calculated according to :
17 Re
v =--
s pdp
The relations (11.68), (II.69) and (II.70) are also valid for gas bubbles rising in
a liquid if they have a rigid surface. Because of the upward direction of the force,
instead of (p P - p) of course (p - Pb) has to be used. Gas bubbles with diameters
< 08 mm are spherical and the drag coefficients for spheres can be applied.
Bubbles with diameters 08 < db < 15 mm form oblate spheroids, the drag
coefficient of which is approximately two-thirds of the C.., value of a sphere
with the same volume.
Gas bubbles in liquids have a rigid surface as long as:
d, < (11.71)
N
"'
0:
it
u
I
24
-c =-
... Re
/
/
1
/;
/
I
I
I
//
t/

I 2 4 6 8 tO 2 4 6 8 100 2 3 4 6 8 103
--Re
Figure 11.41 Drag coefficient and related functions for spherical particles
103
If surface active agents or dirt are present in the liquid. the mobility of the
surface is decreased and equation (II.71) gives too low values of the critical
value of db. Thus, air bubbles in water are rigid if db < 26 mm. For bubbles
with a mobile surface the flow resistance is lower. The stationar y rate of rise is
then given by :
(11.72)
Figure II.42 il!ustrates the rising velocity of air bubbles of various diameters
in water. According to equation (II.72) bubbles with diameters of db = 2j a/g
104
5r--------------------------------------,
I
Equation ( U . 70)
Rigid Mobile
- - d
11
(mrn }
Figure II.42 Velocity of rise of air bubbles in water
have the minimum rate of rise. For very large bubbles (for air in water if d >
15 mm) equation (II. 72) is simplified to: b
(pgdl >>a)
(Il.73)
The rising gas bubbles then have the form of spherical caps. Comparison of
equations (If.73) and (li.70) written for half a sphere shows that for these con-
ditions C"' = ~ a t aRe value of > 12,000 (air in water). This C..., value is compar-
able with that of a circular fiat plate {Ta ble II.4).
11.5.3. Free fall of droplets
The fr iction forces acting on a falling liquid droplet will deform the droplet
from a purely spherical to an elonga ted form until the droplet breaks up. The
surface tension counteracts this deformation and we can assume roughly that
the droplets will no longer break up if this ratio is equal to or smaller than unity:
1t 21 2
friction force C.., 4 dP
1
p
8
v C d _
2
------ - ---..,...- - = w p{JgV ~ 1
105
In the dimensionless group we recognize the Weber number, the critical va)ue
of which should then be of the order of magnitude of:
W -
pgdpmuvz 8
e cr- 19
(J cw
(II.74)
(assuming turbulent conditions). Hence free-falling droplets can only be stable
for diameters smaller than or equal to dpmax .
Experimental results have shown that We 22 for free fall and We 13 if
the droplets are sudden] y exposed to drag forces.
11.5.4. Particles in non-stationary flow
All relations described in this paragraph so far apply only if the stationary
state bas been att ained. If we consider a free-falling sphere, the non-stationary
dv
M dt = M g - CwAtpv
2
(II. 75)
if we assume Cw to be constant over the whole range of velocities. In the station-
ary state (velocity v.J :
M g = CwA!pv;
(11.76)
is valid, as we have discussed in the previous section. Thus equation (II. 75) can
be simplified to :
dv v
2
M- . = M g - fttf g-
dt v
2
s
Integration between v = 0 at t = 0 and v.
1
t yields:
v
1 +-
' V
5
= exp (2gt)
1- Vs
vs
(II. 77)
In Figure II.43 vjv
5
has been plotted against the dimensionless factor 2gtfvs.
It appears that the velocity is stationary if2gtjv
5
> 5. For 2gtfvs < 1 the velocity
can be represented with good approximation by:
1; gt
- = -
i.e. v = gt ; the relation applies to free fall in vacuum (no friction losses). This
means that up to a timet < v
5
/ 2g friction loss may be neglected, which justifies
our of a constant drag coefficient in equation (JI.76).
Velocities of free fall are often measured by measuring the time for free fall
over a distance L. In that way, the m1ean velocity ( v) = L/ t is determined. The
106
10
o-s
0 6
..
"" ......
!>
0-4
02
0
0
I
v I
vs=7 v
5
0
10
2gl
v
12 0 14 16
Figure 0 .43 N on-statjonary fall of spherical particles
mean velocity can be calculated as :
which yields with equation (1177) :
< >
{
2 + exp

+ exp (-
2
gt)'}
.V Vs 1 Vs Vs
- = - n -
Vs 2gt 4
18
(II.78)
This function is also shown in Figure II.43. lt can be used to calculate (via trial
and error, see problem 4) the stationary velocity of free fall from a measured
mean velocity.
II 5.5. Rate of sedimentation of a swarm of particles
The rate of sedimentation of a swarm of equally large particles in a liquid is
lower than that of one separate particle. In a given system of particles and liquid
this rate of sedimentation (v.Js appears to depend only on the fraction 4> of the
volume which is taken up by the particles. This constant is also connected with
the porosity e: this is the volume fraction of the continuous phase so that
e = 1 - qJ. There are various theories about the relation between (vs)s and o
or e. A well-usable empirical relation is that of Richardson and Zaki :
(vJs =

In this equation is the rate of settling of one single particle of the system
question (e = 1) and n a value. dependent on = and on the rei
of the particle diameter and the diameter of the vessel D, in which they
107
For dJD
1
< 01 the authors give the following values of n :
<01 1 10 100 >500
n 465
435 353 280 239
Equation (II. 79) has the advantage that the exponent n is given as a function of
a Re number defined for the free settling velocity of a single particle and not,
as .happens in most liteliture, as a function of a Re number based ori the un
known swarm velocity.
11.5.6. Cylinders perpendicular to the direction of flow
For an infinite cylinder perpendicular to the direction of flow the picture of
Cw = f(Re) is analogous to that for spheres. At Re < 10-
1
, Cw"' 1/Re and for
Re > 10
3
, C.w :::=:: 1 2. If the cylinder is shortened the flow can also pass the ends
and Cw becomes lower (Cw = 062 for L/D = 1 at Re > 10
3
). If the cylinder is
not round but a streamlined crosssection, Cw decreases considerably
to values between 001 and 01. If, as in another extreme case, it is flattened
(perpendicular to the direction of flow), then C.., 2 for LID = co.
In practice, banks of tubes are often used as heating-or cooling--elements
in a liquid or gas flow. In flow perpendicular to the tubes two different ways of
arrangement can be distinguished :
(a) in line and (b) staggered (see Figure II.44).
\\\\\\\\\\\\
- -ooool
0000
v
0
- 0 0 0 0 F
0 0 0 0 .
-oooo
0. \\'\\m\.\%'
{ 0 )

0 0
v
0
- 0 0 0 F
0 0
-0 0 0
. \\\\'V\\\\\\
( b )
Figure U.44 Flow perpendicular to a bank of pipes :
(a) in line, (b) staggered
In case (a) most pipes are in the 'wake' of thei( .predecessor, in case (b) they
are not. Owing to this shielding the resistance of arrangement (a) is lower than
that of (b) and so is the heat transfer. For case (b), as a rule of thumb, we can use
the fact that the resistance coefficient of a tube (related to the superficial velocity)
in the range 30 < Re < 2000 is about 15 times as high as for a single t ube in
the flow if the distance between two neighbouring tubes equals the tube dia-
meter.
Naturally it also occurs quite often that, for example, in the case of heat
the resistance for flow aiong a bank of tubes must be calculated.
For turbulent flow use is then made of Fanning equati on (II.35) and the
hydraulic diameter.
108
ll.5.7. Problems
1. What is the maximum diameter of a rain droplet at 20C?
Answer : d P = 66 mm
*2. When driving his car { 16 m wide, 1 3 m high, 9 50 kg) at a steady 105 kmjh
on a highway on a windless day, Klaus notices that, if he pushes the clutch
down, it takes 5 seconds for the car to decelerate to 95 kmfh.
(a) Estimate the maximum value of the drag coefficient.
(b) Driving steadily, which percentage of the petrol used (101/ 100 km) is
converted into mechanical energy, if the heat of combustion is 11000
kcalfkg (p = 700 kgfm
3
)?
Answer : (a) Cw < 055
(b) l 55 per cent
3. Two metal spheres of equal weight but of different diameter fall through air.
Calculate the relationship of their stationary rate of fall if the ratio of their
diameter is 3 and the flow round the spheres is dynamically similar. Is this
supposition correct?
4. The ' Euromasr tower in Rotterdam bas a height of 104m. It takes a cherry
(diameter 17 em) 65 seconds to fall from the top of the tower to the pleasant
park below.
(a) What would be the velocity of stationary free fall?
(b) What is the drag coefficient?
Answer : (a) 20 m/s; (b) Cw = 045
5. According to a newspaper report the air 1n Mexico City (situated at 2200 m
above sea level) contains per volume unit 25 per cent Jess oxygen than the
air at sea level.
(a) Ascertain that this statement is correct by calculating from a balance
of forces the pressure as a function of the height above sea level.
Given : Air is an ideal gas.
Gas constant R = 8310 J/kmol K
Air temperature = 2rc
' Molecular weight' of air = 30 kgfkmol
This slighter air density in Mexico City causes the sprinters to meet
with lower air resistance.
(b) In what time will a sprinter tneoretically be able to cover the 100m in
Mexico City because oftbis lower air resistance, if he does the 100m at
sea level in 10 s?
Assume: (i) that both in Mexico City and at sea level he develops
the same power.
109
(ii) that he runs at constant speed.
(iii) that the time necessary to pick up speed is negligible.
(c) His actual time in Me::dco City is. however. likewise 10 s.
(c) Assumptions (ii) and (iii) are wrong.
6. The gas bubbles formed at the bottom of a glass of beer are released as soon
as their diameter is ca. 1 mm.
(a) Calculate their stationary rate of rise.
(b) How many times more rapidly would an equally large water drop faJJ
in air?
Answer : (a) 01 05 mj s; (b) 35 times
*7. A horizon tal air duct with rectangular cross-section (H x B = 30 x 30 em
2
;
L = lOm;xj Dh = 2 x 10-
3
) containsanairbeatingelement. This element
consists of six rows of fifteen tubes placed transversely to the direction of
flow. The rows are installed in a staggered position (Figure fl.27b). The
external diameter of the tubes is D = 2 em, the mean temperature in the
duct is 80Cand the mean pressure is 13 x 10
5
N/m
2
. Calculate the pressure
drop in this duct at a mean air velocity of 20 m/ s.
2
*8. Show that John indeed gave the right answer to the problem of the gas
grenade treated at the beginning of this paragraph.
Problem 2
Klaus' data collected in a quiet hour on the ElO motorway will help us to
calculate a drag coefficient that includes not only air resistance effects but also
friction losses. The actual C..., value will therefore be smaller than the value we
calculate from the unsteady-state momentum balance:
dv C 1 '
-M- = A-
2
pzr
dt w
Now, from the measured data at f = 100 kmfh:
and we find :
dv = 10 kmfh = 0-555 mj s2
dr 5 min
950 X 0555 = OSS
C"' malt = 208 X 0-60 X 772
110
If we assume the friction energy loss of the rolling wheels on the road to be of
tbe same order of magnitude as the energy loss due to air friction. we find
C\9 ~ 0-30, which compares well with the value stated in Table II.4.
The total energy content of 10 litres of petrol is:
10 X 10-
3
. 700 . 11,000. 4-2 X 10
3
= 327 X 10
5
Joule
and the mechanical energy for moving the car 100 km:
dv
- M - L =50 x 10
5
Joule
dt
Thus the percentage transformed into mechanical energy is -155 per cent.
Problem 7
The pressure drop in the channel is caused by friction along the wall and by
the flow resistance of the pipes, and we can assume that these two resistances
are additive. The pressure drop in the empty channel is then found to be (using
Dh = 03 m, Reh - 3 x 10
5
, 4f = 0024, equation Il.35):
tlp
1
= 200 N/ m
2
The drag coefficient for one single pipe would be (with Re = 2 x 104, Table
II.4) :
Cw = 12
Thus, for staggered tubes C ~ = 15 C..., = 18. Thus the force on one pipe is:
F = CwAtPV
2
= 266 N
The force on all pipes together is then F.ot = 615 F = 239 N and the pressure
drop:
t:.p
2
= 239/009 = 2660 Nj m
2
Problem 8 John and the tear gas grenades
An unsteady-state momentum balance in the x-direction for this case (see
dvx l 2
M-= -C pv A
dt _ w2 x
-----
/ '
/ "'-..
l -:.. .. ---- v/ ~
I
::0... II I
I
I
I
I
'
x.-
( 1)
111
which yields after integration between t = 0, vxo and t, vx:
1 l
(2)
Now vx = dx/ dl and after substituting vx from equation (2) and integration
between t = 0, x = 0 and te, Xe we find :
x = 2M In (
1

e C A + 2M
wP
(3)
Now, for steady-state free fall :
and thus:
2MfCwpA = v;fg (4)
Considering now the velocity component in the y-direction (and neglecting
friction losses because g Cw{pr.lj2M)) we have :
dv
1
dy
dr = - g ; V
11
= dt = - gt and y = v
110
t - !-gt
2
So, for y = 0 :
te = 2vy
0
/g
Introducing this expression and equation (4) into equation (3)' we
xe = v; In ( 1 +
g vs
Now:
2 .
v"'
0
Vy
0
= v
0
sm a cos ex
which is maximum for rx 45, i.e. v.xo = v>'
0
= 071 v
0
, and so the maximum
distance is :
If you, as did the poor inventor, neglect the air friction, you find for the maximum
distance:
2 2 .
Xe = -V
0
SID 0: COS f1.
g
which yields for Xemax = 202m!
112
11.6. Flow through beds of particles
The police doaor rose and stared chac the man had died by
town gas poisoning not longer than 10 hours ago. ' Remark-
able', John thought. The room where they found the body
was locked from Ihe inside and there was no smell of gas at
all, although concentrations above one-thousandth of the
lethal level can be smelt quite easily. The gas valve was open
but due to an explosion at the gas works the supply ceased
8 hours before. A small fan in one of the windows pressed air
into che room so thaz lhe pressure was 80 mm water column
higher than outside the room. The room had a volume of
80 m
3
2
of wall (20 em thick'h the
porosity of the material being approximately 20 per cent and
the particles which constituted most of the material having
a diameter of ca. 0-3 mm. Lighting a cigarette, John con-
cluded that the man had died at another place.
The fiow through a bed of particles is of importance for all branches of
technology; in chemical engineering for filtrationy catalytic processes, separation
processes in columns with packing materials, etc., and also for petroleum
production, soil mechanics and hydraulics. A great deal of theoretical and
experimental work has been done by various groups. Outwardly the various
formulae we encounter often differ, but on second thoughts they have much in
common
Two types of particle bed are of great technical imponance : the fixed bed and
the fluidized bed. ln the former bed the particles rest on each other and are not
moved by the fluid. If in this bed the velocity of the fluid js increased, a situation
is created in which the force on the bed becomes equal to the apparent weight
of the particles. The flow lifts the bed, as a result of which the particles become
suspended. This is called the fluidized state which is characterized by a thorough
mixing, especially of the The result is that temperature- and/ or
concentration differences in the fluidized bed are rapidly eliminated, so that in
the bed uniform conditions prevail which are often desired. If the velocity of
the fluid is increased still further. the particles are ultimately blown out of the
bed and we enter the field of pneumatic transport. We shall now discuss some
essential points of the flow in the two types.
II.6.1. Fixed bed
This bed may be considered as a collection of particles and also a s a collection
of interconnected winding channels. We stick to the former description in order
to calculate the energy dissipation per unit of mass in a bed of spherical particles
as a function of the superficial velocity v
0
related to the cross-section of the tube
which forms the boundary of the bed.
The actual velocity (v) in the bed is of course higher than v
0
, notably v
0
=
e( v) if e again represents the volume fraction voids (the porosity) of the bed
(for beds of spheres, in practice, 035 < s < 045). The train of thought is now
as follows : we calculate the energy which is dissipated around one particle in a
113
ftow of velocity ( v) ; this energy is multiplied by the total number of particles
in the bed in order to obtain the constant ErlPm, ie. the total energy which is
dissipated in the bed (see also equation II.41). To this end we base our calcula-
tions on the definition equation for Cw 01.67) to determine the force which the
flow exerts on one particle :
The drag coefficient in this relation is self-evidently another from that which is
found for one sphere in an infinite fluid. From the force F we can now calculate,
by multiplication by (v), the energy whic4 is dissipated round one sphere. The
total number of spheres in tbe bed is subsequently calcu1ated from the volume
the spheres take up in the bed, (1 - s) AL, where Lis the height of the bed So:
or:
(1 - e)AL 1t H
3
ErlPm = ErrPVoA = Cw
4
dp,;p(v)

6 p
_J. .!.
2
1-c;L
Err - 2Cw2Vo t 3 d
p
At a sufficiently high velocity in the bed (high Renumber) the flow is turbulent
and, as we know, Cw is constant. Ergun found for this range Cw = 23.
At low velocities in the bed the flow is no longer turbulent, but shows eddy
currents. In this region (which is also called laminary') Cw is expected to be
inversely proportional to the Re The difficulty here is to define the
Re number which is characteristic of this flow. The .characteristi.c velocity is,
of course, (v). However, the characteristic cross-sectional dimension oftheflow
between the particles is not so much -determined by the particle diameter as by
the space remaining between the particles. Now it appears that the hydraulic
radius A/ S of the cavities between the particles can be used for correlating Cw
as a function of Re
11
(= 4p( v) Rnf'1S) for all packed beds. This hydraulic radius
R
11
is connected withe the specific surface area of the particles Sp:
A surface o.f cross-section of cavity volume cavity
S = circumference cavity - wall surface cavity
volume cavitiesfm
3
porosity (m
3
jm
3
) s
= wall surface cavitiesj m
3
= specific wall surface = Su
The specific surface area is, in turn, again a function of e and the particle dia-
meter, namely:
_ 6(1 - e)AL nd
2
_1 _ _ 6(1 - e)
S., - nd3 PAL - d
p p
114
so that :
Re& = PVodp
3 '1(1 - e)
Now it appears that for Re,. < 1, Cw = 152/Re,.. This relation is known as the
Carman formula, but is also named after Blake and Kozeny.
According to Ergun a good description of the flow resistance in the entire Re
range is obtained by adding the relations for laminar and turbulent conditions,
with the result :
fj,p Err 1 - e ( v )
-=- = -
3
170-(1- e)+ 1-75
pL L dP e v
0
d
11
(11.80)
It appears that the flow may be considered as tl,lrbulent if Reh > 700. The
constant v
0
dJv is again a Reynolds number and is often used in the case of flow
through fixed beds, although somewhat incorrectly. Restrictions in the use of
equation (11.80) are that the particles must not deviate too much from the
sphere shape and that the diameter of the particles must be considerably
smaller than the diameter of the bed ( <
2
1
0
) .
For a number of forms of nonspberical particles Efr has also been determined
In such a case, instead of d, in equation (1180) tbe diameter a, of a sphere having
the same volume/ diameter ratio as the combined particles in the bed should be
used, and the pressure drop so found should be multiplied by a correction
factor k. In Table 11.5 the correction factors for a number of agricultural and
horticultural products are given.
Table D.S Factor k with which
equation (11.80) underestimates Err for
some agricultural crops (purified
ducts, random' arrangement)
Product
Peas
Rapeseed
beans
Clover seed. wheat
Rye, summer barley
Maize
Sugar beets. carrots
Oats
k
105
12
14
17
27
32
35
38
For asparagus rather divergent k values are found , just as, for example, for
flow through yarn packet s and yarn filters. Here it is important to find out
whether these cylinder-shaped bodies lie mainly in the direction of flow
(k 075) or mainly perpendicular to the fl ow (k :::::: 11). In these cases there is
hardly ever a question of 'random' arrangement.
115
IJ.62. Filtration through a bed ofparticies
Equation (II.80) can also be used for an idealized description of the filtration
process. which is practicall y always carried out under laminar flow conditions.
Rearrangement yields, if Vis the amount of filtrate at timet:
dV s
3
~ p A d
2
vA-A. --- P
0
- '+' II - dt - 170( 1 - s )
2
Yf L
(ll.80a)
During filtration, the suspended particles will collect on the filter medium (e.g.
filter paper, cloth) and form a cake. If we neglect the flow resistance of the filter
medium, the term Lin equation (II.80a) is t he thickness of the filter cake, which
is a function of the weight fraction f of solids in the suspension and the amount
of filtrate V already produced. A material balance shows:
LA(l - e)pp = fp,V (11.81)
Substituting L into equation (II.80a) and integrating between V = 0 at t = 0
and V, t , we find for the amount of filtrate at any time :
d
j
e
3
flppP
v- sr4 --t
- 85(1 - e) rJPd
and for the rate of filtration:
A. d V J e
3
ll.ppP
'+'u = dt = dPA 340(1 - e) Yfp,Jt
(11.82)
(II.83)
The derivation of equations (II.82) and (II.83) was based on the following
assumptions:
(a) constant e, i.e. the filter cake is incompressible;
(b) constant ~ p . i.e. the filtration is carried out at constant pressure ;
(c) spherical particles; and
(d) no How resistance of !tlter medium.
In practice, these assumptions are not always permissible. For a gjven system
the influence of the flow resistance of the filter medium can be allowed for by
adding, in equation (Il.80a), to the resistance of the filter cake which is repre-
sented by:
(1!. 84)
a resistance Rm for the filter medium. We o btain
dV ~ p A
<Pp = dt = 17(R
1
+ Rm)
c
(II.85)
116
Substituting Lin equation (11.84) from equation (11.81) we find:
R
1
=
1
v pr7170(1 - .s) = fv R
c A Pp e3 A c
With this expression equation (11.85) becomes:
- ~ . = dV = .1pA
'f'v dt ( V )
11 f .F.R, + Rm
,-J
Integration with V = 0 at t = 0 yields:
1 ( V)
2
( Y) 6p
- - fR + - R =-l
2 A c A m r(
or :
(11.86)
(11.87)
(II.87a)
Thus a plot of (A/ V)t versus J1A should yield a straight line with slope
-!<ttfllp)fRr:, whereas the extrapolated value of (F/ V)t for V/F = 0 yields
(71/ llp)R"n which enables us to detemrine the resistances of the filter cake and
the filter medium Rc and R,. .
11.6.3. Fluidized bed
Important factors for the fluidized bed are the velocity vc;. at which the fixed
bed changes into the fluidized bed, the maximal velocity at which the fluidized
bed can only just exist (before pneumatic transport occurs) and the expansion
of the bed as a function of the velocity.
The velocity v
01
can be calculated from the foregoing (Err for a fixed bed) if it
is borne in mind that at this velocity the force of the fluid on the particles is just
equal to the apparent weight of the particles:
tipA = g(l - e)(pp - p)AL
Using the Ergun equation (11.80) it follows:
t,;2 .d2 { v }-1 gpd3
0
'
2
P = 170--(l -c)+ 1-75 _ _ P(p - p)s
3
v v -d r ~
2
p
Ot p 'I
(II.88)
The maximal velocity at which the fluidized bed can just exist occurs in a
highly dilute bed (e--+ 1) and is equal to the stationary rate of fan vs of a single
particle.
The expansion of the bed as a function of v
0
is more difficult to predict
because the flow condition in the fluidized bed is in principle unstable. A gas
stream, particularly. tends to move rapidly in large bubbles through the bed
(aggregative flujdization). A bed fluidized with liquid (provided its length is not
117
much greater than its diameter) is, as a rule, fairly uniform (particulate fluid-
ization). In the latter case the bed expansion can be calculated by using the
relation of Richardson and Zaki for the rate of settling of a swarm of particles .
(paragraph 11.5.5). It should be borne in mind that in the fluidized bed the
relative velocity between the fluid and the particles is (v) = v
0
jc.
The relative velocity between the phases in a. settling test (in a vessel with a.
bottom) must be calculated from the velocity of the swarm (vs)s. The settling
particles displace liquid upwards, so that on settling the relative velocity vr is
equal to the velocity of the swarm plus the velocity of the liquid. Since the
volume of particles which settles per unit of time equals the volume of liquid
which is displaced per unit of time, the velocity of the liquid on settling is equal
to (1 - e)(v.s)Je. So, the relative velocity between fluid and particles is:
1 - e (vs)s .
v, = (vs)s + - -(vs)s = -
e e
Hence the expansion of a uniformly fluidized bed is given by Richardson and
Zaki as follows :
/ < V) \ l f(n- 1)
e = \ v; }
or e- -
_ l /n

I 1.6.4. Problems
* 1. Equation (II.80) can, apart from numerical factors, also be derived from
the Fanning equation (II.35) by regarding the bed as a complex of irregular
channels with total length Land hydraulic diameter 4Aj S. Check this and
calculate that for the bed applies: 4fL'/L = 152/Reh for Reh < 1 and
4fL'jL = 23 for Reh > 700.
*2. In a vertical tube (Di = 10 em, L = 25 em) almost spherical anthracite
particles are fluidized with air. The particle size is :
dp(mm)
Mass( %)
from to
100 (}84 20
0-84
(}71
29
071
(}59
23
059
(}50
25
050 042 3
The densit y of the anthracite (measured with a pyknometer with mercury)
is 1970 kgj m
3
, the temperature is 32C and the pressure above the bed is
10
5
N/ m
1
.
(a) At which mass flow rate does the fluidization start?
118
(b) At which mass flow rate are the particles with 0.42 mm diameter blown
out of the bed?
(c) What is therefore the range of mass flow rates suitable for fluidization?
2
s
(b) 35 kgfm
2
s
(c) 1 < c/>';, < 3 kgfm
2
s
3. A suspension is filtered at a constant pressure difference 6p over the filter.
At t = 0 the filtration begins (the cake thickness is then 0) and after a time
t a volume V has been filtered. The following results are fo und :
100
14l
173
t
(s)
20
40
60
!lp
(at)
02
Q-2
Q-2
Predict the filtered volume after 20 s if the test is repeated under the following
conditions :
(a) at a pressure difference of 0-4 at ;
(b) with a second material, the particles of which are of the same form as
the first but with the linear dimensions smaller by a factor of 2 (at
6p = 02at).
3
{b) 05 m
3
4. A mixture of oil and bleaching earth is filtered over a plate and frame press
of 20m
2
filtering surface. After 4 minutes 1m
3
and after 10 minutes 2m
3
of filtrate are collected. How long does it take to filter one charge of
20m
3
?
5. At reguJar intervals the filter press of problem 4 must be cleaned. This
cleaning operation takes t
0
minutes, irrespective of the amount of cake to
be removed. If the flow resistance of the filter cake is neglected (Rm RJ,
how long should the filtration last in order to obtain the maximum filtration
capacity over a complete cycle (filtration and cleaning time)?
0
*6. Show that John reached the right conclusion in the problem stated at the
beginning of this paragraph.
Problem I
With equation (11.35) we have:
..,-e find :
and :
L '
!J.p = 4f-!p( v) z
D,
4A 4edP
Dh = S = 6{1 - e)'
Vo
(v) =-
e
R
_ 2 pv
0
dP
and eh - -
3 '1(1 - c:)
llp = Err = - e)3
pL L Ldpe
3
4
Comparing this with the expression developed in paragraph 1!.6. 1:
1 zl- e
L - 2 w"'iVo d e3
p
and utilizing the information about -C.,.. given there we find :
L ' 152 L'
4/- = -(Reh < 1) and 4f- = 23 (Reh > 700}
L Reh L
Problem 2
l19
Since we are dealing here with a mixture of particles of different size we must
be careful which diameter to use in equation (II.88). The derivation of this and
of the Ergun equation was based on the specific wall surface Sp given as:
S = surface of cavity = 6(1 - e)
" yolume of particles dP
For a mixture of particles of different size we must therefore use :
a = 6 X volume of particles = L nidt = I M i
P surface of particles L nflf L M 1
d;
where M; = mass fraction of particles with diameter d;. Thus we find :
10
ap = 020 029 023 025 003 = 0-
69
mm
092 + 078 + 065 + 055 + 0-46
120
Estimating c ~ 0-4, we now find from equation (11.88) (17 = 17 x 10-
6
Ns/m2,
p = 13 kgfm
3
}:
rf>':..o; = PV
0
; = 040 kgjm
2
s
From equation (II. 70) we find for the steady-state velocity of free fall of the
smallest particles v = 265 m/ s ; thus <f>':na
11
= 35 kgfm
2
s.
Problem 6 John and the body in the locked room
From equation (11.80) we find the flow rate of air leaking through the wall to
be (dp = 800 N/m
2
, laminar flow) :
{j>,. = VoA = 79 X lO-
3
m
3
fs
Assuming the gas in the room to be perfectly mixed the material balance :
de
vdt=- c11
leads, after integration between t = 0, c
0
and t, c to:
c (j)v
ln- = - -t
c
0
V
Thus, after 8 hours, the gas concentration should still be one-seventeenth of
the lethal level and John should be abJe to smell the gas.
H. 7. Stining and mixing
John saw chat the only man on duty was very busy crying to
speed up the circulation of tile people past the crashed car.
He saw a parallel to a mechanical .stirrer and remembered
that increasing the speed by only 25 per cent would increase
che frictional power by a factor cwo. Therefore, he cold him
not to pay too much attention to the crowd's speed, but, now
it Juu:l started moving, to concentrate on keeping it flowing
smoothly.
Mixing and stirring are two conceptions which partly overlap. Mixing is the
uniform distribution to a certain scale of inhomogeneities over a certain space.
Mixing can be attained by stirring. The object of stirring can also be to promote
heat or mass transfers or to make a suspension or emulsion. Three transport
mechanisms generally occur during mixing :
(a) convection, i.e. the material present is moved ;
(b) dispersion, i.e. different parts of the material move at different speeds; and
(c) diffusion.
11.7. 1. Types of stirrer and fl ow patterns
Figure II.45 shows a few types of stirrer which are applied in practice. Types
(a) and (b) will induce radial flow, type (c) axial flow, whereas type (d) will induce
{ 0 )
I
( b)
I
l c )
Propeller
Figure 11.45 Various types of impellers
121
(d)
flow in both directions. Figure II.46 shows the flow patterns obtained for two
types or impellers.
I
I
I
I
------ ------
{I
I
Radial flow type impeller Axial flow type impeller
Figure II.46 Flow patterns of impellers
In the circulation mainstream of radial tlow type impellers fo_UI different
zones can be distinguished :
(ii) the vertical flow along the wall ;
{iii) the horizontal flow back to the stirrer shaft ; and
(iv) the axial flow along the shaft to the stirrer.
Measurements have shown that in the radial liquid flow leaving the stirrer
rfr = constant
T being the distance to the Stirrer Shaft and Vr the radial Velocity Of the liquid
flow at position r , averaged over the time and stirrer height.
The flow pattern of axial flow type impellers is less distinct. The downward
directed liquid flow (these stirrers are usual1y made to pump downwards)
changes direct ion at the bottom and flows upwards along the wall of the vessel.
122
Here again a circulation pattern can clearly be distinguished in the flow pattern.
A propeller stirrer. however. causes not only axial but also rather strong radial
and tangential ftow in the vessel. The flow pattern of a propeller stirrer is
therefore less systematic compared with that of a radially pumping stirrer.
When a draft tube is placed around a propeller stirrer the flow pattern in the
vessel becomes more systematic.
When the vessel has not been fitted with baffles, vortexing occurs at high
velocities so that gas can be sucked into the liquid. Baffles counteract this effect
Sufficient activity of the baffles is attained when the product of number times
width of the baffles divided by the vessel diameter is ~ 0 4 . This situation
(mostly reached by applying 4 baffles of 01 D width) is called ' fully baffled'.
11.7.2. Power consumption
With dimensional analysis it can be derived that for geometrically similar
situations the power number P
0
must be a function of the Reynolds' number
Re related to the stirrer and the Froude number Fr:
P (pnd
2
n
2
d)
P
0
=
3
d
5
= f(Re Fr) = f -;-
pn '7 g
The Froude number only plays a role when vortexing occurs at the liquid
surface. Figure 11.47 shows a typical power curve for a stirrer. The lower power
consumption in the tank without baffles is caused because air is sucked in, as
a result of which the density of the mixture decreases.
The shape of the power curve ofbaffied systems may be physically interpreted
as follows. If we consider one of the impeller blades moving through a liquid.
then the force exerted on an object in a stream of fluid moving at mean velocity
v is given by :
where A is the crosssectional area of the object and Cw the drag coefficient
Since A - d
2
, and v corresponds to the tip speed of the blade (xnd) multiplied
by a slip factor, the force on the blade is proportional to pn
2
d4.. The powe:
needed to overcome this force is equal to force x arm x number of revolutions
per unit time, or -Fnd. Therefore, the power number P
0
= P/ pn
3
tP is pro-
portional to the drag coefficient Cw. The shape of the power curve is similar to
that of the drag coefficient versus the Reynolds number. F or Re > 10
4
the
power number P
0
becomes a constant. In this turbulent region P
0
is only a
function of the geometry of the system. At values of Re < 10, P
0
is proportional
to 1/ Re, thus the power consumption is given by :
P = P
0
pn
3
d
5
(Re > 104)
(Re < 10)
Table 11.6 gives a survey of some approximate values for P
0
and P'o for variolA
impellers.
123
150
Lomtnor--------Tronsition region -------Turbulent
With baffles
2
Wtthout baffles
10
10
2
10
3
--- pnd2./"f/
Figure ll.47 Power consumption curve for turbine stirrer
The energy consumption for suspending particles in a vessel by means of a
stirrer depends on the minimal stirring speed n* at which the particles no longer
settle. For this minimal rpm applies :
{g(p.;dp)dr =
(11.90)
where dis the stirrer diameter,ftbe weight percentage of the particles and C
a constant of the order 1 (for turbines approximately 15) which depends on the
geometry. The power consumption then is:
p = Popn*3ds
where pis the mean specific gravity of the suspension.
Table II.6 Rough estimations of P
0
and K
Impeller
six blades, width/ diameter- l/ 6-1/8
pitch = diameter
pitch = two times diameter
six blades, width/ diameter - 1/8
Re < lO
Po
70
70
70
42
42
70
6
45
3
03}
1.0
1 5
Re >
K,
13
06
13
04-DS
08
.
123
Lomi nar--...o-----Tronsitioo region------Turbul ent
Wi th baffles
W1thout baffles
10 . 10
2
10
4
10
5
--- pnd
2
!')
Figure U.47 Power consumption curve for turbine stirrer
The energy consumption for suspending particles in a vessel by means of a
stirrer depends on the minimal stirring speed n* at which the particles no longer
settle. For this minimal rpm applies:
{g(p:;dp)dr =
(1190)
where dis the stirrer diameter,/ the weight percentage of the particles and C
a constant of the order I (for turbines approximately 15) which depends on the
geometry. The power consumption then is:
p = Popn*3ds
where pis the mean specific gravity of the suspension.
Table II.6 Rough estimations of P
0
and K
Re< 10 Re > 10
4
Impeller
Po Po
six blades. width/ diameter-l/ 5 70 6
six blades, widt b/ diameter- 1}6-l/ S 70 3
pitch = diameter
42
03}
pitch = two times diameter 42
1.()
six blades, width/ diameter -l/ 8
70 15
Re > JQ4
K,
13
06
13
04-05
08
.
124
I 1.7.3. Pumping capacity and mixing time
If we consider only the biggest velocity component of the liquid flowing from
the impeller, the pumping capacity is given by the product of the mean velocity
of the liquid leaving the impeller and the area described by the blade tips. For
the propeller this leads to (d - diameter of impeller) :
whereas for a turbine we find (W = width of blade):
<f>v = iindW
If we assume ii to be proportional to the tip speed of the blades (1tnd) we find
for a given geometry (for turbines W ,_ d) :
(11.91)
In the turbulent region. Kv is a constant for a given impeller geometry. Some
values of KD are listed in Table TI.6.
In many chemical engineering operations, e.g. the blending of gasoline; the
mixing of reactants in a chemical reactor, etc., knowledge of the time necessary
to reach homogeneity may be very important. Usually the mixing time is
defined as the time needed for homogenization to the molecular scale. Since
measurements on this scale are beyond experimental capabilities, an investigator
is only able to measure the terminal mixing time, being the time required to
attain homogeneity on the scale of observation. In the turbulent region, for
many types of impellers the relation (tm = mixing time) :
ntm = constant
holds, for geometrically similar situations.
The mixing time is about four times the circulation time of the vessel contents.
Thus, if no dead spaces are present:
v
tm ~ 4 tPc
is a good estimation of the mixing time.
1!.7.4. Problems
1. A catalyst precipitation process has been developed on a pilot plant scale
(diameter reactor 025 m). It appeared that using a standard flat blade turbine
(six blades, fully baffled tank), a minimum stirring speed of 700 rpm was
required to obtain a satisfactory product (p liquid ~ 1300 kglm
3
) .
A geometrically similar factory scale reactor (06 m diameter) has. been
designed. Calculate which stirring speed should be applied and what the
power consumption would be if we keep constant:
(a) the power input per unit volume;
(b) the pumping capacity per unit volume;
125
(c) the mixing time;
(d) the size of the partic1es that are only just not suspended in the liquid
Answer: (a) 391 rpm, 685 W
(b) 700 rpm, 4000 W
(c) 700 rpm. 4000 W
(d) 391 rpm, 685 w
2. Show that John reached the right conclusion in the problem described at the
beginning of this paragraph.
ll.8. Residence time distribution
The sergeant on dury in front of the supermarket informed
John that the m.an they we,e looking for had entered the
supermarket together with his wife 15 minutes ago. He said
tluu they seemed to be doing their shopping for the next week
and the man looked relaxed. John, knowing from his own
experience that shopping for the week in this supermarket
cook on the average about 30 niinutes, mode a swift calcula-
tion. He decided that he had a probability of 60 per cent to
catch the man if he searched the supermarket directly with
his platoon. He therefore ordered them accordingly, but they
failed to find the man. At night, he read the following
chapter and came to the conclusion that the probability of an
event and the actual course of an evenl are, ash should know,
two different things.
Flow through an apparatus practically always results in a certain residence
time distribution. Fluid elements entering the apparatus simultaneously will
generally leave it at different times. This distribution in residence time can be
described quantitatively with two distribution functions which are closely
related, the F and the E functions.
Although residence time distribution is of great importance to the design of
process equipment, we will restrict our discussion to the principle. A more
comprehensive treatment of non-ideal flow, including its influence on the
design for chemical reactions, is given by Levenspiel (Chemical Reaction
Engineering).
Il.8.1. The F function
I< 0 , C = 0
I ~ 0, C = Co Somphng point
tracer cone. C
Figure D.48 Measurement of the F curve
Consider the continuous flow system shown in Figure Il.48. From a time
t = .o on, we replace the feed stream by a fluid containing a tracer material
(e.g. colouring agent) at a concentration C
0
. If we measure the mean concentra-
126
T irne 1 -
Ftgme U.49 The F curve
tion of the tracer material in the effluent stream we will find that the tracer
concentration (and also < C) / C
0
, its dimensionless concentration) increases with
time, as illustrated in Figure 1149. Now F(t) is defined as the volumetric part
of the e:fHuent stream which had a residence time in the apparatus of less than
the timet. Since we know that all untraced fluid in the effluent stream has been
in the apparatus for a time t, F(t) is represented by the fraction of traced fluid
( C)/C
0
in the effluent. Consequently 1 - F(t) is the fraction of untraced fluid
in the effluent
The F function (and also ( C) ) is averaged over a small volume taken from
the flow (mean cup values). For measuring, part of the flow is collected, mixed
and subsequently analysed. If concentration gradients or velocity profiles are
present over the cross-section (as is the case during laminar flow) we must be
careful to measure the correctly averaged value.
The F function bas a number of properties, the most important of which are:
F(t} = 0 for t = 0
(i.e. no fluid remains infinitely in the apparatus) or, formulated more precisely,
F(t) = 1 for l = oc
(i.e. no fluid remains infinitely in the apparatus) or, formulated more precisely.
mathematically :
lim (1 - F(t)) = 0
(i.e. all fluid originally present in the apparatus will have left i t ultimately).
Funhermore. from a mass balance we find:
foco f/J
0
(Co - ( C) ) dt = VC
0
and using F = C/ C
0
we obtain :
f
::o v
(1 - f) dt = - = 't'
0 cPv
(Il.92t
127
t--
Figure 11.50 Determinat ion of T
The mean residence time thus can be determined from experimentally found
F(t) curves, as illustrated by Figure II. 50. A line at t = t
1
= constant creates
the areas A
1
, A
2
and A
3
such that :
rl = At + A3
Because:
. T = Lro (1 - F) dt = A2 + A3
this line represents the mean residence time r if A
1
equals A
2
As a check on
the measurements, r can be calculated as -r = Vflf>v from measured values of
V and f/J ,.
Often, instead of the time r. a dimensionless time parameter 8 = t/ -r is used
and we will apply this quantity in all following considerations. It will be evident
that for the F(8) function equation (II.92} becomes:
f oCX) (1 - F) d8 = 1
I 1.8.2. The E function
The residence time distribution function of the fluid in the apparatus is given
by the E function, which therefore represents the age distribution of the fluid
leaving the vessel. Figure II.S 1 shows a typical E curve. The fraction of material
in the exit stream with a residence time between (J and (J + d(J is given by E d8,
from which fol1ows that :
f oO() E d8 = 1
The fraction of effiuent material younger than 8
1
) which is also represented by
the F function, is then given by:
(11.93)
127
t---
Figure 11.50 Determination of -r
The mean residence time thus can be determined from experimentally found
EU) curves, as illustrated by Figure II.50. A line at t = t
1
= constant creates
the areas A
1
, A
2
and A
3
such that:
Because :
this line represents the mean residence time r if A
1
equals A
2
As a check on
the measurements, t can be calculated as r = V /4>., from measured values of
;and tPv
Often, instead of the time c, a dimensionless time parameter 8 = t/ -c is used
and we will apply this quantity in all following considerations. It will be evident
that for the F(8) function equation (II.92) becomes:
f
0
00
(1- F)d8 = 1
11.8.2. The E function
The residence time distripution function of the fluid in the apparatus is given
by the E function, which therefore represents the age distribution of the fluid
leaving the vessel. Figure II.51 shows a typical E curve. The fraction of material
in the exit stream with a residence time between e and 8 + d8 is given byE d8,
from which follows that:
Jo:o Ed6 = 1
The fraction of efiluent material younger than 8
1
, which is also represented by
the F function, is then given by:
(11.93)
128
G
Figure 0.51 The E curve
and the fraction older than 8
1
is :
ra) r8,
1 - F(81) = J ~ Ed(} = l - J ~ E dO
8t 0
The E curve can be measured by injecting a tracer material into a flow system
for a very short time and by measuring its concentration as a function of time
at a downstream point. A typical plot of these concentrations versus time is
shown in Figure II .52. The area under the curve :
LCX) ( C) dt ~ L c L\t
t -
Area ; 0 : total
amount of lf'ljected
tracer materlol
Figure D.52 Response curve to a tracer injection
signal
(11.94)
represents the total amount of uacer substance injected, Q. In order to find
E(t), this area must be unity ; thus all concentration readings must be divided
by the total amount of tracer injected:
( C) ( C) ( C)
E(r) = s ~ Cdt = Q ~ L ( C) lit
(11.95}
129
The mean residence time can then be calculated via :
J
co
00
L t( C)Llt
t = rE(t} dt ~ L tE(t} At = ~
0 0 L ( C) At
(11.96)
Since 9 = t/ t and E{8) = ~ E ( t ) we are now able to construct the actual E curve
in dimensionless parameters.
As we will see later, it is often desirable to know the variance of the distribution
curve, which can be calculated according to :
U
2
= Jco (8- 1)
2
E d8 ~ L 8
2
E il8- 1 = L t
2
( C> L\t- 1 (II.97)
o r
2
L <C> L\t
1 1.8.3. Simple applications ofF and E functi ons
l/.8.3.1. Perfect mixer. A perfect mixer is a vessel in which stirring is so
effective that the composition of its contents is identical at all places. Naturally
the effluent from this vessel has the same composition as its contents.
Keeping in mind that Cout = C a mass balance over this vessel reads :
dC
A. C
0
- A. C = V-
o / ~ o/v dt
(Il98)
For the determination of the F curve from timet= 0 on, the tracer concentra-
tion of the feed stream is changed from Ci n = 0 to Cin = C
0
Integration of
equation (II.98) therefore yields, via:
---- --! dt
J
c dC Jr l/J
o Co - C- o V
F = 5:._ = 1 - e -e
Co
(ll.99)
For the determination of the E curve at time c = 0, a tracer is injected into the
vessel for a short time. The amount of tracer added would, when well mixed,
give a concentration C
0
. Integration of equation (13) for this case yields
. (C,n = Ol) :
fc dC = - (' lf>a: dt
Jc
0
C Jo V
E = C = e - 8
Co
(11.100)
As can be seen, the results obtained agree with equation (Il.93), which can be
written as E = dF/d8.
I / .8.3.2. Plug .flow. During plug flow all parts of the fluid move at the same
speed and there is no axial dispersion and no residence time distribution.
130
Therefore, the F curve is described by :
and for the E curve:
F = 0 fort < -r
F = 1 fort > -r
E = 0 for t # T
but E = tracer input signal at t = r !
I 1.8.3.3. Laminar flow in circular tube. Here, the residence time distribution
is caused by the differences in velocity and by diffusion. If we neglect diffusion,
we can write for any streamline with velocity v and residence time t :
<:> =2(1- :,) =i= 9
(11. 101)
Because the maximum velocity at the centre is twice the mean velocity, both
the F and the E functions are zero for:
v .
(v) < 2 and tf 'C < ! respecuvely
For times > r/ 2 we find for the F curve :
F = = 1 _ ( 1 _ = 1 - ;t: = , - Uol' (11.102)
and for the E curve:
dF 4
E = d8 = (28)
3
(II. 103)
The F and E functions derived for the three cases treated here are shown in
Figure 11.53.
112
-e
Plug flow oreo
Lommor flow
oreo = 1

112
Figure ll.S3 F and E curves for three simple cases
131
I I .8.4. Continuous flow models
The analysis applied in the last chapter to three idealized cases of flow is not
applicable to most practical flow situations. It is therefore desirable to describe
actual fiow conditions by models. Two of these model systems, the 'dispersion
model ' and the ' tanks in series model ', are widely applied, and we will analyse
both models in the followi ng paragraphs.
11.8.4.1. Dispersion model. The dispersion model is based on the plug ftow
of a fluid with a certain amount of intermixing in axial direction. This inter-
mixing can be due to diffusion by Brownian motion or by turbulence (Figure
11.54).

f deol plug f low
X
dx
I I
I I
: :
plug
f low
D.S4 Dispersed plug flow
If we assume the mass flow rate for dispersion in the x-direction to be pro-
portional to t he concentration gradient a mass balance over a length dx shows:
dt x dx
JC
x dx
from which follows:
dC dC d
2
C
-d + V.x-d = D-d 2
t .X X
Here, D is the dispersion coefficient.
x+dx:
(Il.l04)
For the case of an instantaneous tracer injection (at timet = 0, place x = 0)
this differential equation .can be solved to yield :
E = = g exp{- (:8--_:_l}
2 tr.8- 1
vxL L:,r-
(II. lOS)*
(Levenspiel, Chern. Eng. Sci. , 6, 227, 1957). Equation (II.105) describes the
distribution of the tracer concentration as a function of time 8 and distance
from the point of injection x = vxt. Figure 11.55 shows E curves for various
*This model description holds only when the convective veiocity is higher than the velocity of
diffusion (i.e. vx > 20/L). The reason for this limitation is that for lower velocities the diffusion
against the direction of flow presents a problem which can only be solved by specifying the flow
and the dispersion mechanism in detail for the area x < 0.
132
o/v}( L =
l
e
Figure ll.55 Dispersion modeL E curves
values of the parameter DfvxL according to this relation. For DfvxL 1
equation (II.105) can be simplified. For small values of DfvxL theE curve will
be zero but at 8 ~ 1. Now with 8 ~ 1 we obtain:
D
- 001 (II.l06)
VxL
which is the normal Gauss distribution. For low values of DfvxL the maximum
of E occurs at(}= 1. For higher values ofDfvxL (>001) the maximum occurs
at fJ < 1.
In order to fit an experimentally determined E curve to one of the possible
theoretical curves the variances of the curves can be compared. For a flow
system with negligible entrance effects {ie. for small values of D/ v'J:L), the variance
can be calculated from equation (II.105) via:
to be:
2
J ~ (x - x)f(x) dx
a = J ~ f ( x ) d x
(11.107)
which is simplified for low Df v;rL to :
D
.,-2 = 2-
v:r;L
133
(II.107a)
The F curve can be obtained from theE curve by integration according to:
F = s: E(O) dO = I E AO
(II.108)
Integration of equation (Ill OS) is not possible, but the F curve can be constructed from
theE curve by graphical integration of the latter. Equation (11.106), which is valid in the
case of very little dispersion [(DfvxL) 0.01], can be integrated directly. The corresponding
F curve is then :
J
6 1 f
8
{ (1 - 8)2}
F = E(O) d8 = 1 exp - D d9
0 2n: _!!_ 0 4-
v ~ v%L
which can be solved to yield:
where erf y is the normaJ error function:
erf y =]; J: e-tl de; erf(m) = 1; erf(O) = 0
(II.109)
(II.llO)
I 1'.8.4.2. Tanks in series model. For a number n of perfect mixers in series with
wtal residence time -r the E function can be shown to be :
(C) 'lffY" -1
E = -- = e- "'
C
0
(n - 1)!
(Illll)
For large n (n > 5) the .expression n! ~ -,r: e-"fon is valid and equation
(11. 111) becomes :
(n > 5) (Il.llla)
Figure II.56 shows E curves according to equations (11.111) and (II.llla) for
various numbers of stirred tank reactors in series.
The variance of the E curve for n stirred vessels in series is given by :
(II.112)
134
H ~ r - - - - - - - - - - - - - - - - - - - - - - - - ; - - - ,
l
8
Figure ll.56 Tanks in series model, E curves
and analogous to the. development of equation (11.110) we find for the F curve
for n tanks in series :
F - 1 !'e {1 (] (nB) 2 (n0)3 (n9Y' - 1 }
- -e +n +2! + 3J + ... +(n-1)!
(II.113)
F curves for different values of n are shown in Figure II.57.
The analysis of experimentally determined F curves is more complicated
than that of E curves. Two cases can be distinguished : F curves for n < 50 or
Df(vxL > 0.01, and F curves with very little dispersion, i.e n > 50 or D/v;cL <
0.01. Both cases wil1 be treated separately.
(a) High dispersion : n < 50 or Df vxL > 001
There are two possible methods of analysis :
(i) Via the E curve
Plot the measured F values versus time and determine graphically at
different times the slope dF(t)/ dt = E(t). The values determined for E(t}
at the approximate time can then be analysed as described in paragraph
II.8.2.
135
10
e--
Figure ll.57 Tanks in series model, F curves
(ii) Via the slopes of the F(8) curve at 8 = 1
The slope of an F curve at 8 = 1 can be shown to be given by:
(
dF) n"
d8 8= 1 ~ (E)e..,l = (n- 1)! e-n
The relation between this slope and the number of stirred vessels in
series is shown in Figure ll.58.
30.--------------- --- ----,
-
..
n
Figure U.58 Tanks in series model, dFfdO at 8 = 1
136
To analyse a measured F curve, determine the mean residence time as
shown in Figure II. 50. Determine graphically dF fd8 at 8 = l and find the
number of stirred vessels in series from Figure II. 58.
(b) Low dispersion: n > 50 or D(v:r;L < 001
By plotting the F curve on probability paper a straight line is obtained.
The mean residence time is found from this plot on the t-axis at the 50-
percentile point, whereas the time difference between the 16- and the 84-
percentile points equals two standard deviations. Thus:
2(], = tlr=084 - t1F=OI6
(II.l14)
With <1:} = l ju;, the variance and hence the number of ideal mixers can be
determined using equation II.ll2.
11.8.4.3. Comparison of dispersion/ tanks in series models. At high dispersion
(small n, large Djv%L) the two models y1eld E or F curves of different shapes.
At very low dispersion (large n, small D/vxL) the two curves become more
similar and are identical at n--+- oo or Df vxL--+- 0.
The number of vessels n of the tank in series model and D f v : : ~ : L of the dis-
persion model are related via the variance :
2
1 D ( D )
2
{ ( v xL)}
(J = - = 2- - 2 - 1 - exp --
n ~ L v ~ D
2
Tonks in series rnodel
n"' 5
8--
Figure U.59 Comparison of models
(11.115)
137
bich becomes for small dispersion numbers :

(II. ll5a)
Figure 11.59 shows an Ecurve calculated for both models on the basis of identical
(equation Il.115). As can be seen the E curve of the tanks in series
model yields a higher proportion of material staying only for a short time in
the system, whereas the E curve of the dispersion model (which shows also a
smaller maximum E value at higher 0) shows a greater portion of liquid staying
for a longer time in the system.
It is evident from the above that on analysing and presenting the residence
time distribution of a continuous reactor system the model must be selected
carefully in order to present the reactor system as well as possible. If, for example,
the reactor system consist s of a series of vessels or compartments in which
mixing occurs, the tanks in series model should be chosen, whereas in the case
of a tubular reactor the dispersion model will generally represent the system
more satisfactorily.
11.8.5. Dispersion inflow systems
The dispersion of fluids flowing through tubes is satisfactorily described by
the dispersion model. For laminar flow in tubes the dispersion coefficient is
found to be (d, = diameter of tube):
D = 5 x

+ 0 ( if Re < 2000and ReSc Sc = (II.lJ6)
1
r
r
r
Lommor f low
Re Sc 30L/d
1
,..

10
4
10
5
10
6
- Re Sc d,lv 11/D IO
Turbulent flow
-- Re
Figure ll.60 Dispersion number for flow in tubes
138
whereas for turbulent flow :
D ~ Q. 2v ;xtl
1
(if Re > 10
5
) (IJ.117)
Figure 11.60 shows more detailed information about the dispersion coefficient
for flow in tubes.
In the literature the degree of fluid dispersion is often not described by the
dispersion coefficient but by stating the height of a mixing urut, HMU. For the
tanks in series model this length is given by:
L
HMU= -
n
(11.118)
and states the length of pipe that is equivalent to one perfect mixer. For the .
dispersed plug Bow model, applying equation (II.llS), we find:
For fully developed flow through beds of particles (e.g. fixed beds, packed
columns) the height of a mixing unit is found to be equal to one particle diameter:
HMU ~ dP
in the range 01 < Re < 2000. This relation also applies for flow around
cylinders (e.g. in heat exchangers), as shown in Figure II.27, if instead of dP the
cylinder diameter ~ is used. Since for packed columns and fixed beds mostly
dP L, the number of mixers in series is so big that plug flow is approximated.
Knowledge of the dispersion coefficients in pipe flow enables us, for instance. to calculate
the degree of mixing in pipelines, if one fluid is followed by another Jiquid miscible with
the first. The situation is schematically drawn in Figure ll.61. At t = 0 (entrance of tube)
t h e r t ~ is no mixing at all, whereas at t = 't some intermixing has occurred. The problem
is now to determine the volume of liquid leaving the pipe, in which the concentration of A
and Bare lower than the preset values C ~ and Cl. This volume often has to be considered
as a loss.
This problem is most easily solved by using the definition of the F curve: C
8
== F(O) and
CA = 1 - F({}). Hence, the considered volume, which leaves the pipe between 8
1
and 9
2
,
c*\
I e
: Equotion{llIIO)
I
I : 0
0=0
02 I = T G,
0 = I
Figure ll.61 Mixing in pipelines
139
can be calculated from:
c; = F(8
1
) and = 1 - F(8
2
)
If we concur to the model of pipe flow with axial dispersion and if CA and q are nor too
low, we might use a linearization of the F curve around the mean residence time:
F ! + (dF) (9 - 1) "" _!_ + E (8 = 1)(8 - 1)
2 dO s=t 2
Replacing E (8 = l) with t he help of equation (11.106} we obtain :
1 1
F + (8 - 1)

(11. 119)
Equation (II.ll9) could have been derived directly jf we known that the
penetration depth of both fluids into each other is given by 2.j1tDLfv:r (see paragraph IV.2).
Introducing equation (11.119) into the analysis of the contaminated volume, which
occupies a length of pipe L as it reaches the end of a pipe of total length Lo. gives the
following result:
Hence, the contaminated volume:
increases with the square root of the dispersion coefficient ;
increases with the square root of the total pipe length;
decreases with the square root of the velocity of the fluids; and
decreases with increasing c: and c:.
For turbulent flow in pipes (Re > I 0
5
) we flnd, with equation (II.ll 7):
0

-.1
ro-
1
-o

Q
c:
e
0
c
0
v
tl)
=
'ii
0.
5.
0
c:
g
u
c
ct
i0-
4
10-
6
L
2
( C* ,.....)lff2trd
1
1- A - '--B --
Lo L
0
20-80
\ Peent contommoHon
5-95 10-90
10-
3
10-
2
Dlvx L = ( Dlvx d
1
Hd
1
/L)
Figure H.62 Mixing in pipelines
(II.l20)
140
or, roughly:
L ~ (1 - q - q")./"Q,
indicating that the length of the contaminated zone depends on the square root of the
product of the total length of the pipeline and its diameter.
Because of the fact that we have approximated the concentration profiJe by a straight line,
equation (11.120) will give results that can be a factor 2- 3 too low at very low values of
C ~ and c;_ Levenspiel has given an exact solution of this problem and has prepared a
chart for an easy calculation of L/L
0
, as shown in Figure 11.62 (from Petr. Refiner, March
1958).
11.8.6. Problems
1. De.velop an expression for the E and F functions of two perfect mixers in
series (each with a residence time r:/2) and show that the results are in
agreement with equations (II.lll) and (11.115).
2. The feed stream to a continuous deodorizer (a vacuum steam stripper
applied for rendering edible oils tasteless) is at time t = 0 changed from
bean oil to coconut oil and the refractive index of the effluent stream is
recorded. From the refractive index the percentage of coconut oil is deter-
mined:
Time (min) Percentage of coconut oil Time (min) Percentage of coconut oil
30 0 60 705
40 5 65 83
45 165 70 92
50 345 80 99
55 52
(a) What is the mean residence time of the oil?
(b) How many perfect mixers in series would yield a corresponding residence
time distribution?
(c) What would be the dispersion factor if we wanted to descnbe the
residence time distribution by the dispersion model?
(b) 25 vessels
(c) 002
3. Calculate the height of a mixing unit for air and water flowing through a
circular pipe at Re numbers of 1, 100 and 10,000.
Air
Water
25 x lO- s
'2 X 10-
9
17 x 10-s
10- 3
Air
Water
1 100
0-Sd,
SOOd,
10,000
d,
d,
141
4. A relatively small waste water fiow from a factory is bled into a big effiuent
stream of 0.1 m
3
/ s and discharged in to the sea. If the waste water flow is
stopped, the concentration of impurities in the effluent stream at the end
of the drain appears to fall alniost exponentially to nil. The concentration
at the outlet is then halved every 100 minutes. If t he plant is running and
1 kg of very harmful material, uniformly distributed over 10 minutes,
enters the waste water flow, calculate the maximal concentration of this
material at the outlet.
3
s. In order to determine the mixing efficiency in a cascade of six stirred vessels
the following experiment is carried out. A continuous stream of water is
passed through and at time r = 0 an amount of acid is injected into the
first reactor. At regular intervals samples from the effiuent of the last
reactor are taken and analysed :
Time (min) Concentration (g/1) Time (min) Concentration (gfl)
0 0
30 300
5 0-10 35 212
10 163 40 139
15 323 50 051
20 396 60 010
25 371
70 0
(a) What is the mean residence time in the cascade?
(b) How many perfect mixers in series would yield the same residence time
distribution?
(c) What would be the dispersion factor if we wanted to describe the
residence time distribution by the dispersion model?
(d) How many grams of acid were injected if the fluid flow rat e was
100 ljmin?
(b) 5 5 vessels
(c) 010
(d) 102 kg
142
6. Show that equation (II.ll5a), which relates the model of a cascade of
mixers to the model of flow with axial dispersion and which is based on
the criterion of equal variance, also predicts that when both models are
matched in this way :
the E values at (} = 1 are the same for both models and, hence,
the slopes dF/d(} at(} = I are equal.
*7. Show that in the problem stated at the beginning of this paragraph John
calculated the correct probability of catching the man in the supermarket.
Problem 2
A plot of the F values given versus time (see the figure) yields, after graphical
integration, -r = 551 min. At r , we find graphically for the slope of the line:
or :
I dF) = 0-0357 min-
1
\ dt t="t
(
dF) = (dF) r = 196
de 8= 1 dt z=t
From Figure 1158 this leads to n 25 perfect mixers in series. The dispersion
number is then found via
10

0
D 1
()'2 = 2- =-
v L n
;c
S:
v

I
0
'
1.
ll.

0 4
- 1
---Slope =00357 mrn

00

J I I I I I
20 40 60 eo 100
r ( min)
Problem 5
With equations (11.96) and (11.97) we find:
< = 2:tcAt = 2600 = 25-Smin
I cat 101-8
cr = L: 1 = o-t8t
r
2
l:cllt
Using equation (II.107) we find by trial and error :
_Q_ = 010
v;xL
The amount of tracer injected is found as:
L c /j,t = = M; M = 102 kg acid
</>v
Problem 7 John and supermarket problem

143
John regarded the supermarket as a perfect mixer with a mean residence
time of 30 minutes. Since the man had entered the supermarket 15 minutes ago,
the chance that he was still in the shop is given by
1 - F = e-
9
....:. e - o-s = 0-606
or 60 per cent. However, John .neglected to realize that be was looking for one
man only, whereas the concept of residence time distribution is applicable only
to a very great number of elements!
ep ( ifl hob/em Sclv1 n9
4. ft: my\lie ftlf L "'Jn ll !f,Cit!l!.
I! Id )I ? I . r;, D1 / In
' I J)l
('r II. 11
SOlve ror Re "' pvQ
. '7
Solt<.. ror ft
, , VI l:;u( hJ
h , gml tr I{ u't wl otral!r r.:
.,. - L .
c/:
' VJ!:f\J" r.::t ftt
CHAPTER III
T.,..c f t t.J,
rrt , T - LIT; ( 1ern_,)
Jchft r t11t - -llv..n 1..,.
9 U, , l'tldoh fqn 11
In this chapter the transport of energy is studied on the understanding that the
changes in heat content are much greater than the changes in mechaniGa]
energy. In this case the law of conservation of energy is written as :
d:, = lPrninOill - mout0out- lPH + A
where E, = pcPT V and 0 = cPT* (as long as no phase transition
H = heat that is exchanged with the surroundings and f/>.,. = mechanical
energy added. With this law of conservation all problems of heat transpon
in a system can be solved ; the following t herefore does nothing but ascertain
the consequences of this balance.
The transport of heat..plays an important role in many processes in engineer-
ing. It may be that we wish to promote the transport from one medium to
other (generation of steam, heating or cooling of liquids, removal or supply
heat of reaction, cooling of transmitter tubes, removal of heat generated
nuclear reactors, etc.) or to suppress the transport to restrict loss of beat
cold (e.g. insulatiol) of pipes, heated vessels and cold stores).
Three different mechanisms according to which heat can be tr
(invariably from higher to lower temperature) may be distinguished, namely:
(a) Heat conduction
In non-moving media in the presence of a temperature gradient heat
transported by the molecular movement (from high t o low tenmerat1ure1
According to Fourier, the heat flux tP:/t (Jj m
2
s) occurring at this ......... n.er ...
<P'H =
. dT
-,1, -
dx
We have taJcen here for the heat capacity the vaJue which is correct if the process proceeds
constant pressure. Fundamentally, this is not quite accurate, because the presence of a flow
that there are presf. ure differences.. In mcst cases this is not a great problem because either cp
c. are almost equal to each other (e.g. when solid substances and incompressible liquitts
involved) or the process is almost isobaric If we want to proceed in a precise manner, we
read, instead of c, , .c. + T(6p/ f>n9 !JVj ()T, this constant only being calculated for every"'"".;_
meot if it is known which change in the volume per un it of mass ()()takes place within the
at which the temperature changes witb
145
Since cp''g is positive in the direction of decreasing temperature, the minus
ln many correlations for heat transport the beat conductivity A. is found in
the combination AjpcP which, because of its analogy with the diffusion co-
efficient, is called thermal diffusivity a (m
2
/s). In Table 111.1 approximate
values for the heat conductivity and for the thermal diffusivity are collected.
Table W.l Approximate values of land a for various materials
Gases
Liquids
002
a (m
2
/ s)
20 X J0-
6
0-1 x to-
6
Solids (non-metallic)
Metals
0-2 (water 06)
2
20-200
1 x to-
6
s-so x to-
6
(b) Flow (convection)
In this case heat is entrained with the moving conducting medium. The flow
may be the result of external causes (forced fl ow or convection} or spon-
taneous, e.g. as a result of density differences caused by temperature dif-
ferences (thermal, free flow or convection), or of bubble formation (boiling).
This is the transfer of beat by electromagnetic waves between surfaces of
various temperatures which are separated by a me<lium transparent for
becomes important but it can also play an appreciable role at room tem-
perature.
These three mechanisms often occur in combination. In addition, for the
heat transport by (b) the heat conductivity). is a very important property, for a
ftowing medium can only absorb beat from a hot wall if it is conducting.
Temperature-levelling in moving media takes place by convection (mixing of
cold and hot) as well as by conduction.
Just as in the theory of'fiow phenomena, for cases of stationary heat con-
duction knowledge of the relationship between the heat flow and the driving
force is sufficient to rapidly SC?lve one-dimensional problems. In the theory of
flow Newton's law was t he st arting point ; in this case it will be Fourier's law.
ID.l. Stationary beat conduction
John realized char the man found in eire storage room for
deep. frozen foods was dead. Although he wore protectir;e
clothing his body was quite cold and had scarted to freeze.
The doctor informed John that one hour ago the temperature
measured at the arm had still been ffC. John concluded,
realizing that the storage temperature was - J(f C, that the
man must have died approximately Jt hours ago and started
questiqning his colleagues.
146
lll.l .I . Heat conduction through a wall
In the case of stationary heat conduction through a wall (thickness other
dimensions infinite, temperature difference oveT the wall {j. T = T
1
- T
0
) the
beat fl ow through any plane parallel to the end faces will be of the same mag-
nitude (i.e. no accumulation of heat in the wall takes place because of its station-
ary state). Now that the heat flux <f>'H appears to be constant (independent of
x), A. dT/dx will have to be constant (Fourier); so if A. is constant in the tem-
perature range from T
1
to T
0
, the temperature distribution over the wall will
be linear. If the heat conductivity A. of the material is known, the beat flux at a
given temperature difference can be calculated :
</>" - ). Tt - To
H- d
(IIL2a)
Methods for measuring the heat conduction of solid substances are bas
on this equation. If the wall consists of more ta yers (Figure II I.l ), for the sanie
reasons as in the rf>'H must be constant in every cross-section in every
layer and so in every layer dT/ d.x must be constant (and inversely proportional
to the conductivity of that layer!). If for every layer we write equation (JII.2a)
as follows:
A,.H H = flT: = T:- T: . l
'f' A. I I IT
I
and add up all these relations for all layers we get :
AT. = T _ 7: = .J.n (X2- X1 + X3- X2 + X
4
- x
3
)
t ot 1 4 ' B
1
,
ll. t 1.2 1\.3
--K
Figure ill.l Stationary heat conduc-
tion through walls
(111.3)
The terms (x
2
- x .}/ }..., etc., can be considered as heat resistances, and the
above equation says that the total temperature difference (T
1
- TJ equals the
heat flux multiplied by the sum total of the resistances placed between .x
1
and
x
4
(Compare this with Ohm's law in electricity.) The sum total of all partial
resistances. which are also called the partial heat resistances, is described as the
147
total heat resistance for which usually (and also here) the symbol 1/ U is used.
'U' is then the total heat transfer coefficient.
. -
ll 1.1.2. Heat conducti9.n)hr.ough, cylindrical walls
.
The next example is the stationary heat conduction in the thick wall of a
cylindrical tube (see Figure 111.2). In this case the heat flows (per unit of
. t ubt:: length) through the surfaces r-= R
1
and r = R
2
and through all surfaces
in between must be equal:
A-' ,dT
2
'PH :::; '"7 1. dr 1_tr = constant
(III.4)
{ntegraHon yields:
. .
T = constant ln r + c;,
and with tlte boundary conditions indicated in the figure we find for the. tern-
. perature distribution in the wa:Il:
a -
0;:- k d .J / dr
A
Figure m.2 Stationary heat conduction
through a cylindrical wall
(IIL5)
' .
Subsequently, the heat flow J,er unit oflength can be calculated as a function of
the temperature difference a pplied:
, , d T 2nJ.(T
1
- T
2
)
H = -2tul. qr = In R2/ R1
(III.6)
which is indeed independent of r.
Equation (IJl .6) can be used to the optimum thickness of thermal
insulations of a pipe. There are two criteria. The first js purely
technological. With increasing diameter of the insulation the heat transport
res}i;tance increases and so does the outer surface of the insulation. Due to this
148
a situation may arise where, with increasing insulation thickness, the radial heat
loss per meter length of pipe 'n goes through a maximum. If the beat transfer
coefficient (by free convection and radiation) at the outer surface is called a.
0
,
we find from equation (111.6) after addition of the transport resistances in the
insulation and at the outer surface (T
3
= air temperature):
' _ (Tt - T3)
R- ln Rd Rt 1
---+---
2nA. 2nR
2
cx
0
(IIL 7}
From this equation it can be calculated that the maximum heat loss is obtained
for a.
0
R.j). = 1. For values of rxoR) J. higher than 1, 's decreases monotonously
2
of the insulation. for most insulation materiais
l = 0.05 W/ m oc and cx
0
is of the order of magnitude of 10 W/m
2
paragraph IIL6). In those cases the critical value of R
2
at which the above
phenomenon occurs is R
2
= 5x 10-
3
m. With such insulation materials and
under such conditions there is therefore no risk that fitting an insulation jacket
increases the loss of beat. A different situation arises when, for example, for the
insulation of laboratory apparatus asbestos is used (A.= 015 W/ m
0
C}; in that
case the critical value of R
2
= 15 x 10-
2
m!
:The second optimization criterion is connected with the consideration that
for technical installations the choice of the insulation thickness is determined
for economic reasons. At greater thickness the loss of heat decreases but the
costs of insulation rise. By writing down the equation for the total capital loss
as a function of the insulation thickness (loss of beat, interest and depreciation)
and by differentiating this equation to the insulation thickness, the optimal in-
sulation thickness can be calculated_
IIJ.J.3. Heat conduction around a sphere
As the last example of stationary conduction we sha11 treat the heat transfer
from a sphere to a stationary medjurn (sphere diameter 2R, sphere
T, ambient temperature T
00
, constant A. of the medium around the sphere).
Here also applies that the heat flow through each surfacer = constant must have
the same value :
2
=constant (independent ofr)
dr
from which follows the temperature distribution in the medium around the
sphere:
T- T R
cc
(IlLS)
---=-
This result is again sufficient for calculating the heat flow which is transferred
between sphere and medium at a given temperature:
(III.9)
149
Newton described the cooling of bodies of any form with a pragmatic cooling
law in wh.ich he put the heat flow proportional to the outer surface of the body
and to the prevailing temperature difference (T
1
- Tee,). The proportionality
constant, which is a purely phenomenological quantity and contains all that
we do not know of the transfer process, be called the heat transfer coefficient a.
For the sphere from the previous example this cooling Jaw is therefore for-
mulated as follows :
(III.lO)
It follows that in this case !X = J..R. In our j argon this is summarized to Nu = 2.
where the Nusselt number Nu is defined as a2R/ A.. The Nu number represents
the relation between the heat resistance which is estimated on account of a
characteristic dimension of the body {in the above example 2Rj ).) and the actual
heat resistance (1/ a). We may also consider this quantity as the relation between
the dimension of the body (2R) and the thickness of the Jayer over which the
drop in temperature takes place (/.fa).
Analogously for a fiat plate and for a cylinder we find Nu = rxdjl = 1 and
Nu = a2Rj ). = 2/ln R
2
j R
1
respectively. IJu hD ) rth tr -H.11 l,caJ -lr:ntrrttr!d
- tnltt f. lronrr rt
K ,,. lu
1
(lllJ 1
I I I .].4. General approach for the calculation of temperature distributions It '
1
..,. . (i (Q
1
tc "'' rill
The general approach for the calculation of temperature distributions is a
microbalance which here will be given for a Cartesian volume element (A and
cP constant). The approach is completely analogous to that which we followed
for calculation of velocity distributions during laminar flow. again is the
restriction that as soon as the eddies take 9ver the heat transport from the molec-
ular transport, exact calculation is no longer possible, we can
the molecular transport of heat mathematically (Fourier's law). but not the
. eddy transport. This means that pure conduction problems and to
laminar flows can be dealt with but transfer t_o turbulen't flows
cannot.
z
Figure ID.3 The Cartesian volume element
The microbalance is as follows. In the x-direction an amount of heat equal to
( - J.(dT/dx) + pcpvxT)dydz flows in at position x. The net transport ;in the
volume element through the walls x and x + dx is therefore:
. ,
150
Analogous expressions can be written for the y- and z-directions. The beat
content of the volume is pcPT dx dy dz and the production of beat per unit of
time is q dx dy dz (hence q is production per unit of time and per unit of volume).
accumulation _ fi . fl production
. f . - ow m - ow out + . f .
umt o t1me un1t o t1me
now becomes :
dpT dpvxT dpvyT dpvzT _ , d
2
T , d
2
T A. d
2
T
c, dr + cP dx + c" dy + cP dz - 1\. dx2 + A dy
2
+ dz
2
+ q
-
If we use the continuity equation once to modify this beat balance Uust as ~ t h
the micromomentum balance) the result is: .. -
(
dT dT dT dT) (d
2
T d
2
T d
2
T)
peP dt + vx dx + v:v dy + v: dz =A. dx2 + dy2 + dz2 + q (HI.ll)
In cylindrical coordinates this equation reads :
(
dT dT v
8
dT dT) (' 1 d( r ~ n 1 d'T d' Tl)
pep dt + vr dr + r dO + v, dz = ).. T dr + r
2
d8
2
+ dz
2
+ q
(Ill.l2)
The reader may check that the Simple cases treated in the first three sections
of this paragraph can be derived from these equations. !n the case of conduction
. through a cylindrical wall treated in paragraph .JII.l.2, for example, q = 0,
vx = v>' = vz = 0, dT/dt = 0 (stationary state), dTjdfJ = 0 (constant tem-
perature in all radial directions) and dT/d2 = 0 (constant temperature in axial
direction). So we find from equation (III.12) :
~ + ~ l = 0 r ddTr = constant
r dr '
which is identical with equation flll.4).
III.l5. Temperature distribution in a cylinder with uniform heat production
An electric current flowing through a wire will, because of the electric resis-
tance of the material, produce an amount of heat q per unit of volume (W jml)_
By beat conduction the heat produced is transported to the surface of the wire
which has a constant temperature T
0
. From the surface of the wire there is a
heat flow to the surro'undings which, in steady-state conditions, equals the
amount of heat produced Thus :
c/>nlr=R = qnR
2
L (fiLl
151
L
Figure lll.4 Temperature distribution in a .cylinder
In order to find the temperature distribution in the wire we set up a micro-
energy balance over a cylindrical shell ofthickness dr (Figure III.4}. For steady-
state conditions, equation (111.1) can be written as:
dE,
dt = 0 = 2nr L4>8L. - 21t1' L<Pil,. + c1r + 2m' dr Lq
Dividing by 27tL dr we obtain:
. O rcP'HI, - d(rcp'H)
= dr + rq = - dr + rq
Integrating this equation we find :
q . cl
'PH= lT ;-7
'
The integration constant C
1
must be zero because, at r = 0, if>'H cannot be
infinite. Thus, applying Fourier's law (equation III.2):
'H = =
2 dr
A second integration (assuming). to be constant) yields:
. T = _!i.,.:z + c2
4).
and, using t he boundary condition T = T
0
for r = R, we finally obtain:
(111.14)
Thus, the temperature distribution is parabolic, the maximum temperature
being at the centre (r = 0):
(III.15)
152
The average temperature is given by:
foR 2nrT dr s: Tor dr + J.R <R
2
- r
2
) dr
( T) = fR - tR
Jo 2w dr
(Ill. 16)
For the heat flow at the surface we find:
dTI
</JH = -2rcRLJ.-d _ = nR
2
Lq
T r"'R
which is in agreement with the overall balance equation (III.l3) set up for steady-
state conditions.
It is interesting to compare the results just obtained with the equations
developed for the shear force and flow velocity distribution for laminar flow
in a circular pipe (paragraphs IIJ.2 and 111.3). The microenergy balance
applied here is analogous to the micromomentum balance of paragraph Il.l.3
and the correlations for Jarrunar flow found in paragraph 111.2 can be translated
into the temperature distributions just developed if we use the conversions
given in Table 111.2. Apparently, the processes of laminar fiow through pipes
and of uniform heat production in a cylinder are analogous.
Table m.l Analogy between laminar How and uniform heat production
Laminar flow in cylindrical pipe
Trz

[In/s]
'1 [Ns/m
2
]
P 1- PlfL =

Ill.l.6. Problems
Cylinder with uniform heat production
4>11
T- To
A.
q
(W/ m2]
rCJ
[W/m
0
C]
[W/m
3
]
1. An oven wall consists in succession of a layer of firebricks (Ar = 121
J/ m oc s), a layer of lagging bricks {11
1
= 0.080 1/m oc s) and a layer of
bricks (),b = 069 J/ m oc s). Each layer is 10 em thick. The temperature on
the inside of the wall is 87rC, on the outside 32C.
(a) If the surface area of the wall is 42m
2
, calculate how much heat is lost
by conduction every 24 hours.
(b) Calculate the temperature T'" in the middle of the layer oflagging bricks.
Answer : (a) 21 x 10
9
J ;
153
2. The Chemisch Weekblad of 30 September 1966 gave the following news
item:
In Elsa, a mining settlement in the Yukon district, situated only 290 km south of
the polar circle where the temperature sometimes falls to - 70C. the water supply
was seriousJy hampered by freezing. This pro blem has been solved by putting a
floating pumping engine in the nearby lake. The construction is such that the water
inlet tube is well below the level of the ice. The 125 an thick pipe connecting the
pumping engine with the storage tank in town-a distance of ca. 5 Jan-was
insulated with a 5 em thick layer of urethane foam. The flow rate of the water in the
pipe is on an average 6751/min Even at extremely low temperatures water of 2SC
can be pumped from the lake to tbe town without heating being necessary.
What follows from these data for the heat conductivity of urethane
foam?
Answer : 003 W /m C
3. A management considers installation of double windows in a large office
building. According to the bookkeeper the fixed costs (depreciation,
maintenance, insurance) are Dfl. lOOOper year more than for single windows.
Natural gas costs 2 cents per decamega joule (10
7
J), a year has 2 x 10
7
heating seconds and the average temperature difference inside-outside
is l0C. The heat transfer coefficients inside and outside are both
20 w / m
2
C, a pane is 3 mm thick (l glass = 1 w / m C) and betweeii the
double windows there is a 5 mm air space (). air = 0025 w / m C).
(a} Calculate the loss of energy per year for the single and the double
windows.
(b) What would the total window surface area be if the additional expenses
of double windows could be paid for from the saving in heating costs?
Answer : (a) single windows : 194 x 10
7
J/rr/; double windows:
65 x 10
7
J/m
2
(b) at least 388m
2
4. A length of steam pipe (outer diameter 005 m) bas to be insulated. Depend-
ing on the thickness of the glass wool insulation a Dutch firm in November
1970 quoted the following prices (including glass wool insulation, outer
layer of aluminium plate 1 mm thick, labour, transport and tax) :
Thickness of insulation (mm)
10
20
30
40
50
Price Dft.j m
1350
15-00
1700
1950
2250
Which thickness of insulation represents the economic optimum?
154
Data : capital costs per year= 20 per cent (10 per cent depreciation, 3
per cent maintenance, 7 per cent loss of interest) of investment
steam temperature = 180C; mean air temperature = ooc
J. glass wool = 0-05 W /m oc
steam price = Dft. 1().00 per ton
heat of vaporization AH = 480 kcal/kg
The steam pipe is used for 5550 hours per year.
Answer : 30 mm ; total costs per m = Dfi. 5 70 per year
5. The rate of heat production of a Newtonian liquid per unit of volume by
viscous dissipation in a cylindrical pipe is given by q = '1(dvf dr)
2
[Wf m
3
].
(a) If the pipe has a constant wall temperature Tw, develop an
flow.
caused by viscous dissipation can be neglected.
Answer : (a) T- T. =

;:)
1! : 1/Jj Mft!l .. J'6. The hydrogenation of edible oils is carried out with a suspended catalyst.
What is the maximum temperature difference between the catalyst surface
01
'
1
3 .-u' and the oil?
411
rl ; . ,3) 'li m
1
'h
J ;.r 1
, ' r I I'!' 1l / 1 t-
('.,'"'-' 1':1 2 5 10 6 .. .
. Datil : spherical particles, dp = . X - m '-;4' n)
" ,_ ... J 'i
1
apparent specific gravity= 1750 kg/m
3
it f..,} ' rr .. 02 per cent by weight of catalyst applied <ltt dH I n : h!l vT
s H o J hydrogenation rater= 13 x 10-
3
kmoljm
3
s .P h k
i.v1- liPJ 1: , ,
11
,.v! heat of reaction Ali = 25 kcal/mol
2
_ 1.. .
_ r. _
4

molecular weight of the oil M = 880 gjmol h (
1
P . )(
1
..,, )
).:a lyl..a; ' '
40
'" p
0
;
1
= 900kg/m
3
;).oil =02kcal/ mhC < ':'
--
J
-
1
off f
'1 l.'l : Answer : 3 x 10-
4
oc
I
-J
''1'" "Ol0Cd _:rvl
- l ,J , ,:
*7
I '' .
i; {.
A copper pipe (length 05 m, outside diameter 2-5 em, wall thickness 2 mm)
bas a thick heat insulation and the ends of the pipe are being kept at ooc
(i.copper = 400 w / m C). Calculate the temperature distribution in the
pipe wall if an electric current flows through the pipe giving 20 W in the
stationary staLe.

- x
2
) c
*8. Show that John drew the right conclusion in the case of the deep frozen
body, treated at the beginning of tllis paragraph.
Problem 6
A heat balance over one catalyst particle reads:
rAH

n
155
where n =number of catalyst particles per m
3
oil = 125 x 10
14
Now for
these small particles Nu 2; thus a 2ld and & T can be calculated.
Problem 7
A beat balance over a length of pipe dx reads :
dT q dT
r/>
8
=-A).- + -dx = - A).-
dx x L dx x+dx
dJI
L
or :
Integration yields:
. dT q
dx = - J..LA x + C
1
but since, at x = 0, d T fdx must be zero, C
1
= 0. A second integration leads to :
-q
T = x
2
+ C
2A.LA
2
and, since T = 0 at X = +fL, c2 = qLj &}.A. Thus the temperature distribution
is given by :
q {L
2
}
T = 2J.LA 4 - x
2
156
Problem 8 John and the frozen body
If we regard the body as a cylinder (L R) we can set up the following heat
balance per unit length :
A.' T - T,.
'+'H = ln R / R 1
2 1 +---
2nA. 2nR
2
cx
0
= _McpdT
L dt
or, collecting aU constant values into one constant C :
(T d T = - C f' dt; In T - Ya - - Ct
Jro T - 1: Jo To - 1:
The information that T
0
= 8C at t = 0 and T = ooc at t = 1 hat T'g =:= - 30C
allows us to calculate the value of the constant C as C = 0 236 h - I .
So, assuming the poor fellow had no fever, his temperature was 37C when
still alive but ooc after t hours and we find, using the same equation, that death
must have occurred 34 h ago.
ill.2. NoD-stationary beat conduction
When John entered the empty room, he found a hot iron on
an asbestos plate of 1 em thickness. He lifted the plate which
proved still to be cold on the underside. He decided that
somebody had been in the room less than 5 minutes ago and
ordered his men to search the apartment.
If we want t o describe the variation of the mean temperature of a body with
time caused by heat conduction we have to start by setting up a microbalance.
From this balance we can calculate the temperature as a function of time and
place. This temperature distribution then enables us to calculate the mean
temperature of the body. This complicated procedure is necessary because
during beat conduction large temperature gradients occur in the medium that
necessitate a careful averaging procedure to find the mean temperature.
We have already come across the microheat balance in the previous para-
graph (equations III.ll and III.12). For pure conduction without heat produc-
tion (vx = v:v = v:: = 0, q = 0) equation (111. 11) is simplified to:
In order to solve this equation we generally need to know:
the geometric boundaries of the conducting medium;
the temperature distribution at a certain time, e.g. at t = 0 (initial condition):
the conditions for the temperature (or temperature gradient) at the boundaries
of the medium (boundary conditions).
157
The analytical solutions of equation (lll17) are known for a great number of
problems and can be found in the literature {e.g. in Carslaw and Jaeger).
For non-stationary heat conduction the duration of the heat penetration
process is of great importance. If this time is so short that the heat penetrating
into one side of a body has not yet reached the other side we speak of heat
penetration into a semi-infinite medium. Accordingly, if the penetrating heat
has reached the geometric boundaries of the medium we speak of penetration
into a finite medium.
lll.2.1. Heat penetration into a semi-infinite medium
~ t us consider a solid body with a flat boundary at x = 0 and an infinite
length in the positive x-direction, as shown in Figure III. 5. At t = 0, the tem-
perature of the medium is T
0
in all places (for all values of x; initial condition).
At t ~ 0, the wall at x = 0 is brought to a constant temperature T
1
(boundary
conclition). A second boundary condition is given by the assumption that the
medium .is semi-infinite, which means that at t > 0 and x = co, T = T
0
(no
penetration to other wall). For this problem equation (111.17) is simplified to:
with :
dT d
2
T
dt =a dx
2
(111.18)
initial condition : T = To fort= ot X ~ 0
boundary condition 1 : T = T
1
for t ;'i:?; 0, x = 0
boundary condition 2 : T = To for t > 0, x = co
1 r, ~ - - - - - - - t _=ex:. _____ _
0 X
Figure m..s Penetration of heat into a semi-infinite
medium
Equation (Ill.l 8) can now be solved by anticipating that the temperature dis-
tribution can be described by two dimensionless variables:
Via :
() = T - To and ~ = x
T
1
- T
0
..j4cU
de dB d ~ 1 x de
dt = d ~ dt = -2 4ar
3
ct e
158
and :
d
2
B = d
2
B(de)
2
de

_ d
1
e _t_.
dx
2

dx + dx
2
-

4at
equation (111. 18) is simplified to:
d
2
B dB 8 = 0 = oo}
d( 2 + = 0 with B = 1 for e = 0
(111.19)
Integration yields:
and:
e = c,J: + c,
With the boundary conditions we find C
2
= 1 (since e- = 0) and:
B = 0 = C
1
fooo

de+ 1 = ! C
1
.j;t + 1
Thus C
1
= and we find as the final solution :
or:
2
(111.21}
..1. cl .. /c
,/IC.p
l _ fJ = l _ T- T
0
= T
1
- T = f"
1
./4.ii
7;.- To T1- To j; Jo
(III.22)
The right-hand term of this. equation is called the error function, which is often
represented by the symbol erf, defined by:
2 JY
TJ - Ja erf Y = .fi. o ct e
In Table III.3 a few values of this integral are tabulated ; complete tables can
be found in the handbooks.
Equation (III.22) enables us to calculate the temperature distribution in the
medium at any time t and in any place x. The type of curve found is illustrated
in Figure III.6, which shows the temperature penetration into a semi-infinite
slab of stainless steel (a= 7-5 x 10-
6
mz/ s) at t
1
= 40, t
2
= 1 and t
3
= 01 s
for T
1
- T
0
= l0C. It appears that, for small penetration depths x, the tem-
perature curves can be represented by straight lines. The slope of these tina
159
Table 111.3 The error function
y erfy
0 0
f/J: = (1r
0.1 011
I IJ t
02 022
J."

04
().43
- u.. lrr -rq)/I-
06 0-60
08 074

10 0-84
h ..
lli'Jn fet eel
12 091
fS-
.::.
15 ,097
co
100'

I

0 lrrct
1
/rrat, )rrat
3
x-
Figure Ifi.6 Temperature distribution in a semi-infinite
medium
is found from equation (III.22) to be :
dT dO (T
1
- T
0
) d8
= (:(1 - ToF. - !A:: --;
dx x = O dx x=O y 4at d"> ( =0
Using equation (III.20) we find:
d T 2 T
1
- T
0
T
1
- T
0
dx x=O - - ,fi .J4Qi = - Fa

(III.23)
showing that the tangent to the temperature curve at x = 0 goes through the
point T = T
0
, x = Fat. The distance x = is called the penetration
depth, which represents the distance x over which the temperature difference
T
1
T
0
at x = 0 has dropped to 20 per cent of its original value. The last state-
ment can be checked by substituting x equation (III.22).
160
The heat flux through the wall. of the medium (x = 0) is found, using equation
(III.23), as :
(III.24)
Equation (III.24) is one of the most important results of physical technology;
it is known as the result of the penetration theory. The number of successful
applications of equation (Ill24} is enormous, because it is often possible
to define a characteristic time for the heat penetration process. Take, for in-
stance, the heating or cooling of particles in a rotating drum. The contact time
of the particles with the wall follows from the volume of particles and the
rotational speed of the drum In most cases the penetration depth ( ~ f is
smaller than half the smallest dimension (D/ 2) of the particles, so that the
medium can indeed be regarded as-semi-infinite.
The condition for semi-infiniteness of the medium:
is often written as :
D
for 2
at
Fo =- 01
D2
( < 0.05)
(III.25)
where the dimensionless number Fo is the Fourier number. The physical
meaning of this number is :
Fo = ~ = ( penetration depth )
2
* D
2
thickness of medium
The heat fl ux through the plane x = 0 (equation 111.24) can also be described
with a heat transfer coefficient tX:
(III.26)
Thus:
(IIl26a)
If the penetration process lasts a time te, the mean beat transfer coefficie:tit
is given by :
1 f' !ffPC i' 1 !f:PCp
( a ) = - a dt = --f r. dt = 2 -
te 0 Jtle 0 V L 1tt,
(Ill.27)
161
Applications of this fonriula are numerous ; see, for example, problems 2, 4, 6
and 7 at the end of this paragraph.
II 1.2.2. Heat penetration into a finite medium
If the penetration depth is greater than the dimensions of the body (Fo > 01),
the penetration theory proves imidequate, so another description must be found.
The situation then is assketched in Figure III.?. After a certain period of time
tin the figure for t > t
3
, which corresponds with . Fo > 01) the consecutive
:em perature distributions remain geometrically similar, which means:
T - T
0
= (T
1
. - T
0
){1 - A(t)f(x( R)}
The function A(t) only depends on the time; it decreases with progtessing time
(A( oo) 4 ' 0). The temperature distribution is further given as a function of the
place by the time-independent funCtion f(x/R) which must satisfy the following
conditions:
f(+ 1) = 0 and df fdxlx=o = 0
d"
Jl'll ~
-l_qr J' Y -
dx ,- "
F"-.gu:re 01.7 Heat penetration into a finite medium (for
0 < t. < t
2
penetration theory ; for t
3
< t < a:> constant a:}
f :co
To t - - ~ - - - " " " ' t " " ' ' - - - - ........ ---1
- R
0
+ R
-x
The nonstationary temperature distributions in a body of any form.should
therefore satisfy this general relation for larger times (Fo > 01). We shall
now ascertain .what the consequences will be by further analysing the case of
the flat plate. If we apply the general relation for the temperature distribution
of the microbalance (equation III.l8) to the flat plate, the result after a slight
1 d
2
j R
2
1 dA
f.(x/ R) d(x/ R)
2
=a A(t) dt
(IIL29)
The left-hand term is only a function of the place, the right-hand term only a
iunction of the time. Therefore the two terms should be constant, e.g. equal to
- {3
2
(a real, negative number). Only then can the two terms for all x and all t
be equal to each other and only then will A(t) for c -4 XJ asymptotically approach
zero. It follows that :
A ..... exp {- {1
2
at I R
2
) and f(x/ R) "" cos (f3x/r)
162
Thus the differential equation for the flat plate and the boundary conditions
are satisfied, but not, however, the starting condition, which after all could be
expected of a solution which applies only to large times. Now the value of /J
2
has still to be determined. It follows from the consideration that only cosine
functions of certain wavelengths neatly fit the range - R < x < R (just as when
vibrating a string the permissible wavelengths are determined by the length of
the string). The highest possible wavelength is here apparently equal to 4R,
further possibilities in this problem being (considering the boundary conditions
f( + 1) = 0) 4R/ 3, 4R/ 5, etc. The relevant fJ values are: rr/ 2, 3x/2, 5rr/ 2, etc.
(as the reader may establish for himself).
These p values, which indicate which of the cosine functions fit, are called the
of the problem. The lowest P value (here 1t/ 2) is the most important
for long times, for they need an A function decreases less rapidly with
time than the A fllDctions belonging to the other f3 values. So, of all the possible
solutions (which, for example, have to be summed up with certain weighing
factors if also the initial condition has to be satisfied) for long times, only that
with p = 7t/ 2 remains :
exp (-:::)cos(;;)
(Hat plate) (III.30}
The proportionality constant (which is of the order of 1 and here equals 4/n)
cannot bC found in this way. A complete analysis is necessary (it is this weighing
factor which reminds us of the previous rustory oft = 0). It is, however, sufficient
for practical purposes to know that it is almost 1 ; the cumbersome full analysis
is seldom required. For times in the range oft = 0 (Fo > 01) it yields a poorly
converging series of cosine terms as solution. A better usable solution for short
times is .already available (see the previous paragraph).
For longer times we are also often interested not so much in the actual tem-
perature distribution as in the heat transfer coefficient ex, which we define
on the basis of the mean temperature difference :*
cx(T
1
- ( T )) = Q>'fi = -A.dT
dx x=R
From equation (III.30) we find for the temperature distribution :
d T ( 1t
2
at ) tt . 1tX
dx = - C t(Tt - To} exp - 4R2 2R sm 2R
and at x = R :
(1Il.31)
(111.32)
(IIL32a)
The mean temperature ( T ) is given by:
(T) = s: { T, - C,(T, -To) exp (-:::)cos;;} dx
which yields for the mean temperature difference :
( T)- 7; = -C, - T
0
)exp (-
Thus we find with equation (III.31):
or:
A._::_
2R n
2
).
et.=--= - -
2 4 R
cx2R 1t2
Nu = - = - = 4-93 (Fo > 01)
l 2
163
(IIL33)
(III.33a)
(IIL34)
(ill.34a)
Due to the fact that for Fo > 01 the shapes of the temperature distributions
do not change with time, the Nu values become constant. The value of the
Nusselt number depends only on the form of the body. For a flat plate we found
Nu = 493 and for a sphere Nu = 66. For bodies which can be enclosed by a
sphere we can estimate the Nu numbers by applying NUsphcre = a.DtfA. = 66
and taking for D
1
the diameter of the sphere having the same volume as the
body. For a cube (with side D) we find:
V = D
3
= - D
3
Nu =- = -D = 66 - = 53
6 1 cube .. ). Dl 6
(exact solution 49). For a cylinder with L = D we find analogously:
Nu
1
= 6-6
3
{f 58
I cy \}3
(exact solution 5.6). Since for Fo > 01 the-Nu anaa values .are constant a heat
balance over the total body becomes:
d (1)
pc P V dt = et.A(T
1
- ( T ))
(IIL3S)
Integrating, with ( T) = T
0
at t = 0, we find :
In T1 - ( T) = _ = _aD _i_t = _ Nu
T
1
- T
0
pcPV A. pep VD D
(III.36)
This solution is identical to the relation found with equation (JII.33a).
Both relations show that a plot of the relative temperature equalization
164
1-0
I
..._-
-- .......
-- -..
..._
......
I
-Fo=at/0
2
Figwe JD.8 Mean temperature during non-stationary heat con-
duction d..
( ~ - ( T))/ (T
1
- T
0
) against Fo = ! tj D
2
should yield a straight line for
Fo > 01. This is indeed the case, as shown by Figure lll.8. An analogous graph
can be produced for the relative temperature equalization in the centre of the
body, as shown in Figure 111.9.
Since 'H = a(T
1
- ( T )) they-axis of the first graph also represents:
-1-
)J T
1
- <T> 4 > ~
---=--
T
1
- T
0
cr(T
1
- T
0
)
(III.38)
Figure 111.8 can also be used to find the heat flux into the medium at any time.
Il/.2.3. Influence of an outside heat transfer coefficient
In paragraphs Ul.2.1 and III2 ? we have considered cases where the beat
transport outside the medium is very big (au- oo) compared with the beat
conduction in the medium. This situation is, for example, present if a cold piece
._o
I
... -
Flat plote
Square cyl inder, L :oo
\---\--Cylinder, L "00
Cu be
\--\-___.:lr----1\--- C yll nde r, L = 0
\---+---+--- -+----+--Sphere

0 0 1 0 2 0 3 0 4 0 5
Fo-=at (D
2
Figure Ill.9 Centre temperature during
conduction
non-stationary heat
165
of wood (poor heat conductivity) is -placed in a steam atmosphere (good heat
transfer) with a saturation -temperature T
1

The opposite situation is realized if a piece of metal (good heat conduction)
of temperature T
0
is placed in air (poor heat transfer) with a temperatU:re T
1
.
In this case the heat transport outside the medium is so slow that temperature
equalization in the metal is practically complete ; the temperature T inside
the medium is the same in all places and the rate of heating or cooling is in-
dependent of the heat conductivity of the material. A macroscopic heat balance
for this case leads to an exponentially decreasing temperature equalization like
equation (111.36), but with the outside beat transfer coefficient a:" instead of ex:
cxuAt
---
pcpV
(III.38)
166
If both heat transport resistances, the outside (1/ cx
11
) and the inside (1/ <Xi "' Df J.),
are important, it is possible to construct graphs like Figures IIL8 and III.9
for a series of values ofcc.,. These graphs can be found in a number of handbooks
(e.g. VDI Wlirmeatlas). We can, however, approximate the exact solution by
assuming that also during non-stationary conduction the heat resistances
are additive (as is the case during stationary beat conduction).
For the region with constant Nu values (Fo > 01; i.e. long penetration times)
we can then write instead of equation (11!.36):
-ln Tl - ( T ) = + = Nu + t }-1
T
1
- T
0
pcPV au D a
11
(III.39)
or, if we call/the function, shown in Figure 11.8:
ln Tt - ( T) = + -t}
T
1
- T
0
Ciw
(III.40)
Apparently, we can also use Figure II18 to solve problems with an outside
heat resistance if we use instead of at./02 the corrected parameter
at ( j 1
Dz 1 + cr: . !Xu)-
i.e. the times necessary to reach the same relative temperature equalization
with and without an external heat resistance form a ratio of (1 + a) aj : 1.
In equations (IIL39) and (IIL40) T
1
is the outside temperature and not the
surface temperature T! of the medium. Because of the resistance T1 is
lower than T
1
The surface temperature can be calculated from a heat balance:
(III.41)
Equation (III.41) appears to be in good agreement with the known exact
solutions. The heating or cooling times calculated with the above procedure
arc! within 10 per cent of the exact solutions. In the literature other cases are
also known where an exact analysis showed that place- or time-dependent
resistances in series are not exactly additive, but that only a small error is made
by assuming additiveness. We can therefore extend the approach applied to
cases where the internal heat resistance is time-dependent, i.e. to short pene-
tration times, F o < 01. The times needed to reach a certain temperature
equalization with and without external resistance form a ratio of(l + ( a.;)/'XJ: L
In this equation ( !X) is the mean internal beat transfer coefficient averaged
over the time of the heat transfer process te (equation III.27).
We can, of course, also use the relations :
_!_ = -
1
- + _!_ and 'li = U(T
1
- T
0
)
U ( a i) ctu
(III.42)
where ( o:i) is the mean inside heat transfer coefficient, in order to calculate the
heat flux into the medium.
167
111.2.4. Problems
1. A packet of butter of 250 g has a temperature of 20C. It is put in a re-
frigerator of 4C After 120 min the centre of the packet appears to have a
temperature of 6C.
(a) After bow many minutes (from the beginning) is this temperature
5C?
(b) How long does it take for the interior of a 500 g packet (equal in geo-
metry to the former) to reach a temperature of 6C?
Answer : (a) 160 min; (b) 190 min
2. Check that if two conducting media a and b with uniform temperatures
T: and 7b are brought into contact with each other, immediately a constant
temperature (contact temperature) prevails at the surface, which is
given by:
T, - 7;
T:: :- (Apcp)a
*3. A cooling drum (external diameter D
11
= 1m) has a surface temperature of
20C and rotates at an angular velocity of2n min- t. A bituminous product
with a temperature of 80C (at this temperature the bitumen is just liquid)
is brought as a layer onto the drum and after a half-revolution is scraped
off again. The requirement is that the temperature of the cooled bitumen
does not exceed 25C. Properties of bitumen: c, = 920 J/kg oc ; p == 10
3
kgjm
3
; J.. == 017 W/ m oc.
(a) What is the maximum thickness of the layer which still fulfils this
requirement under the con<litions mentioned?
(b) By which factor is the capacity of. the drum increased if the angular
speed of 6 tt/ rnin is chosen?
Heat exchange between the layer and the surroundings is neglected.
On cooling no latent beat is liberated.
Answer : (a) 22 mm; (b) 173
4. Objects are submerged in a hot bed to warm them up. The par-
ticles (diameter 1 mm, ). = 1 W/m oc, cp = 4 X 10
6
Jjm
3
0
C} remain
on an average for 03 son the surface of these objects and ca. 60 per cent
of the surface is occupied by particles. Calculate the heat transfer coefficient
between the fluidized bed and the objects. Assume that every particle
remains on the surface equally long, that the gas with which the particles
are fluidized does not contribute to the heat transfer and that the objects
are submerged in the bed much longer than 03 s.
2
oc.
5. Elsevier's Culinary Encyclopedia recommends the following boiling times
for obtaining hard-boiled, non-crumbly eggs from various birds:
168
Bird
Turkey
Goose
Duck
Chicken (category 2)
Guinea fowl
Partridge, pewit
Pheasant
Pigeon
Egg weight (g)
95
85
70
60
40
36
30
24
How can these data be summarized scientifically?
Boiling time (s)
850
780
620
500
420
390
350
300
6. A solid glass sphere with a diameter of02 m is s'uspended in a room which
at the beginning has the same temperature as the sphere. The ambient
temperature suddenly increases by l0C.
(a) If the heat transfer coefficient is a = 8 W f m
2
C, how long does it
take for the centre of the sphere t o adopt the temperature change for
i? And if a = oc?
(b) Answer the same questions for the mean temperature of the sphere.
Explain the difference between these answers and those to question (a).
In which case is the difference the greatest and why?
A.cJus = ().8 Wfm oc; aglass = 44 X 10-
8
m
2
/s
Answer : (a) 45 and 126 h; (b) 40 and 063 h
1
7. A pan with custard must be cooled from IOOOC to 20C. To this end the
CJ>u(i .. t "' pan is put in a sink filled with water of l0C to the same height as the custard
,,((i
1 in the pan. Water is added to keep the temperature around the pan at
nr.:.r <n 1)1 10C. There are two ways of cooling: .
Ci-t (a) Allow the pan with custard to cool gently.
ht1 (r, -<r) )Cb) Stir the pan regularly and use a pan scraper to scrape the custard off
the cylindrical pan wall and mix it with the bulk. Each part of the wall
- - <T))"

should be scraped once every minute.

:n hA
1
dt ehlculate the times necessary to cool the custard down to a mean tern-
-,, ;lpV ;t perature of 20C according to both methods.
\
AC. c 7 2
Data: Thermal diffusivity oftae custard is 10- m j s and .l = 05 W f m cc .
_ ;:{1) },A ] i. over the entire temperature range. The custard is so viscous that
' - -Fip;l t
0
free convection is fully suppressed. Heat transport through the
l<lj bottom is neglig1ble. The pan diameter is 20 em.
'I
!
Answer : (a) 146 min ; (b) 12 min
8. 1f water droplets (temperature 20C) fall from a given height onto a hot
plate, the droplets will no longer moisten the plate above a certain plate
temperature (Tp) (Leidenfrost In foJJowing table this
_ ) f A \ ! . - ,,.., :.4 !., Zr> :::c - ( s , J r t.
1
m (' -
r ( 7[' J 01.
1

'ol - --;--
VI - 1
.1
\.
169
temperature is given for four materials. Determine whether. this pheno-
menon occurs at a contact temperature characteristic for the droplet.
Material of plate ). [Wj m oq
pcp(10
6
(Jfm
3
C)]
r, r cJ
1 382 351 199
2 140
244 206
3
72 145 219
4 14
205 244
Answer: r;, = 190 + 2C
*9. Tins filled with spinach (cliameter and height 12 em) are put in a room
filled with steam of l20C to sterilize the spinach.
(a) The heat transfer coefficient of the steam to the tin is 10,000 W /m
2
C.
After how many seconds does the heat resistance in the tin determine
the rate of
(b) Make a (schematic) sketch of the temperature distribution in and
around the tin at the moment the temperature of the centre of the tin
begins to rise.
(c) If the temperature of the tins is originally 20C, for how long should
the tins be put in the room with steam; so as to ensure a spinach
temperature of at least ll0C everywhere in the tin?
Properties of spinach: A. = 07 Jf m oc; p = 1200 kgjm
3
; cP = 400 Jfkg oc
Answer : (a) 1 s; (c) 25 h
*10. Show that John took the right steps in the case of the empty apartment
discussed at the beginning ofthis paragraph = 10-
7
m
2
/ s).
(.){::, l l\'IU-1- HJ"2:/r
Comments on problems X "" {trdi x o-107 rn
Problem 3
The situation is as drawn in the left picture; there is non-stationary heat
conduction. The figure on the right shows the temperature distribution in the
film. x P "' 1 /. T- 1q
SO'"C
r. =SO"C
0
X .., 'J
p
- _L. - ---
--
25"C
= 0. 01
170
(a) The temperature equalization after a half-revolution (at time c = nf 2n =
t min == 30 s) is :
T1 - T,. = 20- 25 =
0
_
0833
T
1
-T
0
20-80
and Figure 11!.9 yields a corresponding value (flat plate, since heating only
on one side, 2d must be used instead of D) for the Fourier number of:
at at
Fo = 0-27 =- =-
D
2
4d
2
from which we find d = 22 x 10-
3
m.
(b) Fo stays constant, but c
2
= !t
1
; thus:
and the capacity :
Problem 9
The figure shows the temperature distribution in the spinach tins during
non-stationary heating.
(a) The heat transfer resistance in the tin after a short time is oj A-spinach, with
o = JMt. We can assume that the inside resistance determines the overall
heat transfer resistance as soon as :
which is the case after:
t = 1 s
f 1j =1 200C
1-
\
\ ,,
T. = zooc
0 I: 0

R 0 R
- x
(b) See the figure at t
2
.
...
171
(c) Temperature equalization :
and Figure Ill.9 yields :
and we find:
_T,.=,. l_-_T. ..:;..: m = Q.l
Tl
at
Fo = 009 = D
2
t = 2-Sh.
Problem 10 John in the empty
The penetration depth of heat into the asbestos plate is {) = fo, < 1 0-
2
m,
and therefore t < 5 min, meaning that somebody must have been in the apart-
ment less than 5 minutes before.
ill.3. Heat transfer by forced convection In pipes
It was near midnight when John entered his fiat. An open
poe of water was boiling in the kicchen. John rhought: ' The
pot has a diameter of 20 em and could have contained at most
6 I of waw. The boccom of the pot is covered with a 3 TTUn
layer of scale and the temperature controller of che healing
plate is pur at 15fr C.' He remembered che pages w come and
looked for other signs of whether somebody was at home or
not.
In this paragraph we will treat, subsequently, heat transfer during laminar
pipe flow and during turbulent flow and heat transport from one fluid to
through a solid wall.
1113.1. Heat transfer during laminar flow in pipes
If a fluid moves through a pipe as a plug the penetration theory discussed in
paragraph III.2 can be applied for calculating heat flows if one substitutes
t = xj(v). However, if .radial velOcity distributions occur, the penetration
---- --
X=O
X=L
Figure ID.lO Heat transport during laminar flow
172
theory cannot be applied as such, because the flowing fluid transports heat as
well.
Consider the situation shown in Figure IIl.lO. A viscous liquid passes as a
laminar flow through a circular tube. At x < 0 the liquid and the pipe wall have
uniform temperatures T
0
. At x 0, the temperature of the pipe wa11 is
Tw (Tw > T
0
). Because of this temperature gradient there will be a heat flux q
from the wall into the liquid by conduction in the radial direction. Since there
is also a temperature gradient in the x-direction, there will be conductive heat
transport in the x-direction as well and, furthermore, heat is produced by
viscous dissipation and heat is transported in the x-direction by the flowing
liquid An energy balance over a ring-shaped control volume 2rrr dx dr consists
of the following expressions :
conduction in radial direction: + qrlr21tr - + dr) dx
conduction in axial direction: + qxlx 2nr dx - qxlx + d:r 2nr dx
heat produced by viscous dissipation : -21t dr dxrrx ddxvx = 11 d
2
vx2tu dr dx
dx2
energy transported by fluid :
pcpvx(T- T
0
)lx21tr dr - pcPvx(T- T
0
)1x+d.r2nr dr
Equating heat input and beat output over the control volume, neglecting heat
conduction in the x-direction and beat production by viscous dissipation
(which is allowed in the majority of cases), we find:
or:
d(rq,.) . dT
0 --T-rpc v -=
dr P x dx
(111.43)
Applying Fourier's law (equation 111.2):
qr = - ). dT; d(rqr) = dT)
dr dr dr dr
wre find :rom equation (lll.43) :
d ( dT) dT
-.A- r- + rpc lix- = 0
dr dr P dx
(III.44)
DR wlocity distribution for laminar flow in a circular pipe is given by (equation
v, =2( v) ( I - ; : )
)
n
.173
Introducing this expression into equation (Ill.44) we .find:
d ( d T) ( r
2
) d T
-J. dr r dr . + 2rpcP<v> 1 .... . R2 dx = 0.
(III.45)
This differential equation (the so-called Graetz equation) has been solved
f9r a number of bound,ary For the conditions in Figure 111.9
the solution is, for great distances from the pipe entrance :.
Nu = T = 366 for <v>D
2
> 01
(III.46a)
and for the region near the pipe entrance:
( ax ) -t
Nu = T = 108 ( v)D2
ax .
for (v)D2 < 005
(IIL46b)
The mean Nu number over the entrance region is accordingly given by :
1 ixe ( ax ) -t ax
(Nu) =- Nu dx = 162 <

for <

< 005 (III.46c)
r:. Xe 0 ( 1 = _:. ) V V
U:z "' p< )j 1 , '1 )!
The term axf(v) D
2
is called the Graetz number (Gz); it has here the same
funcfion as the Founer number for non-stat10nary conduction In the entrance
region of the pipe (short contact time) the thermal boundary layer is built up
and therefore Nu "'x--t. At great distances from the entrance the temperature
distributions will not change with x and the Nu number becomes constant
The above Nu values and beat transfer coefficients are related to a mean
fluid temperature .(T) which is defined as being proportional to the heat flux
transported by the liquid. Thus:
1t rD/ 2
<T>
4
D
2
(v) = Jo 2nrv,T,.dr . .
(III.47)
This is the temperature we would measure if the fiow was collected in a vessel
and mixed (mean cup temperature).
Equations (Ill.46a) and (lll.46c) can also be written as:
( Nu) = 162 Ret Prt (;; r t; ( Nu ) m; = 366
as the reader may check. The reader may also check tha1 for laminar flow the
Nu numbers for pipes of varying diameter and the same length are equal if the
volumetric flow rate through the pipes is equal. The ratioS' of the heat transfer
coefficients are then ff. Ifoc
2
= D
2
j D
1
, whereas the total beat fiows in both pipes
are equal.
The Prandtl number is defined by Pr = vja = Cp7f/A and represents the ratio
of the rate of momentum and beat transport.
For other types of velocity distributi ons and other pipe geometries tempera-
ture distributions and heat transfer coefficients have also been calculated
174
theoretically. The results can alwa)'S be presented by an equation like equation
(Ill.46b), but the numeric constants change with the geometry of the flow and
the rheology of the fluid Generally, we can state that the greater the velocity
gradient near the wall the greater the heat transfer coefficient will be. For that
reason, under laminar conditions non-Newtonian (pseudoplastic) liquids often
show higher heat transfer coefficients than Newtonian fluids. Measurements
about temperature distributions and temperature equalization in laminar flow
are scarce. Practical experience indicates that, especially in laminar flow, the
temperature dependence of the physical properties of the fluid (especially
viscosity) has great influence on the rate of heat transport. During heating up
of a liquid the warmer liquid near the wall will have a lower viscosity than the
liqutid near the centre of the tube. Since the shear stress distribution is fixed.
(see paragraph II.1.2), this means that due to the heat transfer process the
temperature distribution will be flattened (more uniform). This increases the
heat transport.
Differences in specific gravity caused by temperature differences may cause
extra flow (free convection) and also tend to increase the heat transfer. An
experimentally checked relation for the heat transfer during laminar flow of
Newtonian liquids is given by Siedt;r and Tate as :
<Nu> = 186 Re Pr( t (111.48]
( Nu)min = 366
where 11 is the viscosity of the fluid at mean fluid temperature (7: - T
0
)/2 and
YJ,., the fluid Viscosity at wall temperature Tw.
nept ll/.3.2. Heat transfer during turbulent flow
nNt ftY' The bestknown empirical correlation for heat transfer during turbulent flow
u. re in circular pipes is: f
7 h l:z) Nu = = 0.027 Re
0

8
{IJI.49)
'I 1- 'lw
T-.. 11 gi 'IU'l , ehl 1
which is valid in the region 2 X 10
3
< Re < 10
5
and Pr 0-7.
We can gain some insight into this type of relation by using the boundary
layer concept (Figure 111. 11). For turbulent flow in pipes the total velocity
gradient can be thought to be located in a layer of thickness lJh at the wall (see
paragraph 11.2.1)- The thickness of this hydrodynamic boundary layer is
defined by (equation 11.24):
( v) j
! w = Yf (>b = 2 p(v)2
which gives for the ratio of pipe diameter to hydrodynamic boundary layer
(equation 11.25):
D f
- =-Re
bh 2
V=O
<V>
- - -v
{T)
T=Tw
---T
Fiaure lli.ll Boundary layers in turbulent pipe flow
175
An analogous model can be set up for the temperature distribudon by
assuming t hat the total temperature gradient is located in a layer of thickness {>T
So we can write :
or:
(III.SO)
We have already found (paragraph Il21) that the ratio of hydrodynamic
boundary layer to thermal boundary layer is given by (equation II.29):
(Jh = Prt
br
Combined with equation (III.SO) we obtain :
Nu = f Re Prt
2
(III.Sl)
and using the Blasius equation (11.37) to substitute for the friction factor we
finally obtain for turbulent flow in pipes with smooth walls:
Nu = 004 Re
0
.
75
Pr+
which is quite similar to equation (III.49) if we add again the Sieder and Tate
During turbulent ftow there is also an entrance region where higher heat
transfer coefficients occur. We can c9rrect for this effect by multiplying equation
(111.49) by (1 + 07 D/L). Since for most practical heat exchangers L D, this
correction term is rarely necessary, however. Equation (IJT.49) will yield
approximate values for heat transfer in pipes of other than circular cross-
section if instead of the pipe diameter the hydraulic diameter of the channel
is used
176
If a circular pipe is wound to a spiral (diameter 2R) the heat transfer co-
efficient compared to that in a straight tube is increased by a factor of:
( 1 + if 10 < < 200 (III.52)
due to the occurrence of systematic eddies.
III.33. Partial and total heat transfer coefficients
Heat must often be transferred from one flowing medium to another, which
are separated from each other by a wall. We then have (e.g. as shown in Figure
III.l2) cx
1
, a wall with thickness t4 and conductivity A,.,, moreover possibly
deposit or dirt layer with dd and .A.d and finally cx
2
The overall heat
coefficient U can now be formally defined according to :
Q,, .
., iTDl l/J':I = U{(T1 ) - ( T2 )}

, _ T ..{III.53)
Ul 1il -Tor
1 r R
-
---
1
h UJ1
of di r t
h
Fl uid I Wa ll Flu!d 2
._ "" dw A\ j. T
- 'l'-!1 -
r,.J 7-
) ,
) -
)j
II -
",; P.lil
_! _
al
x,..
Tor ..
To/l-
A5 az
tr ( I 1 ( Ti
I:J ha
t -1 dd
h4 t('dA,
:=
A\ -\ Ao
----- :2

AL- Av
<T2

--
1--.
) - - - - --- -
1--- 1- -

Ao

Lli
K
Di st ance
..
1T
=
Llr 1
Figure ID.12 Heat resistances in series
- -
x 1JCJ

where < T
1
) and < T
2
) are the mean temperatures of the flowing liquids. 1 ust as
_, with heat conduction through various solid layers {Figure III .1 ), here in the one-
dimensional case the total resistance 1/ U equals the sum total of the partial
resistances. Thus :
(III. 54)
Heat flow - to:
1
from:
Gas (free convection)
~ = 5- 15
Table 111.4 Survey of heat transfer coefficients (W/m
2 0
C)
Gas (free convection)
~ = 5- 15
Room/outside air
through glass
u = t- 2
Gas (jl owing)
IX= 10-JQo
Liquid (free convection)
C( = 50-1000
Liquid (flowing)
water
(x = 3000-10,000
other liquids
(J. = 500-2000
---------- ---------
Superheaters
u = 3-10 .
Oven
--------------'------- ... .. ---- . --------
u = 10-40 +
{X= 10-100
Liquid (jree convection)
(): = 50-10000
Liquid (flowing)
water
0: = 3000-10,000
other liquids
0: = 500-3000
.,
Heat exchangers
for gases
u =: 10-30
--------- --- -'-- ---
heating
u = 5---15 .
Gas coolers
u = 10-50
Gas boiler
u = t0-50
~ - - - - - ~ - - - - - - - - - ----
Oil bath for heating
u = 25-500
Heating coil in vessel
waterfwater
withO\!-t stirring
u = 5o-250
with stirring
u = 500-2000
Cooling coil
u = 500-1500
if stirred
Heat exchanger
waterfwater
-u = 9oo--.zsoo
water/other
liquids
u = 200-1000
-------------------..,.....-----,...------------ -------
Condensing vapour
water
0: = 500(}-30,000
other liquids
0: = 1000-4000
u = 5-20
Air heaters
u = 1()-50
Steam jackets around
vessels with stirrers.
water
u = 300- 1000
otlier liquids
u = 150--500
Condensers
- .steamjwater
u = 1000-4000
other v,apour/water
iJ = 300-1000 :
Boiling liquid
water
C( = 3 5 0 0 - ~ , 0 0 0
other liquids
C( = HJ00-20,000
Boiler
u = 10::-40 +
Evaporators of cooling
units, hrine coolers
u = 300-1000
Evaporators
steamfwater
u = l5()(H)()QO
steam/other liquids
u = 300-200()'
178
When the various surface areas are not of the same size, e.g. in the case of thick-
walled cylinders, the varying size of inner and outer surface areas should be
taken into account.
The value of U is mainly determined by the greatest resistance in series-by
the lowest value of oc or ).j d. In metal apparatus mostly ).wfdw oc (A. steel
45 Wfm ac). However, if a dirt layer is present, A.Jd
4
can have a considerable
influence on the overall U (e.g. in water-cooled heat exchangers A.
4
jd
4

03/ 5 x 10-
4
= 600). The partial heat transfer coefficients increase in the order
gas (free convection), gas (forced convection), liquid (free convection), liquid
(forced convection), condensing vapour and boiling liquid. Table III.4 gives a
rough survey of values for the partial heat transfer coefficient occurring
practice and the consequences for the overall coefficient if two partial
are combined.
In the case of laminar flow equation (III. 54) yields only approximate results.
For an exact solution ofthis problem equation (Ill.45) should be solved applying
the correct boundary conditions.
III.3.4. Problems
L A very long horizontal heat exchanger for warming up liquids consists of
internally smooth pipes which are kept at constant the flow
through the pipes is turbulent
(a) If the pressure drop is doubled per unit of length of the pipes, by which
factor does the heat transfer coefficient increase?
(b) By which factor should the pipe length be increased in order to attain
the same heating up as in the former case?
Now answer the same questions for a laminar flow if the entrance region is
negligibly short with respect to the total pipe length.
Answer: turbulentflow (a) oc 1-39 times as large
(b} L 1-08 times as large
Laminar flow (a) a constant
(b) L 12 times as large
*2. On the outside of a steel tube (outside diameter 1 in, wall thickness
3
1
2
in)
steam condenses, whereas there is a turbulent fiow of cooling water at a
rate v through the tube. From measurements it is found that on variation
of v the total heat transfer coefficient U can be represented by :
1 0004
- = 0001 + - -
u vos
where U is expressed in BTUj ft
2
h oy and based on the external pipe dia-
meter ; v is expressed in ft/s. Using the same tube some time later the same
experiments are carried out and it is then found that :
= 00025 +
0
"
004
U vos
179
The difference should be ascribed to a deposit layer on the cooling water
side. Determine the heat resistance caused by the condensate layer, tube
wall and dirt layer and a relation for the a on the cooling water side. Find
out whether the latter relation agrees approximately with the existing
empirical correlations for this case. The ..l. of steel is 26 BTU/ ft
2
h F.
Answer : dirt layer 00015 ft
2
b FJBTU
tube wall 00001 ft
2
h FJBTU
condensate layer 00009 ft
2
h FjBTU
ex = 250 v
0

8
; agrees within 10 per cent
3. For a completely turbulent gas flow (Re = 10
6
) in a rough pipe (D = 0.10 m,
relative roughness 10-
2
) the following relation is valid between the N u
number and the friction factor:
Nu = f Re
2
Calculate a., if the heat conductivity of the gas is 0.02 W / m oc and of the
pipe material 200 W/m oc.
This rough pipe (wall thickness 2 mm) is surrounded by a steam jacket
which keeps the temperature of the outside at 120. The inlet temperature
of the gas is 20C, the outlet temperature 30C. What will be the outlet
temperature if the gas throughput is doubled, and by which factor is the
total amount of heat, which the pipe transfers per unit of time, increased?
Answer : ex = 950 W jm
2
oc; 4>w = twice as large; ~ u ' = 30C
4. On the outside of a tube a hydrocarbon is evaporated. The heat necessary
for the evaporation is supplied by condensing steam in the tube. During
this process a thin layer deposits on the tube wall. Measurements have
shown that the amount of deposit is proportional to the total amount of
heat transferred. Prove that the heat flux at time t is given by:
A-" ( ) {( 1 d
1
1 )
2
2c ~ T t}- t AT
'fJH t ~ - +- + - + -.,------- u
~ s .A.l < X ~ t ).z
where :
a. - the partial heat transfer coefficient on the steam side,
d
1
= thickness of the tube wall,
i.
1
= heat conductivity of the tube wall,
cxk = the partial heat transfer coefficient on the hydrocarbon side,
c = volume of deposit per unit of heat transferred,
).
2
= heat conductivity of the deposit,
ll T = the difference between the mean temperatures of the steam and of
the hydrocarbon.
180
5. A heat exchanger must be designed for heating oil from 40 to 60C. The
beating medium is condensing steam of l00C. For this situation we can
assume that the heat resistance is entirely on the oi1 side. The oil flow
through the pipe is laminar (Gz > 0-1). For an oil flow of 10m
3
jh.100 pipes
with an internal diameter of 2-2 em and a length of 1m are found to be
necessary.
(a) If 100 pipes with an internal diameter of 3-3 em are chosen, what should
be the pipe length?
(b) If 15 m
3
/h of oil must be heated in the original heat exchanger from 40
to 60C, what should be the pipe length?
Answer: (a) 1 m ; (b) 15 m
*6. For a .flow through a circular pipe (steady-state condition) the velocity
distribution and the temperature distribution are measured to be parabolic
functions of the radius. Calculate the Nusselt number for this flow.
*7. Show that John took the right steps in the case of the boiling water discussed
at the beginning of this paragraph.
Problem 2
The overall heat transfer coefficient is given by :
1 t 1 ds d4
- =-+-+- + -
u rxt' rx.., ~ ~ s A..,
For the steel wall we find d
3
j).
3
= 00001 ft
1
h oF/BTU . . J"l:te first measurement
(dd = 0) yielded:
1 - 0-004 1 1 d
5
- = 0001 + -- = - + - + -
U
08 '
V !Xc !Xw As
Now only ~ w is dependent on the fluid velocity. Thus!Xw = 250 v
0
'
8
BTU/ ft
2
b
0
F.
Further:
i.e. :
..:_ = 00009 ft
2
h F /BTU.
o:c
The second measurement analogously yields :
dd = 00015 ft
2
h F!BTU
;(d I
181
We can calculate ocw with equation (111.49) as :
a.- Pr = 41BOv'wtm' c
= 283v
0
'
8
BTU/ ft
2
h oF
Problem 6
The velocity distribution of the flowing fluid is given by
V - Vm., ( 1 -
We now know, furthermore, that the temperature distribution is parabolic,
i.e. T- r
2
j R
2
Defining that T = Tw at r = Rand T = T.. at r = 0, the tempera-
ture distribution is therefore given by:
T- T;
__ .;:_ = 2
T.,.,-1; R
..____ ..
The mean temperature of the flow in' the pipe can be calculated from:
to be:
The heat tlux through the pipe wall is now given by :
lPHir=R R = i.(Tw- = x(Tw - ( T) )
from which we find :
N = C(2R = 4(Tw -:- 7;.) =
5
u A. Tw- <T >
182
Problem 7 John and the boiling water
John assumed that aD heat transfer resistance is caused by the layer of scale.
Thus :
4>H = ~ ~ T = 157 w
4>
- 4>n - J.A l!:.T - .e.g o- s kgj
--- -\r x l s
Ill !1H d!l.H
In 4 hours. therefore, only about 1 kg of water will evaporate.
m.4. Heat exchangers
John had just poured out a cup of coffee and filled his pipe
when his boss asked to see him for a j ew minutes. ' HeW, John
thought. 'the coffee will be cold before I am back.' Remember-
ing the pages to come, John put milk and sugar in his coffee
and went to see his boss.
Heat exchangers play an important role in technology, not only for condition-
ing liquid ftows but also for attaining a favourable heat economy. On the basis of
tbe demands made on the heat exchange (the amount of heat rPH which must be
t ransferred per unit of time and the available temperature difference) and of
data on the beat resistances, the required exchanger area A of a heat exchanger
can be calculated from the formula:
l/>
8
= U A ~ T
(111.55)
Usually both U and ..1 Tare dependent on the place in the heat exchanger, so that
the formula for the entire heat exchanger should read :
(111.56)
We have already discussed the determination of the overaU heat transfer
coefficient U for simple pipe flow (paragraph Ill3) and we will return to this
subject in later paragraphs (see also Table lll.S). Here, we will find out how the
mean temperature difference can be determined in a number of simple cases
and we will consider the problems which have to be solved when designing heat
exchangers.
111.4.1. Determinacion of mean temperarure difference
Let us consider a simple model of a heat exchanger- two concentric tubes
(Figure III.13). Here, for example, liquid' gives up beat to liquid", either co-
currently or countercurrently. In the former case TIL should invariably be
~ T{ . In the latter case the outlet temperature T ~ can also be higher than T ~
ln the first instance a heat balance can be set up over the entire heat exchanger,
which shows that in the staionary state just as much heat is supplied as removed.
If tne heat exchange with the surroundings can be neglected and no heat is
T il
L
Liquid
1
- _ r. '
.P 'm' To'--.Noc---! -:;I ::::::::::::::::::::::::X::::_.-7,7"!- - L
c:----_..:...__----Jil!
I 1 JJ I
I Liquid I
x=o x=L
c" r,lf
"'m ' p t o
0 -x L 0 -x L
Figure Ill.l3 Heat exchanger and temperature gradients for co- and
countercurrent flow
produced in the apparatus, the heat balance will read:
183
{
+ countercurrent
- cocurrent
(lll.57)
This relation is valid for any heat exchanger which satisfies the above conditions*
and is indispensable for solving a heat exchanger problem.
In order to calcula.te the mean T ovei'the entire apparatus in equation
(II 1.56), we should first know T' and T" as a function of the place. For the type
of heat exchanger in Figure III.13 we .make a heat balance over a small part dx:
-(</Jmcp)' dT' = +(cJ>,.,cP)" dT" = dl/>H
Then, according to equation. (III .55), the following applies:
dn = (T' - T")US dx
(III. 58)
(III.59)
where S = circumference of the exchange area. After introduction of =
T' - T" we find from equation (II1.57):
l\TL - l\1;, = -</J"[(:c,)' + (m
1
c.Y']
* If heat exchangers are considered in which phase transitions also take place (evaporation or
condensation) the heat balance should be given a more general form:
- = - H'L) = >
8
where H = the enthalpy per unit of mass.
184
and from equation (111.58):
d T ' - d T" = d!l T = de/> H [ l + l J
(cf>mcp)' (cJ>mcp'f'
Elimination of the term in brackets and of dc/>H using equation (IIL59) finally
gives the following differential equation f o r ~ T:
d6 T = (A 1i. - .1 T
0
) S
1). T u H dx
(II 1.60)
~ o r a few special cases the integration can be carried out easily.
(a) If U is independent of x or T (often the case with gases). Then integration ~ f
!!.TL- ~ T o
</>y = USLinf!..TJll.To = USL(!!.T'hog (IIL61)
10
2
9
e
7
6
5
4
3
r
2
1 5 ~
10
9
8
7
6
5
4
3
2
15
ld! 8 6 5 4 3 2 15 10 8 6 5 4 3 2 1 5
l!. To[oc] ---
Figure DI.J4 Logarithmic mean temperature difference as a function of ~ T,
and LlT
0
()
0
1-..J
<J
185
We see that for this case T in equation (111.56) equals the logarithmic
mean of TL, both cocurrently and countercurrently.
Figure II I .14 gives a simple graph which can be used for a rough estimation
of the logarithmic mean temperature difference. From this figure we find, for
example, for ATL = 40oC and AT
0
= 100C that (A 11toc = 22C.
{b) If U is dependent on the temperature (hence also on the place). In firs t
approximation U can be assumed linearly in aT:
U = Uo + (U L - U
0
)(AT- AT
0
)/(11TL - AT
0
}
Substitution in equation (III.60) and integration between 0 to L gives :
<f>H = SL U L AT
0
- U
0
!:iTL (III.
62
)
ln ( U L !:i T
0
/ U
0

So here the product U!:iT is used instead of !:iT in equation (111.56). If
the value of U depends on a more complicated way, equation (111.61}
can be applied to a number of sections of the heat exchanger or the integra-
tion can be carried out graphically. Cases where this is necessary are, for
example, those where combined condensation and after-cooling or pre-
beating and evaporation occur. During the transition between the two
mechanisms the value of the transfer coefficient U may change rapidly.
Finally, we have to add that most heat exchangers applied in practice do not
belong to the simple 'single pass' type discussed here but to the 'multiple pass'
principle illustrated in Figure III.l5. Also, for these apparatus a mean tempera-
ture difference can be calculated. In practice, however, we prefer to use the
r"
L.
r"
0
I

ro'
-L
r"
0
Figure (11.15 Principle and temperature gradients
of a multiple pass heat exchanger (one shell, two
tube passes)
186
logarithmic mean temperature difference of a countercurrent heat exchanger
extended by a correction term lj;:
ATL - llT
0
I ll1'c.
n llTo
Values for the correction fa-ctor 1/1 can be obtained from engineering handbooks
where graphs like Figure lll. l6 can be found for various situations. These
graphs show t/1 as a functi on of the ratio X of the temperature drop of one
fluid and the maximum temperature difference, (T. - T
0
}/(T
0
- T
0
), and of
the ratio R of the temperature changes of both streams, (T
0
- T{)/(TL - T ~ J - .
In order to ensure a high efficiency of a multiple pass heat exchanger, we should
try to keep t/1 > (). 7 5.
Figure 111.16 Correction factor t/J for multiple pass heat exchangers (one
shell, even number of tube passes) as a function of R and X
Jl/.4.2. Height of a transf er unit
If we consider a heat exchanger with constant wall temperature Tw we can
set up the following heat balance :
or :
</>n = UA To - ( T ) = 4>rnc
9
(T
0
- ( T ) )
To -T.
In ( T ) - ~ w
T
0
- Tw uA USL
ln ( T ) - T.., = --mc-P = --mc-P
i.e. the mean temperature difference b etween the fl uid and wall decreases by a
factor 1/ e ( =0-37) over a length Le given by:
L = 4>mce
e US
(I Il.63)
187
This length, which is characteristic of the heat transfer process, is called one
HTU, height of a transfer unit', and is used in many books on chemical engineer-
mg.
We will use the concept of HTUs here to show how we can allow for fluid
dispersion when designing a heat exchanger. If fluid dispersion occurs in a
heat exchanger, extra heat transport in the direction of flow will be the con-
sequence, and the actual temperature gradients over the length of the exchanger
may differ from those shown in the previous section (where dispersion was
neglected). This is illustrated by Figure U1.17 which shows the temperature
gradients over the length of a heat exchanger (like Figure III.13) for cocurrent
plug flow and for dispersed flow of liquid. It will be clear that in the latter case
the mean temperature difference between the two fluids is smaller and, con-
sequently, a larger surface area (or greater length) of beat exchanger is required
for the transfer of the same amount of heat. We can easily correct for this effect
of dispersion by using instead of the height of a transfer unit HTU the HETU,
the height equivalent of a transfer unit which is given by the sum of the HTU
and the height of a mixing unit HMU:
1
....
HETU = HTU + HMU
r. ' No di spersion
0
, I
........
......
.......
With di spersi Ofl
\
__ \
-
L
Figure ID.17 Temperature
dispersi ori
T '
L
(III.64)
The height of a mixingunit can be determined by the relations given in para-
graph 11.8.5. From what is stated there we can conclude that for turbulent
flow and for flow around obstac1es HTU HMU (i.e. dispersion can be
neglected) but that under laminar flow conditions a correction for the effect of
dispersion can be necessary.
1 I /.4.3 Design of heat exchangers
With the knowledge now available it is possible to determine the main
dimensions of a heat exchanger. lf the overall heat transfer coefficient, aB
temperatures and the fluid flows are known. we can easily determine the total
surface of the exchanger required.
The task remains to determit}e the number N, diameter D and length L of the
pipes as well as the diameter of the shell Ds and the geometric pattern of the
r--"'
Pressure index (
0
/o)
N
E
'
Shell Tube
-
20 atm 6 2
D
1..-J
30 aim 16 7
d)
.2
40 atm 29 14
...
0.
60 atm 44 23
0
1'-
0>
Shell material Index (%)
304 ss 9
316 ss 12
C r ~ Mo
5 C r ~ Mo 4
Ket tle or bellow + 16%
Length lndel'. (%)
I 2m 4 i
1 6m 27
24rn !2
30m 9
3 6m 5
42 m 2
0
- 3
10 2.0
Total oreo [m
2
]
Tube material
~ ~ - - - - - - - Hastelloy C
50 100 200
304SS
5Cr ~ Mo
cs
Figure m.t8 Costs in 1970 of standard carbon steel shell and tube heat exchanger (3/4 in diameter, 15/ 16 in
pitch, L0/10 atm) in Dfl/m
2
for all materials; cost index CS 65 per cent/year, all others 45 per centjyear-
-
00
00
189
pipes in the shell. For the solution of this problem two momentum balances
are available which restrict the applicable velocities, diameters and lengths
due to limits put on pressure drops in the pipes and in the shell. Furthermore,
we have an economic balance which is used to find the cheapest design for
a given problem. Once we have -chosen the type of heat exchanger these three
balances determine N, L and D, and geometric considerations then further
determine shell diameter, etc.
If we realize that, besjdes the solutions of the above problems, questions
on construction material and a great number of constructional details must
be solved, we can understand why big engineering firms often have departments
specialized in designing heat exchange equipment.
For a rough estimation of the costs of a heat exchanger Figure II1.18 can
be used This figure shows, for instance, that a heat exchanger of 100m
2
with
stainless steel 316 pipes and aSS 304 shell, suitable for 300 psi in the-shell and
600 psi in the tubes, and with a pipe lengtb of 10 feet will cost per rn
2
surface
in 1973:
445 + (9 + 6 + 14 + 9 + 15)% =Oft. 683/m
2
Finally, in Table III.5 some heat transfer correlations are summarized,
which are useful when designing heat exchangers. A few of these relations have
already been discussed in paragraph 111.3 ; others will be treated in following
paragraphs.
III.4.4. Problems
1. Through a horizontal smooth pipe (internal diameter 18 mm) water of
90C is forced up at a rate of 08 m/ s. Determine the distance from the
entrance at which the water begins to boil if the pressure drop is negligible
and the pipe wall over its entire length is kept at l10C.
Answer : L = 15 m
2 A closed vessel filled with 86 1 of water which is stirred thoroughly bas a
temperature of l00C whereas the air temperature is 20C. If the total
outer surface area of _ the vessel is I m
2
and the mean heat transfer co-
efficient U over this area is 25 W fm
2
oc, how long does it take for the tem-
perature to fall to 50C?
3. In a double-tube beat exchanger benzene is heated from 20 to 60oC counter-
currently with water, which during this process cools down from 88 to 48C.
During an overhaul the direction of the water flow is accidentally reversed.
Determine the final of the benzene. (The beat transfer co-
efficient is independent of 6. T.)
190
Table IU.S Heat transfer correlations for heaL exchangers
Flow siluotioll Heat transfer correlation Rtmorks
-
Flow throu9fl S1r119ht
Re" < 2300. 2 5< Pr <4000
Physical properties ot thnds
circular pipes
I I I I
at <T> >

Re'j Pr! (hr y
z
0 .
r
<Nu>,., n3 66 f'or for spiralled
Ftc-., along c.rculor popes
2300<Re< Pr > 07
p1pe see equottcn II 52
I I
without bofflts

OOa7 Re
0
" Pr
3
(.!. )
7
' "
Flow o!Qng circular
dp 4
n number of pipes
pipes witt. bofflts
2 2
3 < Re -- < 2 lliO 0 5 < Pr < 500
<A>., ,
r,.+ Tovt
s
OptniiWJo mi'm Physico properties 01
ui
-o. ... fm
f)"' 0 25
--2-
I !
Mton orto ovolloble for flow
T
<Nu>" <";fl Q-22 Reo.n Pr
3
(t )'
I
-'s j L {Sd r P( 2 2 }!
:
p
<A> -;s!O d l 4 Ond lo
Flow along flot plate
Re = < 3 x 10:1
s equotton em 65)
Cone 5idt)
I I
033Re
2
Pr3 Physical properties at
7in+ Tout
-<' ;;::::>

--2-
I I tO\$<Re <t0
7
L
I
<Nu> =0 037 Reo.e Pr
3
Plate heat

R
In prae11ce 4< a< 6 mm
f ---- .,
I I for rough es11male (within
<Nu>

c 'le
0
Pr
3
( .. )"
foetor 2 ) toke c OS ond
c 011<1 a <ltpendtnl an type and mal<e of plates
057
FCIII1119 film on won
Re .F<r.>CII < 1600
r't+ J i
/
' ' { } !
a ..
- l
\ om w,..zg
::::
<a>20 5 Rt
1
-;;z
0 2
a,.,,b 0 2 a tom
Re

-.! Re>3200
::
:f
I
! r.tr.
{ 2 3} s
Pr
3
Phvsl col properties at
:::
<> 00087 Re
0
'
4
2
.,.
surface htot
I
<a> I 13 t {).pt., nN}
If number at rows al s.craper blodes
..
N numbe r d revolut1ons
-. CofrtciiCn foet or i' I for water

02 < oy< Q-6 for viscous ll q11ids
8oollf19
su Figure m.24
*4. A turbulent liquid stream of 25C flows through a straight pipe, the wall
temperature of which is lOOoC. At 3m distance from the entrance the mean
liquid temperature is 50C.
(a) Determine at what distance from the entrance the liquid has a mean
temperature of 90C.
(b) To what degree does this distance change if the flow rate is doubled?
Answer : (a) 15m; (b) increases by 15 per cent
Table IlLS (cont) Heat transfer correlations for heat exchangers
Flow Slf\XIliOf'l
Notvr41 conveeri o,. Otoun4
honzontal pi pes
Natural COllvediOn on
venrcol swfoces
Ccndensotion en
nor tzontol prpe
Ccnd.nsotton
ver11col ptpe
Ccndei'SOIIOII 11'1 honzonlal
PIPe
Condepsation in
vertical pipe
Heo: tronsfer correlation
1000 < Gr Pr k &'
9
t:>p < tf?
...
I
<NrD <>-SS { Gr Pr}
4
l
1700<Gr Pt l g
11
' < Wle
""
I
Pr}
4
Gt Pr > i0
8
I
t3 { Gr Pt} )
I <a> o-n{t;/'13 AH}4
Dt}AT
Usi! correlation for eoPuenspt ioro on vertical p1pe
If considerable vop04.1r veloci ty,
Remarks
Physocal pn:per1res crt
r w11 + r b .. lk
2
of c ombined roatural /
fa tc.ed corovectrOt\ on < a > su
poroqroph m.s
of lnett CiQS oe
condensotron see m79
PllySICOI prOperti es of fh.ildS
ot T -1 AT
For quk est ornatton of
<a > su Fogure m.23
Index c "' condensot e
tndex c vopOUT
191
*5. A gas stream of 10,000 m
3
jh must be cooled from 200C to 50C. Use is
made of a bank of steel tubes installed in the open air. The gas flows through
the pipes. Water is sprayed onto the bank of tubes and wind blows between
the pipes. In this way the outer wall of each tube is kept at a constant
temperature of 20C. Design the most economical heat exchanger for the
job for the case where the pressure drop over the tubes must be less than
01 atm. State the length, number and inner diameters of the tubes and the
approximate 1970 price of the unit.
Data : aair = 10-
5
m
2
/ s
v. = 10-
5
m
2
/ 2
a1r
material of construction = stainless steel 316
Answer : L -. 4-8 m ; N = 87; D = 005 m ; Dfi. 33,00000
192
*6. Show that John drew the right conclusion regarding his cup of coffee dis-
cussed at the beginning of this paragraph.
Problem 4
A heat balance in the stationary state reads :
'"
1 L2
, . _ - - - - - - ~ - - - - ~ - - - - - - - - - -
<f>n ""' et.rr.D L A T;
08
= t/>
11
pc P tl T
(a) Writing the above equation for two lengths of pipe 1 and 2 we obtain
(cx
1
= et
2
, u. = r:
2
) after division of tbe equations:
L = L ll1;og I A'Ji = 149
1 tlll: AT. m
log2 1
(b) Now the fluid velocity v varies and with it 4>v "' v and a "' v
0

8
. Using the
same procedure we find :
Problem 5
L2 = Ll AT.og lllT2alt!Ju2 = 149 X 20 2 = 171 m
11 'lioa 2 f1 Tl CX2 tP v
1
We can set up the following heat balance:
. </>H = U A llTjog = ifJ.uPC P ll T
Assuming a
1
J .Jdw and applying equation (IIL49) we arrive at :
ND08L = vpcPtJ.Tvos = 385(ml8)
0.027v
0

8
An L\ 7;
01
Figure ITI.l7 shows that costs are minimal for L = 16 feet = 4-8m (length
of standard pipe). lf we choose now three practical pipe dimensions, we can,
with the help of the Fanning equation and Figure II1.17, calculate the following
table :
D(m) (}025 (}()5
010
N 158 87 51
v(m/s) 40-5 162 6-95
4f 0.019 0019 0019
6.p(N/m
2
) 3900 310 286
A (m
2
} 595 . 66 77
Cost (Dfl./m
2
)
520 500 470
Total cost (Dfi.)
31,200 33,000 36,200
193
We see that any diameter above approximately 004 m will satisfy the pressure
drop criterion. The smaller the pipe diameter, the lower the total cost. Thus
we would choose the smallest practical diameter above 0.()4 m, probably
D = 2 in = 0.05 m.
Problem 6 John and the cup of coffee
John wants to keep his coffee as bot as possible. If he added milk and sugar
directly, the heat balance:
Mcp(To- T) = Mscps(T- Ys) + Mmcpm(T- Tm)
shows thit the temperature of the mixture is:
T = McpTo + MscpsT;. + M'"cP'"T'"
McP + Mscps + Mmcpm
The further cooling of the coffee is then described by :
which can be integrated to yield :
T - 1'air ( UAt)
= exp --
To - Yair . McP
~
'
'
' , visit to bos-s
I
.........
.........
.....
.......
.......
-
J....

7
a l r
0
_,
The decrease in temperature is shown in the neighbouring graph. If John had
left the coffee without milk and sugar, the temperature difference, T - T,
1
p
would be bigger. Thus the total system would lose more heat and the final
temperature after addition of milk and sugar would be lower.
194
m.s. Heat t!'amfer by fOI'ced convection around obstacles
The body was lying aLa draughty corner. The police doctor
stated that theman could only have been dead for 20 minutes
and that death was due to a stab in the heart with a pQinted
weapon of approximately 3 em diameter. The murder
weapon had not yet been found. John noticed that the jacket
and shirt of the victim were wet. He estimated the air
temperature tc be 20 C and the wind velocity at the comer
tc be 8 mfs. He made a quick calculation and concluded that
the man could have been stabbed with an 'icicle.
l/1.5.1. Flow along ajlat plate
This situation is drawn in Figure 1!.37. A thermal boundary layer is present,
the thickness f>r of which increases in the x-direction. This causes the local heat
transfer coefficients to decrease in the positive x-direction. For the calculation
of the heat transfer coefficients we can apply the same technique as used
in paragraph 111.3 for the calculation of heat transfer coefficients for turbulent
flow in pipes.
We first calculate the thickness of the hydrodynamic boundary layer from
equation (11.64):
Using our knowledge about the ratio of the hydrodynamic and the thermal
boundary layer thickness (equation 11.29):
~ = ( ~ ) t = Prt
oT a
we can now calculate the Nusselt number as:
Nu = rtx = _:: = .:._ Prt = 0332 ~ Prt
). {)T bh -v --;
(111.65)
This equation is valid for Re = vxf v < 3 . 10s.. At higher Revalues the boundary
layer is of constant thickness.
111.52. Heat transfer to falling films
The derivation of a relation for heat transfer to falling films follows the lines
set out in paragraph III.3 when discussing laminar and turbulent pipe flows.
During laminar film flow the Graetz number is practically always:
ax
Gz = ., < 005 -
<v) D- .
i.e. the Nusselt number is constant. Under this condition we find for flow
between two parallel plates from an analysis identical to the one applied in
paragraph 111.3.1 to circular pipes :
(X 2<5
Nu=-= 376
l
Consequently. for a laminar film on a fiat plate :
Nu = rxDh = a4<5 = 2 x 376 = 752
l .A.
Substituting fJ by means of (see problem 2 paragraph ll. l):
0
= '{31jf. =
vi>iW pg
we find, after some rearrangement :
-
Thus the heat transfer coefficient to a falling film is :
{
p(v)4b} - t{gp2 J.. 3}t
a = 205
2
17 tT
for :
Re = p (v)40 = 4<Pm < 1600
Y/ w,., .
Equation (ll1. 66a) is in agreement with experimental results.
195
(III.66a)
Above Re = 3200 the film flow is turbulent and the heat transfer coefficient
can be described by the empirical correlation :
(J. = 00087 -

;._3} t{v}t
17 rf a
(IIL66b)
. f.I1.5.3. Flow around spheres and cylinders
During flow around cylinders, spheres and similar bodies a boundary layer
is similarly formed which is thicker as x (the distance measured along the surface
from the edge) is larger. As has been discussed when treating the fiow resistance
of bocties (paragraph 11.5.1), the boundary layer becomes detached somewhere
behind the greatest thickness of the body ; behind it there is the 'dead water,
with eddies. The mean heat transfer coefficient (a ) over the entire surface is
mainly determined by what happens at the flow side of the body. The relations
for ( a:) for such bodies therefore have a form which very much resembles
equation (II1.65). Of the great number of empirical rela tions for cylinders and
spheres the following are as a good approximation for gases and
liquids (Pr > 07).
196
(a) Long cylinders with circular cross-section perpendicular to the Bow :
(a)D
11
( y) 020 . _ (v,.Du) 050(") 033
). = 042 ~ , 0=>7 - v- ~ (1 < Re < 10
4
) (ITI.67)
(b) Flow around spheres:
( )D ( )O ls ( D loso( )033
a).
11
= 20 + 13 ~ + 066 Vrv
11
~ (1 < Re < 10
4
) (111.68)
In both equations the physical properties of the fluid should be taken at the
average film temperature unless the correction term (tt/ t7w)
0
.
14
extra term containing the Pr number is connected with the curved shape of these
bodies. At Re >ca. 10
4
the exponent of Re is slightly higher than (}5. More
detailed information can be found in various handbooks.
At Re = 0 we find for a sphere Nu = 2, the value for pure conductive trans-
port. At high flow velocities the convective transport is much greater than the
conductive transport. For this situation we can estimate the thickness of the
hydrodynamic boundary layer on the basis of the penetration of momentum
{paragraph 11.1.6) as :
O ~ r = foi
Further, reafuing that lJJ !Jr = Prt and that the timet available for creating the
boundary layer is r ~ Djv we find:
Nu = - -= - ~ 056 - Prt
rxD D ~ v
A lJT v
which is in complete agreement with equation (III.68).
The same argumentation applied to cylinders perpendicular to the dircx:tion
of flow will yield equation (III.67). For cylinders parallel to the direction of
fiow the boundary layer will have a constant thickness at any place and we can
apply the relations for heat transfer in laminar and turbulent flow in pipes
discussed in paragraph III.3. In the latter case we can apply equation {111.43)
~ w e use the hydraulic diameter D
11
instead of the pipe diameter D.
Tbe two given relations (equations III. 67 and III.68) are applicable in situa-
tiom v.here the heat transport by convection (in the x-direction) is much more
imponant than the conductivity in that direction, which means:
dT .d
2
T
pcpvr dx /, dxl
'I.e .. _
197
The left-hand side of this inequality is a dimensionless constant which is called
the Peclet number (Pe = The right-hand term is a value which follows
from the temperature distribution_ This value differs for spheres and cylinders
but is of the order of 1. So it could be said that the above relations hold if
Pe ::.::. 1.
Both equations must be applied with care at low flow velocities. In these cases
we have to check whether heat transfer by free convection participates sig-
nificantly (see paragraph II1.6}.
lll.5.4. Heat transfer in packed beds
The heat transfer between the fluid and the particles in a packed can
also be described with the relations for a boundary layer flow (equation 111.65).
provided the hydraulic diameter Dh is taken as the characteristic dimension
of the successive short channels of which the bed is built up (see paragraph
Il.6.1). The result for a bed of spheres is :
( a) Dh _ ( u. ) eDu _ ( 2v
0
Du )t t
). -
3
(l _ e)). - constant x
3
v(l _ e) Pr
{111.69)
For the constant for which approximately the value 066 is expected, a value
of (075 + 015) is found in the range 10 < Reh = 2v
0
D.,/3v(l -e) < 10
3
.
Many authors state that for Nu number of a packed bed (e = 040):
= (1-8 +
(III.69a)
applies, pro,ided 30 < v
0
DJv < 3 x 10
3
and Pr 1. The reader may ascertain
that this corresponds with the above more fundamental result So it is again
Reynolds number is used (just as in paragraph 11.6.1 for the pressure drop
relations in fixed beds the choice is: Rehor v
0
Djv, of which the first is fundamen-
tally correct)-
The heat transfer between the tube wall and the fixed bed (i.e. particles+
which, for example, is important for catalytic reactions in this type of
reactor, can still only poorly be described. The problem is that of the total
-temperature difference between the wall and the centre of the bed an appreciable
part is found in the outer layer of particles (thus at the wall) but that consider-
able temperature differences also occur over the diameter of the bed_ It is there-
fore obvious to try a model description in which two heat resistances occur :
one in the boundary layer at the wall and one in the bed itself. The resistance
at the wall {1/(a.)) can, in principle, be described with the same relation as that
between particles and fluid (equations III.69 or III.69a) because the-wall can
be considered as a (somewhat large and strange) particle. In practice this
appears to be the case, only the constant is ca. 30 per cent lower than in equations
(III.69) or (111.69a). The reason is that the mean dimension of the channels at
the wall is slightly larger than in the heart of the bed, where the particles lie
198
closer to each other. The resistance in the bed is usually described with the
aid of an effective heat conductivity. Regarding the heat transfer the bed
(particles + fluid) is taken as a solid body with a heat capacity equal to
(1 - + t(pcJnuid
and an effective ).en This heat conductivity is then, of oourse, dependent on the
porosity of the bed, the of the particles, the of the ftuid and of the flow
situation in the bed. A rule of thumb for the effe.ctive heat conduc-
tivity is :
(III.70)
where ;.:r is the effect ive conductivity of the bed if no flow occurs. It can be:
.calculated with the help of Figure llt19 in which J.:cr/A.r has been plotted against
J...,J A.r for various values of r..
-
* ';
-<
10 20 50 100 200 500 1000
Figure lll.19 Effective heat conductivity A:Hjn a packed
bed without flow
At high temperatures. the heat transport from particle to particle is controlled
by radiation transfer. This mechanism works parallel to the heat conduction
through the fluid Even then, Figure Ill.19 can be used provided the con-
ductivity ;., of the fluid is increased by 4 eaT
3
D
11
(the latter term can be under-
!11.5.5. Heat transfer in fluidized beds
The problem here is that the fluid usually moves through the bed in the form
of bubbles (see paragraph The contact between the fluid and the particles
is therefore poorer than expected for a bed in which bubbles are a bsent (par-
ticulate fluidization). Usually the transfer resistance between the bubbles and
199
their surroundings is greater than the transfer resistance between the particles
and the fluid that flows round the particles.
In the case of 'particulate fluidization' the situation is analogous to that in
the fixed bed. For not too high porosity (e < 0.75), when the channel model
can still be used the transfer between the particles and the fluid is described
with equation (III.69), whereas at higher porosity the particles ' notice' so little
from each other that the transfer is described with the relation of the single
sphere, provided the actual speed in the bed is used as the characteristic speed.
Also, the heat transfer between a wall and the bed can be described by the
considerations discussed for fixed beds, but owing to the mobility ofthe particles
the effective conductivity in the bed is so great that we may assume the tempera-
ture in the bed and outside the 'boundary layer' at the wall as being uniform.
The heat transfer coefficient at the wall is again given by equation (III.69) with
a constant factor which lies ca. 30 per cent lower than indicated there. Because
of the presence of the solid wall the channel model remains valid up to very
high porosities ( e < ().9).
111.5.6. Problems
* 1. A stirrer (power consumption 4 kW) stirs a fluid in a horizontal cylindrical
vessel (D = 1 m, L = 3 m) with an insulated bottom Air is blown at a rate
of 9 m/ s perpendicular to t b ~ axisof the vessel. If the vessel is completely
filled with water, how will the temperature of the water change with the
temperature if at t = 0, T = T
0
= 'Fair?
0
= 415 {1 - exp ( -98 x 10-
6
t)}
3
)
to fall from a height L through air. The temperature of the air is 20C.
What should L be to ensure that drops of 2 mm diameter reach the bottom
completely solidified? The drops reach the stationary rate of fall very soon
after being released The heat conductivity of lead is so high that in the drops
a uniform temperature can be assumed equal to the melting temperature
3
Jjkg).
3. Two equally heavy metal spheres of good heat conductivity fall stationary
through air. The ratio of the diameters of the spheres is 2: 1. With the help
of dimension analysis it can be proved that the product of the stationary
rate of fall and the diameter of the one sphere is equal to that of the other
sphere (vD = constant). lithe initial temperature of the spheres is T
1
and that
of the air Teo, calculate the ratio between the distances the spheres should
travel to be cooled to temperature T
2
The product of the density and the
specific heat is the same for both spheres.
Distance large sphere =
2
Distance small sphere
200
4. If a thermometer with temperature T
0
is suddenly put in a flowing fluid with
a temperature 1f, the mean temperature T of the thermometer will, in course
of time, approach the value T,. The time constant 1: of a thermometer is
defined as the time necessary to reach a value of (1f - T), which is equal to
(1f - T
0
)j e, where e = the base of the natural logarithms. If the ther-
mometer conducts the heat infinitely well, has an external surface area A
and a total heat capacity c (Jr C), and if the heat transfer between
the flow and the thermometer is o:, tben derive that :
c
7: =-

We now consider a thermometer consisting of a mercury-filled iron cylinder
(D,. = 20 mm. length L = 200 mm, wall thickness d = 15 mm) under the
following conditions:
(a) in air of 20C at a flow rate of 10 m/s, perpendicular to the cylinder;
(b) in water of 20C at a flow rate of 2 perpendicular to the cylinder;
(c) in condensing steam with a mean beat transfer coefficient (X = 15 x 10
3
W/m
2
C.
Calculate t in all cases, both for a clean thermometer and for a thermometer
covered with a 05 mm thick dirt .layer (A.dirt = 06 W / m oc). Neglect the
effect of the ends. The specific heat of Fe is:
that of Hg is :
cP = 13 x 10
2
Ji kgC
'
Answer : (a) clean: 1 = 150 s; dirty: -r:::;: 159 s
(b) clean:-r = 17s
(c) clean: r = 065 s; dirty: t = 88 s
*5. Show that John drew the right conclusion in the case of the wet body dis-
cussed at the beginning of this paragraph.
Problem 1
The outside heat transfer coefficient can be calculated with equation (III.67)
to be a = 1 0 26 W j m C. A heat balance over the entire system reads:
dT
<f>H = 4 x 10
3
= vpcP dt + e>:A(T - T
0
)
and after integration between t = 0, T
0
and t, T we obtain :
T - T
0
= 41-5{1 - exp ( -98 x 10-
6
t)}
201
Problem 2
The velocity of free fall is found to be (equation II.69) v = 24 m/ s and with
equation (III.68) we find for the outside heat transfer coefficient a = 469
W /m
2
1t
l/>Hl = (;d!p Ml& =and; !l Tt
from which we find t = 0616 sand L = 148 m.
Problem 5 John and the body at the comer
The heat transfer coefficient to a cylindrical icicle is given by equation
(IIJ.67) and, using A.= 0025 Wf m oc, v = 142 x w-
6
m
2
js, a = 20 x 10- 6
m
2
/s, we find :
(X= 95 v-t (Wjm
2
C)
The heat balance per unit length of icicle reads :
~ ~ dD A
'I'H = - rr.D dt p llH m = rxnD u T
Substituting for rx the value found above and integrating between t = 0, D = D
0
and t, D = 0, we finally obtain (with Mlm = 334 x 10
3
Jfkg) t = 610 s, i.e.
an icicle would melt away in approximately 10 minutes.
ID.6. Heat transfer during natural convection
Yes,' John thought, 'the alibi of the suspected farmer seems
to be waterright.' The only way by which he could have
returned home in time for the police from the house of the
murdered banker on the other side of the lake .would have
been to cross the lake. The air temperature had been (f C
for weeks and had dropped quite suddenly w - N C 15 hours
ago. An ice layer of 3 em thickness was necessary on this
lake to carry a man. John remembered an old lesson, made
a swift calculation and arrested the farmer.
II 1.6.1. Heat transfer during natural convection
Heat transport by natural convection is of importance in cases where there
is no forced fl ow, e.g. during cooling of solids suspended in quiet air, or heating
or cooling of liquids through walls v.ri thout forced flow caused by pressure
differences, or stirring.
Consider, for example, a vertical surface with wall temperat ure Tw in a stag-
nant fluid which has a temperature Ta) at a great distance from the surface
(Figure III.20). Near the wall the fluid is heated so that the specific gravity
there is (usually) lower than at a greater distance from the wall. In the stationary
state a velocity distribution adjusts itself (in Figure III.20 drawn for a height
x), which is partly determined by the heat transport from the wall. The beat
202
X
f
Width of velodly and
z ltmperaue field
/
/
I
/
/
-Y
Figure m.20 Free convection along a vertical
wall
transport in tum is influenced by the velocity distribution. In this case. therefore.
the equation of motion (Navier-Stokes) and the energy equation are coupled.
For laminar natural convection, which in practice occurs most frequently,
the velocity and temperature distributions can be calculated for simple geomet-
ric situations. As a measure for the heat transfer, once again the local partial
beat transfer coefficient ex can be introduced. It appears that in the case of natural
convection along surfaces in general :
applies, where A., a and v should be taken at the mean boundary layer temperature
and the coefficient of expansion pat the ambient temperature T ':X> This relation-
ship shows that ex decreases with increasing x because the boundary layer thick-
ness increases with greater x.
For a vertical surface with height L, the mean heat transfer coefficient ( ex)
can be represented by:
The quantity .BITw - Tool can also be expressed in densities:
The dimensionless group which determines the heat transfer can then be written
as:
-z- ti(Hot)te
1
rrn ;nk:j .J.ltJ Y'fl
G, .: .9 X p iJT
)J.\$ - --
203
Gr is the dimensionless Grashof number* which is often encountered when
describing transport during- fiow caused by density differences
(although in heat transfer to liquids (Pr 1) the product Gr Pr almost always
occurs).
For vertical surfaces applies (theoretically and experimentally):
( a)L
-
1
- = <Nu) = ().55(Gr Pr)-l (10
3
< Gr Pr < 10
8
) (111.71)
By approximation this relation can also be used for horizontal pipes, if L is
replaced by D. Equation (III.71) can be understood by realizing that the force
per surface area (gc5h J1p) in the positive x-direction (Figure fii.20) must be in
equilibrium with the shear force at the wall, which can be estimated to be:
Thus:
- Vm.ax
gbh &p = rJ !>,
(I1I.72)
The thickness bh of the hydrodynamic boundary layer can now be estimated
with the help of the penetration theory as:
c::; /2vx
bh = y /LV = -
Vma:tt

Substituting vmax using equation (III.72), we find for tbe thickness of the hydro-
dynamic boundary
[) = {7tV2Xp}*
h llp g
Remembering that the ratio of the thermal and hydrodynamic boundary layer
thickness (see paragraph !1.2.1) is given by {equation II.29):
c5r = - t = Pr--t
o
11
a
we find for the Nu number:
Nu =- = - Prt =
2
Prt = 075 Gr+ Prt
X X {X
3
t::.p g}f
br b
11
1tV p
(III.73)
Hence we see that with the simple model chosen we obtain a relation quite
similar to the exact solution given by equation (III.71).
Gr can be regarded as a combination of Re = vLJv and the Froude number v
2
p/Lg Ap where
the velocity no longer occurs:
Gr = Re
2
/Fr
------- - -----
204
If Gr Pr becomes > 10
8
the flow along the surface is turbulent and ex no
longer depends on x and by approximation (ex) no longer depends on L. It is
clear that in this case in a relation of the type of equation (111.71) the exponent
of Gr Pr must be equal to ! . For this case has been found :
( Nu) = 013 {Gr Pr)t (III.74)
A similar situation occurs on cooling a horizontal plate. Above the plate
thermally generated eddies occur. If the dimension of the plate Lis great with
respect to the size of the eddies (condition: Gr Pr. > 10
7
) (ex) is no longer
dependent on L. In that case:
( Nu) = 0-17 (Gr Pr)t (111.75):
During heat transfer by forced convection, in principle free convection is
always involved, because there are temperature differences in the medium.
The combination of both effects is very complicated and as yet little is known
about it. The effect of free convection on heat transfer is negligible compared to
that of forced convection if the liquid flow velocities are :
Vrrce coovcc:tion < Vcorced convection
From equations (III.72) and {III.72a) the maximum ftuid velocity can be
calculated and we find:
fiitilP
Vcreeconvc:ct ion = V -p < Vrorccd conwclion
and, by multiplying both sides of this equation by x/v, we finally obtain:
~ < XVrorced convection
yp;r- y
or:
.jitGr < Re (III.76)
So we see that free convection heat transfer effects can be neglected if Re > for
but that convective heat transport is most important if Re for. In the
region Re ~ f o r we can add the fluid velocities due to free and forced con-
vection and use the sum total to calculate N u as a function of the Re number.
111.6.2. Problems
1. Water of 50C is passed through a non-insulated horizontal pipe in air
of 20C. Through the wall an am.oun t of heat of 100 W is lost. The heat loss
is completely controlled by free convection. Later on water of 80C is
passed through the pipe. Calculate the heat loss.
205
2. Liquid condenses in an air-cooled pipe of 10 em diameter and 2m length.
Calculate the ratio of the condensation capacities if the pipe'1s put in the
horizontal and vertical positiQns respectively and if
(a) g = 3 x 10
6
vpwa
and
(b) g Ap = 3 X 10
8
vpwa
Answer : (a) horizontal : vertical = 212: 1
(b) horizontal : vertical= 27:1
*3. A metal thermometer well (see Figure III.21) (outside diameter D, wall
thickness b) is inserted over a length L into a gas . pipe which a
Figure UI.21 Thermometer well
warm gas with temperature Yg flows. The tempe_rature of the well
the wall is lower than T,, so that by conduction through the metal (A.nJ heat
flows through the well to the wall. In the stationary state this beat is supplied
by the gas (heat transfer coefficient ex). Show that for the temperature of the
206
sleeve applies :
d
2
6.T a
dx
2
- Ab AT = 0, !iT = T - T,
X= 0, d !iT= 0
dx
X = L, /i Tw = Tw -
The solution of this equation is:
"T __ ",..,..cosh px h jfg
u u 1 . w ere p = -
"'cosh pL A.<5
There is no heat flow through the plane x = 0. Calculate T
0
, = l00C,
Tw = 60C, D = 20 mm, lJ = 1 mm, L = 015 m, .A. = 100 W/m oc and
ex= 25 Wj m
2
oc.
0
= 926C
*4. Show that John took the right steps during the case of the dead banker
described at the beginning of this paragraph.
Problem 3
A heat balance over a short length dx of the thermometer well reads in the
stationary state:
0 = - <f>'HnDJix + dx - etnD dx( T - 7;)
or:
d"
0 = - nDb_____!!_ - anD(T - T.)
dx
1
Now th e heat ftux in the x-direction is also given by 'H = - A. dT/ dx. Thus:
We simplify this equation by writing it in the following form :
d
2
!:iT __ a fl.T __
2
" T fa
dx
2
...t<S p u , where p = "V M
207
This differential equation must now be integrated between the boundary
conditions :
X = L, T = T..., ' 6 T = T..., - T,
X = o. c/J'H = 0, d
11
T = 0
dx
Multiplying through with d 11 T/dx yields :
2
d 11 T d
2
0. T _ d (d ~ Tl
2
2
d ll T
2
d
2
dX dx2 - dx ---ax- = (IXP aT = dx(p 6T)
ctaT
-- = +p6.T
dx -
and thus :
LlT = Aepx + Be- P"
Using the boundary conditions stated, the values of the constants A and B
can be found to be :
A = B and T. - T. = A ePL + B e- pL
,., 8
So we find, finally, for the temperature distribution along the thermometer
well :
T - T, epx + e- px cosh px . ~
----::.. = = With p = -
T., - T, e PL + e- pL cosh pL .At5
The temperature measured at x = 0 is thus given by :
and with the data given we finall y find for the temperature measured
T
0
= 9265C instead of the actual gas temperature T, = l00C
This example shows that considerable mistakes can be made in temperature
measurement if thermometer wells are applied The deviations increase with
decreasing p and thus with decreasing beat transfer coefficients (gases, free con-
vection) and increasing heat conductivity and wall thickness of the well material.
Problem 4 John on the ice
The figure shows the situation during freezing up of t ~ e lake. Assuming
Yc ~ T..., (we will check later!) we can calculate the heat transfer coefficient due
to free convection by equation (III.75) to be ( ex) = 875 W/ m
2
oc (using
A = 0025 Wim oc, v = JO-
5
m
2
is, Pc = 129 kgjm
3
, Pa = 1385 kg/rn
3
, a =
173 x 10- () m
2
j s).
208
Air
lee
Woter
For setting up a heat balance over a thin layer dd we can neglect the specific
heat of the ice formed because at most:
:::::; 42 X 10
4
61fice = 335 X 10
5
Jjkg
Thus our heat balance becomes:
</>'H = a.( 'I; - = 1rw - == - Ml
Solving for the ice surface temperature 7; we find:
(for conditions here 4 = -2lC; thus, indeed, r;:::::; Tw). Substituting r;
in the heat balance leads to:
.
(
rJ.d)dd Tw- Ta
1 + - - = -ct---
;. dt tili p
and, after integration between t = 0, d = 0 and t , d, we find that it takes roughly
135 h to form an ice layer of 3 em thickness. Thus, the farmer could have been
back in time, as John rightly concluded.
ill. 7. Heat transfer during condensation and boiling
Condensation and boiling heat transfer can be regarded as special cases of
convective heat transfer. The heat transfer is attended by a change of state
(vapour - liquid or vice versa} and the great difference in density between the
two phases causes flows which greatly promote the heat transfer. Therefore
much higher heat transfer coefficients can be obtained than in the case of free
209
convection without phase transition and even in forced convection. The beat
transfer coefficient further depends in principle on the temperature difference,
just as was the case in free convection, in one phase.
II/.7.1. Film condensarion
We shall use the classic example of Nusselt for the calculation of the heat
transfer coefficient over Lin film condensation of a pure vapour
on a vertical plane with height L (see Figure III.22). We assume that the con-
densate film covers the surface entirely and flows down laminarly and that the
difference ll T between the condensation temperature 7; and wall temperature
T.., is constant over the entire height. When {J is the thickness of the liquid film
the local heat transfer coefficient is oc )./b (J. = conductivity of condensate).
In order to calculate ( oc) over the height L, we must as a function of x .
Wolf
x=O
Conde ... sate
- df.ll
-f
tl>v =<v>WS+d (<v> WS )
Vapour
L
l!
X =L

liL
Figure IU.22 Fihn condensation
When on a surface element of height dx and width W an amount con-
denses, the condensate flow 4>v increases over d:X by an amount

which is
equal to d((v) Wb) (mass balance). Furthermore, on the condensate surface an
amount of heat tv! oP d" is released which in the case of laminar flow should
be transported to the wall by conduction only; so the energy balance will be :
).
l:lH,.p d<f>v = W dx
Now it is assumed that for each value of x the -relation between 4J., and is
given by:
0
. J 3q( v)lJ
pg
(see problem 11.1.2) although strictly speaking this equation applies only to
laminar vertical films of constant thickness. So :
cf>v = pgb
3
W/371, and dcf>v = pgb
2
W db
17
--- - --- - - - - -
210
Substitution into the energy balance gives:
!lT l dx = llH
11
p
2
g
03
db
71
After integration over x between 0 and Land over b between 0 and OL we get:
J.llT L = AHoplgo1_
4'7
.
The average heat transfer coefficient (a) is defined by:
Miup
2
gWo1.
( ex) AT WL = MlvP4>r:lx=L =
371
From the last two equations we find:
4 A.
<(1. > = --
3 c5L
but in this form the solution is hardly usable because oL is not an easily measur-
able constant. Elimination of bL from the three previous equations gives :
( a ) = (}94[AHu Pl ).3gJ * (111.77)
LYf6T
So ( (1.) is inversely proportional to .jL; for this reason horizontal condensers
often applied and an increase in the capacity is sought in Wrather than in
L. Usually horizontal internally cooled t ubes are used. For one horizontal
tube a formula analogous to equation (ll1.77) but with a coefficient of 072 can
be derived if instead of L the external pipe diameter Du is used. This relation
can be extended to the case where n horizontal pipes are placed under each other.
The general equation for this is then :
[
llH &1 p2). 3g]t
( ex) = 072 nD,J7 6 T
(III.78)
The values of ( a: ) found in practice for laminar condensate films are on an
average 20 per cent higher than follows from theory. Figure III.23 gives a
nomograph from which partial heat transfer coefficients for c;:ondensation can be
rapidly found. The validity is limited to those cases where the condensate film
still flows laminarly < 10
3
) .
The presence of a non-condensable gas in the vapour greatly reduces the
rate of condensation. The vapour must now diffuse through the gas which has
accumulated at the condenser surface. The partial vapour pressure at the con-
dc!nser surface is then co,:15iderably lower than the total pressure. The presence
of relatively small amounts of gas can therefore decrease the temperature at
'*The constants p, rr and should be taken at a mean temperature T
1
o f the condensate film which
equals 1f = Tc - f

c 4
-30
0
.30

go
120
ISO
180
210
liquid
( I ) HzO
( 2) CH
3
0H
( 3) C
2
H
5
0H
' ( 4 ) c
6
H
6
(5) NH
3
{ l!.H,:'i
3
}'
14
1 (Ov) ' (4H) .
r
. ,.
. :
' .
1
Lt:.T
ntXlT
10
.
Figure ffi.23 Nomograph for the calculation of heat transfer coefficients of
condensing liquids if 4(v)oLfv = 4cpnJW1] < 1000
211
which the vapour condenses .considerably., so that.a dif-
ference AT is avaUable for the heat transport through 'film. If
001 < PtnerJPtotal < 04 we can correct for this ,effe,ct by *e following empirical
rule: ..
CXmixture :;::::;; Ol rP::;;;_
C!pure vapour . V p;:::
. ..
(III.79)
-Equation (III. 79) shows that the presence of, for 20 per cent of air
by in a condensing vapour will the heat coefficien:
by a factQr 4. If P inerJPtotal >. 04 a.heat transfer coefficient should be calculated
assuming no condensation hutjust cooling of the gas mixture.
' <
. .
I I 1.7.2. Drop wise condensation
[f the cooled surfac.e is poorly wettable by .the condensate, so-called dropwise
condensation occurs. Parts of the surface are then almost dry and have a very
.high local transfer coefficient . . As s,oon as condensate has formed a
drop collects which after having grown sufficiently rolls down and sweeps
part of the surface clean. The average heat transfer coefficients can be up to
ten times higher than on film condensation.
212
I 11. 7.3. Boiling
The boiling phenomenon is complex. There are eight principal situations,
as shown by the following table of possibilities (a), (b) and (c):
(a) J.Y all (b) Liquid (c) Medium
horizontal at boiling temperature flowing
vertical undercooled stagnant
Most is known of the heat transfer from horizontal walls to a stagnant medium
at boiling temperature (water kettle on the fire), the so-called rree boiling'.
The greater part of the literature deals with this situation. Dependent the
degree of superheating (6 T with respect to the 1ocal boiling point) four boiling
regimens can be distinguished : ,
( I ) Low superheating (6 T < ca. 2C), free convection (with or without forced
convection), no bubble formation yet. A slight superheating (and consequent-
ly overpressure) is necessary to form bubbles because of the following
effects:
tlp due to surface tension (=Zaf R and so tlp is rugh for small bubbles),
ll.p so that the growing bubble can displace the liquid, tlT to trans-
port the heat of evaporation through the liquid to the bubble and-
if no allowance has been made on selecting the boiling temperature
(reference for \T)-a !lp due to the hydrostatic pressure on the immersed
wall.
(2) Nucleate boiling (for water : 2 < l'i T < 25 - 75C) : in certain 'active'
places (small indentations in the wall)vapour bubbles (nuclei) are formed.
which in an undercooled liquid 'collapse' because of condensation-as soon
as the tops of the bubbles reach into the cold liquid as a result of growing
(very high <XS). In a liquid at boiling temperature they grow until, because of
their buoyancy and momentum, they become detached from the indentation.
Every active place yields bubbles of a size characteristic thereof. At low
/'iTs those places first become active which yield relatively large bubbles.
If superheating is increased the at which nuclei are formed in
these places becomes higher until the maximum rate of discharge of the
bubbles formed has been attained for the relative active place. Next, smaller
indentations which produce smaller bubbles are produced at a higher
frequency and bring about more 'agitation' in the interface, so that a
increases rather strongly with ll T (approximately proportional to T
3
).
It appears from the foregoing that in this regime the s urface conditions of the
wall (nature and size of the irregularities, wettability) play an important
role. It is therefore difficult to carry out reproducible experiments, particular-
ly because when a wall has been exposed to boiling once its surface changes
(ageing).
213
-
(3) Leidenfrost regime: the bubbles of neighbouring active places coalesce and
form a closed vapour layer. As soon as this takes place the heat transfer
coefficient decreases considerably with respect to the values which were
obtained in reginie (2), because the vapour layer has a fair heat resistance.
If instead of a given wall temperature a given thermal load is applied (electric
heating element, nuclear reactor), the temperature of the wall rises con-
siderably on transition from regime (2) to (3), as a result of which the heat
transfer coefficient decreases further until finally the heating element melts.
In these cases one should therefore stay at a safe distance from the critical
T where the Leidenfrost phenomenon occurs.
(4) A stable vapour layer has formed from which bubbles with a dimension
characteristic of the liquid are released (for stagnant, clean water approx-
imately of 15 mm) ; in the vapour layer _the heat transfer takes place by con-
duction and particularly by radiation (tiT for water > 500C).
III. 7.4. Heat transfer in evaporators
Evaporators are used for concentrating solutions, if necessary combined
with crystallization of the dissolved substances. There are a great many versions,
from 'boiling pans' to pipe evaporators, in which the solution flows through
banks for example, vertical pipes . which are heated by condensing
steam.
During the evaporation in vertical pipes a combination of heat transfer
mechanisms occurs. In the lower part of the pipe the incoming liquid is heated
to the boiling point (heat transfer by forced fiow). In the upper part the evapora-
tion begins. After a short boiling height so much vapour has been produced
that in the upper part of the pipe mainly a vapour flow occurs which entrains
the liquid which has not yet evaporated. In this range the heat transfer is deter-
mined by the flow conditions of the two-phase mixture rather than by the
boiling phenomenon.
In certain types of pipe evaporator the Hquid is pumped into the pipes
(forced circulation). there are also many evaporators with natural
circulation (thermosiphon action), the rate of circulation irt them depending
Qn the vapour production and on the heat transfer. The latter is influenced
by the rate of circulation.
Pipe evaporator design is largely based on experience supplemented by
experimental data which have been obtained with model set-ups. For dilute
aqueous solutions the as usually lie between 2000 and 5000 W /m
2
oc and for
hydrocarbons in the range from 500 to 1000 Wj m
2
oc. Figure III.24 gives a
survey of experimentally determined heat transfer coefficients during boiling
as a function of heat flux. This figure can be used for quick estimation of boiling
beat transfer coefficients. In order to avoid the 'Leidenfrost' regime j ust discussed
it is advisable to limit the vapour velodty at the evaporator surface to roughly
Q,2 to 0-3 m
3
j m
2
s.
214
1
u
0
N
E
....
3
.........
0
Free convection Nudeott boiling
Figure ID.l4 Heat transfer coefficients during boiling of liquids
II 1.7 .5. Problems
*1. A chemical reaction is carried out. in ethanol solution under r ~ f l u x . The
reflux condenser consists of a straight inner glass tube (A.,ta.ss = 1 W/ m C)
of 12 mm diameter arid 2 mm wall thickness and an outer pipe of 30 mm
inner diameter and 40 em length.
(a) If 15 1/min cooling water of l5C are available, calculate the amount of
heating energy that can be supplied to the reaction mixture so that just
no vapour leaves the condenser.
(b) What would be the amount if chemists could be persuaded to use stain-
Jess steel equipment?
Answer : (a) 315 W ; (b) 700 W
2 A condenser consists of a number of copper pipes (Dj = 228 mm, wall
thickness d = 125 mm). On the outside of the pipes 9080 kg/h Freon 12
condenses at 322C. The heat of condensation (1375 kJfkg) is removed
by cooling water in the pipes. The flow rate of the water is 122 m/ s, the
inlet and outlet temperatures are 155 and 178C respectively. and the
total heat transfer coefficient Freon-water related to the external pipe
215
surface is 1010 Wj m
2
'0C. Calculate the number of pipes n the condenser
consists of and the length L of these pipes.
Answer : n = 72 pipes ; L = 39 m
3. A packet of 6 x 6 horiZontal tubes of 30 mm outer diameter and 15 m
length is.surrounded by saturated steam of 180C.
(a) If the surface of the tubes is kept at 120C, calculate the heat fiux
through the wall.
(b) What would it be if the pipes were in a vertical position?
2
; (b) 236,000 Wj m
2
Problem 1
The maximum amount of heat that can be removed is given by cPH = U_A llT.
Now AT and A are known. Thus, after determining the overall heat transfer co-
efficient U,
8
can be calculated.
For the various partial heat transfer coefficients we find <lwater = 2220
W /m
2
C, a
1
.tass = 500 W / m
2
C, acondeMate = 950 W/ m
2
oc {from Figure ITI.23).
Thus, U = 286 W /m
2
oc and <Pn = 315 W. If the inner pipe is made from stain-
less steel listed = 13,000 Wj m
2
oc and U = 630 W/ m
2
oc , so
8
= 700 W.
ill.8. Heat ttansfer in stirred vessels
Heat transfer in stirred vessels is strongly dependent on the flow pattern
produced by the stirrer. As we have already seen in paragraph II.7:1 (Figure
II.46) ~ propeller-type stirrer creates two systematic eddies which take up
nearly half the vessel, the flow in which is 'parallel to the wall and to the shaft
of the stirrer. A paddle-type stirrer mainly produces an eddy perpendicular to
the stirrer shaft. whereas a turbine-type stirrer produces two eddies. one above
and one below the stirrer.
For the flow along the wall of the vessel thls means that a propeller or turbine
stirrer produces a flow in the axial direction, whereas a paddle-type stirrer
.induces flow in the tangential direction. The axial flow along the wall is caused
by a jet of liquid which comes from the stirrer and hits the wall at a well-defined
point. From this point a boundary layer is built up along the wall. The thickness
of this boundary layer increases with distance from the stirrer. Also during
tangential flow a boundary layer along the wall of the vessel is formed, but this
layer has a constant thickness.
It is understandable that it is not possible to produce a simple heat transfer
model on the basis of this complex hydrodynamic picture, especially if we realize
that the rate at which the boundary layer during axial flow along the wall is
formed is dependent on the energy of the eddy and thus on the Renumber of
the stirrer nd
2
/ v. For relatively low Renumbers tbe momentum transport from
the eddy to the wall boundary layer is small and we can make a rough guess of the
216
heat transfer coefficient at the wall on the basis of momentum penetration:
l A A. A
a. ,....., - = --= Prt
<>r Dh Dr b" TCvxjv..,
where xis the distance along the wall from the point of contact with the liquid
jet and vw the mean velocity of the flow along the wall (which is proportional
to nd). This model predicts that :
or :

Nu Pr>\d \x
where the diameter of the stirrer d and the diameter of the vessel Dare introduced
because ex is independent of v.
For relatively high Renumbers-the momentum transport between the core of
the eddy and the boundary layer determines the boundary layer thickness which
is practically uniform (analogous to the boundary layer thickness during tur-
bulent pipe flow) if the flow situation in the vessel is completely turbulent. This
results in (see paragraph 111.3.2):
/D]OS
Nu - Re
0
'
8
Prt \ d
For turbine-type stirrers wbich.induce mainly axial flow, the experimentally
found exponent of theRe number between 05 (low Re, Re < 10
4
) and
07 to 08 (very high Renumbers, Re > 5 .x 10
5
). A rough description of the
dependency of the value on theRe number in the region of practical interest
(10
3
< Re < 10
6
) leads toRe exponents of 065 to 070.
The dependency of Nu on Pr
1
as predicted by both q1odels (for high and low
Re numbers) has been experimentally substantiated by a number of authors.
The influence of the parameters Dfd and Df x has been studied by only a few
scientists and no agreement has been reached. Both theories predjct too strong
a dependency of Nu on Df d or on Df x. This is due to the fact that the ratio
v...,jnd (which we a ssumed to be constant) is a function of d/ D and of xfD. For
turbine stirrers, for example, we might prove that v,..,/rui ,_ df D, because the
width of the jet of liquid leaving the stirrer remains constant until it reaches
the wall (see paragraph II.7.1); it is easily proven that in this case :
Only a few results of measurements of local beat transfer coefficients (see
Figure 111.26) are available because measurement of these data is complicated
217
and, anyway, for design purposes a knowledge of heat transfer coefficients
averaged over the height of the vessel is sufficienf .
For flow creating a systematic eddy perpendicular to the axis, as is the case
with paddle-type stirrers, we practically always find :
Nu...., Rel Prt
which is another reason why for turbine- and propeller-type stirrers (which
always produce tangential flow) the mean Re exponent is found to be ca.
065 for 10
3
< Re < 10
6
.
In the foregoing we have considered heat transfer to the wall of a stirred
vessel. The presence of a coil makes no principal difference but complicates
the picture because the fiow of liquid is mare complicated. The presence of a coil
will decrease wall beat transfer coefficients (by up to 25 per cent for normal
configurations) because of decreased flow velocity along the wall. Coil heat
transfer data are correlated with a coil-Nusselt number and are generally
found to be:
rt.dc t .J..
Nu = - -. Re
l
Because the combination of vessel and stirrer is not completely standardized-
which is partly due to the fact that a stirrer is often not selected for heat transfer
performance only but also for other tasks (e.g. suspending, mixing,
chemical engineer designing a stirred vessel has to search literature in order to
find heat transfer data whlch can be applied for solving his problem. In doing
so, the chemical engineer will have to answer the following questions:
Which type of liquid flow is required?
Which Re range is required?
Are the literature data found applicable to the specific geometric configuration
of the design?
Are local or average heat transfer data required?

---o
0

!
0
de
1.
I
D
:
3
0
H
I

.
H
0
"
"
l
l
0
"i
:
5
0 '
I
1
0
d
4
6 - I
( 4 baffles)
0
- m
Figure ID.25 Standard configuration of stirred vessel
218
In the following, a few .experimental correlations for heat transfer to a 'stan-
dard turbine stirred vessel' are given (see Figure III.25).
The average beat transfer to the wall Qf a standard vessel according
to Figure III.25 is:
(Nu) J <a;o =

(111.80)
for 10
3
< Re < 10
5
. For slightly different configurations the relation:
( Nu) = - - = 101 - Prt - - -
(rx)D (nd
2
lt ('7w)014(d)013(L)Ol2
A. v t1 D D
(III.80a)
can be used (017 < d/D < 075; 01 < L/D < 07). For the standard coli;.
figuration equation (III.80a) yields equation (111.80) with a constant of 077.
Local heat transfer data to the wall can be calculated from the relation
-(standard configuration with h = vertical distance from stirrer):
Nu = = 017Re+Pr('1:)"' .. if < 15
or:
( )
014
Nu = 15 Ret
17
:
if 10
3
< Re < 10
6

D
if - > 15
h
Figure Ill.26 illustrates the dependence of the ratio of rx
1
ocaJ(a.) on the height
of the vessel. The average heat transfer .values calculated from the local values by
integration over the vessel height are in agreement with the values found from
Height of
0
,--- stirrer
a/(a) - ---
Figure 111.26 Local heat transfer coefficients in
standard vessel
219
equation (111.80), i.e. the surface area under the line a./ ( rx) = f(Dfh) in Figure
III.26 is unity.
The average heat transfer coefficient to coils in a vessel with the configuration
of Figure III.25 has been found to be:
< ) d . ( d) 0 1 (d ) 0 5
( Nu) = a .A. ' = 017 Re
0
.
67

D ; (III.81}
for
d d ' .
025 < - < 058 003 <...!. < 015 and 400 < Re < 2 x 10
5
D , D
A much-applied scaling-up rule for geometrieally similar vessels is to keep
the stirrer energy input per unit of volume of the vessel (which is "' n
3
D
2
)
constant. In this case, for double-walled the amount of heat transferred
per unit of time and of volume (which is -r:zD
2
j D
3
) is proportional to D-4
(foi low Revalues) to v -
1019
(for average Revalues). For geometrically similar
vessels. therefore, the heat transport decreases with increasing size of the vessel
if the scaling-up rule of constant power per unit of volume is applied Dy.ring
scaling-up steps have therefore to be taken to increase the heat flow (extra
surface, larger temperature difference) or higher power consumption of the
stirrer has to be
If vessels with heating or cooling coils are used and the heat transfer surface
per unit of volume is kept constant, the heat transferred per unit of time and
of volume is proportional to D- t to D-t if the above scaling-up rule is applied.
Under these circumstances only a slight correction of stirrer speed (and thus of
power consumption) is required to keep the heat transport during scaUng-up
constant. ,
lll.8. 1. Problem
1. Two water! ike liquids which react instantaneously with each other are pumped
through a continuously stirred tank reactor. Pilot plant experiments with
a double-walled tank (diameter of vessel = height of double wall = D
= 04 m ; stirring speed n = 360 rpm) according to Figure 111.25 showed
that the production capacity of the reactor is determined by the rate of
removal of the heat of reaction. It appeared that the heat transfer coefficient
was independent of the cooling water flow rate.
(a) By what factor would the production capacity per unit volume of the
pilot plant reactor be increased if additional cooling by means of a coil
(O.J m diameter. 001 m tube diameter. 15 turns) was applied?
(b) The reaction is to be carried out on a factory scale in a double-walled
reactor, according to Figure. III.25, of 3 m diameter. By what factor is
the production capacity per unit reactor volume increased if the scaling-
up rule of constant stirrer power consumption per unit reactor volume
is applied?
220
(c) What stirring speed would have to be applied in situation (b) if the same
production capacity per unit volume as in the pilot plant reactor is
desired?
(d) The chief engineer of the factory wants to use a vessel with a cooling
coil instead of a double wall A coil of 25 m diameter, made from pipe
of 5 em diameter, is available. How many loops of coil are needed to
assure the same production capacity per unit volume as obtained with
the reactor of situation (b) if again the scalingup rule of constant power
consumption per unit reactor volume is applied?
(e) What is the ratio of the cooling surfaces of cases (b) and (d)?
Answer : (a) 213 x ; (b) 011 x ; (c) 2700 rpm;
(d) ca. 15 turns ; (e) AwarJAcoil = 16.
'Heat rays' are electromagnetic waves (05 < ). < 10 pm)* which are radiated
by a body owing to thermal agitation of the composing molecules. Between
two surfaces radiation exchange may take place if the interspace is more or
less diathermanous for the entire or part of the wavelength range. There is,
heat transport in both directions and, if the temperatures of the surfaces differ.,
the difference between the opposite heat flows does not equal nil
By means of thermodynamics it was derived that the heat flow which is
emitted per unit of surface cJ>; by a so-called black surface with temperature T
is given by the Stefan-Boltzmann law:
(III.82)
where the radiation constant (J = _575 x 10-
8
W(m
2
K
4
and T the absolute
temperature of the surface. .
For a black body the intensity distribution over the wavelength range is
given by the Planck radiation law. The wavelength at which this intensity 1s
maximal is inversely proportional to T (Wien's displacement law):
Amax T = 2880 (micron K)
(III.83)
At room temperature )max = 10 ).lm (fa{ infrared) and at 6000K = 05 J1rt1
(yellow).
A completely black surface emits the maximal amount of radiation relevant
to its temperature; it has the emission coefficient e = 1. According to Kirchhoff's
law (see equation III.84) the absorption coefficient for the radiation on the sur-
face is then also a = 1. A body which absorbs all radiation and does not reflect
any is indeed black.
In general, technical surfaces have an emission and an absorption coefficient
smaller than 1 the value depends on the material and the roughness of the
surface, on the temperature and on tbe wavelength of the radiation. If in a
*Other of electromagnetic waves are radio waves (.4 > J<t Jllll)l , visible light (04 < ). <
o-8 Jllll), X-rays (J. ;:::: 10-
4
.um) and 1-rays {i. R: w-
6
pm).
221
certain wavelength range preferential absorption occurs, then just as in the
case of visible light the surface is 'coloured'; -
In engineering, mean absorption coefficients over the entire wavelength
range will often suffice. In other words, such surfaces are considered as
between the extremes 'black' -and 'white'. Of most non-metallic surfaces
a > 08, so they are almost black for the infrared radiation. Metals reflect the
heat rays only satisfactorily if polished and clean; the coefficient a
lies then between 005 and 020. Oxidation and contamination increase a.
Figure 111.27 shows absorption coefficients fo_r a number of
. .. . .. '
I . , . I . . ..
I Oxidized metals I Non-.me!als
: I {
Asbestos
I I Gypsum
I Soot
1---:----.:._-4
1
------...J.I_ Refractor y bricks
Pol ished metals
f-
rr
1-l-
l
1
08 r
r--------+----[ron
., 06
Bross
I
__ __.__Cu
I
I
I
I
I
I
I
I
I
I
04
r-------+
1
-AI paint
r
r
I
r--------r: -Ni
iron .1
02- I
!""--Cu N i
1

t- Pt,Zn
1
1
0-Ag
Figure Ill.27 The absorption. coeffiCient for heat radiation
' ' '
I '
-,
The amount of heat which is transported between two or' m<?re
per upit of time depends not only 011 the temperature but also on :
. . . . .
'the spatial configuration (how fat can the surfaces 'see' each other?);
the absorption and emission capacity of these surfaces ;
the possible absorption -by the intermediate medium (water v.apour and
carbon dioxide, for example, have strong absorption bands in the infrare) ; .
the emission of the intermediat'e medium-(luminous flames)'.
Only the first two will be further treated here_ For calculating the net transfer
by radiation a good balance of the radiation transport has to be drawn up.
This can best be done by using PoJjak,s scheme. Poljak assumed
that for every wall the total energy'flux 4>" of the radiation which is emitted by the
wall equals the sum total of the energy flux of its own emission' euT
4
( = e;)
and the energy flux of that part of the foreign radiation (</>")on the wall which is
refiec;ted by it ((1 - a)i'): . .
(HL84a)
222
the energy flux of the net transport which is emitted by the wall(;;) equals the
difference between its own emission and that part of the foreign radiation which
is absorbed by the wall:
4>: _= - iuM .
(III.84b)
Much can be concluded frpm these relations:
(a) If the wall is completely enclosed by black then in the case of ther-
modynamic equilibrium = 0 and <Pi = = uT
4
So according to
equation (III.84b) e = a, i.e. the emission coefficient numerically equals the
absor.piion coefficient (Kirchhoff's law).
(b) From what has been said under (a) it follows that, in the case of equilibriulll,
in a closed system through each plane the total energy flux of the radiation
equals 4>; = <1T
4
.
The Poljak system can be formulated even more practically by converting
equation (111.84) algebraically as follows :
(I II.85a}
(IU.85b)
using Kircbhotrs law. The practical course of calculation is then that with
equation (lli.8Sb) for each wall the total radiation is expressed as a function of
the temperature of that wall = .aT
4
) and of the still unknown net transport
With the relations for the total radiation of each wall thus obtained the
foreign radiation f>'( on each waiJ can be calculated. These ;' values used in
equation (UI.85a) then give the net energy fluxes for each wall as functions of
the wall temperatures.
In order to calculate the foreign radiation </>'(" for a wall with number k, we
must know the fractions Ejk which indicate which part of the total radiation of
wall j directly reaches wall k. These view factors Fare calculated by cumbersome
integrations (summation small surface elements) and we have to take into
account that the radiation intensity depends on the angle cjJ between the direc-
tion of the radiation and the to the surface (approximately proportional
to cos </J, Lambert's law). However,for a number of geometrically simple cases-
to which we shall restrict ourselves-the view factors F can be found in a simple
way. The reader may check that for infinite parallel walls with numbers 1 and
2 F
12
= F
21
= 1. In general, it can be proved that for two walls with surfaces
A
1
and A
2
A
1
F
12
= A
2
F
21
applies. This can be understood by realizing that
in equilibrium the wall energy flux from 1 to 2 between these two planes

must be equal to the total energy flux from 2 to 1 (=A
2
F
21
</J'; J
It follows that if wall 2 fully endoses wall! and walll does not emit radiation
(e.g. concentric cylinders or F
21
= A
1
/ A
2
, because F
12
shouJd be equal
to 1 (all radiation by wall 1 should reach wall 2. but only a fractioa
223
A
1
/ A
2
of the radiation which is emitted by wall 2 reaches wall 1 ;the fraction
2
) again reaches walll).
With these data we can now calculate the net radiation transport between
two bodies of which the one (1) is completely enclosed by the other (2). The
foreign radiation over wall 1 ( =A
1
</>f'
1
) is then equal to F
21
times the total
radiation of wall 2 ( = F
21
A
2
cf>;). It follows that the energy flux of the radiation
Qn the smallest wall ( = rfJi'
1
) equals the total energy flux of the radiation emitted
by the larger surface ( = cf>;). So, using equation (III.85):
.A._, = A., II - = A., II - = .A._II - l - a l <P" - .A._ JI + 1 - Q2.A._II
'111 o/1 '+'12 o/1 o/2 'f'zl a nl o/:2 a '1'12
1 2
Between the net radiation transport of body 1 (;
1
A
1
) and that of body 2

the relation exists:

cp;1A1 + = 0
because the energy lost by wall 1 most be taken up by wall 2 (law of energy
conservation). From the two latter relations we find for the net radiation
transport from body 1 :

l + 1 - a
1
+ A1 1 - a2) cpll = cp" . _ </J" = e1('r. _ 1'!) (III.S6)
a A a d zt z2 1 2
1 2 2
This is an important relation which is often used for calculating radiation
transport
If the temperature difference {T
1
- T
2
) is not too great, (Ji - can be
expressed approximately as 4 P(T
1
- T
2
) . In these cases a heat transfer co-
efficient for radiation Cl.r can be defined, which according to equation (Ill.86)
equals:
.A

1
1 - a 1 A 1 1 - a
2
) -
1
a,.="W l + + - - --=
a
1
A
2
a
2
(III.86a)
This means that at room temperature a,. is about 6 Wj m
2
oc and can therefore
not be neglected with respect to free convection. At higher temperatures a,.
considerably and hence so does the radiation contribution to the total
heat transport.
II 1.9. I. Problems
1. A long rod (diameter l l em, emission coefficient 08) with a temperature
of 327C is suspended in quiet air of 2rc. Calculate the contribution of
free convection and radiation to the total heat transfer coefficient. Assume
the air to be an ideal gas with a density of lOkgjm
3
at 27C.
Answer : horizontal rod: 46 % free convection
vertical rod : 29 % free convection
224
*2. A small ceramic sphere (diameter 1 em), e.g. of a thermocouple holder, is
situated in the middle of a long tube with a diameter of 20 em. Air of l50C
is passed through the tube at a mean velocity of 20 m/ s. The temperature
of the tube wall is l00C. Calculate the temperature of the sphere.
3. Check that if between two large flat walls with equal emission coefficients
a third equal waU is placed the net heat transport by radiation is halved.
An example of this type of radiation shield is the layers of aluminium foil
in the walls of refrigerators and cold store rooms.
4. The maximum intensity of sunlight occurs at a wavelength of approximately
05 x 10-
6
m. Calculate the surface temperature of the sun.
Problem 2
equal convective transport to the sphere. Thus : -
( ex) (T, - TJ = a.,(7; - Tw)
Now Re = 7000, thus ( a) = 183 Wfm
2
oc (equation 111.68). Assuming
7; ~ T, , 'f = (T, + TJ/2 = 125C ~ 400K. Thus,employingequation(lll.86a)
we find (a
1
= 09, A1/A2 1):
CHAPTER IV
Mass Transport
In this chapter the transport of matter will be treated. The starting point of
our analysis is the law of conservation of matter, which reads for a substance A
(see paragraph 1.1):
(IV.l)
The following pages are no more than an elaboration of this law.
Apart from agreement in the description of the transport of matter and heat
transfer, there are distinct differences as, for example :
(a) The 'production' of mass (by chemical reaction) is usually a compJjcated
function of the molar concentrations of the reactants (and of the tem-
perature), whereas the production of impulse and heat is seldom a strong
function of concentration or temperature.
(b) The interface between two media between which matter is transferred (e.g.
between two liquids or a gas and a liquid) is usually mobile, contrary to the
interface between two media between which heat is transferred (usually
a solid wall).
(c) The order of magnitude of the d i f f ~ s i o n coefficients ID of liquid and solid
substances is much smaller than, for example, the thermal diffusivity or the
kinematic viscosity (for gases at 20 C and 1 atm, D ~ 10-
5
m
2
/ s ; for
liquids at 20C, D ~ 10-
9
m
2
j s ; and for solid substances D varies widely
between 10-
10
and 10-
15
m
2
f s). This means forliquids and solid substances
that in the first place the penetration of matter in a given time takes place
over (much) smaller distances than the penetration of heat or momentum ;
that in the second place already with convective currents of low velocities,
wruch frequently occur in the case of mobile interfaces, the convective
transport exceeds the diffusion transport in the direction of flow.
From the foregoing there appears to be a dilemma in the description of the
mass transport. We formulated a law of conservation of mass because mass is
the most characteristic property of matter. Accordingly we defined the diffusion
coefficient in paragraph 1.2, equation (I.13), as the proportionality constant
226
between the mass flux and the gradient in the mass concentration:
, ~ . , = -ll} dpA
'+'mA, :r. A d;c
(Fick) (IV.2a)
In addition, production rates in chemical reactions are functions of the molar
concentrations, which we. shall indicate by cA, etc. (kmolfm
3
). So we can
imagine that for many calculations it offers advantages to define the diffusion
coefficient as the ratio between the mole ftux (kmoljm
2
the molar concentration:
, . ~ , . , n dcA
'+'mol A,x = - LY A dx
(Fick)
(IV.2b}
Both thus defined diffusion coefficients are in principle different (although in
practice they are often almost equal) and in applying them we should realize
on which concentration the description is based.
The diffusion coefficient has been defined in such a way that the net transport
through a fixed plane (in Figure 1.6 the plane n = constant) is zero. If we
restrict ourselves to binar'y systems (with components A and B) for practical
purposes two cases can be distinguished in which the net transport is zero:
A For every unit of mass A which diffuses through a fixed plane a unit of
mass B moves in the opposite direction by diffusion ('barocentric' system).
B. For every mole A a mole B moves in the other direction through the plane.
sub A. Net mass flow = 0. p = constant
For the diffusion transport of substance A we can write :
':rrA,:r = - [)AB :A
whereas for the transport of substance B applies:
A.."
0
dps
'f'w.B,x = - BA dx
where ':nA.x is the mass flux of A (kgjm
2
s) in the x-direction for a net mass
flow = 0; PA is the mass concentration of A (kgfm )).
Because the conditions have been such that:
,/,." A-,1 0
"f'rnA.x + 'f'mB.x =
whereas :
PA + p
8
= p =constant or dpA + dp
8
= 0
it follows from these relations that DAB . DnA. In other words, the definitions
have been determined in such. a way that the diffusion coefficients of A in B
and of B in A have the same numerical value. These definitions are used for
describing the dlffusion transport in solid substances and liquids ; here, in the
227
case of not too high concentrations of, for example, A in B the density may be
assumed to be constant.
sub B. Net mole ftux = o. the total c =constant (kmoljm
3
)
Now we can write for the mole flux. (kmol/m
2
s) in the x-direction:
and :
A." n dcA
'f'mol A..x = - 11..11 AB dx
A." 0) dc
8
'YmolB.x = -:- BA dx
cA and cB are molar concentrations (lcmoljm
3
); for ideal gases and constant
pressure and temperature their sum total is constant and equals c. These defini-
tions are very suitable for describing the diffusion transport in gases. In the case
of equimolar diffusion :
A." A." 0
'f'mol A.x + '#'mol B, ;r =
Also, here, the diffusivities have been defined in such a way that :
_[]) AB = . O)BA
According to the two definitions the dimension of II) is (length)
2
/ time, so that
Din the MKS system is expressed in (m
2
/s). This was likewise the case for the
two other transport coefficients v (kinematic viscosity) and a (thermal diffusivity).
However, in the case of diffusion it is .advisable to distinguish between two
cases: p = constant and c = constant For these cases []) has to be defined
using respectively the gradient of the mass or of the molar concentration . .
IV.l. Stationary diffosion and mass transfer
IV.J.l . Stationary diffusion
John noticed that, before m had left his office, the lawyer
who had disappeared had prepared a pot of coffee. A cup on
his desk was still half .fuU and from the scale on the sides.
caused by evaporacton ofrhe wacer, he could see that the cup
had been .filled to 1 em below the rim. He estimated rhe cup
co be 8 em high. remembered the pages to come and concluded
that the lawyer muse have left his office approximately 17
days before.
During most practical cases of diffusion neither the net mass flux nor the
mole flux discussed in the foregoing paragraph will be zero. A practical example
in which the net mole flux is almost zero is the distillation in which, for every
mole of the light component which passes into the gas phase, approximately
one mole of the heavy component enters the liquid. However, on absorption
and extraction of one component from a mixture the net transfer is invariably
I :_ t;A'
. '
I ij'
,,
I .
,
It

228
unequal to zero. guring a chemical reaction e.g. at a catalyst surface, the
mole fluxes of-the be the mole fluxes of the product to be
discharged are in
When describing til ll1l!st for
the of fP b.Y tnm&fer as SIJCh.
Fqr J1lO.Vliig fl velocity v relative to .a plane X = constant,
t}le steady-state diffusion flux of component A relative to this velocity is:
de A
( v A - v )c A = - 0 dx
rate of molar transport relative to the x = constant is then, since
VC = </>':not A t B :. . . . . . ..
' " ! : t I \,
</J" . [DdCA ((jJ" if>" . CA
mol A,;.; = - dx + mol A,x + mol B,.x) C
{1V.3a)
:' . , ,.. \, . , ;. , ....
whereas'!ihe mass tra:nsport is given by :
first right-hand teqn in these equations the mass flux resulting
from diffusio'n, whereas the secopd term is of the mass flux
flow, Equations and (IIIJb) can only be solved if the
of A and B ;s kl}gwn. We cari two
.. .. . i I
. ' . .. ' . . . i l :
(a) 4>';.. + <Pi = 0, i.e. diffu\$!Qn \ '-
(b). ,= 9, .i .. e: o,f A ,
) . :, ::-: :\ ... :;-;, . . ..'i : !. : \ .. . .. :... I :.:\ 1 \, \ \ :l. ' ..
In; practice,. sit.'"l;lti.on.s; be,tw.een. these ofteq o9cur, but sine
' - -. - - . " ' ,.. ' .'i
(a). it is

to>ptoduce :a good es:timate of the:molar flu,x by making a ahoice between these
:.ln)thefollowing, t:Wo in.greater ,_
: .. . ,,
:1 '': ... ,( . - :-
(aj d.iif:J!..siPn, <P'k+ 4>; = 0. .
can now be simpljfied :
:we consiq.r. one-dimension.al diffusion in the x-direction, the
mass balance (IV.l) yields (there is no production of A) :
' . . :
. '
or ':nolA,;= coqstant for every plane X =constant. Solution' of this
.equation yields, ; with the boundary conditions c A = c A; at x ' x
1
and
! .
c:
. ~
~
c:
&I
u
c:
0
u
l
-- Distance
Figure IV.l Concentration profile during equimolar diffusion (drawn
line) and during diffusion through a stagnant fluid (dotted line)
cA = cA
2
at x = x
2
, the concentration -distribution :
CA- CAl X - X
1
We see that the concentration distribution (Figure IV.l) is linear.
229
(IV5t
The mole flux of A is found after substitution of dc.Jdx in equation
(IV.4) by the value calculated'from equation (IV.5) to be :
(IV.6)
So, during equimolar diffusion the mole flux (and analogously also 'the mass
flux) is directly proportional to the concentration gradient llcjAx. We
notice that mass transport by equimolar diffusion is analogous to the trans-
port of heat by conduction. For one--dimensional beat conduction we had
found in paragraph 111.1.1 (equation II1.2a):
A-" T
2
- T
1
p e P ~ - pcPT2
'+-' H = - /, = - Q--"'-----.___
x2- Xt x2- xl
This equation is identical to equation (IV.6) if we replace the molar diffusivity
10 by the thermal diffusivity a and the molar concentrations c (moljm
3
) by
the 'heat concentrations, pc P T ('W j m
3
). We can therefore also adapt the
relations found for stationary heat conduction around cylinders and spheres
to diffusion. Using equations (lll.6) and (Ill. 8) we find for equimolar diffusion
230
between coaxial cylinders with diameters D
1
and D
2
and concentrations
CAl and CAl :
(IV.7)
and for equimolar diffusion between concentric spheres (diameters D
1
and D
2
, concentrations cA
1
and cA2) :
2-nD
1
D
2
</>mol A = R) D D (cAl - CAz) (IV.8)
2- 1
(b) Diffusion through a stagnant body. cb':notB = 0
This process can be observed, for example, during selective absorption oi
extraction of material A from a mixture of A and B. For this situation
equation (JV.3a) can be simplified to:
A." _
0
c dcA
'PmolA,x- - C- CA dx
(IV.9)
Application of the mass balance (equation IV.l) yields, again, that under
stationary conditions and no production of A the molar flux of A through aU
planes x =constant must be the same. So :
de A
-=constant
dx
The above equation can be integrated and, applying the boundary conditions
x
1
, cA
1
and x
2
, cA
2
, we find for ~ h e concentration distribution:
C - CA =,1 C - CA>) (x-.xl)/(xl -.x!)
C - CAl \c- CAl
(IV.lO)
This concentration distribution is also shown in Figure IV.l for materials
A (e.g. solvent) and B (e.g. solute). Wi th the aid of equation (IV.lO) we find
from equation (IV.9) for the molar flux for diffusion through a stagnant
body :
, ~ , . , . _ Oc In c - cA
2
'f'moiA,x - X2 - X 1 C - CAl
(IV.ll)
If cA is at all places much smaller than the total concentration c, the logar-
ithmic factor in this relation can be approximated by :
}n C - CA2 ~ _ CA2 - CA l
C - CAL C
and we find :
(IV.12)
the same result as for equimolar diffusion.
231
Mass transport through a stagnant fluid is a factor of
I' _ C } C - CA2 _ 1 1 CAt + CA2
JD- n - + 2 + . . .
CAl - CA2 C - CAl 2
bigger than during equimolar diffusion. The correction factor fD is named
after Stefan; if trus factor is applied to equations (IV. 7) and (IV .8) these
equations will also correctly describe, for the geometries considered, diffu-
sion through a stagnant fluid.
I V.J.2. Mass transfer coefficients
Analogous to the introduction of the heat transfer coefficient ex in paragraph
111.1.3 whlch was defined by:
dT
<Pi A dx x=O
ex = T ... - ( T) = - Tw - ( T)
we can now introduce a mass transfer coefficient k which is defined as the ratio
of the mass flux and the concentration gradient (which constitutes the driving
force) in one phase :
A.." dx
k
'+'mol,A __ rn. xO
A = - !LJIA
cAo - (cA) cA, - (cA)
(IV.l3)
Here, 4>':no
1
A is the mass flux of A into phase Band cA, - ( cA) is the difference
between the concentration at the interface and the average concentration of A
in phase B (mean cup concentration). For diffusion from a sphere into an
infu1ite stagnant medium we find, combining equations (IV.13) and (IV.8)
(the latter for the case D
2
~ ex:>):
The dimensionless number kAD/ 0 is called the Sherwood number Sh, which
plays the same role in mass transfer as the Nusselt number in heat transfer. For
diffusion from a sphere into a stagnant infinite fluid Sh = 2, analogous to
Nu = 2 for heat conduction. The dimension of the mass transfer coefficient
is the same as that of velocity, distance{ unit of time.*
If a certain component is exchanged between two mobile phases (e.g. gas
absorption, extraction, distillation, etc.) we encounter an overall mass transfer
coefficient K which is composed of the partial transfer coefficients in the different
* In mass transfer with gases the partial pressure instead of the molar concentration is often used
for calculations. A mass transfer coefficient is theo defined as:
4J:C<>1A = kr{pA.w - PA.r)
since for ideal gases c = p/ RT, k, = kJ RT.
232
phases. Contrary to heat transfer (paragraph 111.2.2), however, in the case of
mass transfer it is not possible to calculate the overall coefficient by summing
up the partial resistances (1/ k) to the total resistance (1/ K). This is due to the
fact that the concentrations in both pbases are generally not equal if the two
phases are in equilibrium.
Figure IV.2 shows an interface between two fluid phases through which a
stationary mass flux </l' of a certain component occurs ; the mole concentration
distributions drawn on either side of this interface are representative of the
mass exchange between two turbulent flows. According to the definition of the
mass transfer coefficient k we can write :
= - = k"(c;._.J - (IV.l4) :
It may now be assumed that there is equilibrium between the two phases at the
interface, so that c:...w and cA,w are related according to the equilibrium relation
for this system.

0
...
c

u
c
0
u
I
Phase
1
I
cA,f
i'-lnterfaee
I
c'
A,w
P
. 10
nose
I

A,w
Distance
Fig .. e IV .2 Concentration profiles near the
interface of two liquid phases
In principle, with this equilibrium relation cA,w and c;.,w can be eliminated
from the previous relations, so that a relationship between <P;o, A, k' , k", c'A./
and c}.J remains. An analytically simple relation is only obtained if the equi-
librium can be represented by means of a relation of the form:
where m represents the distribution coefficient. At low concentrations the
distribution of a component over two phases approximately satisfies such a
relation.t
Elimination of c:.. .... and with this distribution law gives:
A." = (2_ + _
1
_)-
1
(c" - cA,JJ = K"(' c" - ci.JJ (IV.l5a)
If' mol A k" mk' A ,f m A,f m
t Henry's law for gas-liquid and Nernst's distribution law for liquid-liquid systems: the relation
between the distribution coefficient m and the Henry coefficient He is m = HefRT.
233
or also :
II - !I ! t If I
l
m 1) -t
moi A - k + k' (meA,/ - cA./)= K (meA./ - c,..,
1
)
(IV.15b)
It appears that the total mass transfer resistance (1/ K", if related to the phase J'
and 1/ K' is related to the phase ' ) is composed additively of the partial resist-
ances 1/ lt and 1/ kH, but that at the same time the relation of the equilibrium
concentrations is involved. The reader may check that the partial resistance
of the phase in which the equilibrium concentration is lowest is relatively of
the most importance.
Further. it appears from these relations forK that the concentration difference,
or 'the driving force' with the help of which k
11
and k' have been defined, rep-
resents the deviation from the equilibrium between the two phases. If we write
instead of cA..rfm., (i.e. the concentration phase " would have if it was in
equilibrium with cA..r ), and instead of meA./ , cA:
1
, we get the more general
definition for k' and k" :
A." = K '(c'* - c' ) = K"(c" - c" *)
'f'mol A A./ A,/ A,/ A./
When in a certain apparatus (mass exchanger) a mass flow </>mol A of a certain
component must be transferred from one medium to another, the required
exchange area A follows from an equation of the form:
where K and c are related to one of the two phases. If K does not vary too
much over the apparatus, in the case of cocurrent or countercurrent flow of both
phases, for ( cA- the logarithmic mean between the driving forces at the
inlet and outlet of the apparatus can be taken (see also paragraph
IV.l.3. General approach for the calculation of concentration distributions
The partial differential equation which forms the basis for calculating con-
centration distributions is obtained if the law of conservation of matter is
applied to a small (e.g. Cartesian) volume element of the medium. If the fiow
which may come from the diffusion itself is neglected, so that Fick's law
(equation IV.2) can be used for a fixed surface element in the medium, the deriva-
tion is completely analogous to that of the general differential equation for the
heat transport (see paragraph
The resulting equations which we can now derive are for a component A:
(IV.16a)
234
where r A = production of A in kg/m
3
s (barocentric system) and :
dcA dcA dcA dcA
dt + Vx dx + Vy dy + V: dz =
( d
2
c d
2
c d
2
c }
ff)Al dx; + dy; + dz2A + RA
(IV.l6b)
whereRA = production of Ainkmoljm
3
s. Strictlyspeakingt these two equations
only apply at constant IJ) A and if PA p orcA c. In that case they can also
be used if more than one component is present in a low concentration in the
medium.
TI1e diffusion equation for low concentrations with constant ID and without .
chemical reactidn (also called Fick's second law) corresponds entirely with
Fourier's differential equation (equation I 1 1 . 2 ~ The solution of a great many
diffusion problems can therefore be found in books on heat conductivity (see,
for example, Carslaw and Jaeger).
The distribution of the neutron concentration in a nuclear reactor is in
principle also described with an equation such as (IV. l6), where rA, the neutron
production by nuclear fission, is proportional to the neutron concentration.
The production terms r A and RA usually depend on the concentration in the
mixture in a way which is determined by the chemical kinetics of the reaction.
For simple kinetics and for simple geometric situations solutions of equation
(IV.16) are known, especially for the stationary statet We shall return to this
subject in paragraph IV.S.
IV./.4. Film theory
Analogous io the boundary layer theory in heat transfer (see paragraph
111.3.2) Lewis and Whitman introduced the concept of a boundary layer to
mass transfer. They assumed the mass transfer resistance to be located in a
stagnant boundary layer of thickness be (see .Figure IV.3) through which mass
transport has to take place. The mass balance (equation IV.16b) reads for this
case (stationary state):
(IV.l7)
At the interface (x = 0) the equilibrium concentration cAo is present, whereas
on the other side of the film (.x = <5,) the concentration equals the bulk con
centration cAt Solving equation (IV.17) with these boundary conditions
yields the concentration distribution:
CA -CAl = 1 - X
CAo-CAI 8,
(IV. l8)
235
I
---x
Figure IV .3 Film theory eoncentration distribution
With the aid of this concentration distribution we can now calculate the mass
transfer coefficient for which we find:
deA l
k = - DA dX x=O [)A
CAO - ( CA) = t J ~
Thus, the Sherwood number is given by :
kD D
Sh= -=-
[)A ~ c
(IV.l9)
(IV.20)
and we see the formal analogy between the Shand the Nu number (Nu = D/ br)
and the dimensionless expression jJRe ( = D/ll,J, which .all represent the ratio
between geometrical scale and the thickness of the boundary Layer for the
transport process concerned We may not conclude from equation (TV 20) that
the mass transfer coefficient is linearly dependent on the diffusivity because
!Jc is also dependent on [)1, as will be discussed later.
The thickness o f ~ ' of the mass transfer boundary layer cannot be determ.ined
directly but can be calculated if the mass transfer coefficient is known. So,
the film theory does not actually help us to predict k values. The advantage of
this theory is the simplicity of the model and the possibility to estimate the
order of magnitude of the film thickness over which the mass transfer process
occurs. A further advantage of this theory is the fact that the influence of a
chemical reaction on the mass transfer process can easily be studied: We will
therefore use the film model extensively in paragraph IV.5 when treating mass
transfer with chemical reaction.
236
I V.l .5. Problems
*1. The vapour pressure of naphthalene at room temperature is approximately
005 mm Hg. If a mothball (diameter 1 em) consisting of this material
is suspended in still air, calculate the initial rate of vaporization. How long
does it take before the diameter of the mothball is reduced to half its initial
value([)) Of naphthalene in air = 0-7 X 1 o-S m
2
/ S, p - 1150 kg/m
3
)?
Attswer : 4>m = 115 x 10-
10
kgfs; t = 23 years
2. If in a gas phase a reactant A diffuses to a catalyst surface for dimerization,
the mole flux of A is given by:
Prove this.
..1,.11
=
_[])A 2c de A
2c- cA dx
3. If in a desiccator a wet plate is at a distance of L = 0.1 m fTom a layer
of a drying agent, calculate the drying time if:
total concentration c = 4-5 x 10-
2
kmoljm
3
;
equilibrium vapour concentration of water under the prevailing condi-
tions C = 11 X 10- J k:Jnolfm
3
;
diffusion coefficient of water vapour 0 = 25 x 10-
5
m
2
/ s;
surface area of the plate = 12 x 10-
4
m
2
;
amount of water to be evaporated = 12 x 10-
3
kg.
The desiccator can be considered to be isothermal and convective :flow
can be neglected.
4. In a closed gas burette two equal volumes of a pure gas and a pure liquid
are brought into contact with each other under atmospheric pressure and at
room temperature. After shaking vigorously for some time the pressure
appears to have become two-thirds atmospheric, whereas the temperature
has not change<l Calculate the distribution coefficient of the gas in the
liquid
Answer: m = (concentration in liquid/ concentration in gas phase) = 0.5
* 5. Small bubbles (original diameter d
0
) of a pure gas are brought into a liquid
in which they ascend very 'Slowly. During ascending the bubbles disappear
by absorption (Sh = 2). It appears that the time of solution t and the
original bubble diameter satisfy the relation = x t where the
constant has the value.2S x 10-
9
m
2
/ s. Give a theoretical explanation for
this -relation and calculate from the constant the diffusion coefficient of the
I
gas in theliquid,ifthesolubility of the gas is given by m = 05.(concentration
in the gas phase divided by concentration in the liquid phase). Assume that
237
the pressure and consequently the concentration in the bubbles remains
constant.
Answer: fi) = 25/ 16 x 10-
9
m
2
f s
6_ A bottle filled with dietbyl ether is connected with the surrounding air
(T = l5C) by a long vertical tube (L = 50 em, D = 2 em). The diffusion
in the tube is determinative of the rate of evaporation of the ether. Per
second, 165 x 10-
9
kg ether evaporates.
(a) How much ether t/Jm would evaporate per second if the tube were twice
as long and twice as wide?
(b) What is the diffusion coefficient of ether in air?
(c) Make a rough estimate of the time one should wait after filling the bottle
before the diffusion coefficient can be measured with this set-up.
The vapour pressure of. ethyl ether at l5C is 485 x 10
4
Nj m
2
.
8
kgjs
(b) 130 X w - s m
2
js
(c) ca 120 min
7. The air pressure (15 atm) in the tyres of Klaus' car decreased by 005 atm
in 6 months. What is the diffusivity of air in rubber?
Data: volume of tyre 251
surface of tyre 07 m
2
wall thickness 1 em
solubility of air in rubber m = 007 m
3
j m
3
1 1
m
2
j s
8. Somebody measures the diffusion coefficient of water in oil by pouring a
layer of oil onto a layer of water in a glass vessel (cf> = 68 em). The vessel
is put in a thermostat (21C) and the space above the oil is kept dry with
P
2
0
5
The weight decrease of the vessel is measured as a function oftime.
The results are as follows :
Layer thickness (em)
000052
013
0-37
0.63
t-35
226
Weight loss glass vessel (gfday)
3-14
0.0116
().0042
0.0032
().0046
00053
Calculate the diffusion coefficient of water in this oil and explain the results.
The solubility of water in the oil at 20C is 0004 i>er cent by weight ; p
oil = 880 kgfm
3

Answer: 0 = 14 x 10-
9
m
2
/s
238
*9. Show that John drew the right conclusion in t he case of the missing lawyer
discussed at the beginning of this paragraph.
Problem I
Since the air is stagnant. Sh = 2 is valid (paragraphs IV.l.2 and IV. l.4-)_ We
can assume the partial pressure of naphthalene to be equal to zero at a great
distance from the mothball and be equal to the saturation pressure p* at the
surface of the ball (no resistance at the interface). So, the molar flow rate is
given by :
..l..
1
= kAc* = 2 nD
2
_.!!_
'l'mo D RT
which leads to mol = l- 18 x 10-
12
kmoljs and, with the molar weight of
naphthalene being M = 106 kg/kmol :
4>m = MQ>mol = 125 X 10-
10
kgls
A mass balance of the mothball shows:
dV
2
dD
cPm = - P dt = -4npD dE= Mc/Jmol
Substituting for mol the expression found above and integrating between
t = 0, D
0
; t, D we obtain :
p(D6- D
2
)RT
t =
8M[)p*
which leads tot = 712 x to- ss ~ 23 years. In reality, the evaporation of the
mothball does not take so long. This is because, in reality, the air is not st agnant.
Problem 5
The concentration of the pure gas A in the bubbles is cA.t = 1 m
3
/ m
3
and
we can assume the concentration in the bulk of the liquid to be cA,L = 0. We
further assume the concentration at the interface cA..i to equal the equilibrium
concentration c! = cA.Jm (no resistance of interface, pure gas, therefore no gas
phase resistance)_ The volumetric flow rate of A from the bubble is then :
dV
, ~ . = Akc* = - -
'l' v - A dt
and, using Sh = 2 to substitute fork, we find after integration that :
constant m 25 o- 9 21
[0= =- x l m s
8[}cA,g 16
239
Problem 6
We can assume diffusion of ether vapour through stagnant air and equation
(IV.ll) can be applied. The partial vapour pressure of ether at the liquid surface
is the equilibrium pressure p* (no transport resistance at the interface); at the
end of the tube the partial p r e s s u r ~ is zero. In order to roughly estimate the
time before the vapour starts to leave the tube we assume a sharp interface
ether vapour /air which moves at a velocity dL/dt. A material balance then shows :
..
r ~ . * dL
o/mol = C A dt
and after combining this with equation (IV.ll) and integrating between t :::; 0,
L = 0 and t, L we find t = 121 minutes.
Problem 9 John and the case of the disappeared lawyer
The situation is illustrated by the sketch. At the top of the cup, the water
vapour pressure is zero {assuming dry air in the office); at the coffee surface
the equilibrium water vapour pressure p* is present (,.., 20 mm Hg at room
temperature). Water vapour diffuses through stagnant air and equation (IV.ll)
yields the molar flow as a function of the distance L from the top of the cup (total
concentration c = pj RT = 10
5
/ 82 x 10
3
x 298 = 0042 kmol/m
3
).
A mass balance, on the other hand, gives for the mass flow of water vapour:
_ dV dL
Combining both equations we find (M = molar weight of water):
AOp p - 0 </>m pA dL
<f>mot = LRTln p- p* M = M dt
Integrating with the boundary conditions t = 0, L = L
1
= 1 em and t, L 2 =
4cm, we find, with IJ) = 25 x 10- s m
2
/s, t = 533 h ~ 17 days and, again,
John is right.
240
IV .2. Non--stationary diffusion
The skin diver, busily filling rhe cylinders first with oxygen
and then with nitrogen at the desired pressure, cold John
that the search for the drowned man had been delayed by the
fact that no bottles with pressurized air were available.
He had found, however, cylinders with pure oxygen and
nitrogen and he would start the search in a few minutes.
John thought: the diver's cylinders are standing upright and
each hasadiameterof20anand a length of60 em. In order to
avoid another casualty, he suggested rhat the diver fill the
cylinders when they were in the horizontal positicm and wait
half an hour before using them.
of the analogy between the micro balances for mass transfer (equation
IV.16) and that for the heat transport (equation III.4) the non-stationary
penetration of matter into a medium can be treated easily and shortly. If in a
rigid medium the concentration is cAo and at time t = 0 the interfacial con-
centration is brought to cAh the mass transfer is given as a function of the time
with the solutions of paragraph II12 if the following substitutions are made :
a = ).J pc P --+ I!) A
(pcpT)-+ CA
(pcpTo)-+ cAo
(pcPTl)-+ cAi
,/,.11
'Y H -+ o/mol A
For shorter times (Fo = 'D AtfD
2
< 005) the penetration theory applies and
we find for the mole flux:
(IV.21)
(analogous to equation Ill.24). For longer times (Fo = 'OAtf D
2
> 01) Sh =
constant applies and Figures 1Il8 and 111.9-after the correct substitutions-
can be used. (It is customary to indicate the characteristic number for non-
stationary heat (and for mass penetration) as the Fourier number, Fo.) We see
that in the case of short times (equation IV.21) the mass transfer coefficient
kat any timet is g]ven by :
(lV.22)
If the total mass transfer operation lasts a time te the average mass transfer
coefficient is found as:
(IV.23)
241
Since the penetration depth for mass transfer ( = jrl)i) is usually very small,
because of the low values for the diffusion coefficient and the short contact
times characteristic of mass transfer (usually 1 s), the penetration theory is
even more frequently used for the description of mass transfer than for heat
transfer.
Especially in cases where the interface (because it flows) is refreshed, the
penetration theory is usable. For ascending bubbles in a liquid the maximal
contact time of the surface elements is given byte = d/v (d = bubble diameter,
v ~ rate of ascension, the 'jacket' of the bubble is renewed every time it has
moved a diameter). For a C0
2
bubble. which ascends in water at a rate of05 m/s
and which has a diameter of 10-
2
m, te = 2, x 10-
2
s. Further. fflco
2
= 2 x
10-
9
m
2
/s (20C, 1 atm). so in this case the m ~ a n mass transfer coefficient in the
liquid phase (k) = 2 ~ = 36 X 10-
4
mjs (mass transfer coefficients for
aqueous solutions under normal conditions are invariably in this order of
magnitude).
Based on the idea that in many apparatus for liquid- liquid or gas-liquid
contact the interface is refreshed because there is flow, it has been proposed to
introduce an age distribution for surface elements in the apparatus, and on this
basis to calculate mean mass transfer coefficients ( k) with the penetration theory.
If 1/f(t) dt is the fraction of the surface with ages between t and t + de, according
to this theory :
( k) = f <Xl t/J(t) (0 dt
Jo y;t
(IV.24)
Usually the age distribution t/l(t) is unknown. In the literature two theories
about f/l(t) can be found. Higbie assumed that the probability to nnd a surface
element of age t for all times between t = 0 and t = te (the maximum possible
age) is equally great, so:
and :
dt
t/l(t) dt = - for 0 ~ t ~ t,
te
t/l(t) = 0 for t > te
..
This is a supposition which is correct for, for example, the absorption of a gas
in jets of liquid, in laminar liquid films and in swarms of falling, rigid liquid
droplets.
In these cases the mean mass transfer coefficient is :
as we had already found with equation (IV.23 ). Danckwerts supposed that every
surface element, independent of its previous history, has at any moment a
242
chance s dt to disappear in the subsequent period of time dt. In this case:
This can be realized by writing this relation as follows:
-dlfr(t) = t/t(t)s dt
which in words means : 'The chance that the surface element disappears in the
time between t and t + dt equals the chance that it is still there at timer multiplied
by s dt. ' Using this supposition we find with equation (IV.24):
<k> = foS
(IV.2S)
The constants is called the surface renewal frequency (unit : s-
1
) . The practical
use of Danckwerts theory is restricted to those cases where a priori something
can be said about the renewal frequency; those cases are rare, unfortunately.
This theory can be applied, for example, for calculating the mass transfer
. coefficients for liquids flowing over a packing material. If the liquid has at any
contact place between two packings a chance p to be completely m i x e ~ the
frequency s is given by the ratio of p and the residence time on one piece of
packing.
The same is valid for a dispersion of small, separate droplets which have a
certain chance (independent of their previous history) to leave the dispersion.
The droplets must stay in the continuous phase for a short time only, so that they
are not completely extracted. For small droplets staying in the continuous phase
for a long time, k = 2rJ/dP describes the mass transfer coefficient in the continu-
ous phase;
IV.2.1. Problems
1. A jet of water (diameter 2 mm, temperature 20C) with uniform velocity
distribution (v = 5 mjs) falls vertically through practically pure C0
2
gas
under atmospheric pressure. How much 00
2
is absorbed per unit of time
by the first 10 em of the jet? The solubility of C0
2
in water at 20C and
1 atm C0
2
pressure is 173 kg C0
2
/ m
3
. The diffusion coefficient D = 17
x 10-
9
m
2
fs.
1
kgjs
2 A waJl is coated on one side with a 02 mm thick layer of paint consisting
of a very volatile component and a heavier, non-volatile component.
The paint is aUowed to dry, i.e. the volatile component is caused to evaporate.
(a) Calculate with a simple physical model how long it takes before the
drying front has reached the interface paint wall.
(b) How long does it take before 99 per cent of the volatile component has
evaporated?
243
The diffusion coefficient of the volatile component in the paint is. D = 22
x 10-
11
m
2
/ s.
(b) 57 h
*3. A layer of water and a layer of toluene are brought together at time t = 0.
Both layers contain 10.kg/m
3
of an iodine compound. The ratio of the
equilibrium concentrations of this iodine GOmpound in toluene and in water
is 10: 1 ; the ratio of the diffusion coefficients of the iodine compound
in water and in toluene is 4. In what direction is the iodine compound
transported? Calculate for relatively short times the concentration at the
interface for both phases, assuming that the transport takes place by diffu-
sion only. Sketch the concentration distribution in both phases. .
3
; cit= 25 kgjm
3
*4. A short laminar jet of pure water (temperature 20C) falls through pure
S0
2
. The gas has a temperature of20C and a pressure of 1 atm. With what
theory can the rate of absorption of the gas be described?
(a) Calculate the surface temperature of the Vfater jet if under the prevailing
conditions the solubility of so2 in water is 154kmolfm
3
, the heat of
solution Is 67 kcal/ mole and the Lewis number (Le = aj D) is 90. In
this calculation neglect the heat transpon to the gas phase.
(b) How much higher wi11 the rate of absorption be if the jet falls twice as
fast ?
(c) What is the influence of the jet diameter on the rate of absorption?
(In practice it appears that the jet contracts because it is accelerated by
gravity. What will be the influence of the contraction on the rate of
absorption considering the answer to the previous,question?)
(b) fi higher
(c) no influence if flow rate is constant
5. The discontinuous phase of a continuously operated emulsion reactor
shows the residence time distribution of an ideal mixer (63 per cent leaves
the reactor within 5 s). Give an estimate of the lower limit of the partial
mass transfer coefficient in the conti.nuous phase. Can the partial mass
transfer coefficient in the discontinous phase also be estimated ([} > 05
x 10-
9
m
2
/ s)?
5
mfs; yes, ~ 10-
5
mj s if dp > 2 X 10-
4
m
6. For large bubbles with a free-flowing interface (model: a half-sphere with
diameter D) the rate of ascension is :
v,. = iJgf)
244
Prove that the mass transfer coefficient for the free-ftowing interface of
these bubbles is given by :
( k) = !6,f!!f
7. A water droplet with a diameter of 2 mm fall s 4 m through surrounding air
(20C). Originally the droplet contains no oxygen. What is its mean oxygen
concentration after 4 m? If it contains originally 5 mg 0
2
/ l, what would
have been its mean oxygen concentration after 4 m? The distribution co-
efficient of 0
2
in H
2
0 is 0033 (20C).
2
/1; 534 mg OJ.!]
*8. The liquid hold-up H, in a packed column at very low gas velocities (below
loading point) can be estimated from the relations for a free-falling liquid
film to be :
H, = a f ; ~ T
1 3
(m
3
fm
3
)
where v
0
= the superficial liquid velocity (m
3
fm
2
s). Prove this.
So the mean residence time of the liquid is given by :
L
t
1
= H
1
-
vo
Estimate the mass transfer coefficient in a column packed with 1 inch Raschig
rings (a= 200m
3
fm
3
) at a liquid load of v
0
= 10-
3
rp.
3
jm
2
s of water.
Answer: k = 42 x lo-s m/ s
*9. Show that John gave the right answer to the diver in the case of the drowned
man discussed at the beginning of this paragraph.
Problem 3
Since ct fcw at equilibrium = 10, the iodine compound is transported into the
toluene phase. For short times we can apply the penetration theory and we find :
,ff;
l
;, = (c,., - c;..,) = (c;, - C
1
) -
1tl
Assuming equilibrium at the interface, thus c;Jc; ..... = 10, we fi nd with the given
ratio of diffusion coefficients :
cil = 25 kgjm
3
; c,.,., = 25 kgfm
3
The penetration depth is given by :
b = JnOi
and since Dr= !Dw, we find :
A-U
V r - 2Vw
245
which enables us to draw the concentration distribution curves in both phases.
Problem 4
Applying the penetration theory the mass flux of S0
2
into the jet at any time
is given by:
If we assume that :
aU heat of solution is produced at the surface of the j et ;
there is no heat loss into the gas phase ;
the temperature in the centre of the jet: remains constant
we can describe the transport of heat into the jet by the penetration theory :
l/J'H = pc l!T /a
p y;i
Further. the heat to be transported is related to the mass flux of S0
2
into the
jet by:
and we find :
4 T = ci {[ . 109C
pc, va
The total mass flow into the jet is given by (using t = L/v) :
( </>
50
,) =

So, if the jet velocity is doubled, is doubled and c/Jso
2
jncreases by a factor
.of .fi. We further see at constant cPv the diameter has no influence on the
mass flow rate.
Problem 8
For the thickness of a falling laminar film we found in paragraph II.l :
where :
d = f 3y(v)d}t
.. g
cPv <f>vL <f>vL Vo
<v)d = width= -A- = -aV- =- a
- ------- - - -- -
246
Since H
1
H
_ {3v
0
v}'i

ga
If we assume the probability t hat the liquid .is completely mixed after each
packing particle to be p = 1 we find for the frequency of surface renewal:
1 L v0 1
s= -r:P = dpTL = Hrdp = adP = 177s- l .
and with Dankwaert's theory we find:
i[) .' k = .jii)s

x 1-77 = 42 x to- s m/s
Problem 9 John and the case of the drowned man
. , !.
The contents of the bottles will be practically homogeneous if:

Fo- -
2
02
4L
(see Figure Estimating the diffusion coefficient of 0
2
in N
2
at 2 x. lo-s
m
2
j s, John thus finds t = 4 hours for the bottles fil led while standing upright.
If the bottles are filled when they are in the horizontal position the distance
L over which concentration equalization has to occur is 1 and so t!'te time
needed is of 4 hours.
IV.3. Mass transfer with forced convection
John was busy doing work he hated, namely cleaning his
pipes. He realized that the rate of growth of rhe tar layer in
the stem was completely determined by mass transfer.
Contemplating that air flow in the pipe was laminar, he
concluded that he could decrease the growth rate of the
layer by a factor of 0-79 if he halved the rate of air flow.
Wondering whether this measure would influence his
pleasure, he started to fill the pipe from the humidor on his
desk.
IV.3.1. Analogy with heat transfer
For mass transfer between a wall and a convective flow the same analysis
can be followed as for heat transfer between a wall and a forced flow (paragraph
III.3) provided mass transfer is considered at low concentrations and without
chemical reaction. This can be understood from the fact that with the following
'transpositions' the entire description of the mass transfer (without chemical
reaction) is also valid for the description of the heat transfer.
Heat transfer
pcPT
i./peP = a

a/peP
Mass transfer (component A)
247
Thus, if for a certain geometric case of heat transfer with forced convection it is
concluded that, for example:
Nu = f
1
(Re,Pr, Gr, Fo, Gz, ... )
it follows from the above that for the same case with the same type of boundary
conditions applies:
Sh = /
2
(Re, Sc, Gr, Fo, Gz, ... )
We have already come across the Sherwood number, Sh = kD/ 0, in para-
graph IV. l. The Schmidt number, Sc = vj [[), performs in mass transfer the
role which is played by the Pr number in heat transfer; it gives the ratio
between the thickness of the hydrodynanric and of the concentration boundary
layer. For gas mixtures Sc is about 1, for liquids Sc is considerably higher
(10
2
- 10
3
for normal liquids), because [b is so much lower than v. Whereas for
gases Sc depends hardly on the temperature, Sc decreases strongly for liquids
the temperature increases.
Before it can be concluded that the results for beat transfer (i.e. the function
f
1
) can be transposed to a reliable relation for mass transfer (function f
1
= f
2
),
it should be ascertained whether the mass transfer process does indeed proceed
entirely analogously to fhe (assumed to be known) heat transfer process. This
means that the functions /
1
and f
2
are only equal to each other if the Re range
used for mass and heat transfers is equal and if the Sc range in which the relation
has to be applied equals the Pr range in which the heat transfer relation applies.
For gases where Sc Pr this 'transposition' can be carried out frequently ;
the relation for heat transfer for liquids cannot always be transposed to usable
relations for mass transfer because for liquids Sc Pr. On the basis of the
analysis for heat transfer between a wall and a turbulent flow (paragraph.III.3.2)
the reader can now establish for himself that in the range 2 x 10
3
< Re < 10
5
,
Sc > Q. 7 a usable relation for mass transfer between a wall and a turbulent flow
is given by :
Sh = 0027

Sc
0
'
33
(IV.26)
With these analogy relations mass transfer values, for example, can be
from t heoretical or experimental data about heat transfer and
vice versa. Now, particularly if Re is not too low, the local or the mean transfer
coefficients over a surface can be calculated with good approximation from
formulae such as:
Nu = CRem Pr
11
Sh = C Re'" Sen
The exponent m varies from ca. j (for the laminar entrance region in a pipe)
via 05 (in the case of flow around spheres) to ca. 08 (for turbulent fl ow in
pipes), but in all these cases (except with liquid metals) the exponent n of Pr
(and also of Sc) is about equal to t. Chilton and Colburn have used
this for representing the analogy between heat and mass transfers in the following
----- - - - -
248
form. They defined a heat transfer :
j
8
=

Pr-t
and a mass transfer value:
....
j D = Sh Re-
1
sc-i
Using these definitions the analogy for geometrically similar cases and not too
low values of Re becomes :
(IV.27)
For turbulent flow through pipes and along fiat plates this analogy can be
extended to:
. . 1j
)g = }D = 2
(IV.28)
where f is the so-called Fanning fricti on factor (see paragraph 11.2, equation
11.22). This analogy with momentum transfer, however, applies only in those
cases where f is caused by wall friction. The flow resistance of a body caused by
eddies which can be an important part of the overall flow resistance is not
attended by analogous effects in t:he. beat or mass transfers. So for these cases
equation (IV.28) does not apply, but equation (lV.27) does.
The analogy between heat and mass transfers also applies to free convection.
The density difference ap which occurs in the Grashof number (see paragraph
Ill.6) might originate from concentration differences as well as from temperature
differences. If there are differences both in concentration and in temperature
(combined heat and mass transfers) the influence of both on Gr can be
into account.
IV.3.2. Mass transfer during laminar fl ow
The mass transfer process can be calculated accurately for a number of
simple fl ow situations, the results then being collected in a catalogue. Although a
certain problem is hardly ever completely equal to a case from this catalogue,
it is possible by proper estimation to approach the act ual problem with a
catalogue case. We can tell the engineer' by this 'estimation'.
It should be borne in mind that three data on the velocity field near the
boundary layer are of decisive importance for the mass t ransfer. These data
are, in order of importance :
(1) the velocity of the boundary layer and that of the fluid (li; and vx);
(2) the velocity gradient perpendicular to the boundary layer and,
(3) the velocity perpendicular to the interface near the surface of the mass
transfer boundary layer (v
1
) .
We shall discuss here five cases which are represented in Figure IVA. These
examples contain the characteristics for the flow field just mentioned. The
transfer process is described with the following mass balance (see equation
249
Case t
JC
Case 2.
Case 3.
v.., = wy+v;
Vy:: 0
Case 4.
vx = vQ;)} Ou1side
vy = 0 boundary layer
v.r * vQ;)l vy *OJ Inside
vi = 0 boundary Ioyer
case 5.
Figure IV.4 Five cases of laminar ftow situations
IV.I6b) and the boundary conditions:
b
2
c be oc
0 = 1[}- - v -- v-
by
2
x5x
1
by
(IV.29)
with c = c
0
for y = 0 and for all x (at the interface the equilibrium concentration
prevails), c = 0 if x = 0 for all y (the infiowing liquid has the concentration
zero) and c = 0 if y = co for all x (no accumulation of substance very far from
the interface).
Case I. Uniform velocity parallel to the interface
Here vx = vi and v, = 0, and the boundary layer also moves at a velocity vi.
Tbis case bas been discussed by Higbie. By moving with the ftow, it becomes a
250
case of non-stationary diffusion (penetration theory wi th t = xjv
1
) . The solution
is:
or :
k = (0 = ro;;;
v -:m -.J-;;
kx = Sb =

= 0-565 Ret Set

[j) X Jnv [) X
The mean mass transfer coefficient is given by :
( k ) = .!._ t k cbc = 2kL
LJo
(IV.30)
The absorption of gases in liquid jets and in laminar falling films proceeds
according to this description.
Case 2. Flow with a constant velocity gradient
Here vx = wy and vY = 0, and the boundary layer is stationary. Leveque*
has solved this problem:
- = Sh = 0539 -
kx (wx
1
)t
0 X [)
(IV.31)
and :
( k) =
For a laminar, Newtonian pipe flow (diameter D) :
dv; s<vx>
W=- -
dy y= O
D
The reader can compare this result with the analogous heat transfer problem
of Graetz (equation III.46b). For mass transfer this case is not so important
but it is a borderline case of the following case in which both an interface
velocity (vi) and a velocity gradient (w = dvJdyly=o) play a role.
Case 3. lnteiface velocity and velocity gradient
Here Vx = vi + wy and vy = 0. The complete solution of this case is known.
In pracrice, approximations of chis solution are sufficient. These approximations
apply tow sufficiently small and vi s ufficientl y small. The characteristic dimen-
sionless value is w
2
'0xfvt. If this value is lower than 1, the penetration theory
(case 1) gives to within 15 per cent the correct mass transfer coefficient. If this
value is higher than 1, the Leveque theory (case 2) gives the correct result to
within 15 per cent
"' Ann. des Mines, 13, 201, 305, 381 (1928).
251
In general, w is of the otder of magnitude of vifL.. The dimensionless group
mentioned can then be written as fD/v
1
L (Peclet number). If we consider bubbles
and droplets with a diameter of at least 1 em and a velocity of at least 01 mjs,
both for gases (D 10-
5
m
2
/s) and fot liquids (Ul 10-
9
m
2
fs), this value is
lower than I. So in these cases the penetration theory can be used. For gas
bubbles with small diameters and velocities the penetration theory is no longer
applicable.
Case 4. Stationary iitterface and a velocity vx Vex; sufficiently far from the
intelface
. For the velocity and shear stress along the interface the following
appllies, according to SGhlichting's boundary layer theory :
Also applies :
where the thickness of the hydrodynamic boundary layer. Further applies:
Sh = kx = x
:r [[} (J
t:
Jf the mass transfer boundary layer {Jc is much thinner than the hydrodynamic
boundary layer for the relation of the two boundary layer thicknesses applies:
from which follows:
S
h = 0-332 Ret
X X
(IV.32)
and:
-Now only applies when Sc 1. But for Sc = 1 equation (IV.32) also
yields reasonably accurate results.
For separate solid particles, e.g. in a fluidized bed or in a not too dense swarm
of rigid liquid droplets, equati9n {IV.32) describes the mass transfer satisfac-
torily if sc 1 has been satisfied (L is in this case the particle diameter d p).
In case of particles in a suspension or rigid droplets in an emulsion eq uat]on
(IV.32) can also be used. The problem in these cases is that the relative velocity
v cc of the particle with respect to the fluid is difficult to determine. This velocity
is mainly determined by the type of stirrer, the number of revolutions, the
diameter of the stirrer D and the diameter of the particle dP.
252
Case 5. Moving interface and a velocit y far from the interf ace
This problem lies between case 4 (v; = 0) and case 1 (v
1
= v
00
). The complete
solution is known.* In this case no simple expression is found for the mass
transfer coefficient
The result of thls analysis can be seen in Figure IV.5. This figure shows that
for Sc > 200 the penetration theory can be applied if the interface velocity
v; is higher than 10 per cent of the velocity of the bulk u::o (case 1 at velocity vJ
From this figure it also follows that for Sc = 1 (e.g. with a gas fiow through a
column with packing over which liquid drips) the boundary layer theory (case 4)
can be applied to predict the mass transfer in the gas phase, if we use as the
characteristic velocity the sum total of the effective gas rate and the sudace _
rate of the film (v
00
+ v;). This result has been experimentally.
0600.----------------- -----,
I
Cow 1. wi1h v,-= "co
Cose 1. wi th cttOI'octeri stic
veloci ty v;
Figure IV .5 The mass transfer coefficient in the flow sit uation for
case 5
Appf_ Sci. Res. lOA, 241 {1961).
253
IV.3.3. Mass transfer during turbulent .flow
; .
If the laminar boundary layer exceeds a certain thickness, the boundary layer
flow becomes turbulent (approximately for VxlJJv > 5 X 1 0
4
). The eddies then
penetrate into the area where the mass transfer takes place. The {statistical)
velocity components in the y-direction contribute, together with the diffusion,
to the transport perpendicular to the interface. Quite near the interface where
the eddies hardly penetrate (laminar sublayer) transport takes place exclusively
as a result of diffusion. At larger distances from the interface the {statistical)
transport prevails by entrainment In between lies an area where both processes
are equally imponant.
Since the transport by turbulent motion is statistical, just as the diffusion,
this transport is described by means of an eddy mass diffusivity E which depends
on y. Equation (IV.30) then becomes:
</>':nol.y = -{0 + E)d:;
{IV.33)
Since the transport of matter and substance proceeds analogously because of
the eddies, the momentum flux can be described with the same diffusivity E:
dv.x
"t = - p(v + E) dy
{IV.34)
The next thing to do is to measure vx(Y); to assume t in all layers near to the
interface to be constant and equal to the shear stress on the wall ; to determine
E(y) using equation (1V.34) and then to use this result for calculating k using
equation (1V.33). To this end equation (IV.33} is written as follows:
Sh = kx = = dy }-
1
X [J) ro
[) J o de A . o \ 1 + tO} .
(IV.35)
A drawback to this method, of course, is that particularly close to the wall
where the resistance to the mass transfer is found v:x:{y) is difficult to measure.
As a consequence the E(y) values derived from the velocity distribution are
inaccurate, which causes errors in the estimation of Shx by this route.
Most data are known about steady-state turbulent flow along a solid flat wall.
The number of suggestions for E(y) is large. All these suggestions can be sum-
marized as follows:
where :
_k = St = __ f_/_ 2 -==
v:n 1 + g(ScVJ72
St (Stanton number)= Sh/Re Sc
f = Fanning friction factor for the solid wall = 2tw/
g(Sc) = function of Sc
(IV.36)
254
In the case wbere Sc = l , g == 0 applies and 2k = fvr:t> (Reynolds' analogy).
Further, g is the following function of Sc:
Sc 1 5 lO
g 0 27 52
100
330
( 10%)
500
1000
( 15 %)
1000
1800
(20%)
2000
2800
(30%)
The greater error in the function g for higher values of Sc is mainly caused
by the various approximations of E(y) near the wall DeissJer, for example,
finds g = 90 S c ~ (Sc > 200); Reichardt finds in the same Sc range g = 88 Sc-i
and Ling = 18 S c ~ .
These results can be used in the description of mass transfer between a pipe
wall and a turbulent flow (e.g. the free space in a packed bed can be considered
as a system of irregular channels). The friction factor fin equation (IV.36)
should then exclusively describe the friction resistance and not the form resis-
tance. This factor should also be related to a Reynolds n umber in which the
hydraulic diameter of the channels occurs as the characteristic length.
I V.3.4. Problems
1. An air freshener is suspended vertically in still air. Calculate the thickness
which evaporates each day.
Data: molecular weight : 200
v = 15 x lo- s m
2
j s
10 = 0.5 x to-
5
m
2
/s
vapour pressure: 10-
3
atm
L = 25cm
p = 1500 kgjm
3
2. The inner wall of a tube has, in the course of time, been caked with a solid
layer of brine of 2 m.m. The heat transfer coefficient ~ at the wall is
1000 Wj m
2
0
C. Calculate the time necessary to dissolve the salt if the tube
is fi us bed under the same conditions with dis tilled water. (PsaJt = 2500 kglm
3
,
the Lewis number under these conditions is Le = a/ fD = 100 and the
solubility of the sa1t is 300 kgjm
3
.)
3. A metal foil has t o be etched. To this end it is drawn through an etching
liquid which flows at the samespeed (1 m/ s)with the foil. The etching process
is fully determined by mass transfer. By what factor does the velocity of
the etching process change if the foH is stopped and the liquid continues
to flow (Sc = 10
3
)?
*4. The Obersalzberg, a mountain in the German/ Austrian Alps near Salzburg,
consists of relatively pure sea salt.. The salt is mined by drilling and digging
255
out chambers of about 60 x 60 m and 1 m height some hundreds of
metres deep into the mountain. These chambers are fil led with water.
Fresh water is added and a 90 per cent saturated salt solution is withdrawn
continuously. Part of the mines can be visited and the guide will tell you that
the height ofthese chambers increases at a rate of approximately 1 em/day.
What speed of solution of the. salt would you predict1 (p
5111
= 2165 kg/m
3
,
solubility = 300 kg/m
3
I Psaturated solution = 1200 kg/m 3, [}Nact = 1 5 X lO-
9
m
2
/ s and v = 10-
6
m
2
/s).
*5. In.a diluted emulsion the rate of circulation of the surface of the droplets
is one-quarter of the relative velocity between the two phases. After a smal1
amount of surface-active substance bas been added, the droplets have
become rigid. By what factor has the external mass transfer coefficient
changed after the addition (Sc = 10
3
)?
6. In a cylindrical vessel {diameter 2m) a 5 mm thick crust ofKMn0
4
crystals
has deposited on the side wall in the course of a crystallization process.
The crystals are removed by filling the vessel with a diluted K:Mn0
4
solution (c
0
= 3 kg/m
3
) of 200C and by stirring vigorously. It is known that
under the same conditions the heat transfer coefficient liquid wall is
1100 Wjm
2
C. After what time has the entire crust dissolved? The solu-
bility of KMn0
4
at 20C is 632 kgjm
3
; the density of crystallized
KMnO.._ is 2700 kgjm
3
and the diffusion coefficient of KMn0
4
in water
is 10-
9
m
2
fs. Let all other substance properties be equal to those of water
at 20C.
7. A liquid flows turbulently along a rough wall. The temperature of the wall
rises by 5C, thus changing the Sc number from 10
3
to 500. How does the
mass transfer coefficient change?
*8. A small cooking salt crystal is dissolved at 20C in a large amount of clean
water which is stirred thoroughly. At what value does the difference between
the water and crystal temperatures adjust itself?
Data : D NaCI = 1-5 x 10-
9
m
2
/ s
AH
5
= 93 X 10
4
Jjkg
solubility = 300 kg/m
3
*9. Show that John drew the right conclusion in the case of his pipe discussed
at the beginning of thls paragraph.
----- - ----
256

We assume the bulk of the salt solution to have a uniform concentration of
90 per cent of saturation, so c<Xl = 270 kgfm
3
. At the ceiling, a thin film of
liquid is completely saturated, ci = 300 kg,lm
3
Because of the difference in
specific gravity free convection occurs, which enhances mass transfer.
By analogy with equation (HI. 75) we can write :
( Sb) = (k)L = 017{GrSc}1 = 017{L
3
g fJ.p
10 v
2
p l 0
Estimating p co :
p<XJ = 1000 + ().9 x 200 = 1180 kg/m
3
we find:
( k) = 123 X 10-
5
mjs
A material balance shows that :
and we find :
dH
--= 148 em/ day
dt
The salt contains small amounts of gypsum and iron oxide which do not dissolve
but settle at the bottom of the chamber. Apparently, these insoluble impurities
slightly lower the speed of dissolution.
Problem 5
The drawings show the flow situations around the stagnant and the mobile
droplet. These flow situations were discussed in paragraph IV.3.2 (Figure
IV.4, cases 4 and 5). So for the stagnant droplet equation (IV.32) is valid:
sb = Set
vCD constonl
VQ)
:
cons toni
I
/
0

v, I
0
Stognont droplet
MobWe dr oplet
257
whereas for the mobile droplet, because Sc > 200, the penetration theory,
-equation (IV.30) can be applied:
Consequently :
--'--- = -- - Sc = 037
k stagnant 0332(v"") t -t
kmobile 0565 Vj
because V; = !vco
Problem 8
We assume that all heat of solution produced at the interface crystal-liquid
is transported .into the liquid with uniform bulk temperature T
00
A simple
heat balance shows :
'H = ct(J; - . ,.._ 4>'/n b.Hs = kci AHs
Now, because of the analogue transport mechanism of heat and mass :
= or =
so:
Problem 9 John and the case of the pipe
In the stem of John's pipe there is laminar flow with flow velocity zero at the
boundary (case 1 of paragraph IV.3.2). For this case the mass coefficient
is proportional to the third root of the velocity gradient, w!- (Lefeque, equation
IV.31). Since for laminar pipe flow w = 8(vx)/D, the mass transfer coefficient
is proportional to ( vx> t and the rate of accumulation oftar is also proportional
to ( Vx) 1-. If, therefore, the volumetric flow rate is halved, the rate of accurnula tion
: is decreased to (t)t = 079 of its initial value.
N .4. Mass exchangers
John put down his cup of tea with distaste and thought about
the detrimental effect of water pollution on his drinking
habits. He knew that his secretary used a coal filter to
decrease the concentration of impurities in the water by a
factor of 10. Remembering chat because of the long dry period
his secretary to i_ncrease the length of the coal column by
48 per cent.
258
The calculation of mass exchangers in which no chemical reactions occur
proceeds analogously to the calculation of beat ex<:hangers. Mass transfer
usually takes place in apparatus in which the two exchanging phases are brought
into close contact with each other. To this end, the one phase is finely dispersed
in the other so that a Jarge exchange area is formed
For selecting the correct type of mass transfer apparatus and for determining
the main dimensions, the following data are of importance :
(a) the thermodynamic equilibrium between the phases in question under the
prevailing conditions (temperature and pressure) ;
(b) the rate of mass transport in the chosen apparatus as a function of the
operational variables (e.g. flow rate, stirring intensity, etc.).
IVA.J. Thermodynamic equilibrium
During heat transport under equilibrium conditions the temperatures in all
phases are equal, i.e. T., = ~ . A T = 0. During mass transport under equilibrium
conditions the activities of a given substance in both phases are equal. These
activities are seldom known and therefore the more practical concept of a
use here exclusively the simplest correlation possible, cA = mc
8
. This simplifies
the calculations considerably but it is not essential for the approach chosen.
For a. large number of cases them value and consequently the slope (gradient)
of the equilibrium line is constant at constant temperature. In such cases there
are sufficient thermodynamic data for calculating apparatus for physical
absorption if the value of m, which is sometimes also called the Bunsen absorp-
tion coefficient, is known as a function of the temperature. Table IV.l gives
a survey of distribution coefficients of some gases in water.
Table IV.l Distribution coefficient m for various gases in
water at 20C. In 1m
3
water at 20C m m
3
gas is dissolved
(reduced t<> ooc and 1 bar)
Gas
Air
Ammonia
Carbon dioxide
Chlorine
Hydrogen
Hydrogen chlonde
Nitrogen
Oxygen
Sulphur dioxide
m
0018
685
086
225
0.018
430
0015
0031
386
In older handbooks the thermodynamic distribution law over these media is
described as foll ows : cA = He P
8
, where P
8
is the par tial pressure of the gas
to be absorbed at the interface. Using the ideal gas law it foJJows that m = He RT
259
(T in oK), He being the Henry coefficient. Table I V.l shows that solubilities may
differ by a factor of 10
4
and that a small difference in chemical structure of the
component to be absor:bed may have a great influence on the solubility.
As soon as the thermodynamics are known. some considerations regarding
the attainable equilibrium should be made prior to selecting the type of mass
exchanger. The thermodynamic equilibrium between the phases under the con-
ditions prevailing in the apparatus indicates what could be achieved in the
case of very intense contact between the phases and at 'very long contact times.
To this end, we take as an example the case where a gas flow ijJ
with inlet concentration c
0
comes to equilibrium by absorption, with a
well-mixed liquid flow 4J
1
with inlet concentration zero.
From a mass balance (mass flow in = mass
cpg(c
0
- c) = mcL
it appears that the concentration in the emerging liquid is given by:
co
c=---
l + m</h
c/>g
(.JV.37)
To ensure the highest possible outlet concentration in the liquid, the ratio
mJJ4>
8
, which is called the extraction factor, should be as small as possible .
. e.g. >, = IOm<f>L The desired ratio between the flows r/>
8
/t/>L is therefore deter-
mined by the distribution coefficient. For poorly soluble gases such as oxygen
it is sufficient to choose r/>g/c/>L 03. For readily soluble gases such as C02,
<P:/<PL should be 10. .
These statements which are based on thermodynamics apply irrespective
of whether absorption or desorption is involved; with readily soluble gases a
. r . .
high Q>g/<PL ratio is necessary in order: ', .
(a) not to exhaust the gas too much and.;. .
(b) in the case of desorption not to build up too high a back pressure of the com-
ponent to be desorbed in the gas phase.
:IV.4.2. Choice of the apparatus
. The type of apparatus with which the mass transfer process has to be carried
out is chosen on . the basis of a number of semi-quantitative considerations.
lf one is not good at selecting the right type of apparatus, further development
of the apparatus does not help much. The .final result will then be the best ap-
paratus of a poor type. _ .
As a first step in the selection procedure the of is considered.
If mr/JL/
8
is high, the g phase contains ultimatelyonly little of the substance
to be transferred, which can be. the aim of the process. If, on the oth.er hand,
thedegree of exhaustion is low; the L phase becomes almost. saturated under the
conditions of the in going gas, which can be the aim of the process.
260
Since when choosing physical absorption apparatus it bas first to be decided
whether the liquid or the gas has to be distributed, it is clear on the basis of the
above considerations that readily soluble gases can only be transferred efficient-
ly at high values of. 4>,/t/JL, so that the liquid will be distributed For some types
of gas-liquid contact apparatus approximate values are given in Table IV.2,
with which the semi-quantitative considerations, on choosing the apparatus,
can be supported.
Table IV.2 Some approximate values for gas-liquid contact apparatus. The flow rate is
v
0
L= 2 x J0-
2
m
3
/ m
2
s
Maximum gas Maximum inter- Liquid Power consumpti<;m
Type of Bow rate facial area fraction per kg liquid
contactor
lio, (m
3
/m
2
s) a(m
2
/m
3
)
(ml/ m3)
(W/kg)
Bubble column 020 300
>06 2
Distillation stage
().3
75
>06 25
(with bubble caps)
Stirred tanlc
reactor
0.15 1000 >0.5 6
Sieve plate 1 150 > 015 l
Falling film 2 100 <005 1
Packed column
l 250 < 005 15
(two phases)
If on the basis of thermodynamic considerations the preference of some types
of apparatus has been expressed, some other consideration may play a part, such
as the pressure drop over the apparatus, the heat transfer, corrosion and the
question of whether a countercurrent way of contacting the phases is possible.
IV.4.3. Size determination of the mass exchanger
If on the basis of the complex of qualitative and quantitative considerations
discussed in paragraph IV.4.2 the type of apparatus bas been determined, we
have to know the mass transfer coefficient and the specific interface as a function
of the operational conditions of the apparatus, so that this apparatus can be
designed.
In the case of physical gas absorption the resistance to mass transfer is
entirely on the liquid side and the total transfer coefficient related to the Hquid
phase equals the partial transfer coefficient in this phase. The values of the
mass transfer coefficients can be calculated from the data of paragraph IV.4.2
and appear to be in the order of 10-
4
m/ s for all practical cases of gas absorption.
The same relations can also be used for calculating mass transfer coefficients
between two liquid phases if their flow behaviour is known.
261
The area of the interface between two phases is more difficult to estimate
and Table IV.2 only gives rough indications. For gas-liquid contacting in a
packed column, the surface of the packing material can be taken as the .actual
interface _(accurate within + 20 per The size of the gas- liquid interlace in
a bubble column or in a. stirred is strongly dependent on power
gas flow rate, gas hold-up and properties of the liquid. Reliable correlations
for the estimation of the size of the interface over a wide range of operating
conditions are stilllackjng. If the volume fraction of gas (I - s) and the average
diameter of the gas bubbles is known, the specitic interface can be calculated
from:
6(1 - s)
a = --- - (IV.38)
dp
This relation can also be used for calculating the size of the specific interface
for dispersions of Jiquids. Having determined the-size of the mass transfer inter-
face we can find the necessary dimensions of the mass exchanger from a mass
balance (equation IV.l).
As a first example we take the case in which two phases 1 and 2 with flow
rates 4>v
1
and 4>v
2
are contacted iii countercurrent plug flow in a column. For
this case we can set up the following mass balance (see also Figure IV.6) for
phase 1 (K
1
= total mass transfer coefficient based on phase 1) over a differential
height dx:
I
j l4>v1

0
1
. Q dx
0 0
--- _ ........ _ .. __ _
I
Figure IV .6 Contact between two phases in a column j f 4> v2 .
Integrating over the total apparatus we obtain:
-
1
J
cl (.x - L) de JL Kla JL dx L
c.<.x=o> c
1
- mc
2 0
<j>v
1 0
HTU HTU
(IV.39)
The terms of equation (IV.39) are dimensionless. The left-hand term is called the
number of transfer units (n). The right-hand term is a quotient of two lengths :
the length of the column (L) and the height of a transfer unit (HTU). The HTU
is a measure of _the efficiency of -the column (small HTU -i> good transfer).
The HTU can-be defined as that length of the column for which the integral of
the left-hand term of equation (IV.39) has the value of 1.
262
The HTU as used in equation (IV.39) is called an overall ' HTU because the
overall driving force (c
1
- mc
2
) has been used (here related to the '2' phase).
If we had only considered one pliase (e.g. the '1' phase) with driving force = c
1
- C1 1 (where C1; is the concentration at the interface of phases 1 and 2) we would
have obtained a partial height of a transfer unit (htu)
1

It can easily be derived from equations (IV. 15) and (IV.38) that :
(HTU)
1
= (btu}l + E(htuh (IV.40)
where E = m<f> tf
2
is the extraction factor. So the HTU is dependent on the
extraction factor.
In fact, equation (IV.39) is the same as the design equation for a plug flow
heat exchanger obtained in paragraph III.4. This will now be shown. lntegratio11
of equation (IV.39) gives :
where
elo - ell., In t1co = L
6 c
0
- L\c L Llc L HTU
llc
0
= c
10
- mc
20
11eL = elL - mczL
If we set up a material balance over the entire apparatus we obtain :
0 = r/Jr;lCi o - 4>vle1L + l/Jm
(IV.41)
where <l>m is the mass flow from phase 1 to phase2. Substituting the concentration
difference c 1
0
- e
1
L with the help of equation (IV.41) we find for the mass flow
rate, if both phases are in countercurrent plug flow:
L ~ c
0
- llcL L\c
0
- llcL
4>m = </Jv1 HTU ~ = K 1aAL ~ = K
1
a AL(6 c)
108
(JV.42)
l n ~ l n ~
AcL llcL
which can be compared with equation (111.61).
In the foregoing we have neglected the influence of liquid dispersion. If
dispersion or longitudinal mixing occurs in the flows, the overall driving force
in the apparatus will decrease (as we have already discussed in paragraph
111.4.2). Equation (IV.39) d oes not adequately describe such a case. However.
for practical purposes, cases in which dispersion has to be taken into account
can still be written in the form of this equation:
[
t(X.,L) d Cl l
cl(x=o) ct - mc
2
- HETU
(IV.43)
which is the definition of the height equivalent of a transfer unit' (HETU).
This HETU is, in general, the sum of the HTU and of forms which account for
the mixing in both phases.
263
As a second example we treat again the situation of Figure IV.6, but now for
the case that the two flows can be treated as plug flows in combination with a
dispersion process with dispersion coefficients 0
1
and [])
2
respectively. For this
case the following result is obtained :
(IV.44)
The determination of these dispersion coefficients has already been discussed
in paragraph II.8.5. There we saw (equation Il.118) that the height of a mixing
unit is given by:
L
HMU=-
n
where n is the number of perfect mixers in series yielding a corresponding
residence time distribution. For small dispersion we found (equation Il.115a):
1 D
_......,2-
n vL
and consequently in this case we find (analogous to equation II1.64) :
HETU = HTU + HMU)
1
+ !(HMU)
2
(IV.45)
For the system water- air in a packed column (HMU)watcr ~ dP and
dP < (HMU)air < 5dP.Sincetheheightofatransferunit is8dP < HTU < 40dP.
in practically all cases plug flow of the liquid phase can be assumed
Of course, for the exceptional case that one of the phases is perfectly m i x ~
the foregoing analysis (which is restricted to small dispersion coefficients)
is not of much help. In such a case equation (IV.43) and the HETU can no
longer be used. However, the approach here is again very simple. Considering
that the bulk concentration of the well-mixed phase is uniform (and equal to
its exit concentration) we can write down the mass balance for the other phase
over a differential height dx. This equation can then be integrated in a straight-
forward way, because it contains only one concentration which is dependent
on the location. If, as our third example, the other phase flows in plug flow,
:equation (IV.39) results again, the index 2 now indicating the perfectly mixed
phase and c
2
being constant, e.g. c
2
(x = L).
I V.4.4. The concepc of theoretical plates
Another frequently used expression in the calculation of continuous mass
exchangers is the 'height equivalent of a theoretical plate' (HETP). One HETP
is that column height for which for the outgoing flows applies that their con-
centrations come up to the equilibrium condition. This has been elucidated
in Figure IV.7, which shows a material balance over a segment of a mass ex-
changer with the length of 1 HETP. So the equilibrium condition is:
c
1
(x) = mc
2
(x + HETP)
264
c
1
( K+ HETP);cp
.. t
x+ HETP
4> ;c(x}
v 1 1
Figure IV.7 Material balance of a mass
exchanger over a height of one HETP
In order to determine the correlation berween HETP and HTU use is made
of the definition equation (1V.40) of the HTU:
-
HETP -- Jr+ HETP de 1
HTU x c
1
- mc
2
(IV.46}
The integral on the right-hand side of equation (IV.46) can be calculated if it
is borne 'in mind that a relation between c
1
and c
2
can be found from a mass
balance. For the case of plug flow this mass balance reads:
- tPv 1C1 + <f>v2C2 = COnstant
Equation (IV.46) can then be elaborated to :
HETP
HTU
1 In /(1 - E)c1(x + HETP)- mc
2
(x + HETP) + Ec
1
(x + HETP))
1 - E \ (1 - E)c
1
(x)- mc
2
(x + HETP) + Ec
1
(x + HETP)
and using the definition of the HETP, c
1
(x) = mc
2
(X + HETP), that can be
reduced to:
HETP _ 1 Jn {.!.. c1(x + HETP) - mc2(x + HETP)} _ ln E (IV.47)
HTU - 1 - E E c
1
(x + HETP) - mcz(x + HETP) - E- 1
So only if the extraction factor approaches the value of 1, which will be chosen
in many practical cases, will HETP ~ HTU.
For packing materials such as Raschig rings and Berl saddles the HTU is of
the order of magnitude of 6-60 em, 15 em being a good average. The reader
may check for himself what this means for the HETP if E is respectively 025,
050, 075 and 1-5.
265
Because the HETP is a stronger function of the extraction factor E than the
HTU, the Ia tter should be considered as a funclC!J!Ientall y more correct expression
for describing the operation of an extraction column.
IV.4.5. Problems
1. A column is filled with a layer of silica gel to a height of 30 em. Air with a
moisture content of 1000 ppm is passed through. When the air flows very
slowly the moisture content after passing the column is 01 ppm, while
under working conditions the moisture content of the air after passing
the column is 10 ppm. Calculate the layer of silica gel necessary to obtain
air with a moisture content of 2 ppm.
- 2. A H
2
S-containing water stream (concentration of H
2
S is 50gjm
3
, tem-
perature 25C) is treated countercurrently with air in order to desorb
the H
2
S. At 25C the distribution coefficient of H
2
S (=concentration
in air: concentration in water) is 0-44.
(a) Calculate the minimum amount of air necessary for fully degassing
1m
3
water.
(b) What is the lowest concentration of H
2
S which can be achieved in the
discharged water if 1m
3
air is used per m
3
water?
3
/ m
3
(b) 28 gfm
3
3. Ajr flows through a column filled with benzoic acid spheres. The outgoing
flow appears to be for 90 per cent saturated with benzoic acid vapour.
Now, instead of air, water is passed through the column in such a way that
the Reynolds number is the same .for both flows. What is the .degree of
saturation of the effluent water with benzoic acid?
Data:
Water
Air
Kinematic
viscosity
1 x 10-
6
m
2
{ s
2 x 10-
6
m
2
j s
Diffusion coefficient
benzoic acid
1 x 10-
9
m
2
f s
2 x 10-
6
m
2
fs
*4. A water stream of 1000 1/ h must be freed from Ca
2
+ ions. The initial con-
centration is 500 mgjl while the required final concentration is 10 mgfl.
This process is carried out in a column filled with ion exchanger consisting
of almost spherical particles. The Ca
2
+ ion concentration which is in
equilibrium with the exchanger is negligibly small. The
step during ion exchange is the mass transfer from liquid to the particles.
266
Determine the product of length and diameter of the column required for
this process.
Data: dp = 2 X w-
3
m, volume fraction voids e = 0-4
DeaH l ollS = 10-
9
m
2
is, v = 10-
6
m
2
/s
5. An extraction is carried out by causing droplets of water to descend
through an oil layer. The resistance to mass transfer is entirely in the oil
phase. The mass transfer can be described with the penetration theory by
introducing as contact time the time a droplet needs to cover a distance
equal to twice its own diameter. The flow around the droplets can be
considered as a laminar flow round a rigid sphere (Stokes' law applica"Qle).
(a) By what factor dQes the mass transfer coefficient increase or decrease
when the droplets are made twice as small ?
(b) How much does the product kS (i.e, the product of transfer coefficient
and total surface area of all droplets in 1 m
3
of the extraction column)
increase or decrease if at this droplet size reduction the total water
flow (m
3
/ s) remains unchanged?
Answer : (a) 071 times as great
(b) 56 times as great
*6. An organic liquid has to be saturated to 90 per cent with a certain substance
A However, only a saturated solution of A in water is available. The
organic liquid which is insoluble in water is now caused to ascend in small
rigid droplets (diameter : d
11
= 2 mm) through the saturated solution of
A in water (rate of rise : v = 10""
2
m/ s). What should be the height of the
water column to attain a 90 per cent saturation of the organic liquid with
A? The diffusion of A both in water and in the organic liquid is 10-
9
m
2
/ s
and the solubility of A in the organic liquid is 200 t imes as low as in water.
The continuous water phase is present in excess and is not exhausted.
The initial concentration of A in the droplets is zero.
7. In order to saturate an air stream with water vapour, the air is passed
through a wash bottle filled to a height of 10 em with pure water of 2SOC.
The air is divided through a sieve plate into bubbles with a diameter of
3 mm. The bubbles behave as rigid spheres. The volume fraction of water
in the dispersion is 08.
(a) Calculate the rate of rise of the swarm of bubbles using the relation of
Richardson and Zaki (paragraph II.S.5).
(b) What liquid height is necessary to ensure a relative humidity of 999
per cent in the outflowing air'?
(c) If the water is originally air-free how long does it take to saturate the
\iqui.d for 99 \)et: cent of air.
Data: [)) H10inair = 25 X w-
5
m
2
/ s
DairinH
1
o = 5 x 10-
9
m
1
/s
Answer : (a} 0-175 m/ s
(b) 11 em
(c) 121 s
267
8. A dry air stream must be saturated with 99 per cent water vapour. This
bas to be attained by passing the air flow through a wetted wall column
(a pipe in which a liquid film fiows downwards along the inside wall).
The thickness of the liquid film is very small compared with the column
diameter.
Data : <f>oair = 69 X 10-
3
ID
3
/ S, D internalc:otumn = 5 X I o-z m
Pair = 1 kgjml, 'lair = 17 X 10
6
Ns/m
2
D HzOioair = 25 X 10-
5
m
2
/ s
For the mass transfer coefficient in wetted wall columns applies, according
to Gilliand and Sherwood (Ind. Eng. Chem, 26, 516, 1934) :
Sh = 0023 Reos3 sco44(2 x 103 < Re < 3-5 x 104)
060 < Sc < 25
(a) What is the required length of the column?
(b) A physicist maintains that the column becomes undoubtedly much
shorter if a recirculation of the gas flow over the column is applied.
Show quantitatively whether his statement is cor rect for a recirculation
ratio 1 : 1 (i.e. just as much gas is recirculated as leaves the apparatus).
(c) What column length would be calculated as the answer to question (a)
if instead of the given correlation for Sh use was made of the
between mass and heat transfers?
Answer: (a) 96 m ; (b) 92 m; (c) 105 m
*9. An S0
3
- air mixture containing 7 mole por cent S0
3
is introduced at the
bottom of an absorption column which is filled with 1! inch Raschig rings
(a = 140 m
2
/ m
3
, e = 0.92). The gas rate calculated on the empty column
is v
0
= 2 m/s. Concentrated

is passed over the Raschig rings

downwards. The S0
3
equilibrium pressure of strong H
2
S0
4
is almost
zero.
(a) What length of column is required to absorb 98 per cent of the S0
3
passed through if plug flow of the gas phase is assumed?
(b) What length of column would be required if measurements on a 2m
long column had shown that the residence time distribution resembled
that of ten perfect mixers connected in series?
Answer: (a) 365 m (b) 405 m
268
*10. Show that John reached the right conclusion during his deliberations about
his cup o f t ~ discussed at the beginning of this paragraph.
Problem 4
A mass balance over a short height dx of the column yields after integration
( c = c
0
at H = 0) :
c { kanD
2
H}
- =exp - - ~ -
Co 44>"
With cjc
0
= 0.02 and a = 6(1 - s)/dp = 1800 m
2
/ m
3
we find :
kD
2
H = 765 x 10-
7
m
4
/ s (1)
The mass transfer process is analogous to bea.t transfer during flow through
a packed bed, so we can adapt equation (III.69a) for the estimation of the mass
transfer coefficient :
and we find :
kD = 6-3 x 10-
6
m
2
/ s (2)
Combining this with equation (1) we obtain for the desired product :
DH = 012m
2
So a small column of, for example, D = 02 m and H = 06 m would be suitable
for this purpose.
Problem 6
The only difficulty of this problem is to decide which mass transfer coefficient
is important. This problem has already been discussed in paragraph 1V.1.2,
so we can be very brief here. The figure shows the concentrations that play a
Water
c*
2
Or-gonic liquid
269
role and, since there is no accumulation of matter at the interface, Y'fC can Write :
<b':, = - cl } = k
2
(c
2
- cf>
Eliminating lhe unknown we find for the mass flux :
cf>" = mc
2
- c
1
nc m 1
-+-
k2 kl
Since the diffusion coefficient of the transported compound is the same in both
phases k
1
= k
2
, but because m = cf!c; = 1/200 the mass transport in the drop-
let is rate-determining and equation (1) becomes :
cf>':n = k
1
(mc
2
- c
1
)
The concentration equalization in the droplets is :
c1 - ( c)= 09c! = Ol
c
1
- c
0

and we find from III.8 that Fo = Otfd; = 005 and L = vt =2m.
Problem 9
The rate of absorption is determined by the gas phase mass transfer resistance
because the S0
3
vapour pressure of concentrated H
2
S0
4
is practically zero. The
gas phase mass transfer coefficient can be calculated with the help of the bound-
ary layer theory (equation IV.32; see . . paragraph IV.3.2, case 5) by using
instead of vCX) the sum V; + vCX). Here we can assume v, V
00
and we find
(vCX) = vj = 218 rnjs) :
k = 0765 x 10-
2
.m
2
/s and . ( k) = 153 x 10-
2
m/s
A mass balance over the differential height dx of the column yields after iii
tegration (c = c
0
at H = 0):
_c = exp (- kanDz H) = exp ( __ kaH) = exp
c
0
4</Jr; . v HTU
Now cfc
0
= 002 and HTU = vfka = 0935 m, so we find:
H = -HTUln002 = 365m
If we want to account for longitudinal dispersion, we first have to calculate
the height of a mixing unit :
L 2
HMU =- = - = 02m
n 10
So we find for the height equivalent of a transfer unit:
HETU = HTU + !HMU = 1035 m
270
and, writing more correctly :
we find H = 405 m.
Problem 10 John and his cup of tea
We assume the mass transfer resistance to be situated outside the coal par-
ticles. A material balance over a differential height element dx of the coal
column then leads (with c = c
0
at H = 0) to the expression :
= exp (- kmtDz) = exp (- _!!_)
c
0
4
0
HTU
The height of a transfer unit stays constant and from the information about the
normal operation of the column (c = 001 c
0
) we find that H
1
= 23 HTU.
With the impure water, for the same final concentration (now c = 00033 c
0
)
a column height of H
2
= 34 HTU is i.e. 48 per cent more.
IV .5. Mass transfer witb chemical reaction
John looked at the fiercely burning haystack. The police
agent informed him that at 5 o'clock in the nwrning, when
he discovered the fire, the stack appeartd to be 4 m high.
Now, 3 hours later che stack was still 25 m high. John
thought: the mitial height of the stack was 6 m and therefore
the fire must have started at I o'clock or later. Lighting his
pipe and carefully extinguishing the nuuch he inquired for
what sum the hay was insured.
Mass transport can be increased if the substance transported into another
phase reacts there chemically, but a chemical reaction can be slowed dow.n by a
transfer resistance (the transport rate is then lower than the conversion
rate). We distinguish between homogeneous and heterogeneous reactions.
During homogeneous reactions the chemical conversion occurs at aU places
where the reactants are present. The reaction rate term has to be included in the
mass balance, equation (IV. l ), which for the situations we shall discuss can be
simplified to :
(
d
2
cA)
DA dxl + RA = 0
(IV.48)
(no concentration gradient in the x-and z-directions, no tlow velocity,stationary
state). Since the transported substance is converted at all places, the concentra-
tion gradient is increased and a sufficiently fast reaction will therefore increase
mass transfer.
During heterogeneous reactions diffusion and reaction are separated. The
reaction occurs at a reaction interface only (e.g. at the catalyst surface) and the
mass transfer coefficient is not influenced by the chemical reaction.
271
The reaction rate term in equation (lV.48) is, for a first-order chemical re-
action, given by :
RA = - k,cA
For a nth-order reaction this expression becomes:
RA
_(IV.49)
(IV.50)
whereas the reaction rate for a second-order bimolecular reaction .between
reactants A and B is given by:
(IV.51)
We will restrict ourselves here to irreversible first-, second- and nth-order
homogeneous reactions and to first- and second-or,der heterogeneous re-
actions. For analysing these problems we will make use of the film theory
(paragraph IV.l.4), i.e. we will assume that the bulk of the liquid is well mixed
and that all mass transfer resistance is localized. in a thin boundary layer o_f
thickness Thus, the mass transfer coefficient is given by k = 0 /bc. . .
For homogeneous reactions we can distinguish two extreme cases : the chem-
ical react_ion is so slow that mainly in the bulk of the liquid and the
reaction is so fast that it occurs partly or completely in the boundary layer.
IV.5.1. Slow r:eactions
If the chemical reaction is slow, we neglect the reaction occurring in the
thin boundary layer and assum-e tha.t the reaction takes place in the bulk
of the liquid: In that case we will find a gradient over the laminar
boundary layer as shown in Figure IV.8. At the interface the concentration is
cA; and on the other side of the film-the equals the bulk concen-
tration. The assumption. of negligible reaction in the film is justified, if the
/CAi
I
I
I
I
I
1
I.
I
I CA
1,)
-x
Figure IV.8 Concentration distribution
during mass transfer with slow homogeneous
first-order chemical reaction
...
272
amount transported through the film (""' AkAcAi) is much bigger than the
amount of A converted in the film by chemical reaction (,.., Aok,.cAJ} Thus, if:
or, making use of the .:-. Of kA, if :
This is genera!Jy written as:
[fi;.k, = Ha < 1
Vk.f"
(in practice < 03)
The dimensionless group Jo Ak.Jki is called the Hatta number which
will be discussed in greater detail in the next section. The reader should further-
more realize that in most practical situations the film volume is very small as
compared with the bulk volume_ If we take. for exampie, a stirred gas-liquid
reactor with a= 1000m
2
/ m
3
, we find l> = OjkA lo- s and the total film
volume is then 10-
2
m
2
j m
3
, i.e. just 1 per cent of the total volume.
Returning to the situation shown in Figure IV.8 we can now set up a macro-
mass balance, realizing that in the stationary state all reactant A that diffuses
through the boundary layer has to react in the bulk volume. Thus:
(IV.52)
where A is the size of the interface and V the bulk volume. Eliminating the un-
known bulk concentration we find for the mass flux through the interface:
(IV.53)
If the reaction is very slow, i.e. Vk, AkA or AkAfY k,. >> 1, the bulk concen-
tration will be almost equal to the concentration at the interface, cA cAI.
From equation (IV.53) we find, then, that the mass flux is given by:
n V
cf>A = Ak,cAi
(JV.54)
This means that under these conditions the rate of the transport process is
completely determined by the rate of the {slow) chemical reaction.
If, on the other hand, the chemical reaction is very fast, Vk, AkA or
AkAfli k, l, we find:
(IV.55)
273
because the bulk concentration cA is now very small (due to the fast conversion
of The rate of the transport process is in this case determined by the ma.Ss
transport through the boundary layer.
IV .5.2. Fast homogeneous first-order reactions
If the reaction is so that all substance A is converted in the boundary
layer, we can easily solve the problem of diffusion with chemical reaction by
using the mass balance (equation IV.48):
where RA = - k,c A. The solution of this differential equation with the boundary
conditions cA = cA; at x = 0 and cA = 0 for x = ro is:
cA = cA, exp ( -xJft )
(IV.56)
This concentration distribution is shown in Figure IV.9. From equation (IV. 56)
we find for the mass flux through the boundary layer :
t
I

.
I
CAi I
I
I
I.
I
I
l
I
=exp:{-
AI I l\
l
I
-----X
Figure IV .9 Concentration distribution
during mass transfer with fast homogeneous
first-order chemical reaction
(IV.57)
In this case the mass transfer coefficient is given and is solely depen-
dent on the diffusion coefficient and the reaction rate constant k, . The ratio
jD Ak,fk! which represents the ratio between the mass transfer coefficients
with and without chemical reaction is called the Hatta (Ha) number, which
plays an important part in mass transfer with chemical reaction.

274
From the concentration distribution, equation (IV.56}, we see that the pene-
tration depth of the concentration distribution with chemical reaction is
given approximately by:
(IV.58)
Consequently, the ratio of the thickness b = 'JJfkA of the boundary layer and the
penetration depth b, is given by :
!__=
8,.
which shows another physical meaning of the Ha number.
ff the reaction occurs partly in the boundary layer and partly in the bulk of
the liquid (Figure IV.lOd) (where. because of the fast reaction, cA is very Jow),
we can show that the mass flux is given by:
(1V.59)
If the reaction is very fast, i.e. Ha = j k,.IO 1 (for all practical purposes
Ha > 3), equation (IV.59) can be simplified to equation (IV.57), which we have
already derived In this case the total reaction occurs in the boundary layer.
If the reaction is relatively slow, Ha = J krDAik! 1 (Ha < 03), we find from
equation (IV.59) the relation for simple physical adsorption :
lj>A = kACA i
and only a negligible part of the reaction occurs in the boundary layer.
The Ha number is therefore the criterion for whether the reaction occurs
completely in the bulk of the liquid (Ha < 03) or completely in the boundary
layer (Ha > 3). If 03 < Ha < 3, reaction in both the bulk and the film is
important. Figure IV.lO shows the five possible cases of mass transport with
homogeneous first-order chemical reaction which we have discussed. If we
want to determine to which class a certain problem belongs we start by cal-
culating the Ha number. If Ha < 03, as a second criterion the expression
Ak.JVk, has to be determined (compare equation IV. 53).
Equation (IV.59} and Figure IV.IO also show that the rate of physical absorp-
tion is increased by a factor :
(IV.60)
by the occurrence of a homogeneous first-order chemical reaction. This increase
in adsorption rate by a factor J 1 + Ha
2
in the case of a chemical reaction is
valid under all flow conditions. The thickness b, of the boundary layer where the
chemical reaction occurs is so small that it is not influenced by the flow con-
ditions. Also. during turbulent flow, the eddies cannot penetrate this thin layer.
The enhancement factor Fe= j 1 + Ha
2
is therefore universally applicable.
_,...-cAi
~
:
CA-.::.CAi
( 0 )
t
u
I
I
I
I
-
J
(b)
( c)
(d)
(e)
Ak
Ha < 0.3 ; ~ >> l
Vk,
No reaction in boundary layer
Ak
Ha < 0.3 ~ ~ I
' Vk,
t / J ~ ::::::: k;.CAi{ l }
A.kA
1+-
Vk,
No reaction in boundary Jayer
Ak
Ha < 03; _ A J
Vk,
No reaction in boundary layer
Ak
0.3 < Ha < 3; ~ 1
Vk,
'A = kAcA;ji + Ha
2
Reaction partly in boundary layer
AkA
Ha < 3; -k 1
v ,
;. = kAcAIHa
Reaction c:Omp1etely in boundary layer
275
Figme IV.JO Five cases of mass transfer witb homogeneous irreversible first-order
reaction
IV.5.3. Homogeneous nth-order reactions
The analysis developed in paragraphs IV.5.1 and IV.5.2 can also he applied
t o nth-order chemical reactions, where the production term of equation (IV.48)
is given by (equation IV. 50):
276
The results of this analysis are completely analogous to that of first-order
reactions and, again, the five situations summarized in Figure IV.l 0 can be
distinguished.
For this type of reaction, the thickness of the reaction boundary layer
(equation IV.58) becomes :
(IV.61)
and the definition of the Hatta number is now:
1 2 n- l
Ha =-
1
krCAt [)A.
kA n +
(IV.62)
With the above Ha number, the relations developed in paragraphs IV.5.1
and IV.5.2 for first-order reactions also describe nth-order reactions. There is
one important practical difference: for first-order reactions b, and Ha are only
(through k,) dependent on temperature, but for nth-order reactions these para-
meters are also dependent on the concentration of A at the interface.
IV.5.4. Homogeneous second-order reactions
A component A is absorbed and reacts with a component B which is present
in the bulk and which is not desorbed The production term in equation (IV..48)
is given by (equation lV.Sl ):
RA = -krcAcB
If the concentration of component B is uniform and equal to c
8
CXJ, the bulk
concentration, the solution to this problem is simple : the reaction can be re-
garded as being of a pseudo first-order type with a first-order reaction rate
constant k ~ = k,c
8 00
. Thus :
(JV_63)
The solution to this problem has already been treated in paragraphs IV.S.l
and IV.5.2. There are two conditions under which the assumption of a pseudo
first-order reaction is justified :
(a) If rhe reaction takes place in the bulk so hat :
H
- ~ ~ O A _
a- -- -
k2
A
(b) If the reactant B can be transported quickly enough into the film where the
reaction occurs so that the concentration of B in the film can be assumed
to equal the bulk concentration (see Figure IV.ll), i.e. if:
(IV.64)

- -x
Fieure IV .11 Concentratjon distribution during
mass transfer and second-order reaction if Ha > 3,
kBcsaJf kAcAi > 10 Ha
In practice this condition is fulfilled if:
kBcBa:; > IOjk,.c
8
CX)IDl A CAo
or :
277
(IV.65)
If equation (IV.65) is fulfilled, the correction factor Fe for the acceleration of
mass transport by chemical reaction is given by:
Fe = j l + Ha
2
= 1 + . (IV.66)
If the reaction is so fast that insufficient B is transported into the boundary
layer, the transport of B will influence the reaction rate in the film. If:
the reaction is so fast that A and B cannot exist without reacting instantly.
A reaction front is formed at x = xR (see Figure IV.l2) where cA = c
8
= 0.
The film theory yields for the transport rates of both components to the reaction
front :
(IV.67)
Hence, the chemical acceleration factor is found to be, bearing in mind the
definiti on of the Ha number given in paragraph IV.5.2 and realizing that for this
278
0
-x
Figure IV .12 Concentration distribution during
mass transfer with secondorder reaction if Ha > 3,
kaescrlkAcAi < 01 Ha
case always Ha > 3:
(IV.68)
where f>A and <5
8
are the thicknesses of the boundary layers during physical
absorption only. The penetration theory predicts that :
bA- SbB - (OA) t . (IV.69)
f>a - ShA - lOa
(see case 1 of paragraph IV.3.2) whereas the boundary l a y ~ r and the Leveque
theory (cases 2 and 4) predict a relation of the form :
{)A= (
0
A) -t (IV.70)
<>a De
If D A and 0
8
are of the same order of magnitude (which is often the case) the
enhancement factor Fe is only slightly dependent on the physical model chosen
and we can use equation (IV.68), together with equation (1V.69) or (IV. 70) to
produce a reasonably accurate estimate of Fe-
In the intermediate area between a fast reaction (k
8
c
8
d kAcA
1
> 10 Ha)
and an instantaneous reaction (k
8
c
8
a>f kAcAi < 01 Ha), so in the area where
(see Figure IV. l3) :
kaeaa> ~ Ha
kAcA
the F, value is a function of the Ha number and of
l
-x
Figure IV.13 Concentration distribution during
mass transfer with second-order reaction ifHa > 0 3
k
8
c
8
.,jkAcAi Ha
279
where the exponent m again is a function of the physical model chosen (m = -!
for penetration theory, m = t for boundary layer The dependency
ofF, on these two expressions is shown in Figure IV.l4 which can be used
for estimating the ch.emical acceleration factor.
Table IV.3 gives a summary of the various solutions for mass transfer with
second-order reactions discussed

10
10
Ha
100 1000
Figure IV.14 Values for the enhancement factor Fe for second-order reactions if
Ha > 0.3, k
8
c
8
J/kAcA
1
Ha
280
Table IV .3 Summary of mass transfer situations with irreversible second-order reactions
kacaao
Ha
kACAc
Solution
<03
> l ORa
Pseudo first -order reaction with = k,cBrr;
solutions of Figure IV.lO apply
>03

>0.3
< 0.1 Ha
Situation given in Figure IV.13
= FckAGAi
with Fe from Figure IV.l4
Instantaneous reaction, Fjgure 1V. 12
= {( + ( r-l C;:} kACAO
where!< m <!
IV 5 .5. Mass transfer with heterogeneous chemical reaction
The chemical reaction occurs at the interface solid- liquid or solid-gas.
If we consider an irreversible reaction :

the material ba,lance for any of the five flow situations discussed in paragraph
IV.3.2 becomes:
The boundary condition s are :
X = 0, y 0, C A = C A>, Cs = CBCD , CR = Cs = 0
x > O,y = oo, CA = CAoo CB = Ca.:c c
1
= c
5
= 0
(IV.71)
dcA n
8
dc8 = dcR rn. des R ( )
x > O,y = 0, O A dy = u..v dy - [) R dy = -ILls dy = A c,.., c8 , ca,cs
The last boundary condition implies that there is no accumulation of reactants
at the interface. The reaction rate RA is related to the area of the interface (in the
case of porous catalysts to the outer catalyst surface).
Because the reaction kinetic term is found only in the boundary condition
and not in the mass balance, the chemical enhancement factor F, for hetero-
geneous reactions is unity. Under steady-state conditions a mass balance at the
interface shows (no accumulation of reactants at interface) :
kA(CAq; - CAi) = kB(CBCD - CsJ = kRCRi = KsCs; = - RA(CA; , CB; , CR, CsJ
(IV.72)
281
These equations are, in general, easily solved. The .rate of conversion is in all
cases given by:
(IV.73)
We will now discuss the special cases of first- and second-order reactions.
(a) First-order reaction. Equation (IV. 73) for tbis case reads :
= kA(cAoo - cAi) = - RA = T<.cAi
(IV.74)
The reaction rate constant is here related to the outer size of the interface
(ma:croscopic reaction rate constant) and has the dimension (m/ s). It is related
to the well-known chemical reaction rate constant k, (s -l) via the specific
interface area a (m
2
j m
3
) as follows:
k, = ak', *
From equation (IV.74) the ratio cAJ cAoo can be calculated to be:
CAi kA
1
-= ----
+ KA 1 k;
+-
kA
and the overall conversion rate is found from equation (IV.74) as:
(IV.75)
Apparently, here the quotient fulfils the role the expression Ak.JVk,
played during homogeneous reactions (paragraph IV.S.i). We can distinguish
three principal situations :
A.
B.
The reaction is so slow compared. with the mass transfer process that mass
transfer does not decrease the reaction rate and the concentration of A is
uniform and equals cA-:c
kA, CAi 0, = kACAoo
The reaction is so fast that the of A at the interface is practically
zero and physical mass transport determines the overall reaction rate.
* In the case of heterogeneous catalysis often a microkinetic reaction rate constant k; related to
the total pore surface of the catalyst is used; 'K,. is related to k;. and k, by:
k
= k" = ak'
r ap r r
where a = outer specific surface of cat alyst particles and aP = specific pore area of catalyst (m
2
/m
3
).
282
c.
Both chemical reaction and mass transfer determine the overall conversion
rate.
Figure IV.lS gives a summary of these three situations and shows the con-
centration distributions which occur. The reader should compare this figure
with the first three cases of Figure IV.lO, which refer to homogeneous first
order reactions.
A.
-)(
B.
c.
= kAcAco
Ftpre IV.l5 Three cases of irreversible heterogeneous first-order reactions
(b) Second-order heterogeneous reactions. Equation (IV.73) for this case
= kA(CAce - cAIJ = - Csi) = k:,CA;Cs ;
Elimination of cs; leads to the following expression for cAJcAJ:
x(CAi )2 + {.x(l- </J)+ 1} CAi- 1 = 0
CAJ
where:
(IV.76)
283
Calculation of c,.JcAoo from equation (lV.76) enables 'us to find the rate o_f
conversion as :
= kACAo 1 - -'
( CA)
CAo
(IV.77)
The reader may check that for rjJ-+ 0 second-order heterogeneous reactions
can be regarded as pseudo first-order reactions.
At this stage, the authors had Gonsidered inserting -a section on the design of
simple chemical reactors. This would have been logical in the lay-out of the
book The main features of the .design of a heat exchanger (paragraph ITI.4)
fo-Ilowed the fundamental considerations on heat transfer in the preceding
sections and the design of a mass exchanger (paragraph IV.4) concluded the
theory of mass transfer without chemical reaction 1n the beginning of this
chapter. However, simple and isothermal chemical reactors seldom occur in
practice and we prefer to refer the reader to a more elaborated text on the subject
of chemical reactors, e.g. Levenspiel 's Chemical Reaction Engineering. Simple
calculations can already be made on the basis of what is offered here, as the
following problems will show-it is only and again applying the law of mass
conservation.
IV . .5.6. Problems
1. On the bottom of a cylindrical vessel (diameter D = 1 m, height H = l m)
with stirrer is a 1 em layer of solid substance. To remove this layer the
vessel js filled with a liquid in which the solid dissolves. In solution the
solid decomposes according to a first-order chemical reaction.
Data: mass transfer coefficient (without chemical reaction) k = 10-
4
mfs
solubility of the solid substance in the liquid = 50 kgjm
3
density of the solid substance p = 2000 kgjm
3
diffusivity [} = 10-
9
m
2
js
reaction rate constant kr = 10-
2
s-
1

(a) Show that the mean concentration of the dissolved solid in the liquid
is much lower than the solubility of the solid substance.
(b) Show that the conversion by chemical reaction in the .boundary layer
is negligible.
(c) How long does it take for the layer of solid substance to be
completely?
(d) A chemist states that the process can no doubt be accelerated by adding
a catalyst which makes the reaction rate constant 1000 times as great.
Demonstrate quantitatively whether his statement is correct.
Answer: (a) c = 001 c,
(b) be = lo-s <0-3 per cent of the total conversion in film
(c) 4000 s
(d) j2 times as fast
284
2. If under otherwise equal conditions pure oxygen or oxygen -from the
surrounding air is absorbed in a strong Na
2
S0
3
solution to which a
large amount of cobalt salts have been added, the difference in rate of
adsorption appears to be a factor of 10. Determine the order of the reaction
between 0
2
and Na
2
S0
3
with respect to the oxygen (the reaction is very
fast, Ha > 3).
3. In a laminar liquid film time Ol s) C0
2
(m = l) is from
the pure gas (20C. I atm). The liquid film contains 004 kmolfm
3
NaOH.
Under the prevailing conditions the second-order reaction rate constant
between C0
2
and NaOH is 6 x 10
3
m
3
jkmol s. Calculate the chemical
enhancement factor and the mole flux in this situation and draw the oon-
distribution in the film. Assume []lc
02
= DoH- 2 x 10-
9
rn
2
/s;
because 1 mole C02 uses two moles NaOH, for CsQ) in this "case
tcNaOH must be used.
Answer: F, = 15, q,;or = 985 x 10-
6
kmoljm
2
s
4. What will be the chemical enhancement factor and the mole flux if in the
situation of problem 3 a temperature of 70C is taken (k.- 2 x 10
5
m
3
fkmol s) and the concentration ofNaOH is chosen equal to 04 kmol/m
3
(m = 042)?
Answer: Fe = 13 5; mo
1
= 56 x 10- " kmo1/m
2
s.
*5. In a packed column of catalyst particles a component from the liquid
stream is converted by a chemical reaction. The reaction is an irreversible
heterogeneous reaction of the first-order and proceeds at the surface of the
particles. How high. should be the catalyst bed to ensure a degree of con-
version of063 ( = 1 - e -
1
)?
Data: reaction rate constant k, = 4 X ro-s m{s (macrokinetic)
diffusion coefficient II) = 10-
9
m
2
js
superficial fiow rate in the bed v
0
. = 4 cmjs
kinematic viscosity of the liquid v = 10-
6
m
2
/s
bed porosity e = ().40
particle diameter d P = 10 mm
*6. During fat hydrogenation pure H
2
is absorbed in liquid fat containing
catalyst particles at which the reaction between H
2
and unsaturated fatty
acid takes place. The rate of conversion per unit liquid volume related to
H
2
is independent of the degree of conversion of the oil and appears to be
linearly proportional to the solubility of H
2
In a certain experiment
the ratio between these two constants appears to be 0.03 mi n-
1

285
If the interface between gas and liquid is increased by a factor of 2, the
specific rate of conversion will be 20 per cent higher. If the catalyst con-
centration is increased by a factor of :2. the specific rate of conversion will
be 50 per cent higher. Calculate for this experiment the gas- liquid inter-
facia] area per unit liquid volume and the {macro )reaction rate constant
of the reaction between H
2
and unsaturated fat.
Data: the partial mass transfer coefficients in the liquid at the interface
with gas kl = 1 o-
4
m/s and at the interface with the catalyst k2 =
w-s mfs
the interface of the catalyst per unit volume of liquid = 150 m
2
fm
3
2
/m
3
; lo-s m/s
7. One of the components (A) of a gasmixture is absorbed in a solution in
which it reacts very fast with a component B which is present in excess
according to a secondorder reaction of the type R = kcAcs (i.e. the reac-
tion takes place entirely in the boundary layer). At a certain concentration
of B the overall mass transfer coefficient is found to be K = 10-
4
mjs.
In order to find for this case the partial mass transfer coefficient in the
gas phase, the overall mass transfer coefficient is measured again at a 4
times higher concentration of B. The result is K = 15 x 10-
4
m/ s.
Calculate the partial mass transfer coefficient in the gas phase if the given
mass transfer coefficients are related to the gas phase.
4
m/s
*8. Reith carried out the following tests:
Two equal streams of 0
2
and N
2
were mixed and passed through a
stirred gas- liquid reactor containing a solution with an excess of
sulphite and sufficient cobalt ions to realize chemical absorption in the
range of the fast reaction (Ha > 3).
Subsequently, he passed under otherwise the same conditions the two
streams 0
2
and N
2
separately into the gas-liquid reactor.
In the second experiment he found absorption rates which were 40
per cent higher than in the first test. How do you expiain this?
Answer : There is complete segregation in the bubble phase under the
conditions of this experiment.
9. The chemical absorption of C0
2
(A) in aqueous NaOH solutions can in
first approximation be described by a second-order reaction :
286
ln a laminar jet 066 mm, t>i = 5 mjs, T = 20C) the following
absorption rates</>, are measured at various jet lengths L :
L
(m) .
2.() X 10-
2
45 x w-
2
tPm
(kg/s)
2.()4 X 10-
8
4-6{) X 10-
8
where 2

= (OH--)ro = 105 kmoljm

3
, cAo = C0
2
= 0238 kg/m
3
,
U)A = 10-
9
m
2
/ sand lOB= 16 x 10-
9
m
1
/s. Calculate the reaction rate
constant k, (at 20C) from these data.
Answer : k, = 5-75 x 10
3
m
3
fkmol s (k,.. = 4-48 x 10-
4
mfs Ha = 46)
*10. C0
2
has to be washed out of a continuous stream of hydrogen at 30C
and 7 atm absolute pressure in a packed column by means of a counter-
currently flowing monoethano]amine (MEA= HO-C
2
H
4
-NH
2
; one
mole C0
2
uses two moles MEA} which is reactivated and re-
cycled. If the C0
2
concentration in the hydrogen has to be decreased
from 20 to 0-01 per cent by volume and if the initial concentration of free
MEA in the wash solution is 3 kmoljm
3
, how high should the packing in
the column be?
Data: column diameter D = 1m
packing : Raschig rings, dP = 75 em (a = 60 m
2
/m
3
, s = 095)
gas flow rate c/>g = 4000 Nm
3
fh
liquid flow rate 4>L = 30m
3
fh
solubility of C0
2
in MEA solution m = 1
second-order reaction rate constant between C0
2
and MEA,
k, = 7500 m
3
jkmol s
p, = 0-56 (at 7 atm), PL = 1000 kgjm
2
rJ
1
= 93 x 10-
6
Nsj m
2
Dco
2
inH:z = 55 x 10-
5
m
2
f s
. - 9 2
Dco
2
inH:zO = 2 x 10 m j s
IOMEAlnHl o = 10- 9 m2/s
vL = 10-
6
m
2
/s
*11. Show that John drew the right conclusion in the case of the burning hay-
stack related at the beginning of this paragraph.
Problem 5
A material balance over a short height dh of the packed column reads (A =
cross-sectional area of column):
- cf>L del..= ka(cL - ci)A dh = krac;A dh
287
and isolating and substituting the unknown interface concentration ci we
obtain :
or, after integration between h 0, cLo and h, cL :
CL { - ah }
cLo = exp . !)
Vo k + k
r
Now the degree of conversion was given as (cL
0
- cJ JcLo = 1 - e-
1
, so
cJcLo = e -
1
and therefore :
We find the mass transfer coefficient with the aid of equation (III.69a) (making
use of the analogy between heat and mass transfer) to be k = 114 x 10-
4
mjs
and with a = 6(1 - e)jdP = 3600 m
2
fm
3
we finally obtain h = 0376m.
Problem 6
The reaction rate between hydrogen and given by a relation of the form
(cL = H
2
concentration in liquid phase):
Now it is stated that the reaction rate is linearly dependent on the hydrogen
concentration, so ex = 1, and independent of the degree of conversion of the
fat, hence f3 = 0, i.e. the reaction is of the pseudo first -C?rder type. We can write :
(kmoljm
3
s)
Hydrogen has to be transported from the gas phase c,. mass
transfer coefficient k,. specific interface a, ) to the liquid phase (concentration
cL) and from there to the surface of the catalyst particles (concentration c;,
mass transfer coefficient kL , specific interface ai) where it reacts chemically.
Under steady-state conditions all these mass transfer steps proceed at the
same rate (no accumulation of H
1
in any phase) and we can write (k; = a!<,,
where k, = mascroscopic rate constant in m/s and k; = chemical rate constant
in s -
1
) that:
(1)
288
From this equation we isolate and eliminate the unknown concentrations cL
and c; and obtain :
_c/J_m_ol = ___
V 1 1 1
(2)
- +--+-
kgag kLai k,.a
1
This equation completely describes the overall hydrogenation process. With the
aid of the information given we can evaluate the various terms of this equation
as follows:
given : 4>mol =003 min -
1
= 5 x 10-
4
s-
1
=
1
(3)
_1_ + _1_+_1_
k
8
a
8
kLa
1
k,a,
given: if is doubled, 4>mor increases by 20 per so:
1 1 1
--+-+-
12 = k,ag kLai k,.a
1
(
4
)
1 1 1
--+ -+-
2kgag kLai k,.a;
given : is doubled (via catalyst concentration), aw
1
increases by 50 per cent,
so:
1 1 1
-+-- + --
15 = k,ag k Lat k,.ai
1 1 1
(5)
-+--+--
kgag 2kLai 2k,a
1
When this equation is worked out it appears to be identical with equation (4),
so the information given is consistent.
From (3) and (4) k
8
a, can be calculated and is found to be k
8
a, =
15 x 10-
3
s-
1
, so that with the given k
8
value we find a
8
= 15 m
2
/m
3
(which
is very low for a gas liquid reactor). Using the values given forkLand ai we find
from equation (4) that the macroscopic rate constant kr = w-
5
m/ s, so the
chemical rate constant is given by k; = 15 x 10-
3
s-
1

Problem 8
Sodium sulfite is oxidized in solution to sodium sulfate by oxygen if cobalt
ions which act as catalysts are present:
0
2
+ 2S0
3
2
- 2S0
4
2
-
At a pH of 7-8 and 15-35C the reaction is :
first-Order in C0
2
+ if 3 X 10-
6
< CeoH < 10-
3
km.olj m
3
,
second-order in 0
2
(see problem 2),
first-order in S0
3
2
- if

< 04 kmol/m
3
,
zero-order in SQ3
2
- if Cso
3
2- > 04 kmo}jm
3

For the above range of Co
2
- concentrations and

> 04 kmolfm
3
the
pseudo second-order reaction rate constant is :
2 x 10\$ < k, < 10
8
m
3
fkmol s
The only explanation of the phenomenon Reith measured is segregation of the
gas bubbles which influences the driving force of the mass transfer process.
H we neglect the concentration decrease in the gas bubbles we can write for
Ha > 3:
>mol = Ac;Jik,c;ID = c't
In Reith's first experiment c, = 50 per cent by volume in all bubbles, so all A.
In the second experiment c, = 100 per cent by volume in half of the bubbles.
so 50 per cent of A if complete segregation occurs. The ratio of the two mass
flux rates would be :
4> 1 3
mol sepuate streams = I ; 1
2
= J2 = 141
4>mol {1) X 1
Theoretically for complete segregation of the bubbles we would expect a 41
per cent higher absorption rate during the second experiment, which indeed
was measured. This strong segregation (no coalescence of bubbles) is caused by
the high ion concentration of the sulfite system. In distilled water and in most
organic solvents considerable coalescence and redispersion of bubbles would
occur.
Problem 10
The sketch of the absorption column shows the gas and liquid flows and the
corresponding concentrations in all streaii?-s. To begin with we assume that the
reaction between C0
2
and MEA proceeds in the boundary layer, i.e. that
Ha > 3. We will check this assumption later on, when we calculate the Ha
number.
r#JL = 30 m
3
/h
cL
1
=30 kmol/m
3
3
<cL> log =153 kmol /m
t
'
t
P
9
=0 56 kg / m
3
.P
9
=507

(7ot m l
Cg = 00 I vol % = 28 x 10-
5
kmol 1m
3
- 4 3
<c
9
>
109
= 7 37x 10 kmol/m
P
9
= 2 31 kg / m
3
1/>9 =635 m
3
/h ( 7 otm)
c
9
=20 vol % =0 056 xmol/m
3
290
Since the .Qeight of the will be greater than 10 packing particles, we
can assume countercurrent plug tlow of both fluids and the mass flow rate of
C0
2
into the MEA solution is given by equation (IV .53) as:
4> = K aAL !leo - Jlcl
'" L llc
ln-0
llc
1
where K.L. is the overall mass transfer coefficient related to the Liquid phase and
Ac
0
and Ac
1
the concentration differences at the column entrance and exit.
Thus:
.lle
0
= me,
0
-

. 0056 - 0 = 0056 kmo]/m
3
llc
1
= mcg
1
- eL
1
= 28 x 10-
5
- 0 = 2-8 x 10..,.
5
kmoljm
3
since the C0
2
concentration in the bulk liquid is zero.
Consequently, the logarithmic mean driving force is given by :
b.c - t:.c
(lleh =
0 1
= 737 x 10-
3
kmolj m
3
01
Ac
ln-0
Ac
1
We shall now calculate the overall mass transfer coefficient KL which IS
given by (see equation IV.15b) :
{
m 1 }-
1
KL = k + k F
g L c
since the mass transport is enhanced by chemical reaction. The partial gas
pha.se mass trans.fer coeffici,ent kg is with the aid of the boundary layer:
theory (case 5 of paragraph IV.3.2; see also problem 9 of paragraph IV.4) to be
(with v
0
= 020 mjs, v = 166 x lo-s m
2
js):
( kg) = 10-
2
rnjs
The partial liquid phase mass transfer coefficient KL is found analogously to
problem 8 of paragraph IV.2 (see comments on problem 8, page 246) to be
(v
0
= 105 x 10-
2
m/ s, s = 62 s-
1
)
kL = 11 x 10-
4
m/ s
We shall now determine the enhancement factor F,. For the Ha number we
find using the mean MEA concentration :
Ha = -
1
.jk Cs [)A=
1
4
/7500 x 153 x 2 x 10-
9
= 435
K r a:> l l X 10- V
L
This confirms our initial assumption that Ha > 3 was correct and the reaction
proceeds in the boundary layer.
At the column entrance .(see 'p,aragraph tV .. 5.4):
k BcBa:> = [Jl B cBco = 062 = 5-53 Ha
kAc Ai [Jl A CAi 2 ().056 ..
and we have to use Figure IV.l4 in order to find .. Fc. With:
. /DA) t +Csoo([JJA) - t = 14 + . 062 = 93
\lOs c Ai Da 0-0.56 x .14
we find from this figure 90. At the column exit (top) :
kBCBct:l 1 3
kACAt = 2 28 X 10- S =
53
'
500
291
which is much bigger than 10 Ha, so here we have a pseudo first--:order .reaction
with Fe = Ha = 435. For further calculation we will use as a rough approxima-
tion the logarithmic mean F, value of <Fc>tog= 208.
With the foregoing results we find for the overall mass transfer coefficient:
{
m 1 } -J { 1 . 1 } - t
KL = k
8
+ kLFc = 10-
2
+ 11 X 10
4
X 20 X 8
= 186 x to-
3
mfs
Since the mass flow of C0
2
from the gas to the liquidphase is l/Jm = fb
8
llc, =
635 x 0056/ 3600 = 99 x 10-
3
the necessary height of. the column is
found as: .
L l/>m . - 99 =154m
K LaA(Ac)
1
og
3
1t 3 ....
1-86 x x 60
4
7-37 x to-
.
',
' ' ' ' J' I,
Problem .11 . ,
Let us assume that the'haystack has the form of a half-sphere of radius r tthat
all burning occurs at the surface. of the stack and that the rate of .burning is
completely controlled by transfer of oxygen froin the air. With these assump-
tions the rate of volume decrease of.the haystack is dependent on the oxygen
transport from the air :
dV
2
dr
2
- = 4r - ,_ k 2nr ( c - c.)
dt dt g g I
or, assuming cg - ci = constant :
dr
dt = constant, r.- c
1
t + c
2
Evaluating the constants with the information given, viz. r = 6 m at t = 0,
r = 4 m, t = x and r = 25 m, t = x + 3 we find x = 4 h, so the fire must have
started at 1 o'clock . .
292
If the haystack has a form other than that of a half-sphere, it will have a higher
surfacejvolume ratio and will therefore bum down more rapidly.
IV.6. Combined heat ud mass tramport
The transport of matter from one phase to another is, in principle, attended by a
heat effect. When, for example, on absorption a certain component from the
gas phase dissolves in the absorbing liquid, heat of solution is released; in
addition, an extra heat effect may occur if the dissolved component reacts with
the liquid or with the reagents in it
The temperature of the interface adjusts itself at a value which depends on the
possibility of heat removal (or supply in the case of a negative effect) through
the two phases. The absorption rate, however, is influenced by the interfacial -
temperature (e.g. via the so that in such a case the mass transfer
and the heat transport are coupled.
Examples in which the influence of the beat transport on the mass transfer rate
could be neglected were given in paragraph IV.2, problem 4, and in paragraph
IV.3, problem 8.
IV.6.1. Drying
A special case of combined heat and mass transfer is obtained if, for example,
a drop of liquid (or an inert object moistened with liquid) is put in a gas flow
which has not been saturated with the vapour of the liqwd Assume that the
liquid originally has the same temperature as the gas. Because the gas is not
saturated at that temperature, evaporation occurs from the liquid surface
(mass transfer). As a heat of evaporation is withdrawn from the liquid
so that its temperature begins to fall The lower the temperature of the liquid)
the greater the temperature difference between gas and liquid and the more
heat will be transported) from the gas to the liquid phase. some time a
stationary state is reached in which the temperature and hence also the heat
content of the liquid no longer changes with time; all heat required for the
evaporation is then supplied. by tile passing gas. In this situation of dynamic
equilibrium the following energy balance applies (see Figu.re IV.16) :
'JJ = t/J':nol AflH e,A
where tllfe,A =the molar heat of evaporation of A at Tw.
II
4> 1r \ c; mol A
H \ \
\ \
.
Go\$ flow
Figure IV .16 Evaporation from a wet
surface
(IV.78)
For the local heat fiux <I>'H to the surface can be written :
t/>i - Tw)
293
For the local rate of evaporation per unit of area applies, if PA.w is sufficiently
small with respect to the total pressure:
A,. ll k (p
'PmolA = RT A,w - PA,g)
(IV.79)
where (or Ttbe mean between T; and T., can be taken. By substitution of these
equations for and in the energy balance {equation IV.78) we find :
PA,w- PA.t = RT
- Tw tllfe,A k
(IV.80)
In this relation PA,w and Tw are coupled via the vapour pressure curve of liquid A.
If the evaporating drop is so small that the values of PA.r and T, do not change
appreciably, rxj k in this relation can be replaced by the quotient of the mean
transfer coefficients over the entire area ( rx)/(k). The value of this quotient can
now be calculated with the help of the analogy between heat aDd mass transports.
Both ( ex} and ( k) refer to the same geometric situation and to the same flow
condition (the same value of Re). If Re is not too small, the analogy relation
(paragraph IV.3.1) can be applied, which leads to:
( ex) = c o(Le)!
(k} pr
(l\".81t
where the product Cp!J must be taken at the mean temperature T and Le =
aj D = ScjPr is the dimensionless Lewis number. For gas mixtures the value of
Le lies between 05 and ca. 2 and for liquids, appreciably higher than 1.
With this relation is found as ultimate result:
PA,w - RTcpP
- J:, flli e,A
(IV.82)
The temperature Tw is called the wet-bulb temperature because this temperature
is reached by a moist sphere if it is placed in a gas flow.*
This temperature is an important constant in the description and calculation
of drying of materials in an air stream. Another application is found in spray-
drying towers in which hot s,olutions in the form of droplets or liquid layers are
brought countercurrently irito contact with the outside air; the droplets cool
down because of evaporation but their outlet temperature cannot be lower
than the wet-bulb temperature, belonging to the temperature and the partial
water vapour pressure of the ambient air.
The wet-bulb temperature takes its name from psychrometry, according to
which T, and Tw are measured with a 'dry' and a 'wet' thermometer respectively
*In the same effect occurs in Jt}ass transfer in a liquid flow, e.g. on dissolving crystals.
Ascertain that jn this case usually very ' small" temperature differences occur.
----- -
294
in a gas stream. Using equation (IV.82) PA,g can then be calculated. The psychro-
meter is usually used for the system air/ water, but application to other liquids is
in principle also possible. The method is particularly useful because T.., is
independent of the gas rate. However, in such a measurement we should make
sure that the heat flow to the 'wet' sphere by conduction or radiation is small
with respect to the heat flow from the gas.
The wet-bulb temperature must not be confused with the so-called adiabatic
saturation temperature This is the temperature a gas (e.g. air with a tem-
perature T, and water vapour pressure PA.J attains if it is saturated with vapour,
the gas itself supplying the required heat of evaporation by cooling from Yg
to J;. The calculation of 7:; is based on an energy balance over a unit of mass
of the gas, in which. tb.e transport of heat and mass does not play a role. Derive
that for 7; (and the relevant equilibrium -vapour pressure PA,s) applies :
PA,s- PA,g _ RTcPp
- 1; - filie,A
(IV.83)
Except for the factor (Le)i, (69) corresponds with (68). For all gas
Le lies near 1 and for the system air/water vapour (Le)i = 095. Therefore,
sometimes (wrongly) no distinction is made between Tw and 7, , although with
regard to mechanism, they are of completely different origin.
IV.6.2. Problems
*1. Air of20C and atmospheric pressure with a relative humidity of79 per cent
is heated to 50C. What will be the partial pressure of the water vapour
and the relative humidity?
This warm air along a completely moist r:t being 35 W /m
2
C.
How much water </>';,.' evaporates per unit of a;rea of the . material if the
heat of evaporati.on is entirely supplied .bY a1r?
Temperature 7; \C)
20-0
22-0
25.0
27.0
'30.0
340
37.()
400
50-0
Vapour pressure Ps (10
3
N/m
2
)
238
269
323
364
433
513
642
750
12-33
3
N/m
2
; 15 per 034 x 10-
3
kglm
2
s
2 A wet object of good heat conductivity is put vertically in dry stagnant
air to dry. After a short time a stationary state has set jn and the object
has reached the temperature T..,..,. The heat transfer air-object is described
295
by ( Nu) = constant (Gr p, and change very little during the
drying process, Pw is small with respect to the total pressure and there occur
no dry spots on the object Find a relation forT.., .
p - p RT
Answer: w g = --peP (Le)t
T,- Tw Mle
Problem 1
The partial water vapour pressure in air at 20C and 79 per cent saturation is
PH
2
o = 0-79 x 238 x 10
3
= 188 x 10
3
N/m
2
. If this air is heated to 50C,
PJi
1
o stays constant, but the saturation becomes 188 x 10
3
/ 1233 x 10
3
= 15
per cent The wall temperature is found via:
Pw- Pg = Pw- 186 x 103 = RTcPp Let= 69
T, - Tw 50 - Tw !lHe
by trial and error to be Tw = 26C and Pw = 337 x 10
3
N/m
2
The mass flux
of the evaporating water is given by equation (IV.79)and with the aid of equation
(IV.81) we find :
.p::,,, = :T(pw - p,) = = 186 x 10- kmol/m
2
s
A - -.,cdktJ tt4lt. 'r J-ktnlw rirtJ a 335 x 10-4 kgfm2 s 1 lA.
"7 11JQ-mal Cf r
wj ID o -62f4-ma)t1J ID o u1 ,:_,. cc-,,,d a /al'Cf 'f aJruw
(K- O
2
133 .!!.... I r1rr 't-Ie T
1
. !OJ ) . K q= P'ft If .rn k "'"cl a-Jfrlch T r illf\A'altr
/J c 311) , J . r,-r r1 d . ]OJ }eJ'4-/tJ 'i )' I I
H.t h.Jfl rJ . @ -lilt !nt , me}a) C/ d, , ,' Jrf/J(. lucri /oJr an(l oii'(J
r - 1 ,, tnfl.4 cuw
\
h
()_
(I
/11 . rb- .
r
"
t:iftUW. k = 6. 'l<f H
rf, ..... 0 o SM
dz. = t -JGI6
Tr -
3
" i K.
nlt ...-i)eJl J'ic/f
Gj+l'cl tnttrf
l"oi
tii ; OL rJ
I
. ,. r
f ' ' l,l ,v
,, q
l

<!."
?l c
!7:rl
..

e
oc .

II'U:tl{)
1
r -:--- ..
I
1. ,_. ( ft - To . )) { T ,, - Ia , '
.....!.. > Ti. ..
a ;
.:= {J .3 II
I
- -
Index
I
""-- T - 7
11
op
.' 0 -"() r...
'I 'I
D T \l
1
= _( LJ 1, - L1 h)/ In ( LJY, f A17 )

t- '' = q,l- c
=--1 -qo) - K 70S-qo)/(44 =- lJ -qqoC
;::.- 2 - '3
\
{jh ) . (
Crt-dtd )/J!JTrll :. o.q ,nqq'c) ==-
Absorption coefficient for heat radiation, Condensation
221 ' effect of peonanent gas,.i11
Absorption of gases, 258, 269, 285 heat transfer coefficients, 210
Analogy Conservation laws. 3 .
between heat and mass transfer, 240 Conversion factors. 22
between heat and momentum transfer,
152
Annulus
laminar flow in, 42
with moving inner cylinder, 46
with rotating inner cylinder, 47
Average temperature difference, 182
Average velocity
in annulus, 43
in falling fitms, 40
in tube ftow, 41
Bernoulli equation, 65
Binary distillation, 32
Bingham fluid, 38
Blake Kozeny equation, 114
Blasius formula, 59
Bohzmann's constant, 220
Boundary layer thickness
concentration, 234
hydraulic, 50, 58
thermal, 58
with chemical reaction, 274
Bubbles
flow of, 104
mass transfer to, 241,261
Catalytic reactions, 280, 286
Chemical reactions
heterogeneous, 280
homogeneous,270, 273,276
Chemical reactors, 283
Concentration boundary layer, 59, 234
Concentrations, definition, 226
Diffusion
coefficients, 12, 15, 227
equimolar, 228
through stagnant fluid, 230
Dimensional analysis, 21
Dispersion, 131, 137
Distillation, binary, 32
Distribution coefficients, 232, 258
Drag coefficient, 100
Droplets, 104
Eddy diffusivity, 58
Eddy viscosity, 56
Effective thermal conductivity in granular
materials, 198
Energy balance
macroscopic system, 7, 144
microscopic
Ergun equation, 114
Evaporators, 213
Extraction factor, 259
FalJing film
in condensing systems, 209
laminar flow, 52
Fanning friction factor, 57
Fick's law, 12
Film theory, 234,272
Filtration, 115
Fixed bed
heat transfer, 197
pressure drop, 112
Flow around obstacles, 98
Flow meters, 94
Flow through orifices, 80
Flow through weirs, 85
Fluidized bed
heat transfer, 198
pressure drop, 116
Fourier number, 160
Fourier's law, 12
Free convection
heat transfer, 201
mass transfer, 248, 254
Free faU of droplets, 104
Friction Joss factors, 57, 62, 68
Gas absorption
from bubbles, 241, 261
in_ packed columns, 242, 261, 265, 286
wttb chemical reaction, 270
Gas bubbles
mass transfer, 241, 261
movement o( 104
Graetz number, 173
Grashof number, 203
Hagen- Poiseuille law, 41
Hatta number, 272
Heat conduction
non stationary, 15.6
penetration, 159
146
Heat exchangers, 182
Heat transfer .
analogy with momentum transport, 152
by natural convection, 201
coefficients, 177, 190
combined with mass transfer, 152
during boiling, 212
' d uring condensation, 208
in fluidized beds, 197
in stirred vessels. 215
on a flat plate, 194
to falling films. 194
to laminar ftow, l71
to spheres and cylinders, 195
to turbulent flow, 174
Height of a transfer unit, 186, 261
Henry coefficient, 259
Hydraulic diameter, 61 , 63
Hydrodynamic boundary layer, 58
Kinematic viscosity, 12, 15
Laminar flow
between fiat plates, 37
by moving surfaces, 46
in circular tubes. 40
in non-circular tubes, 48
non 49
througn an annulus. 42
Laplace number, 27
Lewis number, 293
Macroscopic baJances, 3
Mass balance, 6
Mass exchangers
driving force, 262
extraction factor, 259
gas- liquid, 260
liquid dispersion, 262
packed oolumns, 242, 263
size determination, 260
unpacked columns, 266
Mass transport
analogy with heat transport, 246
combined with heat transport, 292
during laminar flow, 248
during turbulent flow, 253
297
in packed columns, 242, 261, 265, 286
overall coefficient, 232
stationary diffusion, 227
to bubbles, 241, 261
with heterogeneous chemical
280
with homogeneous chemical 270
Microbalances, 16
7
Mixing
in pipes, 137
in vessels, 120
Momentum balance, 8, 17
Newtonian fluid, 38
Newton' s law, 12
Non stationary
flow, 49
free faU, 105
mass transport, 240
momentum transport, 49
Nusselt number, 149
Orifice ftow, meter, 93
Ostwald fluid, 38
Overall heat transfer coefficient, 176
OveralJ mass transfer coefficient, 232
298
Packed columns
mass transfer. 242. 261. 265, 286
pressure drop, 112
Peclet number, 25
Penetmtion theory
derivation, 157
diagrams, 164
Physical properties of materials, 24
Power consumption of stirrers, 123
Prandtl number, 58, 173
Pressure drop, 60, 64
of heat, 220
Reaction force on pipe bend, 10
Residence time distribution, 125
Reynolds number, 26, 57
SchnUdtnUJDber,59,247
Sedimentation of particles, 106
Sherwood number, 231
SI-Units, 21
Simultaneous heat and masS transport, 292
Spheres
flow around, 101
free fall of, 104
heat transfer. 196
Spheres--continued
mass transfer, 246
Stoke's law, 101
terminaJ velocity, 102
Statistical transport, 12
Stefan Boltzmann law, 220
Stirred vessels, 120, 215
Stoke's law, 101
Tanks in series, 133
Temperature equalization diagrams, 164
Thermal boundary layer, 58
Thermal conductivity, 145
Thermal diffusivity, 12, 15, 145
TurbulC?nt fiow, SS
Units and conversion factors, 22
Velocity distribution
in laminar flow, 41
in turbuJent flow, 60.
Venturi meter, 92
Void fraction, 112
Weber number, 104
Weirs. 85
Wien's displacement law, 220
r r J I 1- I j_ '-- j
..: l'"r I !LlL'lGl ! "r li1.(.L L! Mr 't 1- em rhrCJc .rerw af a pd
kt d/1
1
bvtt- tJ>+. uJe4 14 J
411

then Mil 2:rc .. M :"- prfl 1-)wv .Jik. through fh.t;
tnal- a.rrum;nj -1-httl-- bcrlbrr1 vva! aJIJhmt e r !
l 't- 4J 1J tJ. llJ 'fll/ r (J-= str kJ f'I J C?t1 a Sp.
(91<

(
(
Xp
;::
) . 6'f
w
d..
==.
L
P4
== o IT3
---
(>-rt) (!o)