Vous êtes sur la page 1sur 13

Applied Thermal Engineering 22 (2002) 789–801

www.elsevier.com/locate/apthermeng

Design of shell-and-tube heat exchangers when the


fouling depends on local temperature and velocity
David Butterworth *

HTFS, Hyprotech, The Gemini Building, Fermi Avenue, Harwell International Business Centre, Didcot,
Oxon OX11 0QR, UK
Received 22 September 2001; received in revised form 22 November 2001; accepted 1 December 2001

Abstract
Shell-and-tube heat exchangers are normally designed on the basis of a uniform and constant fouling
resistance that is specified in advance by the exchanger user. The design process is then one of determining
the best exchanger that will achieve the thermal duty within the specified pressure drop constraints. It has
been shown in previous papers [Designing shell-and-tube heat exchangers with velocity-dependant fouling,
34th US National Heat Transfer Conference, 20–22 August 2000, Pittsburg, PA; Designing shell-and-tube
heat exchangers with velocity-dependant fouling, 2nd Int. Conf. on Petroleum and Gas Phase Behavior and
Fouling, 27–31 August 2000, Copenhagen] that this approach can be extended to the design of exchangers
where the design fouling resistance depends on velocity. The current paper briefly reviews the main findings
of the previous papers and goes on to treat the case where the fouling depends also on the local temper-
atures. The Ebert–Panchal [Analysis of Exxon crude-oil, slip-stream coking data, Engineering Foundation
Conference on Fouling Mitigation of Heat Exchangers, 18–23 June 1995, California] form of fouling rate
equation is used to evaluate this fouling dependence. When allowing for temperature effects, it becomes
difficult to divorce the design from the way the exchanger will be operated up to the point when the design
fouling is achieved. However, rational ways of separating the design from the operation are proposed.
 2002 Published by Elsevier Science Ltd.

Keywords: Design; Fouling; Heat exchangers; Shell and tube

*
Address: 29 Clevelands, Abingdon, Oxon, OX14 2EQ, UK. Tel.: +44-1235-525-955; fax: +44-1235-200-906.
E-mail addresses: davebutterworth@compuserve.com, dave.butterworth@hyprotech.com (D. Butterworth).

1359-4311/02/$ - see front matter  2002 Published by Elsevier Science Ltd.


PII: S 1 3 5 9 - 4 3 1 1 ( 0 2 ) 0 0 0 2 5 - X
790 D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801

Nomenclature

D tube internal diameter, m


Da flow diameter with asymptotic fouling, m
E energy of activation in Ebert–Panchal equation: 68 kJ/mol
f friction factor
kF thermal conductivity of fouling, W/m K
r tube-side fouling resistance: m2 K/W
R universal gas constant, 8:314  103 kJ/mol
Re tube-side Reynolds number
t time, s
Tf film temperature, K
u tube-side velocity, m/s
uT threshold velocity below which fouling occurs, m/s
Greek
a constant in Ebert–Panchal equation: 8.39 (m2 K/W)/s
c constant Ebert–Panchal equation: 4:03  1011 Pa (m2 K/W)/s
da asymptotic fouling layer thickness, m
l viscosity of tube-side fluid, Ns/m2
q density of tube-side fluid, kg/m3
s shear stress on surface of the fouling or at wall for a clean exchanger, Pa

1. Introduction

There has been a recent rapid development on the science of fouling leading to better prediction
methods. Many of these methods treat the fouling buildup rate as dependent on the local con-
ditions of say velocity and temperature within the exchanger. At present, these equations are
rarely used in design for the reasons that

• knowledge is needed about the way the exchanger will be operated up to the point when the
design condition is reached.
• The design process can become very complicated.
• We do not yet have any consensus between the purchaser and the designer about how to use the
equations.

This contrasts markedly with the customary simple approach of specifying a single, constant
fouling resistance that the designer incorporates into his/her design.
The current paper firstly reviews recent past work by the author in which the fouling resistance
is assumed only to vary with velocity. It then goes on to consider ways of applying the equations
which do allow for temperature effects. In doing this, some ways are suggested of addressing the
above barriers to using the recent equations.
D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801 791

In this paper, simplified cases are deliberately used in order to illustrate the phenomena clearly
without getting distracted with details. However, the main points that are made will apply also if
more complex and realistic cases are considered.
This paper is an update of one presented as an ‘‘Unpublished’’ paper in a topical session at the
AIChE Spring Meeting, April 2001 [4].

2. Definition of design

The word design covers a wide range of activities. In this paper, the restricted form of the
definition is used, which is that used invariably in the chemical and petroleum industries when a
purchaser asks for bids for an exchanger for a specific application. In this situation, the purchaser
specifies the flow rates and inlet and outlet temperatures of each stream. This also defines the heat
load or thermal duty for the exchanger. In addition, the purchaser also specifies the maximum
allowable pressure drops for the two streams. There will, of course be other items in the speci-
fication, such as avoiding flow-induced vibration, but these are not relevant to the current paper.
The design process is then one of determining the size and configuration of heat exchanger which
will achieve the thermal duty within the imposed constraints.

3. Summary of previous work

In previous papers [1,2], the author considered the design consequences of allowing only for the
effect of velocity on the fouling resistance. The case considered was of a hydrocarbon that is being
heated inside tubes using condensing vapour on the shell side. For simplicity, the shell side
was assumed not to be controlling (i.e. to have a very high coefficient and to have no important
pressure drop limit) and to be isothermal.
The design process is explored using the ‘‘design-envelope’’ concept which is described briefly in
the appendix and discussed in more detail by Butterworth [5]. This method has been incorporated
by HTFS into a piece of educational software known as DEVIZE, and into the rigorous air-
cooled exchanger design program ACOL. A similar method has been developed by ESDU In-
ternational and is described in their various data items.
The following design cases were considered, and are illustrated in Figs. 1–5 respectively.

1. No fouling: this gives an optimum design with 303 tubes 4.65 m long.
2. Addition of constant fouling resistance to the above example. This displaces the heat transfer
locus to the right thus causing the optimum to move to a point with both longer and more
tubes. The broken line on the curve shows what happens if the purchaser were to increase
the specified maximum allowable pressure drop from the that originally specified. This is a pos-
sible way of reducing the size of the exchanger finally designed.
3. Step change in fouling. This shows the heat transfer locus obtained if the fouling resistance is
zero at high velocity but has a high but constant value below some threshold velocity. In the
conditions chosen here, two design envelopes are obtained, one for a clean design and one
for a fouled design.
792 D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801

Fig. 1. Design envelope for a clean exchanger.

Fig. 2. Envelopes for a clean and fouled exchanger.

Fig. 3. Envelope with a step change of fouling thermal resistance.


D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801 793

Fig. 4. Envelope with a step change of both fouling resistance and friction factor.

Fig. 5. Envelope with a continuous increase in fouling resistance with reducing velocity––including the effect of in-
creasing the allowable pressure drop.

4. The last case with the additional effect of the fouling increasing the friction factor. Again, this
gives two design envelops but that for fouled exchangers is displaced to a higher number of
tubes.
5. Fouling increasing smoothly with decreasing velocity. A strong dependence of fouling resis-
tance with velocity was assumed here which causes the heat transfer locus to bend over to
the right at high numbers of tubes. This is because the overall coefficient falls rapidly with de-
creasing velocity (because of the strong fouling dependence) thus requiring lengthening of the
tubes to provide the required heat transfer area. As with case 2 above, this figures explores, in
addition, the possibility of increasing the allowable pressure drop.

The main conclusions from these studies were:

1. Allowing for the variation of fouling resistance with velocity at the design stage enables the de-
sign engineer to provide an exchanger with less likelihood of fouling.
794 D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801

2. The inclusion of a fouling resistance has the combined effect of adding the fouling resistance
itself and of lowering the tube side, stream coefficient because more tubes are needed. Hence,
the effect of adding a fouling resistance has a more deleterious effect on the performance than
just the addition of the resistance itself.
3. If we then allow also for the fact that fouling increases pressure gradient, this forces us to a
design with more tubes still further leading to an additional decrease in tube side, stream coef-
ficient.
4. When the fouling resistance increases with reducing velocity, this combines with the two previ-
ous points to give marked reduction of overall coefficient.
5. There are situations where step changes in fouling resistance can lead to two very different de-
signs––one for a clean exchanger and one for a dirty exchanger.
6. Some of the above problems can be avoided if the purchaser could sanction an increase in al-
lowable pressure drop.

It is sometimes said that, when providing the additional area needed to allow for fouling, the
designer should go for longer tubes rather than more tubes. However, point 2 above means that
this is rarely possible unless the purchaser sanctions a higher pressure drop.
Taking points 2–4 together tells us that over-specifying fouling resistance, ‘‘Just to be on the
safe side,’’ has far more serious consequences than is often realised. Not only do we add area to
allow for the fouling but we also have to add area to allow for the lower stream coefficients.

4. The effect of velocity and temperature on fouling

4.1. Fouling equation used

The equation used is that of Ebert and Panchal [3] for crude oil streams, which is as follows:
 
dr 0:88 E
¼ aRe exp   cs: ð1Þ
dt RTf

It is recognised that there are other forms of this equation and that the various constants in the
equation need to be optimised for a given crude stream. The point of this paper is however to
explore methodologies rather than to predict a particular situation. This equation is therefore
considered to be representative of other similar equations.
In the calculations described here, the shear stress, s, is calculated as follows:

f 2
s¼ qu ð2Þ
2
with the friction factor calculated from the Blasius equation:
f
¼ 0:0396 Re0:25 : ð3Þ
2
D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801 795

4.2. Fouling threshold

It is well known that there will be certain conditions of film temperature, Tf , and velocity, u,
where the second term in Eq. (1) is greater than the first ensuring no growth of fouling. Provided
that the whole of the exchanger can be operated under these conditions at all times, the exchanger
will remain clean. Designing to avoid fouling is therefore possible by considering the hot end of
the exchanger and ensuring that the velocity there is greater than the threshold velocity for the
onset of fouling at for the prevailing film temperature.
Setting dr=dt as zero and solving for u, gives the threshold velocity uT as a function of film
temperature as
" #0:38
a expðE=RTf Þ
uT ¼ : ð4Þ
0:0396cqðqD=lÞ0:63

Fig. 6 shows the design envelope for a typical case of a two-pass exchanger. Also shown on this
figure is the locus for fouling to occur at the end of the second pass (called here threshold 2, with
the 2 being for the second pass). To explain, designs above this latter line will be subject to fouling
because the velocity at exit of the exchanger is below the threshold velocity for the outlet film
temperature. Since designs are only valid which are above the pressure drop and heat transfer loci,
there is no fouling-free design in this case. This approach to design has been discussed by Polley
[6].
However, in the current paper an additional locus, threshold 1, is also included on this figure.
This threshold is for fouling to commence at the end of the first pass. This is well above the point
where the pressure drop and heat transfer loci cross and so shows that there is a useful margin
here if only we could find a way to exploit it. One way is to have more tubes in the first pass than
the second pass. Fig. 7 shows the same design but now with the ratio of tubes in pass 2 to pass 1
being 0.9. Both the heat-transfer and pressure-drop loci have moved but only by a very small

Fig. 6. Inclusion of fouling thresholds with the same number of tubes in each pass.
796 D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801

Fig. 7. Inclusion of fouling thresholds but with more tubes in the first pass than the second.

amount. The two fouling thresholds have however moved significantly with both now being high
enough to allow a design satisfying all criteria. Essentially, we have saved pressure drop on the
first pass to give more pressure drop on the second pass to enable the velocity to be increased
above the threshold.
While this is a possible technique, there are few situations where it is likely to be of great
benefit. In order to work, the stream temperature at the end of the second pass must be very much
higher than that at the end of the first pass. Even with a very large stream temperature rise (and 75
K was used in this case) the difference in temperatures at the ends of the passes are much less
because there is a higher temperature rise in the first pass.

4.3. Asymptotic fouling

In some systems, the fouling resistance rises to an asymptotic value after operating for some
time. On the face of it, however, Eq. (1) predicts a continuously rising fouling resistance with no
asymptote. Nevertheless, we can apply the equation in such a way that it will predict an as-
ymptote. We can do this by observing that a finite thickness of fouling builds up. For a given flow
in the tube, this increase in fouling thickness increases both terms in Eq. (1), but the first term
increases more slowly than the second. We can therefore determine a fouling thickness at which
both terms cancel out thus giving no further growth and hence the asymptote.
The condition for zero growth gives the following equation for the internal diameter, Da , inside
the asymptotic fouling layer.
 0:216
Da 0:0396c Re0:63 qu2
¼ : ð5Þ
D a expðE=RTf Þ

In this equation, Re and u are the superficial values (or the values which would be obtained with
the same mass flow in the tubes but without any fouling present).
D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801 797

Fig. 8. Calculated fouling resistances by two methods.

The fouling layer thickness, da , can then be calculated from


 
D Da
da ¼ 1 ð6Þ
2 D
and, hence the asymptotic fouling resistance ra from.
ra ¼ da =kF : ð7Þ
Fig. 8 shows as broken lines the asymptotic fouling resistances calculated by this method. The
fouling layer thermal conductivity, kF , was assumed to be of 0.5 W/m K in calculating these
curves.
One benefit of the asymptotic method is that the resultant fouling resistance depends only on
the end condition and not on the fouling history. It can therefore be applied easily in design. The
current analysis may however be criticized on the grounds that it is stretching the basis of the
Ebert–Panchal method too far.

4.4. Fouling buildup

We could also design on the basis of the fouling at the end of a specified operating period after
which the exchanger will be cleaned. Fig. 8 shows also as solid lines, the fouling resistance pre-
dicted after one year assuming that the velocity and the film temperature remain constant over
this period. Except at the threshold, any agreement between the two methods is fortuitous because
of the arbitrary choice of the time period in the buildup case and the choice of fouling thermal
conductivity in the asymptotic case.
We need to consider the fouling buildup case further to see what operational history should be
assumed in the calculation. We start with a clean exchanger that will over perform unless we take
some steps to cut back on the performance. Given that the fouling rate increases rapidly with
decreasing velocity, we certainly must not cut down on the velocity in any way. The only pos-
sibility is therefore to lower the temperature difference. In our supposed case here, we have to
lower the condensing temperature at the start and then increase the temperature to give us the
798 D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801

Fig. 9. Simulated buildup of fouling with time.

same performance as the fouling builds up. If we suppose that the cold stream is isothermal, this
approach would result in a constant film temperature over the whole time and hence in a constant
rate of fouling.
Unfortunately, this method of doing the calculation neglects the fact that the conditions of
temperature vary throughout the exchanger and therefore the fouling builds up in a complicated
way. We therefore need to investigate this effect. A series of calculations was therefore carried out
to show the buildup of fouling with time and position when there is a realistic temperature rise in
the cold stream. The hot stream temperature was raised steadily as the fouling built up to keep the
heat load on the exchanger constant. The results of this calculation are shown in Fig. 9.
This is a two-pass exchanger with 6 m long tubes, so the distance along the tubes follows the
cold stream from the first to the second pass.
A case was deliberately chosen where, at first, the cold end of the exchanger was not undergoing
fouling. The point of onset of fouling moved upstream with time until fouling was occurring over
the whole tube length. Clearly, there is a strong increase in fouling from the inlet end to the outlet
end of the flow path (from the cold to the hot end).
Although we could run such simulations while designing heat exchangers, it is a difficult and
time-consuming exercise so the question is posed whether we can determine a reasonable design
fouling resistance in an easier manner. A convenient hypothesis is that we calculate the fouling
rate at the design condition and assume that this rate prevails over the whole operating period. A
rating calculation was done on this example using this hypothesis and the resultant heat transfer
area was changed by less than one per cent. Fig. 10 shows the result of the end-rate hypothesis
compared with the full buildup calculation. Clearly the agreement is good. We would need to try
more examples to confirm that this is a universal finding but we can have some confidence in the
method since, as noted above, the buildup case simulated was quite complicated. 1

1
Figs. 9 and 10 are corrected versions of those given in Ref. [4]. An error was found in the earlier calculations but the
difference between the two sets of results is hardly noticeable.
D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801 799

Fig. 10. Comparison of methods of calculating the design fouling resistance.

5. Conclusions

Introducing the effect of temperature on fouling into the design of heat exchangers is shown to
be quite complex because the final, design-fouling resistance depends on the history of operation
of the exchanger. Three methods are considered for overcoming this problem. The first is to design
using the fouling threshold to ensure a clean exchanger. The second is to use the asymptotic
fouling resistance, which can be done but may not be fully compatible with the Ebert–Panchal
method used in the current calculations. Nevertheless, it would be a sensible way to proceed with
design in those cases where a good equation exists to predict asymptotic fouling. The third
method is to analyse the fouling buildup using an assumed operational history. The calculations
involved in this are quite time consuming but a simple hypothesis is shown to work remarkably
well. In this hypothesis, we assume that the fouling rate over the operating period is equal to that
at the end. There is some justification for this hypothesis.

Appendix A. Outline of envelope method

The objective of this method is to design an exchanger that achieve a given thermal duty which
means achieving the required stream outlet temperatures for the given inlet temperatures and
given flow rates of the two streams. The final design must also have pressure drops for each stream
which are below those specified. The full method is explained elsewhere by the author in Ref. [5].
As is stated in the main text of the current paper, a simplified version of the full envelope
method is used here in order to concentrate on the effects of fouling without overcomplicating
matters by introducing subsidiary issues.
Consistent with these simplifications, the following equations are used to calculate the heat
transfer and pressure drop during the design process.
The heat load, Q, is calculated from
Q ¼ UADTm ;
800 D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801

where A is the heat transfer area and DTm is the mean temperature difference which in this case is
the logarithmic mean temperature difference.
The overall heat transfer coefficient, U, is determined from
1 1
¼Rþrþ ;
U h
where R is the thermal resistance between the shell-side stream and the tube inside wall. A con-
stant and high value of this resistance is assumed throughout this paper. The value is consistent
with the shell side being clean and heated by condensing steam. The tube side coefficient, h, is
calculated from the simple Dittus–Boelter equation
k 0:8 0:4
h ¼ 0:023 Re Pr ;
D
where k is the thermal conductivity, Re the Reynolds number and Pr the Prandl number for the
tube-side stream.
Note that h, U and A are based on the tube inside diameter, D, (rather than, as is the con-
vention, the tube outside diameter) because this paper is concentrating on the tube side.
The tube side pressure drop is given by
 
L qu2 qu2
Dp ¼ np 4f þK ;
D 2 2
where np is the number of tube-side passes and K is the number of velocity heads lost per pass due
to entrance, exit and turnarounds. A representative value of 1.8 has been used in the calculations.
The friction factor, f, is given by the Blasius equation (see main text, Eq. (2)).
For a given number of tubes and number of passes, the tube-side velocity is first calculated.
From this, using the above equations, the tube length is calculated which will just use up the
specified pressure drop. This may be plotted as a point on a graph of number of tubes verses tube
length. Repeating this for different numbers of tubes gives a locus of exchanger configurations
which will just satisfy the tube-side pressure-drop limit. This is illustrated in Fig. 11. All ex-

Fig. 11. Illustration of the envelope concept.


D. Butterworth / Applied Thermal Engineering 22 (2002) 789–801 801

changers above the line will satisfy the pressure-drop criterion and are therefore valid designs,
whereas those below the line will not.
The process may be repeated for heat transfer, giving the line shown in the figure. Again,
configurations above this line will satisfy the heat transfer while those below will not. Hence, one
may construct an envelope, shown shaded of those exchangers which will satisfy both the heat-
transfer and pressure-drop criteria. Normally, the cheapest design is that with the least number of
tubes, which is easily determined from where the lines cross. This is referred to as the ‘‘optimum
design’’ in the text.

References

[1] D. Butterworth, Designing shell-and-tube heat exchangers with velocity-dependant fouling, 34th US National Heat
Transfer Conference, 20–22 August 2000, Pittsburgh, PA.
[2] D. Butterworth, Designing shell-and-tube heat exchangers with velocity-dependant fouling, 2nd Int. Conf. on
Petroleum and Gas Phase Behavior and Fouling, Copenhagen, 27–31 August 2000.
[3] W. Ebert, C.B. Panchal, Analysis of Exxon crude-oil, slip-stream coking data, Engineering Foundation Conference
on Fouling Mitigation of Heat Exchangers, California, 18–23 June 1995.
[4] D. Butterworth, Design of shell-and-tube heat exchangers when the fouling depends on local temperature and
velocity, Paper 46a, AIChE Spring Meeting, Houston, Texas, 22–26 April 2001.
[5] D. Butterworth, Visualize your design of shell-and-tube heat exchangers, Chemical Technology Europe 3 (4) (1996)
20–24.
[6] G.T. Polley, Towards Heat Exchanger Design that Takes Direct Account of Fouling, Paper presented at the 2nd
International Conference on Petroleum and Gas Phase Behavior and Fouling, Copenhagen, 27–31 August 2000.

Vous aimerez peut-être aussi