Sie sind auf Seite 1von 6

F

rom time to time, a heat ex-


changer is designed carefully
yet fails to achieve the desired
performance by a wide mar-
gin, achieving, say, only half
the duty. With an understanding of
some of the more common reasons
why this might happen, design-
ers can avoid these problems in the
first place, and troubleshooters can
recognize the root causes quickly.
Exchangers for single-phase opera-
tion, condensing and boiling are con-
sidered in that order here; but as we
shall see, exchangers often handle
a combination of these, and it is not
always obvious which process is caus-
ing the problem. In fact, some of these
problems are quite unexpected and
can even take experienced designers
by surprise.
It must be recognized that the most
important cause of problems in ex-
changers is excessive fouling. Other
articles, books, and conferences have
been dedicated to this problem, so
fouling will not be addressed here.
Instead we consider those exchangers
that have failed for some reason other
than fouling.
SINGLE PHASE
Good flow patterns are key
In a single-phase exchanger, most
problems arise when the unit is de-
signed to meet an unrealistically high
thermal effectiveness. See Realistic
Expectations, p. 45, for the upper lim-
its of common heat exchangers. Oc-
casionally, people inadvertently try to
exceed these values. While not neces-
sarily impossible, such an approach
really pushes ones luck.
To achieve realistically high ther-
mal effectiveness, countercurrent
flow is normally essential. Also, both
streams must be distributed evenly
across any flow path and among all
parallel flow paths. Furthermore,
there must be no axial mixing. Flow
that achieves these characteristics to-
gether is normally referred to as plug
flow. Plug flow, while necessary, is not
a sufficient condition on its own to en-
sure high thermal effectiveness. All
parallel flow paths must also undergo
identical heat transfer processes.
In general, shell-and-tube heat ex-
changers are not good at achieving
plug flow and identical heat transfer
in parallel paths, which is why they
cannot achieve as high a thermal ef-
fectiveness as other types. The main
problems arise on the shell side.
In trying to achieve 90% effective-
ness with a shell-and-tube exchanger,
it is essential to ensure good arrange-
ment of nozzles and headers so that
equal flow occurs in all tubes. On the
shell side, care is needed in the choice
of baffle pitch and cut to minimize re-
circulation paths behind baffles and
thereby minimize the resultant axial
mixing. It is very important to avoid
bypass flows. It may even be necessary
to go to a special design like a twisted-
tube exchanger. For a practical example
of this scenario, see Example 1, p. 45.
The Z and U flow arrangements
Plate-heat-exchanger manufacturers
know well that of the two flow ar-
rangements shown in Figure 1, the
U arrangement gives a better flow
distribution in the plate pack than
the Z arrangement. One might think
that the Z arrangement would be bet-
Feature Report
Sometimes fouling is not
the problem. To get closer
to your design duty,
consider these practical
design tips
Feature Report Feature Report Feature Report
44 CHEMICAL ENGINEERING WWW.CHE.COM DECEMBER 2004
Heat Exchanger Duty:
Going for Gold
David Butterworth
Heat Transfer and Fluid Flow Service
FIGURE 1. In the plate exchanger with a Z arrangement, momentum changes
throughout the exchanger give rise to a larger (unwanted) variation between its
header pressures (graphs above and below each diagram) than that of a U ar-
rangement. Therefore, the U arrangement is preferred for its more identical behav-
ior of parallel channels
ter, since it gives equal-length flow
paths for each parallel stream. (See
Example 2, p. 47, for an illustration of
how misleading this concept can be.)
However, we have to look at the mo-
mentum changes in the header as well
as the frictional losses.
The small graphs over and under
each figure show the pressure change
in the headers when momentum
change effects are taken into account.
It can be seen that the Z arrangement
gives rise to a larger variation in the
pressure difference that drives the
flow in the parallel channels.
CONDENSATION
Venting
One of the main causes of problems
with condensers is the failure to vent
noncondensable gases. This results in
the depression of the dewpoint in the
condenser and hence the loss of tem-
perature driving force. High noncon-
densable-gas concentrations also lower
the heat transfer coefficient. Consider
these rules-of-thumb for good venting:
Vent from the cold end of the con-
denser, where noncondensable con-
centrations are highest
Avoid any direct paths (from inlet to
vent) that do not cross any tubes
Keep the pressure drop per unit
length as uniform as possible along
the flow path to drive the noncon-
densable gas toward the vent (thus
avoiding formation of noncondens-
able gas pockets)
Keep the vent clear of the conden-
EXAMPLE 1.
SHELL-AND-TUBE GAS HEATER
CHEMICAL ENGINEERING WWW.CHE.COM DECEMBER 2004 45
T
his case involves a cross-flow shell-and-tube exchanger with
two tube-side passes formed with U tubes (TEMA AXU type).
The shell side has nitrogen gas, which is to be heated from
21.5

C to 135

C by condensing steam at 151

C inside the tubes.


Under the design operating conditions, almost all the thermal resis-
tance is on the shell side around 98%. While this arrangement
is apparently a simple design, a quick calculation shows that the
thermal effectiveness is 88%, a high enough value that any prob-
lems arising will have serious effects.
The original design was as shown in Figure A (right). However,
during construction of the plant, problems were found with the pip-
ing layout so that the nozzles and bundle layout were rearranged
as shown in Figure B. Nobody thought to block the interpass by-
pass flow path on the shell side, so the exchanger performed far
below the design duty. The outlet gas temperature achieved was
only 97

C as compared with the design value of 135

C. The actual
heat load was only 66% of the design value, and the overall heat
transfer coefficient appeared to be only 42% of the design value.
The first attempt at a solution was to immediately insert a metal
box to block the interpass gap, as shown in Figure C. Unfortu-
nately, while some improvement was achieved, the problem was
not completely solved. The gas outlet temperature was raised to
116

C from 97

C but still fell short of the design value of 135

C.
The overall heat-transfer coefficient had increased to 66%.
At one point, the designers decided to replacing the square tube
layout with a triangular configuration to increase the number of
tubes. It is good that this option was not actually implemented,
because it would not have solved the problem.
The problem was instead with the condensing on the tube side.
The way the U tubes were arranged meant that there was really a
single pass on the tube side rather than two passes. To achieve the
design condition, the first tubes encountered by the gas flow would
experience a temperature difference of 129.5

C, and the last only


16

C. This is a ratio of 8.1 to 1. As the design overall coefficient


was constant, the same ratio was being expected for the heat
loads to the first and last tubes.
In the final and ultimately successful solution, the tube bundle
was rotated as shown in Figure D. Calculations proved that to
achieve the design condition, the outer U tubes would have to
handle about twice the duty of the inner U tubes. This seemed
better than before, but we were not sure if this alone would solve
the problem. So we went for both a rotation of the bundle and a
change to a triangular pitch. The solution was successful, giving
an outlet gas temperature of 140

C and thermal effectiveness of


91.5%, an improvement for both values over the design value.
Incidentally, after all of the attempts to troubleshoot this ex-
changer, it would seem that the original design (Figure A), would
have worked perfectly.



Exchanger type
Shell-and-tube 90%
Plate-and-frame 95%
Plate-fin 98%
Printed circuit 98%
REALISTIC EXPECTATIONS
One important cause of problems is when
the exchanger has been designed to have
an unrealistically high thermal effectiveness
(). The definition of thermal effectiveness is
(1)
where Q is the heat added or removed
from the stream, and Q
max
is the theoreti-
cal maximum amount of heat that can be
added or removed.
The most important stream in this context
is the one with the highest value of . For
single-phase streams with constant specific
heat, Equation (1) becomes, for the hot and
cold stream, respectively (of a two-stream
exchanger):
(2a)

(2b)
When one is trying to achieve a high ther-
mal effectiveness, any flaw in exchanger
performance will have serious conse-
quences. Some exchangers are capable
of higher thermal effectiveness than others.
Table 1 indicates the maximum percentage
that can be realistically achieved for differ-
ent types. To achieve the values in Table 1,
great care is required in design. Going be-
yond these values is not recommended.
sate layer to prevent flooding
Be sure that parallel flow paths
have identical duties
Figure 2 shows a vertical, shell-side
condenser with a sensibly positioned
vent that is at the cold end of the
condenser and high enough up not
to flood with condensate. It has been
known for people to put the vent at the
top of the shell, using some mistaken
reasoning that noncondensable gases
rise. The problem with that arrange-
ment is that the whole of the con-
denser must fill with noncondensable
gas before the vent starts to operate.
More information on the good loca-
tion of vents is given elsewhere [1].
Here are explanations of some slightly
unexpected venting errors that have
caused serious problems.
Parallel condensing
paths with different duties
Figure 3 illustrates a problem that can
occur with air-cooled condensers. In
this particular case, the condenser has
one tube-side pass and two rows, but
the problem can occur with other ar-
rangements as well. In the illustrated
case, the temperature difference of the
bottom row is greater than that of the
top row. Because of the difference in
heat loads, all of the vapor entering
the bottom row condenses, while some
in the top row does not. In compensa-
tion, the bottom row sucks in the vapor
from the outlet header that has not
been condensed in the top row. As the
vapor enters the bottom row from both
ends, the noncondensable gases cannot
be vented and will be trapped, causing
a deterioration in performance of the
bottom pass.
The problem of parallel paths with
different duties can occur in many
situations (also illustrated in Exam-
ple 1). Additionally, Figure 4 shows a
TEMA J-type shell with condensation
on the shell side. The problem in this
arrangement is that there is only one
tube-side pass for the coolant flow.
Thus, one half of the exchanger has a
higher heat load than the other, caus-
ing a pocket of noncondensable gas to
form away from the vent. Incidentally,
the vent is actually in a sensible posi-
tion for designs with many tube-side
passes, which would cause the two
halves to have nearly the same duty.
Zero pressure drop
condenser with dead zones
It is possible, particularly in vacuum
condensers, to have zero pressure
drop, because the frictional losses bal-
ance the momentum recovery that
results from vapor deceleration. The
problem with this condition, of course,
is that there is nothing to drive out
the gas. Increasing the venting is not
a good solution, because rather than
clearing the dead tube, it tends to suck
out vapor from the operating tube.
The problem may not manifest itself
at first, since time is needed for the
noncondensable gases to accumulate.
In fact, the problem could go unno-
ticed for years causing a significant
loss of thermal efficiency throughout
the plant if, say, only one of two con-
densers in parallel is affected.
Feature Report
46 CHEMICAL ENGINEERING WWW.CHE.COM DECEMBER 2004
FIGURE 2. The placement of vents in
vertical, shell-side condensers should be
at the cold end of the condenser, just high
enough to avoid condensate flooding
FIGURE 3. A lesser heat load in the top row of a parallel path inhibits complete
condensation of vapor, which ultimately limits performance of the bottom pass
FIGURE 4. With only a single tube pass for coolant flow, this exchanger has a
higher heat load on one half than on the other. This causes uneven condensation and
inefficient heat transfer
FIGURE 5. With this configuration, the large vapor inlet flow and rapid condensa-
tion achieve a large enough deceleration and pressure recovery to overcome the
frictional pressure drop. To be stable, the zero-flow tube has to be filled with non-
condensable, inert gas
So, with vacuum operation, never
design a condenser to have zero pres-
sure drop. Also, follow the good vent-
ing rule that the pressure drop per
unit length through the condenser is
kept as uniform as possible. For con-
densers in parallel, always have sepa-
rate vents, rather than relying on a
manifold system.
In Figure 5, for instance, all flow
is running through the bottom tube.
With this configuration, the large
vapor inlet flow and rapid condensa-
tion achieve a large enough decelera-
tion and pressure recovery to over-
come the frictional pressure drop. To
be stable, the zero-flow tube has to be
filled with noncondensable, inert gas.
Condensate not draining
Condensation gives a high heat-trans-
fer coefficient, whereas a stagnant pool
of condensate gives a very low heat-
transfer coefficient for the tubes sit-
ting in the pool. So, if the condensate
is not draining properly, the efficiency
of heat transfer will be compromised.
The problem is somewhat common
during plant startup, because debris
that is generated during construction
gets lodged in the outlet line. If that
is not the cause, the problem may be
arising because the condensate outlet
line is too small.
Failure to drain condensate is a
problem that sometime occurs with
vertical thermosiphon reboilers heated
with condensing service steam. Usu-
ally, the reboiler has been designed
with an unrealistically high fouling
resistance, so that, initially, while
clean, it will over-perform.
To bring the reboiler back to the de-
sired performance, the common reac-
tion is to reduce the steam pressure,
which in turn lowers the condensing
temperature. However, the steam trap
may not have been designed to oper-
ate at the lower pressure. If not, the
condensate will not drain properly,
and the shell will fill with condensate
until the pressure from the liquid head
is high enough to force the steam trap
to work. This problem manifests itself
periodically, as a slow deterioration in
performance and a jump back to good
performance. Such unstable behavior
in the reboiler can be violent enough
to cause instabilities in the distillation
column to which it is attached.
BOILING
Unpredictable nucleate-boiling
heat-transfer coefficients
Nucleate boiling is strongly depen-
dant on the microscopic nature of the
surface and the wetability of that sur-
face with the liquid being boiled. Such
properties are variable and difficult to
predict. It has been shown [2] that all
nucleate boiling correlations can be
transformed to the following form:
(3)
where is the heat transfer coefficient
and q is the heat flux. The quantities
A and n are constants for a given fluid
boiling on a given surface at a given
pressure. Typically, n is around 2/3.
Various correlations are available to
give A, but they are not very accurate.
Errors in the prediction of A of a factor
of two or more are common.
The heat transfer coefficient is de-
fined in terms of the wall superheat,
T
s
, as follows:
(4)
Taking n as 2/3 and eliminating be-
tween these two equations gives the
heat flux as follows:
(5)
Hence, for a nucleate-boiling con-
trolled design, any error in A will re-
sult in a large error in local heat flux,
and therefore, in the total heat load.
Admittedly, most boiling equipment
is not entirely controlled by nucleate
boiling. Yet, the magnitude of unpre-
dictability can still lead to serious
CHEMICAL ENGINEERING WWW.CHE.COM DECEMBER 2004 47
EXAMPLE 2. FLOW MALDISTRIBUTION
IN AN AIR-COOLED HEAT EXCHANGER
I
n the initial specification of a two-
pass air-cooled heat exchanger, the
design engineer was told that even
flow distribution on the tube side would
be vital to the performance of the ex-
changer. The designer therefore came
up with the design shown in the top il-
lustration to the right. By having one
inlet nozzle and two outlet nozzles, and
by putting vertical baffles in the return
header, the designer ensured that all
the flow paths were the same length,
believing that this would give a good
flow distribution. The problem was that
he had set up a Z flowpattern, which
gives rise to large variations between
the header pressures and ultimately
does not maintain identical behavior
in the parallel flow channels.
The obvious solution was to convert the Z flow arrangement to a U arrangement (actually
two U arrangements in parallel). This could easily be achieved by replacing the one cen-
tral inlet nozzle with two one at each end of the header. When other plant personnel
weighed in, however, they insisted on a change that did not affect the pipework.
In the end, a rather complicated solution was found, as shown in the bottom illustration.
The vertical baffles in the return header were replaced by a horizontal baffle, and piping
was integrated inside the unit to take the flow from the center upper compartment and
feed it back into the two ends of the lower compartment. The resultant flow arrangement
was therefore two U arrangements in series (and two in parallel).
An important lesson to learn from this experience is that achieving good flow distribution
is often more complicated than it seems. Momentum changes and their downstream effects
on header flow (vis--vis header-pressure variation) must be taken into account.
FIGURE 6. Variation of temperature
along a vertical thermosiphon reboiler
errors if corrective measures are not
in place. Fortunately, the solution is
quite simple. From Equation (5), it is
clear that only a small adjustment in
temperature difference is needed to
overcome any problems due to a bad
prediction of A. There is usually some
control on the hot-fluid temperature,
which is often condensing service
steam, and hence on the overall tem-
perature difference and wall super-
heat.
The lesson to take away here is that if
nucleate boiling plays a significant role
in an exchanger performance, make
sure that the temperature difference
can be controlled. This lesson was forgot-
ten for a period in the 1980s when pinch
technology was becoming popular as a
method of saving energy. People were
replacing service steam with another
process stream to do the heating. This
replacement caused a loss of control of
the temperature difference and resulted
in failure of some boiling equipment.
Superheat required
to initiate boiling
A wall superheat is required to initi-
ate bubble nucleation and, therefore,
to enable boiling. At ambient and
higher pressures, this critical super-
heat is small and will be exceeded
by the available superheat. However,
with increasing vacuum, the required
superheat for nucleation increases. So,
for example, the superheat for water
at atmospheric pressure is typically
around 4C for a modest heat flux.
However, at a tenth of an atmosphere
it will be around 10C.
Again, the actual superheat is dif-
ficult to predict and depends on fine
features of the surface. Care must
therefore be taken in vacuum boil-
ers that there is enough temperature
difference to initiate boiling. Alterna-
tively, the need for nucleate boiling
can be avoided by, say, using a falling-
film evaporator.
Loss of temperature
difference in reboilers
Figure 6 shows a sketch of the tem-
perature variation versus height in
a vertical thermosiphon reboiler. The
liquid entering the reboiler must be
at a higher pressure than at the sur-
face of the liquid in order to meet the
hydrostatic head needed to drive the
natural circulation. This means that
the liquid is subcooled at the inlet, and
there must be a liquid-heating region
in the bottom portion of the tubes.
Moving up the tube, the liquid tem-
perature rises and the pressure falls
until saturation is reached and bulk
boiling commences. From there (with
a relatively pure fluid), the boiling
temperature falls with the falling
pressure. From the figure, it can be
seen that the average temperature
difference is lower than the inlet and
outlet values. In addition to having a
lower temperature difference, the heat
transfer coefficients in the liquid heat-
ing region are lower than if boiling
were to start at the entry to the tube.
So, the heat transfer performance is
reduced by these two effects.
It is important to note here that these
hydrostatic effects are more dramatic
under vacuum conditions because the
hydrostatic pressure differences are
a larger fraction of the total pressure.
That state of affairs, combined with
the need for sufficient superheat to
give nucleation, can make vacuum re-
boilers unpredictable. In some cases,
falling film evaporators can be a better
design for vacuum operation.
While this example focused on ver-
tical reboilers, the same problem can
occur in kettle and horizontal (shell-
side) reboilers. In the case of horizon-
tal, tube-side reboilers, care must be
taken if the shell diameter is large, be-
cause then, the lower tubes in the bun-
dle will behave very differently from
the higher tubes. Check the saturation
temperature for tube-side liquid at the
top and bottom of the bundle, and go
to a smaller shell diameter if the dif-
ference in saturation temperature is
large compared with the driving tem-
perature difference.
Recirculation in vertical
thermosiphon reboilers
There needs to be enough vapor
leaving the tubes to drag the liquid
smoothly out. If not, unstable opera-
tion will result. It is also important
to get the liquid smoothly out of the
reboiler through the outlet nozzle and
pipework (see Example 3, above).
Problematic operation in this re-
gard is characterized by large diam-
eter reboilers with poor and fluctuat-
ing performance. They are certainly
unstable, and may actually shake as
well as upset the performance of the
distillation column.
That problem is somewhat common.
It occurs more often in reboiler appli-
cations that caused problems in the
past. Upon replacement of the reboiler,
the temptation is to make the next re-
boiler bigger in an attempt to avoid the
problem. The tubes, of course, cannot
be made longer without the expense
of raising the distillation column, so
more tubes are added. But, this tactic
merely makes the problem worse.
To avoid this problem, use a flow pat-
Feature Report
48 CHEMICAL ENGINEERING WWW.CHE.COM DECEMBER 2004
EXAMPLE 3. INSTABILITY IN A
VERTICAL THERMOSIPHON REBOILER.
A
new reboiler designer discovered instability
in one of the plant's vertical thermosiphon
reboilers, in that it had problems getting liq-
uid smoothly out of the exchanger. The designer
concluded that the solution was having high
velocities in the two-phase region, which would
have to be paid for with increased pressure drop
there. Also a better outlet nozzle design was in-
vestigated.
So far so good, but then the designer concluded
that pressure drop in the liquid region was
wasted and decided to install large inlet nozzles
and pipework (figure). Consequently, the reboil-
ers were very unstable and the problem had to
be rectified by putting valves in the inlet lines so
that a liquid pressure drop was introduced and
could be adjusted until the instability ceased.
tern map or other criterion that will
ensure that annular flow is achieved
at the end of the tubes.
Dynamic instability in a
vertical thermosiphon reboiler
As a consequence of the compressibil-
ity of the two-phase flow in a reboiler,
there is a time delay between a change
in liquid inlet flow and the increased
pressure drop that would otherwise
stop the flow increase. For flow oscil-
lations of a particular frequency, the
time delay will be just enough to re-
inforce the flow variation rather than
damp it out. Such cases create dy-
namic instability.
The problem is normally avoided
by having some restriction to in-
crease pressure drop in the single
phase region. This removes the time
lag between the flow change and the
pressure-drop change. Mathematical
models have been devised to predict
the conditions which cause such insta-
bilities.
Temperature difference of wide
boiling-range mixtures
Normally with a wide-boiling-range
mixture, the liquid starts to boil at the
bubble point of the mixture, and the
boiling temperature rises from inlet to
outlet of the exchanger. Temperature
differences can be calculated from the
boiling curve, which gives the rise
in temperature versus the amount of
heat added and is typically based on
the reboilers inlet flow.
However, it is easy to forget that
there is internal circulation in the re-
boiler, so the fluid mixture entering the
bottom of the bundle is much richer in
low-volatility (high boiling point) com-
ponents than the inlet stream. Hence,
the temperature differences driving
the heat transfer are lower than what
a boiling curve that is based on the
inlet stream would predict. Failure to
recognize this effect has been known
to cause some kettles to operate below
the design performance.
Edited by Rebekkah Marshall
References
1. Butterworth, D., Condensers: Thermohy-
draulic Design, pp. 647 678. In: Kaka, S.,
Bergles, A.E., and Mayinger, F., eds., Heat
Exchanger, Thermal-hydraulic Fundamen-
tals and Design, Hemisphere Publishing
Corp., Washington, New York, London, 1981.
2. Cooper, M.G., Correlations for Nucleate Boil-
ing Formulation Using Reduced Proper-
ties, Physicochemical Hydrodynamics, Vol. 3,
No. 2, pp. 89111, 1982.
NOMENCLATURE
A Variable in nucleate-boil-
ing prediction methods,
W/(mK)/(W/m
2
)
n
n Index in nucleate boiling pre-
diction methods
Q Heat transfer rate: W
T Temperature:
o
C or K
Heat transfer coefficient: W/
m
2
K
Thermal effectiveness
T Temperature difference
Subscripts
cold Cold stream
hot Hot stream
in At inlet
out At outlet
max Theoretical maximum
s Superheat
Author
David Butterworth is a se-
nior consultant to the Heat
Transfer and Fluid Flow
Service (HTFS; 29 Cleve-
lands Abingdon Oxfordshire,
OX14 2EQ, UK; Phone: +44
(0)1235 525 955; Fax: +44
(0)1235 200 906; Email:
davebutterworth@aol.com)
After obtaining his B.S. ChE
from University College Lon-
don (U.K.) in 1966, he joined
the heat transfer team at the Harwell Labora-
tory of the United Kingdom Atomic Energy Au-
thority where he helped in the foundation of the
HTFS, which he went on to manage for 14 years.
The objective of HTFS is to undertake research
in process heat transfer and use the results in
the development of design software for heat ex-
changers and fired heaters. For his contribution
to process heat transfer, Butterworth received
the American Institute of Chemical Engineers
D.Q. Kern Award in 1986 and was elected a Fel-
low of the UK Royal Academy of Engineering
in 1991. He is currently the Chairman of the
Aluminum Plate-Fin Heat Exchanger Manufac-
turers Association (ALPEMA), a visiting indus-
trial professor at Bristol University and Senior
Consultant at HTFS, which is now part of Aspen
Technology. His other engineering qualifications
include Chartered Engineer, European Engineer,
Chartered Scientist and Fellow of the UK Insti-
tution of Chemical Engineers. When not engaged
in engineering activities, he enjoys landscape
painting and cooking, is a council member of the
UK Herb Society and is a Fellow of the Royal
Society of Arts.
adlinks.che.com/3649-24

Das könnte Ihnen auch gefallen