Professional Documents
Culture Documents
tube length
Feature Report
Report Part 1 Shellside P.D. Max. tubes for 1
Duty parallel shell
1,600
Heat Exchangers:
1,200 1.08
1,000 1.23
Avoid Vibration
1.42
800
1.68
600 2.08
2.69
400
3.85
design process can save money up front length and number of shells in series or parallel) provid-
ing geometry that satisfies both the required thermal
duty and observes the pressure drop (P.D.) constraints
and trouble later on. But care should be
taken in accepting the computer’s results Tubeside P.D. Max. tube length
Shellside P.D. Max. tubes for 1
Duty parallel shell
Graham T. Polley and M.A.Vidal Farfan,
University of Guanajuato, Mexico
0.5
1,600
Tube count (active tubes)
0.55
M
1.77
echanical integrity is an es- Figure 2. With the initial 1,000
1.89
sential consideration in heat assumption that six tube 800 1.06
exchanger design. The most passes will be used, the
1.31
design space in Figure 1 is 600
common threat to this integ- controlled by the tubeside 1.69
rity is tube bundle vibration. So, the pressure drop. Changing to 400
2.42
computer programs used for heat ex- four passes, we obtain the 3.85
changer design incorporate procedures design space shown here, 0 1 2 3 4 5 6 7 8 9
that assess this threat. However, such which is bound by the shell- Lengh (m)
side pressure drop Number of shells in series = 1
analysis is generally undertaken after
exchanger geometry has been identified.
The result can be that geometry that (sometimes of questionable validity) The approach outlined here illus-
is optimal in terms of thermal perfor- for a poor geometry does not justify trates the primary causes for tube
mance is rejected for a more expensive the engineer in accepting the result. bundle vibration and how their analy-
alternative. This added expense is often The best heat transfer is obtained sis can be considered at the same time
unnecessary. This article shows that for conditions in which there close to that heat transfer and pressure drop
identification of geometry that avoids uniform flow across the tube bundle. analyses are being conducted. The
dangerous vibration can be undertaken Perhaps more significantly, most pro- technique is based on a plot method
as an integral part of the design process cedures used to assess the dangers of for heat exchanger design that is eas-
and provides insight on doing so. (Part damaging vibration actually assume ily adapted to handle tube-bundle vi-
2 of this report, pp. 35–38 addresses the that such uniformity exists. bration analysis and thermal design
challenges of two-phase flow) Recent computational fluid dynam- simultaneously. When such analysis is
Modern computer programs allow ics (CFD) analysis has indicated that placed in a framework where changes
the engineer to explore a very wide uniform distribution of flow is only can be easily made to the design fea-
range of bundle geometry. Baffle cut approximately true for a quite nar- tures that control vibration (such as
can be set anywhere within the range row range of geometry. Therefore, en- inlet and exit clearances, baffle spac-
15 to 45% and baffle spacing up to a gineers must be especially vigilant to ing, and nozzle sizes), it becomes pos-
maximum allowable span set by manu- verify computer results with proven sible to quickly and easily identify
facturer’s standards. Beware, however, design principles and overrule the geometry that both satisfies thermo-
that the fact that a computer program computer when there is disagreement hydraulic performance and avoids
provides a prediction of performance (for more, see box, p. 33). damaging vibration during operation.
30 Chemical Engineering www.che.com January 2012
Vortex shedding First baffle
Tube sheet First row of tubes at entry
Turbulence
A
Vibration amplitude
Onset of
fluid-elastic
instability
Baffle cut
Impingment plate Free flow area A
Vc
Flow velocity View on AAA
Figure 3. Of the mechanisms that can cause Figure 4. The bundle entry area is the first region where vibration is
tube bundle vibration, the most serious and the likely to occur. This is set by the location of the first tube row (set by the
one generally leading to damage of a heat ex- distance between the top of the shell and the first tube row, the “entry
changer is fluid-elastic instability clearance”), the length of the exchanger end zone, the dimensions of the
impingement plate and the layout of the tube bundle
1.5 1.5
1 1
0.5 0.5
0 0
500 1,000 1, 500 500 1,000 1, 500
Figure 7. Figure 2 illustrates that using a single shell con- Figure 8. Based on the results of Figure 7, the sizes of both
taining around 900 tubes would provide a good thermo-hy- nozzles were increased by changing the allowable momentum,
draulic design. However, the vibration analysis (displaying the and the bundle layout was changed from a 90-deg. layout to a
ratio of local to critical velocity) shows that the velocities in the 45-deg. layout. Now the velocities in all of the critical locations
outlet nozzle, at the edge of the impingement plate and in the are well below the critical value
bundle exit row all exceed permitted value IHS ESDU
S
oftware currently in use in the chemical Window/crossflow = 0.79
process industries allows design analy-
sis for geometry that is a long way out- v = 0.15 m/s v = 1.03 m/s
v = 0.30 m/s
side a range that was considered acceptable
by experienced engineers working prior to v = 1.47 m/s
the 1980s. The result is that engineers in a
younger, less-experienced generation accept v = 2.05 m/s
geometry that would have previously been
rejected. The consequences can be poorer- Two fold variation across window
than-expected thermal performance, in- v = 0.1 m/s
creased fouling within heat exchanger shells v = 1.0 m/s
v = 0.44 m/s
and unexpected tube-bundle vibration.
The philosophy behind the current design Poor
v = 1.00 m/s crossflow
approach for avoiding the onset of serious vi-
bration is essentially conservative. It involves behavior
v = 1.32 m/s
comparing maximum velocities encountered v = 0.30 m/s
within the heat exchanger to a critical value
associated with the location where that ve- Geometry rectangle model
locity occurs. However, in the application v = 0.10 m/s Baffle cut: 40%
Crossflow: 6
of this approach it is assumed that veloci- Windows: 24
ties across a tube are uniform. This is only
Figure 10. Although the window-to-crossflow area would be expected to give
approximately true for a narrow range of a fairly uniform flow field, the CFD analysis shows the velocity at the edge of the
bundle geometry. Recent work by Alonso baffle is twice that in the rest of the window. There is a jetting effect between the
Vidal and others [5] uses CFD to determine two baffle edges with the velocity down the side of the facing baffle being three
the relationship between bundle geometry times higher than along the face of the other baffle and over 30% greater than a
and the flow distribution within the bundle. uniform velocity
It shows that with some geometries the fluid
will “jet” from one baffle edge to the other. Window/crossflow = 0.78
Local velocities can be substantially higher
than mean velocities. It can be expected that v = 0.24 m/s v = 1.27 m/s
v = 0.6 m/s
the onset of fluid-elastic vibration will occur
sooner in bundles where the flow is poorly v = 1.5 m/s
distributed than in those in which the flow is
nearly uniform. v = 0.24 m/s
In Figure 10 we show the flow field pre-
dicted within a tube bundle having a 40% v = 0.93 m/s
Reasonable distribution
baffle cut. The ratio of window-to-crossflow
area is 0.79 (a value that would be expected v = 1.04 m/s
to give a fairly uniform flow field). However, v = 1.04 m/s
we observe the velocity at the edge of the
baffle is twice that in the rest of the window. v = 1.00 m/s
There is a jetting effect between the two
baffle edges with the velocity down the side v = 1.04 m/s
of the facing baffle being three times higher v = 1.04 m/s
than along the face of the other baffle, and
Geometry rectangle model
over 30% greater than a uniform velocity. v = 0.46 m/s Baffle cut: 20%
In Figure 11 we show the flow field pre- Crossflow: 18
dicted within a tube bundle having a 20% Windows: 12
baffle cut. The ratio of window-to-crossflow
Figure 11. For a 20% baffle cut it was found that the velocity distributions
areas is 0.78 (close to that for the 40% baffle in the window deteriorated as the area ratio increased above 1.2 and that in the
cut). Here we observe that the velocity varia- cross-flow region they deteriorated as the area ratio fell below 0.7.
tion in the window is just 18%. The velocity
distribution across the crossflow region is
more uniform. For a 20% cut it was found that the velocity distribu- studies, these recommendations are that baffle cut should be in
tions in the window deteriorated as the area ratio increased above the range of 15–30%, with a window-to-crossflow area ratio in
1.2 and that in the crossflow region they deteriorated as the area the range of 0.8–1.2
ratio fell below 0.7. This is close to the recommendations of Gilmour [6], who stated
Our recommendation is that only geometry that provides close that baffle cut should not exceed 25%. According to Saunders [7]
to uniform flow fields should be used. On the basis of the CFD that advice was widely accepted in the industry. ❑
following descriptions we assume that distance between the top of the shell impingement plate, where the flow
the entry nozzle is positioned at the and the first tube row, the “entry clear- area is set by similar measurements
top of the shell and the exit nozzle at ance”), the length of the exchanger end to those for bundle entry (Figure 5).
the bottom of the shell. zone, the dimensions of the impinge- The third region that needs to be
The first region is entry into the ment plate and the layout of the tube considered is the edge of the baffle
tube bundle itself. This is set by the lo- bundle (Figure 4). plate (Figure 6). This is controlled by
cation of the first tube row (set by the The next region is the edge of the baffle cut and spacing.
Chemical Engineering www.che.com January 2012 33
Feature Report