A Design Guide Proposal for

Avoidance of Acoustic Resonances

J. A. Fitzpatrick
Department of Mechanical Engineering,
Trinity College, Dublin University,
in In-Line Heat Exchangers
Dublin 2, Ireland
A variety of methods for the prediction of acoustic resonances in heat exchangers
are recommended in the literature. There is currently no clear guideline available to
designers for the avoidance of these resonances. This paper compares some of the
latest developments in this field, proposes a new procedure for design and outlines a
specific example of the procedure.

Introduction
The prediction of flow-induced noise in gas cross flow heat objective of the present paper is to assess these latest
exchangers has been recognized as a design problem for some developments and to propose a new design procedure for the
years. The first attempt at providing reliable data for a design avoidance of acoustic resonances in heat exchangers.
guide was by Grotz and Arnold [1] who conducted a series of
tests in which they measured the flow velocities required to ex- Prediction of the Critical Velocity
cite acoustic resonances in in-line heat exchangers for a variety
There is much experimental data available for heat ex-
of tube spacings. They recognized that a design guide should
provide both for the prediction of those flow rates at which changers of differing geometrical configurations in the form
resonances might be excited and for some measure of the of an acoustic Strouhal number, S„, defined as:
"acoustic damping" capacity of the proposed unit. Following Sa=fad/Uc (1)
this, both Putnam [2] and Owen [3] suggested methods by where /„ is the frequency of a transverse acoustic standing
which the first of these criteria might be provided for, wave excited in a direction normal to both the tube axes and
although the former regarded vortex shedding as the source, the flow direction at a "through bank" velocity Uc and d is
whereas the latter argued that turbulent buffeting was the the diameter of the tubes in the bank. Although there is no
source. The work of Chen [4] reestablished the concept that a common method of deciding how Uc should be determined
design procedure should consist of two parts. He proposed precisely, the variations introduced by using data from dif-
that potentially resonant flow ranges could be predicted using ferent sources is not considered significant as the general prac-
charts supplied in his paper and that the damping capacity be tice is to measure the upstream velocity and to calculate Uc as
estimated by a criterion based on the flow parameters and the the velocity through the minimum gap in the array. The design
geometry of the heat exchanger. A general method for requirement is for a reliable estimate of Sa for a tube array
avoidance of noise problems in heat exchangers has been with specified tube pitch-to-diameter ratios [8]. From this, an
published by Eisinger [5], but the design procedures recom- estimate of the critical velocity is obtained by applying wave
mended are the same as those outlined by Chen [4], although a theory to calculate fa or the range of /„ for a given bundle con-
novel approach in baffle design is outlined. In a review of the tainment. The calculations for this frequency can take account
general problems associated with flow-induced vibration of the presence of the bank [11] but it has been found that the
analysis in heat exchangers, Paidoussis [6] has highlighted the values calculated-assuming an empty duct - do not introduce
confusion that exists with regard to the provision of a reliable significant error for in-line arrays [12].
procedure for designing against acoustic resonance. A general
Source mechanisms have been classified [8] as vortex shed-
review of the noise induced by both isolated cylinders and
ding, turbulent buffeting, and turbulent noise, each being
groups of cylinders in cross flow by Blevins [7] does not
distinguished by differing spectral characteristics. Chen [10]
resolve any of the difficulties mentioned by Paidoussis [6].
has, however, suggested that jet action in the through-flow
Fitzpatrick [8] has reviewed the literature relevant to noise in
regions controls vortex action and provides the source of noise
tube arrays, compared the methods available to the designer,
with "edge tone" type mechanisms. The objective of this sec-
and suggested a new damping criterion. Although a more re-
tion is to investigate the usefulness of each method for the
cent paper [9] has reported results from experiments per-
prediction of the critical flow velocity.
formed to establish the validity of part of this criterion, Chen
The requirement of the vortex shedding theory is a reliable
[10] has put forward a modified version of his original pro-
estimate of a Strouhal number based on a narrowband peak in
cedure for prediction of flow-induced noise calculations. The
the frequency spectrum of turbulence in the bank. The
Contributed by the Noise Control and Acoustics Division for publication in
hypothesis is that a coincidence of this frequency with an
the JOURNAL OF VIBRATION, ACOUSTICS, STRESS, AND RELIABILITY IN DESIGN. acoustic standing wave frequency will result in resonance.
Manuscript received by the Division, July 19, 1985. Consider the results of a previous investigation [13] in which

296/Vol. 108, JULY 1986 Transactions of the ASME

Copyright © 1986 by ASME

Downloaded 26 Oct 2011 to 10.0.224.98. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 x-Turbulence s p e c t r u m peaks
x - T u r b u l e n c e s p e c t r u m peaks y /S=064
1UUU o - P r e s s u r e s p e c t r u m peaks X
o - Pressure spectrum peaks 1800
- 0 - / x
800 S a = 0 26 " ° o oo xx 1600
X -
—
^.1400 /
o
X
600 c
" ^ _ _ ^ - x S = 0-17 0
= 1200 : /* x ^ ^ g ' X S = 0-36
¥ 5= U
400 _^--^^Ix^' '"x °' ""1000 ^ o
X ' ^

-Sa=0-24 -jjv-^eo 6
800 ^.X^X-S'S=0.23
200 -
1 1 r i l l 600 • ' i • i i 1 t 1 1 1 1 1

10 12 14 16 18 20 22 10 12 14 16 18 20 22
Bank velocity ( m/s)
Bank velocity (m/s)
Fig. 1 Frequency of peaks in turbulence and acoustic spectra Fig. 3 Frequency of peaks in turbulence and acoustic spectra
(X/ = 1.97.x, = 1.97) (x, = 1.97,x, = 1.73)

x-Turbulence spectrum peaks „ X - T u r b u l e n ce spectrum peaks

1 0 0 0-- o - Pressure spectrum peaks 1800 o - P r e s s u r e spectrum peaks

1600 X X X
80 - SQ=0-23 O O OO OO
_ °
N
X
"N
- 1400
- X

x >->
u
^S=0-54

& 600 ^ 1200 X ^

CD ^ - S =0-136
x-x X
cr _ j S ^oS=0 "
CD
^ 1000 -
400
x-- A —x y'o
o^x
-X
- 800 - •'" O ° O 0 X X x S n =0-3

200 600 -
I I 1 1 1 1
1 1 I • i

10 12 14 16 18 20 22 24 26 10 12 14 16 18 20 22
Bank velocity (m/s) Bank velocity (m/s)
Fig. 2 Frequency of peaks in turbulence and acoustic spectra (x( = 1.3, Fig. 4 Frequency of peaks in turbulence and acoustic spectra (x, = 1.5,
x,=1.9) 1.73)

the frequency of peaks in the turbulence spectrum were the use of the vorticity Strouhal number of 0.136 would
monitored in four banks before and during acoustic overestimate the critical velocity with resonance being excited
resonance. Figure 1 shows the results from an in-line square with an acoustic Strouhal number of 0.23. The results for an
array with a pitch to diameter ratio 1.97. The results, array with *,= 1.97 and *, = 1.73 are shown in Fig. 3, and
measured in the first two tube rows, show a frequency/veloc- three vorticity Strouhal numbers of 0.64, 0.36, and 0.23 can be
ity relationship with a Strouhal number of 0.14. Results from identified. The use of the first of these would produce quite
other investigators [9, 12, 15] give a Strouhal number of 0.17 conservative limit velocities and, although some discrete noise
associated with turbulence peak frequencies measured deeper was generated at 15 m/s, the resonance was not sustained until
in the bank for this array configuration. Both of these are a velocity of 21 m/s was reached and coincidence of the
shown in Fig. 1 where the measured acoustic Strouhal number acoustic frequency with the lowest vorticity frequency oc-
was 0.26. It is evident from Fig. 1 that the use of the vortex curred. Here again, although resonance has been excited by
shedding hypothesis would seriously overestimate critical "frequency matching", the use of the measured vorticity
velocities and the resonance appears to have been excited at Strouhal numbers would not provide a reliable method for
twice the measured vorticity in the first few rows of the bank. calculation of the critical flow velocity. Finally, Fig 4 shows
This would seem to indicate some nonlinear interaction be- the results for an array with x, = 1.5 and x, = 1.73. In this case
tween the turbulence and the fluctuating pressure. Figure 2 there is no discrete vorticity source for the acoustic resonance
shows the frequency of peaks in the turbulence and pressure and either of measured Strouhal numbers of 0.54 and 0.44
spectra for a rectangular in-line array with longitudinal and would provide a conservative estimate of critical flow velocity.
transverse pitch ratios of 1.3 and 1.9, respectively. Here again A number of other cases where the use of vorticity (or vortex

c= local velocity of sound s transverse gap (T— d)

d = tube diameter S Strouhal number (fd/U) U = design limit velocity
f = frequency of peak in sa acoustic Strouhal number x, = longitudinal pitch to diameter
spectrum ifad/Uc) ratio
fa = acoustic resonance frequency s„ modified Strouhal number x, = transverse pitch to diameter
L = longitudinal tube spacing ratio
Mc = critical Mach number (Uc/Q T transverse tube spacing
Re c = critical Reynolds number U velocity through bank = damping parameters
(Ucd/v) ur bank velocity at resonance

Journal of Vibration, Acoustics, Stress, and Reliability in Design JULY 1986, Vol. 108/297

Downloaded 26 Oct 2011 to 10.0.224.98. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
shedding) would not be appropriate have been reported and vative value for the critical flow velocity for closely pitched
are detailed in [8]. Thus the vorticity or vortex shedding ap- arrays.
proach is quite unsuitable for a design guide procedure. This approach has been criticized by Chen [10] as being
The buffeting approach, suggested by Owen [3], assumes unreliable for a design guide and he argues that "the
that the turbulence spectrum acquires a frequency longitudinal pitch alone cannot bring the measured points into
characteristic with a wider bandwidth than that associated any orderly relationship." As an alternative, it is suggested
with discrete vorticity. Although Owen derived a relationship that the "through-flow" jet is the dominant fluid mechanism
between the dominant frequency of energetic eddies in the and that this controls the vorticity in the turbulence spectrum.
flow and the geometrical and flow parameters, he noted that, Furthermore, it is argued that a new Strouhal number (Ss)
for the data available from Grotz and Arnold [1], his com- based on the transverse gap between tubes (s) gives a good cor-
plicated relationship correlated the data no better than the relation of the data from [1] when Ss is plotted as a function of
simple relationship given by: (L/s). Figure 5 shows a plot of these data together with results
Sa = l/(2x,) (2) from Fitzpatrick and Donaldson [13], Baylac et al. [12],
McKnight [15] and Jaudet and Hutzler [16]. Although
Although an upper bound, and hence a limit velocity for statistical analysis could be used to provide a correlation, the
design, can be established by extending this as recommended line shown on the graph represents a slope of - 1 and a con-
by Bryceet al. [14] to: stant of 2 such that:
Sa = 1/2(^,-0.5) (3) S, = U[2(JL/s)] (4)
it has been noted [8] that this gives an unnecessarily conser- Although Chen [10] has attempted to use x, as a parameter
to provide for better correlation of the data, most of the
points are fairly well correlated by this line, particularly in the
range of (L/s) from 1 to 3 which is of most interest to
1-0-
designers. Considering equation (4), Ss can be rewritten in
-EquolionA
terms of the acoustic Strouhal number and the transverse pitch
ratio as [Sa{x, - 1 ) ] . Furthermore, the ratio (L/s) can be
05 rewritten as [*,/(*, - 1)] and hence equation (4) reduces to the
same format as equation (2) with the acoustic Strouhal
number estimated from the inverse of twice the longitudinal
pitch ratio. The further attempts by Chen [10] to reconstruct
his vorticity Strouhal number charts are not particularly useful
for design guidelines, especially when it has been shown that
resonances can occur without coincidence of vorticity frequen-
cies and acoustic frequencies.
01 The use of the gap size as the geometric parameter in the
Strouhal number was first suggested by Putnam [2] who ob-
tained good correlation of Grotz and Arnold's data [1] by
0-05 x - ref.[1l plotting it as a function of (2x,-xt). Fitzpatrick [8] has shown
o - ref.[13J that this approach provides the best correlation of available
. - ref.[15]
v-ref.[12] data on acoustic Strouhal numbers. Figure 6 shows S s plotted
A -ref.HS] as a function of (2x, -x,) for the data from references [1, 12,
13, 15, 16]. The use of Ss rather than Sa gives a further im-
provement in the apparent correlation of the data and a linear
relationship could be expressed as:
S, = 0.0487 + 0.0956(2*,-*,) (5)
0.01
05 using a least squares fit of the data. A polynomial relationship
L/S
Fig. 5 Modified Strouhal number as a function of dimensionless would probably reduce the error associated with equation (5);
length (L/s) after Chen [10] but such sophistication is considered unnecessary as the upper

1 2 3 U
2xf-X[ )
Fig. 6 Modified Strouhal number using Putnam's correlation [2]

298/Vol. 108, JULY 1986 Transactions of the ASME

Downloaded 26 Oct 2011 to 10.0.224.98. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 12 x - ref.MJ
o - ref.[13]
.-ref.[15]
A-ref. [16]
10 n - ref [9]

NON RESONANT

m 6

NON RESONANT
0-5 0-6
_1_
0-3 04 05 0-6 07 0? 1/X|
1/X( Fig. 8 Modified damping criterion
Fig. 7 Damping criterion from Fitzpatrick [8]
to the case in which a tube array exists in a small cavity such
that the acoustic resonance frequency is high so that there is
limit, which will give the safe operational velocity limit, can be insufficient turbulent energy in the bank at that frequency. In
written as: a number of tests reported by Fitzpatrick and Donaldson [9]
5 = 0.08 + 0.11(2*,-*,) (6) the tube diameter was increased for a fixed duct height and the
duct height was then decreased so that the slenderness ratio
It seems likely that the degree of correlation achieved by was systematically reduced in an attempt to fix a lower bound
using the gap dimensions in the Strouhal number indicates for the proposed damping criterion. The results are included
that jet noise may be the dominant mechanism in acoustic ex- in Fig. 7 and it can be seen that the final test produced a case
citation in in-line arrays as suggessted by Chen [10]. However, where no resonance was excited. From further analysis of the
his approach does not provide a useful design guide although data, a modified damping criterion was suggested as [9]:
further research may provide this. The suggestion in [8] that a
critical velocity for initial design considerations be estimated A* = ( R e c ) M M c S J [ l/[2(x, - 1)] }(l/x,) (8)
from equation (2) and that a safe limit velocity can be From Fig. 8, this can be seen to produce an improvement in
established using the inequality of equation (6) (which has the correlation of the data points and would seem to be of
been modified somewhat in this paper) would seem to provide potential use to designers. The upper zone would indicate
the most appropriate design procedure concerning flow where resonances would be suppressed since all the parameters
velocities in the unit. in equation (8) will be known for a proposed unit. Further-
more, since the geometric parameters x, and x, provide a value
Damping Capacity of a Unit for Sa, the criterion could also be used to calculate critical
flow ranges and hence duct sizes which would give rise to
It has been reported in a number of papers [1, 4, 17] that
resonances. This would allow a proper design procedure for
acoustic resonances occasionally do not materialize where ex-
the placement of baffles at the construction stage of those
perience would suggest they should. This has generally been
units where potential resonances would not be dissipated by
attributed to the capacity of the heat exchanger unit to
damping.
dissipate acoustic energy, although it was noted by Bryce [18]
that enhanced tube vibration suppressed acoustic resonances.
Originally, Grotz and Arnold [1] suggested, as a parameter, Design Guide Proposal
the ratio of the longitudinal gaps between the tubes to the cav- As outlined in the introduction, a successful design guide
ity height, termed the "slenderness ratio." Subsequent to this, should provide a straightforward method for estimating flow
Chen [4] produced a criterion by which the damping capacity rates at which resonances might arise, for calculating a safe
could be estimated from both flow and geometric parameters. limit velocity for operation and for assessing the damping
It has been shown [8] that both the Grotz and Arnold, and capacity of a unit. From the foregoing sections, it is proposed
Chen criteria do not allow for geometric scaling. As the latter that these requirements be met as follows:
criterion had a wide scatter when applied to available data,
scaling effects were introduced to the Grotz and Arnold Stage 1: For an initial design guideline, the velocity at which a
criterion by Fitzpatrick [8] who rearranged the criterion and resonance might be excited should be determined using the
introduced the Reynolds number to yield a criterion, given by: relationship
Sa = l/(2x,)
A = (Re c )' / V(M c 5 a )[ l/[2(x, - 1)] I (7)
Stage 2: For the final design stages, a limit velocity should be
The Mach number provides a relationship between the calculated using the upper bound provided by Fig. 6 as
velocity of propagation of disturbances across a gap to the
velocity propagation across the cavity and is obtained by Ss = 0.0& + 0M(2x-x)
reworking the Grotz and Arnold criterion [8]. As shown in with
Fig. 7, equation (7) provides a good correlation of the data
available from wind tunnel tests on acoustic resonances. Fur- U=fas/Ss
thermore, the data are seen to be bounded on both sides. It has Stage 3: If the limit velocity calculated in stage 2 is lower than
been suggested [8] that the upper nonresonant zone conforms the heat transfer design requirements, the capacity of the unit
to the slenderness ratio concept (i.e., a large cavity with a to suppress resonance can be estimated by using equation (8)
small gap) and that the lower nonresonant region corresponds in conjunction with Fig. 8. If possible, the minimum cavity

Journal of Vibration, Acoustics, Stress, and Reliability in Design JULY 1986, Vol. 108/299

Downloaded 26 Oct 2011 to 10.0.224.98. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
size can be determined at this stage to ensure that the value for patrick and Donaldson [9] is suggested. The universal ap-
A* lies in the upper nonresonant region. If not, then baffles plicability of this parameter has yet to be proven.
should be inserted such that the value for A* is in the lower
nonresonant zone. References
1 Grotz, B. J., and Arnold, F. R., "Flow Induced Vibration in Heat Ex-
Example of Procedure changers," Dept. of Mech. Eng. Stanford University, California, 1956, Report
No. 31.
Consider the case of a square in-line tube array with pitch to 2 Putnam, A. A., "Flow Induced Noise in Heat Exchangers," ASME Jour-
diameter ratios of 2.0. For the tests reported in [9], the tubes nal of Engineering for Power, Vol. 81, 1959, pp. 417-422.
were 12.7 mm diameter, and a cavity height of the test section 3 Owen, P. R., "Buffeting Excitation of Boiler Tube Vibration," Journal
was 200 mm. Implementing the abovementioned procedure of Mechanical Engineering Science, Vol. 7, No. 4, 1965, pp. 431-439.
4 Chen, Y. N., "Flow-Induced Vibration and Noise in Tube Bank Heat Ex-
would provide the following results: changers due to von Karman Streets," ASME Journal of Engineering for In-
Stage 1: Estimate of critical velocity dustry, Vol. 90, No. 1, 1968, pp. 134-146.
5 Eisinger, F. L., "Prevention and Cure of Flow Induced Vibration Prob-
Sa = \/(2x/) = 0.25 lems in Tubular Heat Exchangers," ASME Journal of Pressure Vessel
Technology, Vol. 102, 1980, pp. 138-145.
Assuming/,, = 800 Hz, critical velocity Uc = 40 m/s 6 Paidoussis, M. P., "A Review of Flow Induced Vibrations in Reactors
and Reactor Components," Nuclear Engineering and Design, Vol. 74, No. 1,
[Experimentally recorded value was 36 m/s] 1983, pp. 31-60.
Stage 2: Calculation of limit velocity 7 Blevins, R. D., "Review of Sound Induced by Vortex Shedding from
Cylinders," Journal of Sound and Vibration, Vol. 92, No. 4, 1984, pp. 455-470.
Ss = 0.08 + 0.11(2x-x) = 0.3 8 Fitzpatrick, J. A., "The Prediction of Flow Induced Noise in Heat Ex-
changer Tube Arrays," Journal of Sound and Vibration, Vol. 99, No. 3, 1985,
i.e., limit velocity, U= 34 m/s pp. 425-435.
9 Fitzpatrick, J. A., and Donaldson, I. S., "Effects of Scale on Parameters
Stage 3: Estimate of duct sizes for avoidance of resonance. associated with Flow Induced Noise in Tube Arrays," Proceedings of Sym-
From Fig. 8, for x, = 2, range of A* for resonance is: posium on Flow Induced Vibrations Vol. 2: Vibrations of Arrays of Cylinders in
Cross Flow (Eds., M. P. Paidoussis, Au-Yang, M. K. & Chen, S. S.), ASME,
10 3 <A*<3.4 X 103 New York, 1984, pp. 243-250.
Assuming air as the working fluid, the critical velocity range 10 Chen, Y. N., "Flow-Induced Vibrations of In-Line Heat Exchangers,"
Proceedings of Symposium on Flow Induced Vibrations. Vol. 2: Vibration in
for this array can be determined as: Heat Exchangers (Eds., Paidoussis, M. P. Chenowith, J. N. & Bernstein, M.
7 m/s < Uc < 82 m/s D.) ASME, New York, 1984, pp. 163-170.
11 Parker, R., "Acoustic Resonances in Passages Containing Banks of Heat
and from this, using equation (1), the acoustic frequencies Exchanger Tubes," Journal of Sound and Vibration, Vol. 57, 1978, pp.
likely to be excited are in the range given by: 245-260.
12 Baylac, G., Bai, D., and Gregoire, J. P., "Study of Flow and Acoustic
150Hz</a<1740Hz Phenomena in Tube Banks," Proceedings of Symposium on Vibration Prob-
lems in Industry, UKAEA/NPL, Keswick, 1973, Paper 219.
From this, duct sizes can be specified so that potential 13 Fitzpatrick, J. A., and Donaldson, I. S., " A Preliminary Study of Flow
resonances might be suppressed. and Acoustic Phenomena in Tube Banks," ASME Journal of Fluids Engineer-
ing, Vol. 99, 1977, pp. 681-686.
14 Bryce, W. B., Wharmby, J. S., and Fitzpatrick, J., "Duct Acoustic
Conclusions Resonances Induced by Flow over Coiled and Rectangular Heat Exchanger Test
Banks of Plain and Finned Tubes," Proceedings of Symposium on Vibration in
A procedure by which the likelihood of acoustic resonances Nuclear Plant, UKAEA/BNES, Keswick, 1978, paper 3.5.
in in-line heat exchangers may be predicted has been outlined. 15 McKnight, W., "Turbulent and Acoustic Coupling in a Cross Flow Heat
No attempt has been made to characterize the likely fluid Exchanger," M.Sc. Thesis, Queen's University of Belfast, 1984.
mechanisms involved, although the data have been correlated 16 Jaudet, A., and Hutzler, D., "Pulsations acoustiques et contraintes
vibratoires dans les echangeurs de chaleur," Institut Francais des Combustibles
most effectively using the transverse gap as the geometric et de l'Energie, Journees Internationales des Echangeurs de Chaleur, Paris,
parameter in the Strouhal number. The proposed method 1971, paper 13, pp. 255-271.
fulfils the criteria outlined in the introduction that both 17 Chen, Y. N., and Young, W. C , "Damping Capability of the Tube Bank
critical flow velocities and damping potential are required to Against Vortex Excited Sonic Vibrations," ASME Journal of Engineering for
Industry, Vol. 96, 1974, pp. 1072-1075.
be calculated. The estimation of a safe limit velocity using the 18 Bryce, W. B., "Experimental Study of Tube Bundle Vibrations," in Ad-
approach originally suggested by Putnam [2] is recommended vances in Thermal and Mechanical Design of Shell and Tube Heat Exchangers,
and the use of the damping parameter introduced by Fitz- National Engineering Laboratory, U.K., 1973, paper 8.

300/Vol. 108, JULY 1986 Transactions of the ASME

Downloaded 26 Oct 2011 to 10.0.224.98. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm