C. R.

GERLACH Vortex Excitation of letal Bellows

Manager, Hydro-Mechanical
Systems Section, Department of
Mechanical Sciences, Southwest
Research Institute, San Antonio, Texas
fluences of bellows damping and acoustical resonances are discussed.
The present paper extends the results presented in. an earlier paper, and pertains to the
prediction of vibratory amplitudes and stress levels resulting from flow-induced vibra-
tions of metal bellows. Included is visual verification of the vortex shedding-bellows
convolution coupling phenomena. Two methods for predicting flow excitation ampli-
tudes are described. One method, called a "Stress Indicator" method is a simplifica-
tion of the other, and allows rapid estimates of bellows dynamic stress values. This
simplified method is complemented by a plot of "Stress Indicator" versus number of
cycles to failure, obtained from experimental data for actual bellows failures. The in-

Introduction amplitudes are described. One, called a "Stress-Indicator"

IN an earlier paper by the writer [ l ] , 1 the initial
results of an investigation of excitation of metal bellows with
internal liquid flow were presented. This work was motivated by
flow-induced fatigue failures of bellows couplings in aerospace
applications. In this earlier paper, studies of flow excitation of
longitudinal modes with some sample bellows showed t h a t the
observed phenomena was very much like the similar problem of
vortex shedding excitation of elastieally restrained cylinders.
Thus it was initially hypothesized t h a t a vortex phenomena was
responsible for the observed bellows vibrations. As noted in the
laboratory, excitation of successive longitudinal bellows modes is
possible as the flow velocity is increased. The range of excita-
tion for each particular mode can be correlated with a Strouhal
number based on convolution root width <r,.; for each mode this
"lock-in" region occurs over a well defined Strouhal number
range.
Prediction of the longitudinal vibration modes of a given bel-
lows can be adequately accomplished utilizing a lumped spring-
mass analogy [1, 2]. Once the possible longitudinal vibration
mode frequencies are known, the vortex-structure coupling can be
anticipated by calculating the Strouhal number for each possible
mode over the anticipated fluid velocity range to see if it falls
within the reported "lock-in" range [2].
The objective of the present paper is to extend the earlier re-
suits, with respect to the vortex excitation process, and the pre-
diction of vibratory amplitudes. Photographs obtained from a
flow visualization study are shown which demonstrate the fluid-
structure coupling. Two methods for predicting flow excitation
' Numbers in brackets designate References at end of paper.
Contributed by the Design Engineering Division and presented
at the Vibrations Conference, Toronto, Canada, September 8-10,
1971, of T H E AMEBICAN SOCIETY OF MECHANICAL ENGINEERS.
Manuscript received at ASME Headquarters, May 12, 1971. Paper
No. 71-Vibr-22.
method is a simplification of the other, and allows rapid estimates
of bellows dynamic stress values. This simplified method is
complemented by a plot of "Stress Indicator" versus number of
cycles to failure, obtained from experimental data for actual bel-
lows failures.
Finally, the paper briefly describes bellows flow excitation when
gases are involved. With gases, the phenomena is complicated
by a radial acoustic resonance which enhances the vortex shedding
process.
Bellows Flow Excitation Mechanism
Based on the results of flow visualization experiments [2],
there is firm evidence t h a t excitation of bellows longitudinal
modes is a result of a dynamic coupling between the bellows
structure and a fluid vortex generation and shedding process.
In every sense, the phenomena is a fluid-elastic instability [3-7],
since a mutual interaction or coupling occurs between the dy-
namics of the bellows (the elastic system) and the nonsteady flow
near and between the convolutions (the dynamic fluid system).
Many factors have an influence on this coupling phenomena, in-
cluding the convolution geometry, the flow conditions, and possi-
ble local cavitation or acoustic resonances.
For flow through a convoluted section of bellows, with the con-
volutions fixed to prevent vibration, a steady turbulent flow
exists throughout the bellows, as illustrated in Fig. 1. This is a
physical picture formulated from our own flow visualization ex-
periments, and from the information reported by Haugen and
Dhanak [8], and Townes and Sabersky [9]. Each convolution, in
general, experiences an internal eddying flow because of momen-
t u m transfer from the free stream. Energy is transferred into
each convolution in a mixing zone or shear layer. The number
and size of the steady vortices which exist in the convolutions are
a function of the convolution geometry. Very shallow convolu-
tions may contain only one vortex, while two or more may exist in
deep convolutions.
T h e physical picture described above is valid only with no

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org

Journal of Engineering for Industry on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
FEBRUARY 1 9 7 2 / 87
Copyright © 1972 by ASME
 Inlet velocity profile
Shear layer or mixing region
Steady vortex motion induced
through momentum exchange
in shear layer.
Possible steady vortex
behavior in different
depth convolutes.
Fig. 1 Fluid flow behavior with stationary convolutions
convolution motion or vibration. When the bellows is flow
excited, the physical picture of internal flow described above is
no longer valid, in general. For this case, a large-scale periodic
vortex formation and shedding process occurs which has been
visualized with an apparatus described in reference [2]. The ap-
paratus consists of a two-dimensional clear plastic channel con-
taining a short section of convoluted metal which is placed so t h a t
the convolution tips are exposed to the fluid flow, as in an actual
bellows. When water is passed through the channel at the
proper velocity, the segment is flow excited; this vibration may
be viewed with the aid of a strobe light. By injecting ink up-
stream from the convoluted segment the vortex shedding process
is easity visualized.
Both still and moving pictures of the vortex shedding have been
made; Fig. 2 shows several of the still pictures which were taken
at random times. The presence of a vortex pattern is evident in
each picture. The movie, of course, shows the interaction of the
fluid motion and the segment much more clearly than do the still
pictures.
Fig. 3 shows the sequence of fluid and convolution events
which have been observed in a frame-by-frame examination of
the motion picture. Note, first of all, t h a t the mode of vibration
of the segment is one where each convolution moves out-of-phase
with the adjacent convolution(s). This corresponds to the high-
est longitudinal bellows mode. The vortex shedding process on
the vibrating segment, as shown by Fig. 3, occurs as follows: I n
Position 1, a large vortex (c) has formed between the convolu-
tions 1 and 2, and is being" pushed" out into the fluid stream by
the pinching action of these two convolutions. A large, well-
formed vortex (a) is moving across the tip convolution 3; the origin
of this vortex will be made clear in the description of other events.
•Nomenclature-
Ap — projected convolute area (~irD„Ji) KA bellows spring rate
CF = vortex force coefficient based on N bellows vibration mode number
convolute height area JV0 number of convolutes
Cm = bellows v i b r a t i o n m o d e factor Np number of plies in bellows wall
Dm = m e a n bellows d i a m e t e r Q dynamic amplification factor, a
h = convolute height measure of damping
/ FEBRUARY 1972
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org
on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Finally, a small vortex (6) is beginning to form on the down-
stream side of convolution 2. In Position I I , vortex (c) has been
pinched from between convolutions 1 and 2, and has now moved
out into the fluid stream, and will soon "detach" from convolu-
tion 1 to be swept downstream. Vortex (b), which is forming
on the downstream side of convolution 2, is gaining in strength.
Vortex (a), which in Position I was beginning to move over the
tip of convolution 3, is now further downstream. In Position
I I I , the space between convolutions 1 and 2 is opening up, hence
fluid is being drawn in, and vortex (d) is beginning to form on the
downstream side of convolution 1. Vortex (c) has been swept
over the top of convolution 2, with the timing being about right
to reinforce vortex (6), which is being pinched from between
convolutions 2 and 3; the combined vortex is labeled (b -j- c).
In Position IV, vortex (d) has gained in strength as it forms on
the downstream side of convolution 1. Vortex (b + c) is being
pinched out from between convolutions 2 and 3, and will begin
to propagate downstream. Finally, Position V is the same as
Position I, so the cycle is complete. Note that vortex (a)
in Position I was a combination of two vortices, as is vortex
(b + c) in Position V.
I t must be emphasized that the series of fluid events described
in the foregoing are probably not exactly descriptive of all bellows
modes excited. The main difference likely seen from a com-
parison with other situations, would be t h a t the phasing of vor-
tices propagating downstream would not always be correct,
with respect to a vortex forming at a point downstream, to give a
reinforcement condition, as shown in the above discussion.
We have some experimental evidence suggesting that there
might be optimum convolution geometries, for a given mode of
vibration, which yield relative peak excitation amplitudes. This
might be caused by a vortex reinforcement phenomena.
I t has been shown in the foregoing that two different fluid flow
regimes exist; one occurs when the convolutions are constrained
and, consequently, no organized large scale vortex shedding is
observed, only a steady-state turbulent situation. The other
flow regime occurs when the convolutions are vibrating as a result
of flow excitation; here a large scale vortex formation and shed-
ding process does exist. Therefore the bellows flow excitation
can be described as a mutual or coupled instability between the
bellows structure and the fluid vortex shedding process. One
cannot exist without the other.
The primary condition which must be satisfied for this coupling
to occur is that the bellows structural frequency and the ideal
vortex shedding frequency be equal, or about equal within cer-
tain limits. This limit of vortex-structure coupling can be
called a "lock-in" range, and it can exist over a rather broad
velocity range. Further details of the relationship between the
fluid velocity and the bellows structural frequencies may be
found in references [1, 2],
General Discussion of Bellows Forced Vibration
A calculated coincidence of the coupled bellows-liquid fre-
quencies and the vortex shedding frequency only indicates a
possible flow excitation problem, b u t does not guarantee that
the bellows will vibrate, or indicate the severity of vibration.
A desirable design goal is to be able to accurately predict anti-
cipated stress levels for a given bellows and given flow conditions.
Fig. 4 illustrates a physical model of the fluid and structural
behavior for bellows flow excitation. The process of periodic
t bellows ply thickness
V flow velocity
X convolute pitch
<r, convolute tip width
<rr convolute root width
pf fluid density
Transactions
of the ASME
 Fig. 2 Series of photographs showing vortex shedding from two-dimensional convoluted metal segment
vortex f<lI'Ination and shedding produces periodic pressures on
the convolutions. The amplitude of this alternating pressure is
proportional to the freestream stagnation pressure (1/'PfV').
So far as the bellows structure is concerned, this alternating
pressure produces a net force which may be considered applied
at the tip of each convolution. The amplitude of this force is
assumell (0 be of the form
(1)
where C'r i" a vortex force coefficient (a dimensionless coefficient)
and A I' is the projected convolutes height area over which the
pressurc nets (the fluctuating pressure producing the force).
If the frquency of the vortex shedding coincides with a bellows
longitudinal mode frequency, then a resonance may occur and
the bellows convolutions will experience a vibratory displacement
of the form
(2)
In equation (2), C'm is a dimensionless factor dependent on the
bellows Illode of vibration (mode factor), F is the vortex shedding
fOrce defined in equation (1), K A is the bellows overall spring rate,
and Q is the dynamic amplification factor (damping). This con-
Journal of Engineering for Industry
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
volution vibratory displacement x will cause a corresponding
stress whose magnitude is dependent on the convolution geom-
etry.
Accepting this simple physical model of bellows flow excita-
tion, the problem of calculating resultant stress levels reduces
to one of obtaining a knowledge of the various "factors" illus-
trated in Fig. 4. The primary factors are: C'p, a vortex force
coefficient; C'n" a vibration mode factO!'; Q, a dynamic amplifica-
tion factor; and C'" a geometric stress factor.' Other factors
may also be introduced to account for various unique situations.
For example, when an elbow is located upstream of a bellows,
higher-than-normal stress levels may occur for a given flow ve-
locity. If an acoustic resonance exists, this can change the pic-
ture, also. The objective of the next several sections is to give
results of studies of these various factors.
Vortex Force Coefficient-Idealized Test Model
The initial study of the nature and magnitude of the vortex
shedding fO!'ce was undertaken with a special single convolution
test model discussed in references [1,2]. A bellows convolution
, The stress model indicated in Fig. 4 results from the analytical
form presented by Salzmann [10], and is only representative of a
number of stress prediction methods, e.g., [11, 12, 13].
FEB R U A R Y 19 7 2 / 89
 r^(y0^

yvy\
n

b+c
yT7
~ h (f4>
nr
w yj ^

b+c

b+c 0.10 0.15 0.20 0.25 0.30

Strouhal Number - f o / V
C^Mai Fig. 5 Vortex shedding force coefficients from tests of t w o simulated
convolution geometries

w w V
Fig. 3 Sequence of coupled fluid-convolution events observed with t w o -
dimensional bellows flow visualization model
Vortex Shedding From Convolutions
is simulated in this test model by a flexible ring which is geo-
metrically similar to a convolution. The test ring, or simulated
convolution, is clamped in a special housing between a pair of
exciter coils; also a displacement probe is built into the apparatus
to allow ring vibration amplitudes to be monitored.
With this single convolution test model, experiments were
performed from which vortex force coefficient values could be
deduced. T3'pieal results are shown in Pig. 5. It may be
readity noted that the convolution geometry affects the value of
the vortex force coefficient.

Vibrations of Real Bellows

Vortex Force Bellows Mode Factor. Using the force coefficient information
2
obtained with the single convolution test model, along with an
F-CFCEA ( - W v ) equivalent bellows mechanical model concept, flow-induced
vibration amplitudes were predicted. This was accomplished
by assigning a vortex force at each convolute tip, and then calcu-
lating the forced amplitude of the mechanical model. Com-
paring the resultant vibration amplitudes with observed values
showed the preliminary predictions to be high by a factor equal
Convolution Displacement
to the number of convolutions involved in each "half wave-
length" of bellows vibration. This suggested that there was
... C m FQ
only one effective vortex force per mode number of the vibration-
Further detailed comparison of calculated and observed vibra-
The resultant stress is tions of bellows confirmed this idea. Also, reexamination of the
CsEtx
Stress visualized vortex shedding activity (illustrated typically in Fig.
ha
In the above equations 2) and reconsideration of the pressure forces involved, showed
Cp = vortex force coefficient how this must happen.
Cf = elbow factor Fig. 6 depicts the action of a bellows with several convolutions
Cm » vibration mode factor
vibrating in the first longitudinal mode. As shown, the vortex
Q- dynamic amplification (damping )
Cs = geometric stress factor
shedding causes fluctuating pressure forces on each side of each
convolution. Because of the phasing of these pressure forces,
however, only one effective vortex force exists per mode number
Fig. 4 Illustration of stress resulting from vortex force

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Transactions of the ASME
90 / F E B R U A R Y 1972
 Vortex Activity On Media - Media -
Bellows Vibrating In Water Air
1st Mode
150 150 _

100 - - 100
Q
50
Resultant Pressure
Forces On Convolutions
Pitch Pitch

150

100
- / " " " 100 -
Net Effective Force Q
On Convolutions 50 50
"'
0
0 10 20 30 0 10 20 30
Media Pressure Media Pressure

300 - 300 -

200 ~"~~\^^ 200

— Q
Equivalent Mechanical Model 100 100

Fig. 6 Effective vortex force on bellows vibrating in the first longitudinal

0
mode o ion 200 300 C 100 200 300
Strain Level - ^ . i n . / i n . St rain Level - / i i n . / i n .

Fig. 7 Experimentally determined trend of Q with various parameters

of vibration. The first mode has one effective force point, the
second mode, two, etc. Each of these effective forces is exerted
at the central convolution of each mode half wavelength, or at
the point(s) of maximum displacement of each mode.
Using the one-force-per-half-wave length concept illustrated
in Fig. 6, the analytical derivation of a value for the mode factor
C„„ which relates displacement and force during a resonant condi-
tion, was undertaken. As shown in reference [21, the resultant
bellows mode factor is of the form
N_
cm = -1 + sin
2 Nj\
HN
and refers to the maximum relative displacement point (maxi-
mum stress point) along the bellows. Thus, from equations
(2) and (3), it is evident t h a i the bellows convolute maximum
relative vibratory displacement is
_FQ_ N 7T_ ^ V
%mr — ~T7~ + Sill
8NKA { N, 2 Arr
In equation (4), F is the amplitude of the vortex shedding force
applied as illustrated in Fig. 6, N is the vibration mode number
(1, 2), and Arc is the total number of convolutes.
At this point, estimates of bellows flow-induced vibration
amplitudes could be made, based on equations (1) and (4) plus
CF data typified by that of Fig. 5, if appropriate values for Q
were also known.
Bellows Damping. Because the value of damping is so im-
portant in the prediction of bellows flow-induced vibration
amplitudes, a large number of mechanical vibration tests were
conducted from which Q values could be obtained for typical
conditions. Each damping test bellows was mounted in a
special fixture which rigidly connected the end flanges to ensure
that the vibration modes excited would be the same as those
observed in the flow tests. Excitation was pi'ovided by an elec-
tromechanical shaker, and Q values were obtained from frequency
sweeps with either input acceleration amplitude, or in some cases,
convolute strain response held constant. T h e standard half-
Power point band width method was employed for calculating
the values of Q.
Fig. 7 illustrates the trends of the experimental damping val-
ues, as a function of various parameters. Note that the bellows
damping is generally most sensitive to peak strain amplitudes,
with the damping increasing as strain increases. Further note
t h a t single-ply bellows of different configurations, tested under
similar conditions (same strain level, same interna] media, etc.)
do not necessarily have the same Q values. For one configura-
tion of bellows, increasing the number of plies from one to two
(3) significantly increases the damping (reduces Q); however, going
to three plies adds little more damping, if any, over the two-ply
configuration. For bellows with a liquid as the contained
media, damping may greatly increase at low absolute pressures.
The fact t h a t damping increases as the strain level increases
means t h a t the damping is a nonlinear function of amplitude.
T h e fact t h a t the bellows damping is nonlinear is really no sur-
prise, since structural damping is often of this nature (see La/an's
(4) discussion of hysteretic damping in reference [14]).
Based on the test damping data, a preliminary damping
model was formulated, and is presented in Fig. 8 and Table 1.
To reflect the change of Q with strain level, the factor
Ar„
called the "Bellows Operational Parameter," has been employed
as the independent variable. To account for different internal
media, the effect of the number of plies, and of the bellows stiff-
ness, the data in Fig. 8 are shown as a number of application
curves of Q versus t h e bellows operational parameter. Table 1
defines the use of each application curve.
Bellows Flow Tests. The validity of the model described in the
foregoing was determined from correlation with a number of
flow-induced vibration tests of bellows. The tests and data
reductions were carried out in such a way that vortex force
coefficient (C'F) values were obtained for comparison with CF
values obtained with the single convolute test model.
For test purposes, each bellows was mounted in a water flow
loop, and vibratory strain levels were monitored as a function

Journal
Downloaded From: of Engineering for Industry
https://manufacturingscience.asmedigitalcollection.asme.org on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
FEBRUARY 1 9 7 2 / 91
 1 j 320

- Curve Numbers Correspond To Certain

| |
For These Applications.
Water- 75° F
280
D Eel lows #102
1 =
O # 105
2 - O #112
3 -
i 240 First Mode
- ^ CD
4 - Q_
^
=p ^>~= - •

C
-?
""- ,.^,
5 -
• - - .

, - • — . . . . .

=___; _T2^z~. S 200

"—:-- *^xV_ cz
6 - i . ! ~~'— ! ""
!
!
•

j^g ji.

E 160
on
ca
E
'-o
€ 120
i

Bellows Operational Parameter - ( CpCppH/Np ) ( h / t I 'g

ro
Fig. 8 Dynamic amplification factors for various betlows applications ^

Table 1 Applications information for use with Q values data in Fig. 8

40

Specific Spring Number Internal Media Curve

Rate (see Note 1) Plies (see Note 2) No.

all ranges low p r e s s u r e gases 1

oLy-
over 2000 l b / i n 2
0 ' 16 20 22 24 26
high p r e s s u r e g a s e s , light liquids 1
over 2000 water, dense liquids 2 Fluid Velocity (ft/sec !
under 2 0 00 high p r e s s u r e g a s e s , light liquids 2
under 2000 water, dense liquids 3 Fig. 9 Peak flow-induced strain for three similar test bellows—see
Table 2 for bellows geometry
over 3 000 2 all 3
2000-3000 2 all pressure gases
under 2 000 2 all pressure gases 5
2000-3000 2 all liquids 5
under 2COO 2 all liquids 6 Table 2 Pertinent physical data for test bellows

Nominal Dimensional Data for Bellows HOZ, 105, 112

over 3000 all 4
2000-3000 3 ail 5 Internal d i a m e t e r , D; = 1.49 inches
under 2000 all p r e s s u r e g a s e s 5
under 2000 all liquids 6 External d i a m e t e r , D(J - 2.22 inches

Live length - 1..3G inches-

Use of Tabic To use table, first c a l c u l a t e beliow^- specific spring rate
then look up application curve number cori-e Gponding to Number convolutes, Nc = 7
this specific spring r a t e , number uf p l i e s , and internal
media. Convolute height, h - 0.35 inches

The specific spring rate is here defined r.s Pitch, , \ = 0. 26 inches

KANC
Convolute root and tip widths, o" ., u. = 0. 125 tuche
DmNp
spring rate per convolute, per ply, per unit of Ply thickness, t ^ 0.013 inches

Low p r e s s u r e gases will be defined here as being those Spring Rate Data
g a s e s below 150 psia. Light liquids will be defined as
having a density, relative to w a t e r , of l e s s than 0. 2.
Bi-llows Overall Springrate

102 454 lb/in

,05 545
112 550
of internal fluid velocity. Fig. 9 shows sample data from the
flow tests. In this particular figure, results are presented for
three nearly identical test bellows (by virtue of having the same
design convolute geometry); Table 2 gives pertinent data for from the strain data of Fig. 9. Note how the reduction process
these bellows. Variations in the vibration characteristics are has produced much better agreement of the test results for these
readily evident, and were attributed to measurable but very three bellows. Apparently the major reason for the variation of
small differences in the actual convolute geometry, and differ- vibration characteristics of the three bellows was a dissimilarity
ences in the spring rate. of spring rate. This was likely a result of different amounts
All flow-induced vibration test data were reduced, using the of metal thinning which occurs in the forming process. Differ-
analysis described previously, so that experimental vortex force ences in spring rate are inherently reduced out by the data re-
coefficient (CF) values were obtained as a function of the test duction, hence, this explains the relatively good agreement
Strouhal number. More precisely, the CF values thus obtained shown in Fig. 10.
were effective force coefficients, as described in reference [1]. Several tests were performed in which the effect of bellows
Details of the data reduction procedure may be found in refer- convolution geometry on the vortex force was determined.
ence [2]. Fig. 10 illustrates typical CF data, in this case reduced This was accomplished by step-wise stretching and compressing

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org

92 / F E B R U A R Y 197 2 on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Transactions of the ASME
 SwRI Test Bellows * 105
Water @75"F

First Longitudinal Mode

0.02

0.24 0.26 0.28

Convolution Pitch - inches

Fig. 11 Response of test bellows as a function of pitch-bellows stretched

to change pitch

0.01
0.25
Strouhal Number -
v stress, and which appears valid for judging the relative severity
Fig. 10 Force coefficient data reduced from strain data of Fig. 9
the test bellows so as to alter the convolution pitch. The ad-
vantages of this method are at least twofold; first, it allows
geometry to be varied rather independent of other property vari-
ations, and second, it results in a hardware savings since more
data can be obtained from a single test item. Fig. 11 gives flow-
induced strain values obtained in this manner. The results
represent the observed peak flow-induced strain for each pitch
setting. Fig. 11 also shows these same data reduced to the vortex
force coefficient form. Notice that the force coefficient generally
decreases as the pitch was increased; this trend is in agreement
with the single convolution test model results and other bellows
tests. Also, Fig. 11 shows an obvious trend of " o p t i m u m "
pitch values which produce minimum local- Op value. This
trend has been observed in several bellows tests and can prob-
ably be explained by vortex reinforcement as noted in Fig. 3.
Summary of Force Coefficient Data. In the foregoing sections,
vortex shedding force coefficient d a t a has been discussed as de-
rived from essentially three different kinds of tests. First, the
results of the idealized single convolution model tests were given.
These tests yielded force coefficient data for two different con-
volution configurations; the two configurations had the same
convolute height h and the same tip width, b u t different pitches.
Second, complete force coefficient d a t a were obtained from tests
of a number of "calibrated" bellows having various convolution
geometries. Finally, a limited amount of test data was obtained
by a bellows stretch-compression pitch variation scheme. Com-
parison of all of this force coefficient data shows consistent re-
sults which gives credence to the bellows flow-induced vibration
model developed in the previous sections.
Tests are currently being performed with a family of bellows
having a wide range of convolute geometries. Results of these
tests should give a complete picture of the variation of CF with
convolute geometry. For the present, however, CF d a t a given
m this paper and in reference [2] are complete enough to provide
good initial estimates of bellows flow-induced vibration levels.
Bellows Stress Indicator and Fatigue Life. A l l p r e v i o u s discussion
has pertained primarily to the prediction of bellows flow-induced
vibration amplitudes. Of more importance from a design view-
point is a proper estimation of resultant stresses, and probable
fatigue life. Using one of a number of existing analysis tech-
niques [10-13], it is possible to relate the calculated convolute
vibration amplitudes (maximum relative displacement values
given by equation (4)) to convolute stress. In general, these
techniques do not allow rapid stress calculations. Thus there
has been a need for a simple and approximate method of esti-
mating bellows stress. Such a method has been derived and is
called the "Stress Indicator" approach. This method involves
calculating a number which is approximately proportional to
of bellows flow-induced vibrations. The derivation of the Stress
Indicator follows.
The basic equations which describe peak flow-induced stress
for a bellows convolute are equations (1) and (2), plus some ap-
propriate relationship between displacement and stress. From
the results of Salzmann [10] it appears valid to write.
CsEtx
Stress (5)
fe2
where C, is a parameter dependent on convolute geometry.
For typical bellows of roughly similar geometry, Cs is approxi-
mately constant. Therefore, we will assume the following in
seeking our approximate stress model:
(a) The projected convolute area A p may be approximated by
Ap » Dmh (6)
(6) T h e mode factor Cm may be approximated by
^ m (7)
4iV„
(c) The bellows overall spring rate KA may be approximated
by (see [10])
KA » DmE(Np/Nc)(i/hy« (8)
(d) The factor C s is approximately constant for bellows of
interest. Based on these assumptions, equations (1), (2), and
(5) may be combined to give:
Stress = 7rC. 9l9 (7-2P/F 2 ) (9)
Since C s has been assumed constant, we can extract a quantity
from equation (9) which is approximately proportional to flow-
induced stress for all bellows; this quantity is
Stress Indicator =
(10)

Journal
Downloaded From: of Engineering for Industry
https://manufacturingscience.asmedigitalcollection.asme.org FEBRUARY
on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use 1 9 7 2 / 93
 100.0
\ "" s I 1 1
\\ _\ 1 1 1
Region Of Very H
\\ 1 \ advantage in designing the convolute geometry to give a mini-
" ;S|
V mum Cp. value. I n general, "open" convolution designs are
,£K /
V'v
<s v— /-Regio l Of Random Failures
I Safe Operati ig
| Region
10.0
: A very important factor in bellows vibrations is the value of
: the dynamic amplification factor Q. Test results show t h a t Q
Notes :
- Obtain Cf from Fi ur e 36 of Reference 1
- Obtain Q f r o m Fig i) re 8
- Use KE • 2.0 f Ul S eve re upstream elbow
- Use E c = 1.0 f or r 0 l pstream elbow
- T h i s data s u ncon ser vative where acoust c reso nanc e oc cu
1.0 L J M l _
100,000 1,000,000
Number'Of Cycles To Failure
Fig. 12 Preliminary bellows fatigue life data
To perform calculations with equation (10), it is first necessaiy
to find values for CF and Q from results presented earlier, or from
reference [2].
To supplement the Stress Indicator, a preliminary plot (Fig.
12) for predicting bellows fatigue life has been prepared. These
data represent a collection of all failure cases which have been
observed in our laboratory, plus failure information reported by
others. These fatigue data must be considered preliminary be-
cause of the small number of sample points used in their com-
pilation, and should be good enough, at the very least, to help
judge relative severity of various bellows installations.
Effect of Radial Acoustic Resonance
All previous discussion has related primarily to t h e flow-
excitation of bellows with internal liquid flow. Tests conducted
with internal gas flows reveal that a similar vortex shedding exci-
tation can occur, but there is an added effect which is likely.
I t has been determined t h a t when the vortex shedding fre-
quencies become high enough, a radial acoustic resonance can be
excited within the bellows. The effect of this resonance is to
greatly amplify the convolute vibratory stresses. At the present
time this amplification has not been well quantified, but investi-
gations are under way to do so. The interested reader can ob-
tain further information from reference [2], or from communica-
tion with the writer.
Conclusions
Flow-induced vibration of bellows is caused b y a fluid-elastic
instability; specifically it is caused by fluid vortex shedding from
the bellows convolutions which couples with t h e convolution
vibrations. I n t h e absence of convolute motion there is no
well defined large scale vortex shedding. T h e fluctuating fluid
pressure force which is exerted on a single bellows convolute can
be expressed in the form
F = CPAfWip,V*)
where GF is a vortex force coefficient, Av is the height area of the
convolute, p{ is fluid density, and V velocity. T h e data ob-
tained for CP from a number of bellows tests show it to be a strong
function of the parameter X/crr, the ratio of convolute pitch to
tip width.
I t appears that, because of a vortex reinforcement phenomena,
94 / F E B R U A R Y 1972
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org
on 06/28/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
t there are "optimum" values of X/crr which result in local mini-
1 mum of maximum values of CF. Therefore there might be some
better than "closed" designs.
The response of a given bellows to flow excitation can be pre-
dicted with fair accuracy with a method described in this paper.
Strictly speaking, and because no supporting data are presently
available, this method is presently limited to bellows flowing
liquids and having a size of about 2-in. I.D. We believe, how-
ever, t h e method can safely be extended to include gases and
larger bellows. One important restriction is t h a t no radial
acoustic resonance be present to cause amplification of the forces.
is dependent on the bellows specific spring rate (KANC/Dm), the
internal fluid media, and the number of plies. A preliminary set
of universal Q value curves has been prepared in terms of known
bellows and flow parameters, and can be used for present best-
rs
estimate purposes.
1 10,000,000
A special parameter, called t h e "Stress-Indicator" has been
derived and is felt to be of value in determining the relative
vibration severity of various bellows. A preliminary curve
giving fatigue failure information as a function of t h e Stress
Indicator has been compiled and can be used to estimate the
possibility of failure of a given bellows.
Acknowledgment
The results presented in this paper were obtained during the
course of research carried out under the sponsorship of NASA
Marshall Space Flight Center, Contract NAS8-21133. Mr. R.
H. Veitch of the Propulsion and Vehicle Engineering Laboratory
was the NASA technical administrator for the program.
References
1 Gerlach, C. It., "Flow-Induced Vibrations of Metal Bellows,"
JOURNAL OF ENGINEERING FOR INDUSTRY, TRANS. ASME, Series B,
Vol. 91, No. 4, Nov. 1969, pp. 1196-1202.
2 Gerlach, C. R., and Schroeder, E. C , "Study of Minimum
Pressure Loss in High Velocity Duct Systems," Interim Technical
Report No. 1, Contract NAS8-21133, Southwest Research Institute,
July 1969.
3 Toebes, G. H., and Eagleson, P. S., "Hydroelastic Vibrations
of Flat Plates Related to Trailing Edge Geometi'y," Journal of Basic
Engineering, TRANS. ASME, Series D, Vol. 83, No. 4, Dec. 1961, pp.
671-678.
4 Man-is, A. W., "A Review of Vortex Streets, Periodic Wakes,
and Induced Vibration Phenomena," Journal of Basic Engineering,
TRANS. ASME, Series D, Vol. 86, No. 3, Sept. 1964, pp. 165-169.
5 Toebes, G. H., "Plow Induced Structural Vibrations," Journal
of Engineering Mechanics, Proceedings, ASCE, Vol. 91-EM6, 1965.
6 Toebes, G. H., and Ramamurthy, A. S., "Fluidelastie Forces
on Circular Cylinders," Journal of Engineering Mechanics, Proceed-
ings, ASCE, Vol. 92-EM, 1967.
7 Protos, A., Goldschmidt, V. W., and Toebes, G. H., "Hydro-
elastic Forces on Bluff Cylinders," Journal of Basic Engineering,
TBANS. ASME, Series D, Vol. 90, No. 3, Sept. 1968, pp. 378-386.
8 Haugen, R. L., and Dhanak, A. M., "Momentum Transfer in
Turbulent Separated Flow Past a Rectangular Cavity," Journal of
Applied Mechanics, Vol. 33, TRANS. ASME, Vol 88, Series E, No. 3,
Sept. 1966, pp. 641-646.
9 Townes, H. W., and Sabersky, R. H., "Experiments on the
Flow Over a Rough Surface," International Journal Heat Mass Trans-
fer, Vol. 9, 1966, pp. 729-738.
10 Salzmann, F., "fjeber die Nachgiebigkeit von Wellrohr-
expansionen," Schweiz, Bauztg., Vol. 127, No. 11, Mar. 1946, pp.
127-130.
11 Clark, R. A., "On the Theory of Thin Elastic Toroidal Shells,"
Journal of Mathematics and Physics, Vol. 29, 1950, pp. 146-178.
12 Laupa, A., and Weil, N. A., "Analysis of U-Shaped Expansion
Joints," Journal of Applied Mechanics, Vol. 29, TRANS. ASME, Vol.
84, Series E, No. 1, Mar. 1962, pp. 115-123.
13 Turner, C. E., and Ford, H., "Stress and Deflection Studies of
Pipeline Expansion Bellows," Institution of Mechanical Engineers,
Vol. 171, Nov. 1957, pp. 526-552.
14 Lazan, B. J., Damping of Materials and Members in Structural
Mechanics, Pergamon Press, Oxford, England, 1968.
Transactions
of the ASME