Vibrations of Vertical Pressure Vessels

C. E. FREESE
Mechanical Engineer, The
Fluor Corporation, Ltd., Los
Angeles, Calif. Mem. ASME

This paper is primarily concerned with the vibration of vertical pressure vessels
known as columns or towers.
The procedure for estimating the period of first mode of vibration for columns which
are the same diameter and thickness for their entire length is outlined. A graph is
included for this purpose which recommends limits between vessels considered to be
static structures and those considered dynamic.
A method for designing vessels considered as dynamic structures is described as well
as a detailed procedure for estimating the period of vibration of multithickness (stepped
shell) vessels and/or vessels built to two or more diameters with conical transitions where
the difference in diameter is small.
There is a brief resume of the Karman vortexes effect and a discussion regarding
vibration damping by liquid loading and the benefit of ladders and platforms which help
reduce the effect of periodic eddy shedding.
The design procedure outlined will be useful to the practical vessel designer
confronted with the task of investigating vibration possibilities in vertical pressure
vessels.

Introduction
For many years it was customary to apply guy wires
to tall ,slender pressure vessels. In, recent years, refinery and
petro-chemical officials have demanded self-supporting vessels
from the standpoint of plant appearance and safety.
In order to design a self-supporting vessel of this type, the
following problems must be carefully analyzed:
1 When it is necessary to deviate from the common practice
of designing a vertical vessel as a static structure and consider it as
a dynamic structure?
Contributed by the Petroleum Division and presented at the ASME
Petroleum-Mechanical Engineering Conference, Denver, Colo., September
21-24, 1958, of THE AMERICAN SOCIETY OF MECHANICAL
ENGINEERS.
NOTE: Statements and opinions advanced in papers are to be
understood as individual expressions of their authors and not those of the
Society. Manuscript received at ASME Headquarters, July 9, 1958. Paper
No. 58 - PET-13.

2 What is the most practical method for designing to meet

dynamic conditions?
3 Does the method used produce consistent results and does it
provide additional strength to resist the force due to the massacceleration resulting from the motion of the vessel ?
4 Is the period of vibration of the dynamically designed vessel
such that prevailing winds are not apt to cause excessive
movement?
5 Are the external attachments ( such as piping, ladders, and
platforms) distributed all around the vessel to guard against
resonance due to eddy shedding in the Karman vortex trail at
critical wind velocities?
These problems will be discussed during the outline of a design
procedure presented in this paper.
Before proceeding, it should be pointed out that vessel
vibrations induced by earthquakes are infrequent in occurrence
and this paper is more concerned with vibrations induced by wind
or other forces which may occur every day or many times during
the day may depending upon the location.

Nomenclature
f = lowest natural frequency of vibration,
cycles per second
T = 1 = period of vibration, sec
f
g = acceleration due to gravity
W = total weight of vessel or vessel section
above horizontal plane under construction, lb
WS = shear load at end of section, lb
w = unit weight; lb/ft
w' = weight of vessel element or internal
part, lb
L = total length, ft
l = length of element or section, ft
D = vessel diameter, ft
d = vessel diameter, in.

Journal of Engineering for Industry

ustrg

t = thickness of vessel shell, in

h = t = thickness of vessel shell, ft
12
y = deflection of element or section, ft
y = distance from e. g. of vessel element or
internal part (of weight w) to seam or
horizontal plane under consideration, ft
F = seismic factor
E' = 4320 X 106 = modulus of elasticity for
steel, Ib/ft2
E = welded joint efficiency
3
D h
I = 8
moment of inertia of vessel shell
cross-sectional area, ft4
V = velocity, ft/sec
3

k = Strouhal number
= end slope of element in bending as a
cantilever beam, radian4 (tan = )
P = internal pressure, psig
S = allowable stress of vessel material, psi
M = moment about vessel seam or
horizontal plane under consideration, lb - ft
MT = moment at end of vessel section
resulting from weight of sections to the right
section under consideration
C = corrosion allowance
R = Reynolds number

FEBRUARY 1959 / 77

Design

Procedure

It is customary for most vessel designers to establish the

minimum vessel shell and trend thickness according to the
pressure temperature conditions and then calculate the thickness
required at the bottom head seam due to bending moments
imposed by wind or earthquake forces [9].' Stresses in the
longitudinal direction are involved nod the following notation may
be used to summarize the thickness required:
t

Pd
48 M
W

C
4 SE 0.8 P d 2 SE dSE

(1)

The terms within the absolute value signs are positive for tensile
stresses and negative for compressive stresses. The first term gives
the thickness required for the longitudinal stress resulting from
internal pressure and is positive for pressures above atmospheric
and negative for pressures below atmospheric. The second term is
the thickness required to resist the longitudinal bending stress and
both positive and negative values exist at the same time. The third
term is the thickness required for the weight of the vessel above the
seam being investigated and, since this is a compressive stress, it
has a negative value. The combination giving the highest value
establishes the thickness required to resist the longitudinal stresses.
Consider equation (1) for a typical vessel operating at an
internally pressure greater than atmospheric:

shell is of constant diameter and thickness for its full length, the
period of vibration maybe easily found from the graph shown in
Fig. 1. This graph is plotted from the general formula for the
period of the first mode of vibration of a cantilever beam [7]:
T

2 wL4

352
. E ' Ig

wL4

1785
.
E ' Ig

12

12

(3)

For a steel cylindrical shell, equation (3) may be written:

T 764
. x106 wL3

12

D h

(4)

By rewriting equation (4) in the form:

T 7.64 x106 L
D

wD

12

(5)

the variables (L/D) and (wD/h) are used as parameters to plot the
graph in Fig. 1.
One of the first graphs of equation (4) was issued by a major oil
company for their refinery work. In its original form, all vessels
having a period of vibration over 0.4 sec were ordered designed
as dynamic structures and those having a vibration period of 0.4
sec or less were ordered designed as static structures. Experience
t 0.275 0.307 0.063 0.125
has shown that a more practical limit for this division is a line
drawn from 0.4 sec at the extreme left of the graph to 0.8 sec at the
The required thickness within the absolute value signs will have extreme right and considering vessels having a period of vibration
two values; namely, +0.519 in. and -0.095 in. Therefore the above this line to require dynamic consideration and those below
minimum thickness required is 0.519 + 0.125 in. corrosion to require designing as a static structure. The reason for revising
allowance = 0.644 in.
the former limit is the fact that many vessels having small (L/D)
Next consider equation (1) to appear as follows for the same
ratios and large values of (wD/h) have given satisfactory service
vessel operating under vacuum conditions:
although their period of vibration exceeded 0.4 sec. In general,
vessels having an (L/D) ratio less than 15 are not apt to be critical
t 0123
.
0.307 0.063 0.125
from a vibration, standpoint. One exception to this statement,
For this case, the two values within the absolute value signs are unofficially reported to the author, involved two vessels operating
-0.493 and + 0.121 in. resulting in a minimum thickness of 0.493 near a railroad whereby they were vibrated by railroad equipment.
Both vessels had a period of vibration considerably less than 0.4
+ 0.125 in. = 0.618 in.
As previously stated, the moment M is the longitudinal bending sec and their frequency probably coincided with the frequency of
moment due to wind or earthquake, either of which may be the exciting force, thereby causing resonance. This type of
combined with eccentric loads imposed by mounting heavy response is difficult, if not impossible, to predict accurately and
equipment on the vessel. All designers are accustomed to should be considered as a special case.
If investigation indicates that the vessel should be designed as a
evaluating moments due to eccentric and wind loads, but there are
a few who may not be familiar with the method used for estimating dynamic structure, the method of seismic analogy is
moments due to earthquake. Therefore, the following brief outline recommended. This method consists of designing the vessel for
is presented because this method is recommended as a design earthquake conditions using a seismic factor F3 = 0.20, regardless
of the geographical location. In most cases, the vessel will have
procedure for vessels where dynamic considerations are required.
thicker shell and head material in the lower section. As an
The weight of each vessel element (shell, head, tray, or internal
example, consider a vessel 10 ft 0 in. diameter by 13/16 in. thick by
part) is calculated. and then multiplied by the vertical distance
190 ft 0 in. high which has an (L/D) ratio of 19, and period of
from the circumferential seam (or horizontal plane) under
vibration (after being designed as a static structure) of 1.65 sec.
consideration to the center of gravity of the element. The
This vessel, when designed as a dynamic structure by the method
summation of the moments so found is multiplied by the seismic
l3
factor for the area where the vessel is to operate, thereby yielding a of seismic analogy, resulted in a shell thickness of /16 in. for the
7
15
moment due to earthquake or seismic disturbance. For vessels, the upper 137 ft 0 in. and three 1ower sections consisting of /8, /16,
and
1-in.
thick
material
(the
supporting
skirt
increased
from
1
to l
seismic factor will usually have a value of 0.03 to 0.12, depending
9
/
in.).
The
period
of
vibration
was
reduced
to
approximately
1.4
16
upon the geographical location. Expressed mathematically,
sec.
Whereas the application of this method actually consists of trial
M F3 w' y '
(2) and error, the experienced pressure vessel designer becomes very
proficient in estimating how far down the vessel he can utilize the
After the vessel has been designed in the regular manner material thickness which is based on pressure-temperature
(considered as a static structure) it should be investigated regarding requirements, as well as the length of successive sections of
thicker material. It is usually unnecessary to carry the seismic
its possible behavior under vibration conditions. If the vessel
analogy into the design of the anchor bolts because this method is
1
Numbers in brackets designate References at end of paper.
78 / FEBRUARY 1959

Transactions of the ASME

AS M E

Fig. 1

applied only as a "yardstick" to provide reasonable protection with a

minimum amount of additional material. However, anchor bolt
stresses should be held low (15,000-16,000 psi) for these vessels or, if
a higher stress is used, the design procedure outlined should be
applied to them. Proper tightening of anchor bolts for vessels subject
to dynamic behavior is of utmost importance and it is recommended
that they should be pretightened to the predicted working stress to
avoid stretching and loosening in service It is definitely unnecessary
to apply this method to the design of the foundation unless the vessel
is operating in a seismic area.
The design procedure just outlined produces consistent results and
also provides additional material to resist the force due to
mass-acceleration of the vessel in motion. A number of years ago,
approximate calculations indicated that the total force due to wind
load plus the force due to mass acceleration was about 1.5 to 1.70
times the static force due to 30 lb/ft, wind load for several different
size vessels. It was found that the recommended design procedure
resulted in shell thicknesses within a few thousandths of an inch of
those obtained by the more lengthy approximation. Many critical
vessels have been successfully installed which were designed to the
seismic analogy method just described.
The same company that produced the first graph of equation (4)
tentatively recommended the seismic design method using a 0.20
seismic factor for their vessels requiring dynamic design in order to be
on the "safe side." Since this company was mainly interested in the
response of vessels and other structures to earthquake induced

Journal of Engineering for Industry

vibration, they later revised their recommendations as follows:

Natural period of vibration

Less than 0.40 sec
0.40 sec to 1.0 sec
Greater than 1 sec

Earthquake coefficient
0.20
0.08 divided by period
0.08

Attention is again called to the fact that this paper is primarily

concerned with vibrations induced by wind or other forces which
occur more frequently than earthquakes and it should be noted that the
vessel reported as Case II under Field Data is well within the later
recommendations outlined here and vibration trouble was encountered.
It is agreed that the current practice is probably adequate for
earthquake design; however, all critical vessels (except the vessel
reported as Case II) designed and installed by our company have been
designed to the seismic analogy method using a 0.20 seismic factor.
Not all vessels designed as static structures have the same thickness
of shell for their entire length and some vessels are of more than one
diameter. These vessels, as well as many designed as dynamic
structures cannot have their period of vibration estimated from the
graph in Fig. 1 or equation (4). It is also desirable to know the change
in the vibration period resulting from dynamic design. Of the several
methods referred to in reference books on vibration [1, 2, 3] the
numerical integration of the equation

FEBRUARY I 959 / 79

 Wy2
T 2
g Wy

(6)

is probably the easiest and safest method for the designer who is not a
specialist in vibration to apply. This equation follows the Rayleigh
method of approximation for finding the fundamental period of
vibration as applied to a shaft or loaded beam on too supports. It will
be shown that this equation is reasonably accurate for estimating the
period of the first mode of vibration of vertical pressure vessels.
Equation (6) will result in an estimated period of vibration slightly
lower than the actual period. The degree of accuracy is dependent
upon the number of sections calculated in estimating the static
deflections when the vessel is considered as a cantilever beam
deflecting under its own weight. As an example, the period of
vibration of a cylindrical shell 3 ft 0 in. diam by 3/4 in. thick by 90 ft
0 in. high was estimated under two separate conditions. In order to
eliminate nonuniformly distributed masses, this shell was considered
to have tray sections at one-foot intervals from the top to the ground
and the heads were omitted. When calculated to equation (4), the
period of vibration was found to be 1.088 sec. Dividing the shell into
nine sections, each 10 ft 0 in. long and calculating the period to
equation (6) resulted in an estimated period of 1.08 sec. which is
0.735 per cent low. On the other hand, when this same shell is divided
into five sections having lengths of 30 ft 0 in., 20 ft 0 in., 15 ft 0 in.,
15 ft 0 in., and 10 ft 0 in., the estimated period of vibration to equation
(6) was 1.068 sec which is 1.84 per cent lower than the results from
equation (4). Most vessels designed as dynamic structures have five to
ten sections similar to the latter division and the weight is not always
uniformly distributed. Field test have shown the calculated period of
vibration to be 1.5 to 4.5 per cent lower than the observed periods for
several different size vessels. This is in good agreement for large
structures and it is reasonable to assume that the period of vibration
obtained by the numerical integration of equation (6) will be
approximately 5 per cent lower than the actual period .
Equation (6) is not difficult to integrate numerically, but care must
be exercised to make certain that all factors affecting deflection are
included. Instead of following a complete numerical integration, some
designers prefer to estimate the deflections at the center of each
section graphically by either the area-moment or conjugate beam
method. The same results will be obtained. The choice of method
depends upon the personal preference of the individual. An outline for
the numerical integration of equation (6) when applied to vertical
pressure vessels is given in the Appendix of this paper.

Discussion of Wind Effects

Tall, cylindrical structures such as pressure vessels and stacks are
subject to being put in oscillatory motion by wind currents. The
motion is at right angles or normal to the direction of the wind. This
phenomenon is usually referred to as resulting from the Karman
vortex trail (4, 5, 6, 10). The relationship between wind velocity and
frequency of eddy shedding is given by the equation.

fD D

k
Tk

(7)

Since we are primarily concerned with the resonant condition

which occurs when the frequency of eddy shedding equals or is in the
neighborhood of the natural frequency of vibration for the vessel the
symbols for the vessel frequency and period of vibration are shown in
equation (7). From this equation, we can estimate the critical wind
velocity for most vessels.
The value for k was first determined in 1878 by V. Strouhal as

0.185 and is known as the Strouhal number [10]. It is assumed by

some authorities to be within 0.18 to 0.27 and dependent upon the
velocity of flow [4]. The reproduced graph shown in Fig. 2 gives the
variation of the Strouhal number with the Reynolds number as
obtained experimentally by Relf and Simmons [10]. Research
engineers, employed by the same company as the author, reported the
following values for k obtained from full size vessels after erection
[12]:
k = 0.133, for a 7.67 O.D. insulated vessel at a wind velocity of
39.6 ft/sec (27 mph)
= 0.189, for a 3.0 O.D. vessel at a wind velocity of 32.25 ft/sec
(22 mph)

The difference between the values reported from field data and the
graph is probably due to the size of cylinders tested and the method of
support. When the velocity of the wind is such that the frequency f in
the equation corresponds to the natural frequency of the vessel,
resonance occurs and the vessel will oscillate at an excessive
amplitude. Since aerodynamic stability theory and calculation methods
are beyond the scope of this paper, the reader should refer to
Steinman's paper [10] and similar publications for additional
information in this subject.
One vessel, not designed to the seismic analogy method outlined
herein, gave trouble due to wind induced vibration. This vessel is
identified as Case II under Field Data. It was found to be free from
vibration when the wind was blowing from a direction such that
nearby equipment disturbed the flow pattern and it is conceded by
some individuals that the external attachments also helped to reduce or
nullify the effect of periodic eddy shedding. It is recommended that
any vessel, where possible vibration trouble is indicated, should have
the external appurtenances located around its circumference and not
placed on only one or two sides as was done with this vessel. The
break in vertical ladder runs demanded by some states helps to
accomplish this because intermediate platforms and ladders are
distributed circumferentially. The same vessel which gave trouble
when empty has been satisfactory after liquid loading. Therefore the
additional damping effect of liquid loading cannot be ignored - on the
other hand, neither can too much confidence be put in it as a cure-all.
Some engineers are also concerned regarding the possibility of the
vessel being vibrated at a frequency corresponding to its second mode
of vibration. The second mode of vibration for cantilever beam has a
frequency of 6.37 times the frequency of the first mode [7]. This
relationship will not necessarily hold true for multithickness and/or
multidiameter vessels and more involved methods of analysis for the
second mode frequency have to be employed. The Ritz method [8]
which is a further development of Rayleigh's method can be used for
these cases. It is sometimes referred to as the Rayleigh-Ritz method
and should be applied by designers specializing in vibration problems.
For the average vessel, it is not unreasonable to assume that the second
mode frequency might occur between five to six times the frequency
of the first mode. If a wind velocity of thirty miles an hour has been
estimated to induce vibration in the first mode, it is reasonable to
conclude that vibrations in the second mode will not be induced by any
wind less than one hundred fifty miles per hour. On the other hand, a
vessel subject to vibration in the first mode by winds of only ten miles
an hour might be vibrated in the second mode by winds of fifty to sixty
miles per hour if external attachments do not interfere with the
periodic eddy shedding. Surrounding structures and terrain will also
have some bearing on the considerations involved.
It is not the intention of this paper to overamplify the possibility of
the second mode of vibration. Some engineers maintain that
vibrations in the second mode could be catastrophic; how

80 / FEBRUARY 1959

Transactions of the ASME

ever, to the knowledge of this author no case of this type of failure has
ever been recorded. In fact, no one has reported a second mode
vibration in a self-supporting vertical pressure vessel and vessels are
in service which have (L/D) ratios in the neighborhood of 40:1.
Steinman's paper refers to the Meier-Windhorst tests at the Hydraulic
Institute at Munich (1939) wherein the hydrodynamic oscillations of
cylinders yielded sharply defined results for (a) the low velocity range,
(b) critical range, and (c) high velocity range, and further states that,
"In these vortex induced oscillations, there is no 'catastrophic range' of
increasing amplification with unlimited increase of steam velocity"
[13].

Discussion of Correction Methods

This paper would not be complete without a brief discussion of
possible remedies if trouble occurs. One of the first things done to the
vessel reported as Case II under Field Data was to try a spring loaded
damping device originally designed by a large process engineering
concern and shown in Fig. 3. This device had practically no effect on
the behavior of the vessel. It can be

Journal of Engineering for Industry

argued that the spring load of 800 to 1000 lb results in a horizontal

force of only about 200 to 300 lb at the top of the vessel and this small
force will have very little effect on a moving vessel weighing 50,000 to
300,000 lb. If the spring load is doubled, the resisting load at the top of
the vessel is still a small factor in reducing vibration or limiting the
resulting deflection. However, if an accurate estimate can be made of
the vertical expansion for the operating temperature involved, the
spring could he designed and installed so that the assembly resulted in
full cable tension during operation. This in turn becomes a hazard
because of possible cable breaking which would endanger personnel. It
has been suggested that the spring could be entirely eliminated, but this
does not appear attractive due to the thermal expansion of the vessel
during operation and the subsequent danger of cable breakage just
outlined. Carrying the wires over sheaves and then straight down to the
foundation as shown in Fig. 4 has similar drawbacks. Aerodynamic
paneling similar to that used on the pipeline suspension bridge [6] has
also been suggested, but is not always desirable from other standpoints,
such as appearance and easy access to all sections of the vessel. If
paneling is used, the sections should be attached by bolting them to
clips which

FEBRUARY I959 / 81

have elongated or oversize holes to provide differential expansion

during operation. In some cases additional rolled plate can be applied
at the lower section which will increase the stiffness and lower the
period of vibration. Vertical beams welded for the length of the vessel
could be used, but are not recommended because of their restraining
effect under thermal conditions and possible discontinuity stresses.
One practical approach is to design the vessel so that there is a
separate section in the top which can he partially filled with liquid
(water or mercury if high density is required). The action of the liquid
will rapidly dampen the vibration and help prevent excessive
amplitude build-up, because at the instant the oscillatory motion has
its maximum acceleration, the liquid is still moving in the opposite
direction thereby creating a damping effect. This is the same effect
(only of greater magnitude) as observed from the tray liquid reported
in Case II under Field Data. Of course the choice of liquid and the
possibility of using this type of damping is dependent upon the
temperature involved. To date, it has not been necessary to resort to
any of these methods for vessels designed to the method outlined in
this paper.

Field Data
54 in. I.D. x 146 ft-0 in. High Vessel Shown in Fig. 5.
This vessel was designed to the seismic analogy method described
in this paper. Field engineers checked the period of vibration by
setting the vessel in motion and observing its frequency and amplitude
with a surveyor's transit sighted on a target rod mounted horizontally
at the top of the vessel. A stop watch was used to time the number of
cycles. This vessel could be oscillated by two men exerting a
back-and-forth motion at the top platform.

Case I

82 / FEBRUARY 1959

Whereas format data were not retained, the engineering records show
that they found the period of vibration to be 1.67 sec. Calculating the
period of vibration by the Rayleigh approximation (6), using six
sections gives a period of vibration of 1.61 sec which is 3.6 per cent
lower than the observed period.
Although the calculated wind velocity required to cause
resonance is only 10-12 mph, this vessel has operated without any
difficulty. The external attachments were well distributed about the
circumference.
Case II 84 in. I.D. x 145 ft-6 in. high Vessel Shown in Fig. 6.
This vessel was not designed to the seismic analogy method. The
thickness of the shell was increased in the lower section to withstand a
high wind loading. During the construction period, this vessel was
observed to be vibrating under certain wind conditions and, not only
was the amplitude great enough to be alarming, but the anchor bolts
stretched and an adjoining reboiler was loosened at its foundation.
Research engineers were sent to the field and made a
comprehensive study of the installation. It was observed that
resonance occurred at wind velocities in the neighborhood of 27 mph.
As previously mentioned, critical vibration was induced when the
wind came from a certain direction and, although the vessel could be
mechanically vibrated from any direction, the vessel was
"frequency-polarized'' due to the orientation of the trays and welded
downcomers. The maximum amplitude was 0.45 ft during resonance.

The vibration was recorded by strain gage-oscillograph

equipment and accurate wind velocity readings were recorded at
several different elevations. The recorded amplitude measurement was
checked with a surveyor's target rod and transit as outlined under Case
I.

Transactions of the ASME

During the investigation, the Strouhal number for this column (7.67
ft O.D. of insulation) and another 3 ft D column was obtained. These
are reported in the Design Procedure section of this paper.
Using the term "per cent decrement," defined as each amplitude
having a swing equal to a certain percentage less than that of its
predecessor, this column was found to have a 31/2 per cent decrement
when the vessel was empty and a 14 per cent decrement when the
trays were liquid loaded
The logarithmic decrement for this column is approximately 0.035
without liquid loading and approximately 0.133 loaded with liquid.
The calculated period of vibration of 1.42 sec is 4.05 per cent lower
than the observed period of 1.48 sec.
Case III 36 in. I.D. x 42 in I.D. x 131 ft-0 in. High Vessel Shown in
Fig. 7.
This vessel, which was designed to the seismic analogy method, has
a calculated period of vibration of 1.61 sec. Field readings were taken
in the manner outlined under Case I. The readings shown

Journal of Engineering for Industry

in Table l were taken before the insulation was applied. The average
observed period of vibration is 1.64 sec.
The readings shown in Table II were taken after the insulation was
applied. The column was also pressurized at 210 psig and had a bottom
temperature of 340 F and a top temperature of 90 F. There was no liquid on
the trays but there was about four feed of liquid in the bottom. Both sets of
readings were taken in still air. The average observed period is 1.69 sec.
The calculated period of vibration is only 1.83 per cent lower than the
average from Table I and 4.15 per cent lower than the average from Table II.
The numerical integration of equation (6) consisted of considering the
skirt to be 51 in. average diameter and the 42 in. D section as extending to
the top of the conical reducer. In addition, the upper 51 ft 9 in. consisting of
3/4 in. thick plate was divided into three sections as was the 46 ft 7 in. of
7/8 in thick plate of the 42 in. D section, making a total number of 9
sections.
Calculations indicate a critical wind velocity in the neighborhood of 8-10
mph for this vessel. No excessive movement has

FEBRUARY 1959 / 83

6 If vibration trouble does occur, careful analysis of any proposed

remedy must be made in order to avoid trouble from some other
source.

TABLE I
Observer
1
2
3
1
2
3

Time Seconds
38
51.5
49
67
50.2
50.4

Number of
Cycles
24
31
30
40
30
31

Amplitude at
Start - Feet
.15
.20
.20
.20
.20
.20

Amplitude at
Finish - Feet
.05
.05
.05
.03
.05
.05
Aver.

Period Seconds
1.58
1.66
1.64
1.67
1.68
1.62
1.64

Percent
Decrt.
4.66
4.52
4.67
4.75
4.67
4.52
4.63

Acknowledgment
The author wishes to acknowledge and express his appreciation for
the co-operation of the Standard Oil Company of California, Salt Lake
Refining Company, and the Research Division of The Fluor
Corporation. The assistance of Mr. R. C. Baird as well as many
individuals of The Fluor Corporation helped to make this paper
possible and is gratefully acknowledged.

References
TABLE II
Observer
3
1
3
1

Time Seconds
31
33.2
38.2
38.2

Number of
Cycles
19
20
22
22

Amplitude at
Start - Feet
.25
.30
.30
.30

Amplitude at
Finish - Feet
.05
.05
.05
.05
Aver.

Period Seconds
1.63
1.66
1.74
1.74
1.69

Percent
Decrt.
8.55
9.00
8.18
8.18
8.48

been reported and here again, the external attachments are well
distributed about the circumference.
The increased per cent decrement in Table II is partly due to the addition
of insulation which would have a more noticeable effect on a small
diameter column. Internal pressure is believed to increase the stiffness
and probably was a contributing factor. Whereas the bottom liquid is
near the base and would not be expected to contribute to the increase in
per cent decrement, it could conceivably have some effect.
The corresponding logarithmic decrements for the two conditions are
0.045 and 0.082.

I
J. P. Den Hartog, Mechanical Vibrations," third edition, McGraw-Hill Book
Company, Inc., New York, N.Y.
2
Stephen Timoshenko, Vibration Problems in Engineering, second edition. D.
Van Nostrand Company, Inc., New York, N.Y.
3
N.O. Myklestad, Fundamentals of Vibration Analysis, first edition, McGrawHill Book Company, Inc., New York, N.Y.
4
Reference [1], p.373
5
R. L. Solnick and R.H. Bishop, Noise, Vibration, and Measurement Problems,
Resulting from Fluid-Flow Disturbances,TRANS,ASME, vol.79, 1957, pp. 10451048.
6
R.C. Baird, Wind-Induced Vibration of a Pipe-Line Suspension Bridge, and Its
Cure,TRANS.ASME vol. 77, 1955, pp. 797-804.
7
References[1], p. 459
8
Reference [2], pp. 370 to 382 incl.
9
E.O. Bergman , The Design of Vertical Pressure Vessels Subjected to Applied
Forces, TRANS.ASME, vol. 77, 1955, pp. 863-867.
10 D. B. Steinman, Problems of Aerodynamic and Hydrodynamic Stability,
Proceedings, Third Hydraulic Conference, June, 1946, Univ. of Iowa Bulletin No. 31,
Page 139. Graph shown in Fig.2 reproduced by permission.
11 R.C. Baird , Aerodynamic Vibration of Tall Cylindrical Columns, Paper No. 58
-PET - 4.
12 Report of Fluor Research, unpublished
13 Reference [10] pp. 144 and 145.

APPENDIX

Conclusion
1 Self-supporting vertical pressure vessels should always be
investigated regarding their possible behavior under vibrating
conditions.
2 If the statically designed vessel has a period of vibration such
that it is necessary to consider it as a dynamic structure, it should be
designed to the seismic analogy method using a 0.20 seismic factor.
It is not necessary to apply this analogy to anchor bolts, if they are
not stressed over 15,000-16,000 psi.
3 The period of vibration of multithickness vessels and most
multidiameter vessels may be estimated by the numerical integration
of equation (6). The length of the sections used for solving equation
(6) should not exceed twenty to twenty five feet in order to have the
estimated period within approximately 5 per cent of the true period.
Complicated units with long, conical transitions (making it
impractical to consider the cone as a straight shell having uniform
properties) require more involved methods of approximation, or
recognition that the estimated period for these units, if estimated to
equation (6), may be more than 10 per cent in error.
4 The evaluation of wind velocity effects should include
considerations pertaining to the distribution of external vessel
attachments as well as the surrounding equipment and terrain. It
should be borne in mind that liquid loading in vessels having trays
will help dampen vibration, but should not be relied upon as a cureall.
5 Anchor bolts must be properly pretightened to a torque which
will prestress them an amount equal to their estimated working
stress, otherwise they may stretch sufficiently to affect the period of
vibration and possibly work loose.

84 / FEBRUARY l 959

I Estimating Period Of Vibration

The weights of the following items are used for estimating the
period of vibration when applying equation (4) (or the corresponding
graph of this equation ) or when numerically integrating equation (6):

1
2
3
4

Weight of Shell and Heads.

Weight of Trays, Caps, and Internals.
Weight of Manways and Nozzles.
Weight of Insulation and Fireproofing.

The total weight (items 1 through 4) of the vessel or vessel section

is considered as a uniformly distributed load acting on the vessel
when it is considered as a cantilever beam, i.e., a cantilever beam
deflecting under its own weight. Note that equation (4) requires the
total weight to be divided by the total length in feet to obtain the Unit
loading in pounds per foot, whereas equation (6) and the deflection
equations shown in Fig. 9 are based on the total weight of each
section under consideration.
The numerical integration of equation (6) is accomplished by
dividing the vessel into the required number of sections - one section
for each different thickness of plate with no section exceeding twenty
or twenty-five feet in length, keeping in mind that the greater the
number of sections, the more accurate will be the estimate.
After determining the weight and moment of inertia for each section,
estimate the deflection at the e.g. of the section either

Transactions of the ASME

graphically or by the numerical procedure outlined herein. When

applying the numerical procedure, it is necessary to find for each
section (except the last one at the free end) the deflection at the center,
the deflection at the end, and the end slope due to (a) the uniformly
distributed load W, (b) the end moment, and (c) the shear load. These
are found from the standard deflection equations shown in Fig. 9. The
last section at the free end requires only the deflection at its center due
to its own weight. Reference to Fig. 8 will immediately disclose that
the last section on the free end does not have an end moment MT or a
shear load Ws to take into consideration and the end deflection and the
end slope are not required to find the total deflection at the center. The
center of gravity of each section is considered to be at its midpoint.
Before proceeding with the deflection estimate, the designer should
find all of the shear loads and end moments as shown in Fig. 10.
It is then a simple matter to find the total deflection of each
section, square the deflection, and then tabulate the weight times the
deflection and the weight times the deflection squared for each section,
so that they may be added to find Wy and Wy2 as shown in Fig. 11.
Equation (6) may be written

Wy 2

T 1108
.
Wy

(8)

to further simplify the arithmetic after the weight-deflection data are

found.
As previously mentioned, many designers prefer to graphically
estimate the deflection, but the numerical method is suggested for

Journal of Engineering for Industry

those who do not regularly use graphical methods. Actually either

method requires about the same amount of arithmetic except those
working graphically usually do not keep a detailed record of their
areas and moments.

II Determination of Per Cent Decrement

Let X = amplitude of first swing.
% decrement

A 1

100

Note: A five per cent decrement

means that each swing is 5 per
cent less than its predecessor.
Then:
first swing amplitude
second swing amplitude
third swing amplitude
or
Nth swing amplitude
and
per cent decrement

=
=
=

X
AX
A2X

A(N-1)X

100(1-A)

The term per cent decrement is of value when comparing the damping
effect of different loading conditions for the same column.
FEBRUARY 1959/ 85

86/FEBRUARY 1959

Transactions of the ASME

DISCUSSION
M. Ludwig

(10)

1 .7 6 E

This paper deals with a problem that has long challenged this writer
and his associates. His recommended procedure for calculation of
dynamic wind forces can, however, lead to much more costly designs
than we have found to be necessary. The author is primarily concerned
with the possibility of forced resonant vibrations stimulated by
transverse cyclic wind forces associated with the Karman vortex trail
and, to deal with this possibility, suggests that flexible vertical
pressure vessels be designed to resist a lateral force equal to 20 per
cent of the gravity force. This lateral force is, for most flexible vessels
(those with shell thicknesses greater than 0.4 in. if the total mass is
twice the mass in the shell) greater than our customary design for
either wind or earthquake. The justification for use of this seismic
force is not presented, either as factor for a seismic design or for
avoiding wind-induced vibrations.
An anomalous feature of the author's seismic analogy method of
design is that it logically leads to the conclusion that a simple vertical
cylindrical shell, such as a steel smokestack up to 0.8 in. thick,
requires no special consideration because of possible wind vibration,
whereas a fractionating column of the same thickness and diameter
must be strengthened because of the added mass due to the insulation,
trays and fluid thereon, ladders, piping, and other appurtenances. Such
a conclusion cannot be supported by past experience; excessive
vibration of steel stacks has occurred, whereas vibration of
fractionating columns has seldom been a problem. The author notes
one case of fractionating column vibration but this stopped when the
column was put into operation; either the liquid on the trays provided
adequate damping or the added mass of the liquid increased the
natural period to a less critical value.
Our own experience is that the fractionating columns can be safely
designed for static wind loads alone. The possibility of excessive
wind vibration simply appears too remote to justify any added expense
to prevent such vibration. Strengthening of the steel shell, by adding
thickness, merely reduces the natural period of vibration and increase
the wind velocity necessary to produce forced oscillations; it is not at
all safe to assume that it would eliminate or reduce the amplitude of
vibrations that might otherwise occur.
Why is it that tall vertical pressure vessels, such as fractionating
columns, are far less severely affected by wind vibration than are
self-supporting steel smokestacks? The greater mass per unit of gross
cross-sectional area cannot alone be responsible since large
above-ground oil pipelines have been observed to vibrate in the wind.
There will be added aerodynarnic damping because of attached
platforms, piping, etc., but it can be shown that the energy absorbed
by this form of damping is probably not enough to limit the vibration
amplitude to reasonable values. The external irregularities due to
platforms, piping, etc., could, however, reduce the applied periodic
wind force.
A sound theoretical analysis for forced vibration of the resonant
frequency will answer the question raised in the previous paragraph.
Actual numerical values for the possible vibration amplitude and
resulting stress can be evaluated if the damping constant for the
column can be determined or estimated. The detailed analysis,
although straightforward, is too lengthy for inclusion in this brief
review, but the final equations are listed. It is assumed that the vessel
vibrates in a sustained wind as a uniform cantilever beam in the
fundamental mode.
Am 3.30 106
2

D2
S In Bt
C

D Am
L2

or
5.81 10 6 E

in which

D3
2

Am = maximum vibration amplitude at top of column

B = ratio of total mass per foot to mass per foot of steel shell
C = coefficient for peak periodic wind force - given as 1.71 by
Steinman (reference [10] of the paper)
D = diameter of shell
E = modulus of elasticity of steel
L = height of column
S = Strouhal number
t = thickness of shell
= amplitude ratio for two successive maxima, for free oscillations
In = natural logarithm of . (This is the "logarithmic decrement."
The "damping ratio" or ratio of actual to critical damping
is equal to the logarithmic decrement divided by 2)
= peak vibratory stress at base of column
Any consistent set of units may be used in these equations.
As a numerical example, let
C = 1.71
S = 0.20
B = 2.00
D = 4.00 ft
L = 150 ft
t = 1 in. or 1/12 ft
E = 30 x 106 psi or 4320 x 104 psf
In = 0.10 (damping ratio = 0.10/2 = 0.0159)
Then, from equation (9), Am = 0.135 ft or 1.62 in.
From equation (l0) or (11), = 1270 psi.
The possible deflection and stress calculated here are hardly large
enough for concern. The assumed damping is reasonably small but
could be much less without serious results. As a matter of further
interest, the natural period of vibration for this column is 2.39 sec, the
wind velocity for resonant vibrations is 8.4 fps, and the Reynolds
number for this wind velocity is 214,000. This is within the region
where C should be around 1.71 and S around 0.20, as was assumed. A
value of 1.71 for C is probably the maximum obtainable. It actually
determines the peak value of the periodic force, which probably
includes higher harmonics; the coefficient for the fundamental
frequency component may be substantially less. Also the coefficient
will be less if the column is not a true circular cylinder or if the
Reynolds number is greater than 500,000.
Note especially, as shown by equation (11), that the bending stress
is, for a given damping ratio, proportional to the cube of the diameter,
inversely proportional to the square of the height, inversely
proportional to the shell thickness, and inversely proportional to the
mass ratio B. The greater values of the thickness and mass ratio for
pressure vessels, as compared to steel smokestacks, are particularly
important in minimizing the seriousness of possible wind-excited
vibration. Thus, both experience and theory lead to the conclusion
that vertical pressure vessels, such as fractionating columns, are not
nearly as likely to vibrate excessively in the wind as are
self-supporting steel smokestacks.

(9)

Standard Oil of California, San Francisco, Calif.

Journal of Engineering for Industry

(11)

S In BtL
2

FEBRUARY I 959 / 87

Earl J. Hicks and J. R. Sellers

The author implies that the vibration of a tall vessel is a case of

forced vibration, with resonance occurring when the Karman vortex
trail frequency corresponds with the natural frequency of the vessel.
This is supported by Baird [14] 5 in. his investigation of a pipeline
bridge vibration. We have had a similar experience with a pipeline
bridge in which a critical wind velocity was observed. A search of the
literature finds an exception to this with experimental data to support
the claim that tall stacks vibrate as self-excited vibrational systems
with no critical wind velocity causing forced vibrational response.
Ozker and Smith [15] make the following statements in the summary
and conclusions portion of their paper which are contrary to this
paper: (1) The stack structure under wind action constitutes a self-excited vibration system": (2) The vibrational frequency is the natural
frequency of the structure and remains constant for all wind
velocities'': (3) The stack is at resonance at all times; (4) There is
no critical wind velocity in the sense of forced vibrational response;
and (5) The amplitude increases with increasing wind velocities. If
this were found to be true for a stack, could it not also be true for a tall
vessel? The bulk of the data seems to support the theory of forced
vibration excited by the Karman vortex trail, but there is this one
notable exception. There may be others.
The basic assumption in calculating the natural frequency of a tall
cylindrical vessel is that the base is fixed at the top of the foundation.
This implies that the horizontal displacement or deflection is due to
the elastic deformation of the vessel and that the horizontal
displacement due to the elastic deformation of the soil is negligible.
This means that the vessel would act like a true cantilever beam fixed
at one end. This is one extreme assumption.
The opposition extreme assumption would be that the vessel and
foundation is a rigid structure resting on an elastic subgrade. This
implies that the horizontal displacements of the tower are negligible
compared to the deflection due to deformation of the soil.
Assuming the latter assumption to be correct, it can be shown for a
vessel on an octagonal foundation resting on an elastic subgrade that
the natural frequency of vibration is equal to the following:
B d
f 0.232 s
H q

where
B = short diameter of the base in feet
H = distance between the base and the center of gravity of the
foundation and vessel
ds = coefficient of dynamic subgrade reaction
q = static soil pressure per unit of area
From this equation the following conclusions can be drawn. The
softer the supporting soil, the lower is the natural frequency and the
frequency may be raised by increasing the area if the foundation base
[16].
In general, observed frequencies have been lower than the
calculated frequencies. This is probably due to the vessel and
foundation acting together in a manner somewhere in between the two
extreme assumptions. If the supporting soil is relatively strong, the
first assumption is more nearly correct. If the
3
4
5

Engineering Department, Phillips Petroleum Company, Bartlesville, Okla. Mem. ASME.

Engineering Department, Phillips Petroleum Company.
Numbers in brackets designate References at end of this discussion.

88 / FEBRUARY 1959

supporting soil is relatively soft, the actual condition tends to

approach the second assumption. Past observations on various stacks
have indicated reasonably close agreement between the observed
frequency and the calculated frequency, the latter being based on the
fixed cantilever beam assumption. If the frequency of a vessel is
determined, assuming a fixed cantilever beam, and the supporting soil
is soft, the error would be on the unsafe side; that is, the calculated
critical wind velocity would be high.
The method of design suggested for wind induced vibration appears
to be arbitrary and most likely finds justification in the number of
successful towers. It would be desirable to have a more analytical
approach to the problem. Apparently, any relation between the wind
induced vibration and the 0.2 seismic factor is purely coincidental.
It appears a more realistic approach would be to equate the energy
input in terms of amplitude to the energy dissipated in terms of
amplitude and solving for the amplitude where the energy input is
equal to the energy dissipated. The vibrations at this point would be
an undamped steady-state free vibration and the resulting amplitude
would be a maximum. Knowing the maximum deflection, the
maximum stress could be readily calculated. If the resulting stresses
are excessive, then other methods to dissipate the energy would have
to be used.
This method would require reasonably accurate knowledge of the
coefficient of lift CL, the Strouhal number S, and damping decrement
. Only meager information on the numerical value of these
coefficients is available at this time. It appears desirable to study the
loads and dynamic response in a series of wind tunnel tests. Does the
author know of any such studies?
Aerodynamic paneling, liquid loading on trays, and liquid chambers
are practical solutions. One other worth mentioning is that of tieing
adjacent vessels and structures together. Vessels arranged in a triangle
or square with common ties have a different vibrational mode and a
higher resonant frequency than a single tower. In addition, more
damping is introduced. A single vessel tied at an immediate level to
an adjacent structure will vibrate in a different mode and higher
frequency. These can be used as safeguards for vessels claimed as
critical.
The liquid chamber suggested is one form of Frahm dynamic
absorber system.. A word of caution should be given on this
application. The Frahm dynamic absorber system has two resonant
frequencies and one frequency of zero amplitude. A properly sized
liquid chamber would decrease the amplitude to zero at one frequency,
reduce it over a limited range of frequencies, but amplify it at two
frequencies outside this range. In other words, at certain frequencies
the liquid may not slosh to oppose vibration, but instead be in phase to
amplify it. Frahm antiroll tanks for ships are dealt with in Den
Hartogs book [17]. A liquid chamber appears to be an energy
absorber of this same type.
In Case III, no excessive movement has been reported although the
critical wind velocity is in the neighborhood of 8-10 miles per hour.
In Case I, the vessel has operated successfully with a calculated wind
velocity to cause resonance of 10-12 miles per hour. Both of these
vessels were designed to the seismic analogy method. In Case II
where the seismic analogy method of the design was not used, the
vessel vibrated during the construction period. Does the author imply
that since Case III and I were dynamically designed and Case II was
not that this was the main reason that vibration did not occur or could
it be that the ladders and platforms were distributed around the former
two vessels such that the formation of the Karman vortex trail does not
materialize? It appears that the ladders, platforms, piping, and
insulation might be the major factor in preventing vessels from
vibrating. No doubt there are many vessels with a height to diameter
ratio greater than 20 which are operating

Transactions of the ASME

satisfactorily and have not been designed as recommended by the

author. The author implies that any vessel with a period which falls
above the line shown in Fig. I should be designed dynamically. Would
this be true for vessels whose heights are less than 100 or 75 or 50 ft?
In general, is there some approximate height regardless of the H/D
ratio where vessels below this height would not have to be
investigated?
Although the author does not say specifically that the vessel should
be designed such that the "critical wind velocity" is high enough to
exclude the possibility of the second mode of vibrations it is
mentioned. If this item is not critical, then why is it necessary to use
such an accurate and time consuming method to determine the period
when it could probably be estimated by other methods. Apparently,
the designers were not concerned in Cases I and III about the low
critical wind velocity since they were designed dynamically.
The author's paper is very timely indeed and is viewed with
considerable interest since the trend in pressure vessel design is to
increase their height and decrease the internal pressure. We wish to
compliment him on a very excellent paper and also thank him for
stimulating our interest in the subject.

References
14

R. C. Baird, "Wind-Induced Vibration of a Pipe-Line Suspension Bridge, and Its

Cure," TRANS. ASME, vol. 77, 1955, pp 797-804.

15

M. S. Ozker and J. O. Smith. "Factors Influencing the Dynamic Behavior of Tall

Stacks Under the Action of Wind," TRANS. ASME, vol. 78, 1956, pp. 1381-1391.
16

Karl Terzaghi, Theoretical Soil Mechanics, John Wiley & Sons, Inc., New

York, N. Y., pp. 454-457.

17

J. P. Den Hartog, Mechanical Vibrations, McGraw-Hill Book Company, Inc.,

New York, N. Y., second edition, pp. 129-135.

18

W. L. Dickey and G. B. Woodruff, The Vibrations of Steel Stacks, Trans.

ASCE, vol. 121, 1956, p. 1088.

19

E. A. Dockstader, W. F. Swiger, and E. Ireland, Resonant Vibration of Steel

Stacks, Trans. ASCE, vol. 121, 1956, p. 1088.

L. Acquaviva

The proposed criteria for dynamic design of pressure vessels

appears to yield more conservative results than the method we have
been using for many years. It appears that the seismic coefficient of
0.20 applied on all vessels with height to diameter ratio exceeding 15
or natural period exceeding 1.0 sec may be too conservative based on
our experience. For example, as an extreme case, there is a tower
within the Esso interests which is 173 ft high., topmost 40 per cent of
height is 6 ft 6 in. in diameter and lower 60 per cent 5 ft 0 in. in
diameter. The approximate L/D ratio is 30, natural period of vibration
3.7 sec and critical wind velocity 6.5 miles per hour for the top
section. The tower is on pile foundations and has given no vibration
difficulties even though subjected many times to the critical wind
velocity for the top section; nor has there been any tendency to vibrate
at its second mode for which the critical wind velocity is about 40
miles per hour. The bottom section shell thickness for this vessel is
7/8 in. The thickness required to conform with proposed dynamic
design would be 1 3/8 in.
Unlike steel stacks, for which there are many instances of excessive
vibration reported in technical literature, Case II in the paper is the
first example to our knowledge of such occurrence in vertical pressure
vessels. In this case, the maximum amplitude of 0.45 ft during
resonance does not appear to be of sufficient magnitude to cause
excessive stresses in the vessel shell. It would be of interest to know
if an remedial measures were taken in this case.
6

Esso Research and Engineering Company, Linden, N. J.

Donald J. Bergman

This paper is particularly interesting because it includes some actual

field data which have been extremely difficult to get. It seems
important to note that minor changes in resonant frequency have little
effect on the over-all situation as this merely results in a slightly
different wind velocity to reach resonance. It was much more
important to note how greatly the liquid on the trays of a column
increased the damping effect. This absorbs the energy of the wind
forces and cuts down the magnification factor, decreasing the
maximum. amplitude of vibration.
We had several instances of stack vibration starting in 1938, all far
removed and all solved by use of permanent guys. The first key to the
causes came from Sir James Jeans book "Science and Music." Here
the comment was made that a set of eddies broke off from a cylindrical
surface every 5.4 diam along the wind stream and that these eddies
caused a push from side to side. The example was given of a ship with
1/2 in. rigging at sea in a 40-mile gale where the frequency comes out
to correspond to middle "C" on a piano.
The policy of providing light guys and spreaders was adopted for
columns with high L/D ratios and springs were provided to take care
of column elongation by heat. Several hundred columns were thus
equipped over past years without having one case of resonant
vibration. Perhaps this was just fortuitous because of presence of
ladders, platforms, etc., to interfere with resonance, and absence of
critical winds. Most of these columns were within the critical range as
defined by the author.
Later study indicated the same conclusions that the author reached:
That the guys and springs were probably not effective and,
consequently, about a year ago provision of these guys was abandoned.
Since then one tower was found to have a resonant period with the
wind and permanent guys were added.
Perhaps it should be noted that the elasticity of fixed guys actually
makes them springs with a very high constant. Limiting the movement
of the tower is the desired result, and this the guys do by resisting the
wind forces.
The lateral forces resulting from the action of the Karman eddies
seem to be somewhat greater than the normal wind drag for any wind
velocity. With a low decrement, or high magnification, say, 20, only a
relatively small force, 1 lb/sq ft, is required to give the deflection
obtained when designing for a 100-mile per hour wind at 20 lb/sq ft,
on the projected area of the column.
The lateral forces generated on cylinders were actually utilized
shortly after World War I by the German Flettner rotor ship which
cruised from Germany to New York using two 10 x 60-ft cylinders
mounted on the deck and rotated in order to furnish a fixed instead of
an oscillating force.
Rouses book Engineering Hydraulics on page 130 gives formulas
for an approximation of the aerodynamic lift caused by the Karman eddies
on fixed and rotating cylinders. Additional information for higher
Reynolds numbers was obtained from Prof. L. Landweber of the Iowa
Institute of Hydraulic Research by correspondence. For the 7.67-ft column
which gave trouble with a 27-mph wind, I calculate a loading of 9.1 lb/ft or
1.18 lb/sq ft of projected area. Stiffening a column and increasing its
frequency will decrease the deflection for a given load, but the wind forces
increase as the square of the wind velocity.
Some comment was made regarding the second mode of vibration. This
will always require a very much higher wind velocity than the first mode,
but the column is so much more stiff because of the complicated second
mode curve that it is probable that the vibration is not noticeable if it does
occur.

Engineering and Development Department. Universal Oil Products Company, Des Plaines, III.

Mem. ASME.

Journal of Engineering for Industry

FEBRUARY I 959 / 89

Patent no. 2604838, dated July 29, 1952, to W. B. Traver, of

Standard Oil Company of Indiana, teaches provision of a roughened
surface near the top of a stack either by small strips ranging from 1/4
to 1 in. depth, or by the use of steel grating. This increases the skin
resistance on the surface of the cylinder to prevent formation of
Karman eddies. However, it does not seem to be a very practical
remedy because of the cost of attaching the large number of small
strips to the stack.

Author's Closure
Before replying to specific questions or comments, the author
wishes to thank the discussers for their review and comments
pertaining to the design of vertical vessels subject to vibration.
It is hoped that sufficient interest has been stimulated in this subject
to result in obtaining more field data than we presently have at our
disposal. Up to the present time very little has been done to obtain
information pertaining to tall, slender vessels. If no complaint was
received from the operators, it has been assumed that no critical
vibration would ever occur. This is a normal and expected attitude;
however, it has not contributed to our knowledge of vessel behavior as
heights have increased.
As pointed out in this paper and described more in detail in the
paper given by Mr. Baird, we have had definite experience with one
vessel which did give trouble and, after completing this paper the
author was informed that this vessel still has to have anchor bolts
retightened at intervals which indicates that some excessive vibration
may still be taking place.
The author does not agree with Mr. Ludwig that we can generalize
regarding the relationship of total mass to the mass of the shell itself.
Mr. Ludwig's use of a mass equal to twice the mass of the shell in his
4-ft 0-in. diam X l-in. thick example would result in very heavy
internals. To me, each vessel should be considered individually during
the design stage. Obviously, there will be those where conditions of
terrain and prevailing winds will cast doubt regarding the justification
of added cost to the vessel.
The cost angle of designing to the method recommended in this
paper has been overemphasized. Taking Mr. Ludwig's example and
designing it to the 0.2 seismic analogy results in the following
thicknesses:
Top
Next
Next
Next
Next
Skirt

104 ft
16 ft
8 ft
8 ft
8 ft
6 ft

1-in. plate (orig. thickness)

1 1/8-in. plate
1 1/4-in. plate
1 3/8-in. plate
1 7/16-in. plate
1 7/16-in. plate

This represents an increase of only 7,500 lb added to a vessel

originally designed as 146,000 lb (when l6,000 lb is subtracted for
insulation). This approximation is based on ASTM A-212-B Material,
S.R. and x-rayed, and includes 1/8-in. corrosion allowance in the l-in.
plate, which means that this vessel would have 7/8-in. thickness for
resisting pressure in the hoop direction. We then have 7/16 in. of
material available to resist the bending in the longitudinal direction.
Mr. Ludwig ignores the pressure stress when he reports his bending
stress due to his estimate of the maximum amplitude of vibration.
Costwise, this vessel would represent an investment of approximately
$36,500 as originally outlined. The added material and labor would be
approximately $1350 which is an increase of only 3.7 per cent. In
this case, the added material accomplishes two things; first, it
increases the section of the vessel where bending stresses occur and
will reduce the possibility of failure from fatigue, a condition which
could exist if the critical wind velocity was prevalent; second, it gives
more resistance to bending if a harmonic condition occurs.
Mr. Ludwig reported a critical wind velocity of 8.4 fps or 5.7
90/ FEBRUARY 1959

mph, and apparently used a 4-ft 0-in. diam with his selected Strouhal
number of 0.20. Adding 3 in. of insulation, and using a Strouhal number
of 0.18 (based on field data for two vessels reported in the paper) gives a
critical velocity of 7.1 mph for the vessel as originally outlined and 7.8
mph for the design as recommended in the paper. (The period of vibration
decreased from 2.39 to 2.19 sec.) It is agreed that the recommended
design does not materially affect the critical wind velocity for most vessels.
It is also agreed that stacks may cause more trouble from vibration than
vessels, because they do not have liquid loading and are usually lighter in
weight. However, the inference pertaining to ignoring a stack of 0.8 in.
thick should be further clarified, because the L/D ratio would determine
whether or not the stack should be designed for vibration. Also, there is no
mention of stacks in this paper and it is not recommended that this design
procedure should be applied to them. Since processing is not involved,
stack designers usually increase the resistance to vibration by a generous
conical section in the lower zone which sometimes runs from
to 1/3 the height of the stack. The pressure vessel designer can seldom
resort to this because of internal construction. In order to clarify the
statements concerning stacks, the following tabulation is based on Mr.
Ludwig's definition of a flexible vessel; i.e., one with a shell thickness
greater than 0.4 in. if the total mass is twice the mass in the shell:

Diam,
in.
36
48
60
72
a

Shell th,
in.
7/16
7/16
7/16
7/16

Vessels
wt/ft,a
lb
360
480
600
720

L/D
20
18
17
15

Max L,
ft
60
72
85
90

L/D
24
21
19
18

Max L,
ft
72
72
84
108

This is double the wt/ft for 7/16-in. plate.

Diam,
in.
36
48
60
72
b

Shell th,b
in.
7/16
7/16
7/16
7/16

Vessels
wt/ft,
lb
180
240
300
360

L/D and max L values for stacks actually are independent of shell thickness

because the value of the abscissa (wD/h) of the graph shown in Fig. 1 does not change
as the material thickness is changed ( w and h are proportional).

A comparison will show that the 36-in-diam vessel can be considered

as a static structure if not over 60 ft high, whereas the 36-in. stack could
be considered as a static structure up to 72 ft high, regardless of its
thickness. Similar values are shown for other diameters. As previously
mentioned, it was not intended to apply the graph shown in Fig. 1 to
stacks. However, there does not appear to be a great discrepancy in this
respect when it is realized that all of the metal put in the stack may be used
for structural strength since there is no internal pressure.
Stacks are frequently built using the thinnest calculated material
thickness, and sometimes local buckling has occurred due to warpage
and/or external forces. I do not believe a true comparison between stacks
and vessels can be made.
The calculation of the periodic force and resulting amplitude is difficult
to make with any degree of accuracy, because we lack field data on
vessels. Past attempts to apply some of the data from Reference [10] in
the paper have not always give results consistent with the field data taken
to date. This does not mean the approximations of this nature should not
be made, but we cannot place too much reliance in them.
Mr. Ludwigs bending stresses are very low because he included all of
the shell material for bending, whereas in actual design work we have to
combine the pressure stresses with the bend-

Transactions of ASME

ing stresses. Also, Mr. Ludwig's stress is based on his estimate of the
amplitude and ignores the fact that a deflection exceeding the normal
amplitude might result if resonance occurs, as reported for the vessel
outlined as Case II and reported by Baird.
Mr. Hicks and Mr. Sellers discuss the possibility of the foundation
and soil-bearing capacity as contributing to the flexibility of the
system. This possibility has not been ignored. However, there has
been no observation to justify this assumption. A check into the effect
of anchor bolt stretch did not affect the period of vibration estimate
sufficiently to bother with itother than to require pretightening to
reduce elongation during vessel deflection. When the total mass of
concrete and the surface charge of earth is considered, it is difficult to
agree that they materially influence the natural frequency. If the soil
bearing capacity is low, it is usually necessary to drive piles to support
vessels and similar heavy equipment.
The desire for analytical approach is appreciated. However,
pressure vessels do not always lend themselves to a true scientific
analysis. In this case, so many variables exist that assumptions made
in order to solve equations are apt to produce misleading results. The
fact that a number of successful vessels have been built using this
sornewhat empirical approach is reasonable justification for its
consideration. If we cannot justify a design method on the basis of
successful operation, then we would be forced to discard many of our
practices which are based on experience, including earthquake design.
It is also noted that some reviewers are equally positive that it is
unnecessary to take any precautionary measures because they have not
experienced any difficulty in the past. Therefore, we have one more
instance where the variables are too many to draw a definite
conclusion and data pertaining to location, terrain, wind currents,
detail vessel design, and vibration for each case are not available. If
we had at our disposal sufficient data taken from existing units, we
probably could work out a more analytical approach.
To my knowledge there have been no wind-tunnel tests pertaining
to vertical vessels except those mentioned in Reference [10].
Investigation of testing models for tall, slender columns will
immediately reveal that, in order to obtain reliable data for L/D ratios
of 30 and 40 to one, the model size will be difficult to work with,
because the diameter is so small in order to avoid excessive height.
Some of the larger wind tunnels could probably handle models large
enough to be practical, but the cost would be very high.
Connecting critical vessels to nearby structures of to adjacent
vessels is always desirable if the location of the equipment permits.
This cannot be accomplished in many instances.
I do not agree that the use of a single liquid chamber is comparable
to Frahm antiroll tanks. The type of motion differs and I believe that
there is very little possibility of the liquid amplifying vibration.
However, as pointed out in Conclusion No. 6 in the paper, any
proposed remedy must be carefully analyzed to avoid additional
trouble from some other source.
The paper does imply that the vessel reported under Case II

Journal of Engineering for Industry

might have been satisfactory if designed to the seismic analogy

method, but there is no proof of this. Recognition is given to the
desirability of distributing ladders, platforms, and other accessories
circumferentially about the vessel to reduce the effect of periodic eddy
shedding.
The heavy reference line shown in Fig. 1 is independent of height
and considers L/D ratios more important than height. A vessel 30 in.
in diameter and 90 It 0 in. high can be more critical than a 200-ft
0-in-high vessel of a larger diameter. This reference line from 0.4 sec
on the left to the 0.8 sec on the right of the graph is empirical. There
certainly are vessels having periods of vibration above this line which
were not designed as recommended that are satisfactory. This line at
one time was a horizontal line at 0.4 sec, and experience indicated that
it could be safely changed as shown. Future data may result in another
revision for this limit.
The method outlined in the paper for numerically estimating the
period of vibration for a vessel having a shell varying in thickness is
intended for designers who are not experts in vibration. We are
programming this work for computing machines along with our other
vessel program, but computing machines are not always available, and
other methods of approximation are not very reliable unless performed
by experts in the field of vibration. It is not always necessary to
estimate this period and frequently it can be approximated by using
the graph in Fig. 1. It is up to the designer to decide the degree of
accuracy he wishes to attain.
The column described by Mr. Acquaviva has considerable difference in
the diameter of the top and bottom section. Our experience has been that
vessels of this type are not as critical as tall, slender columns which are the
same diameter for their entire length. I have some reservation regarding
the accuracy of estimating critical wind velocities for this type of vessel
when the diameter difference exceeds 6 to 12 in.

Here is another example of experience and each individual will

have his own idea of :a correct solution.
If I were compiling data pertaining to existing vessels, I would
definitely attempt to include all possible information on surrounding
structures, wind currents, and geological data, in addition to details of
vessel construction. A successful installation in one locality may not
be trouble free in another.
The vibration dampener shown in Fig. .3 in the paper was applied
to the vessel which vibrated at an excessive amplitude. This had
practically no effect on the behavior of the vessel, but was not
removed. Liquid loading helped to prevent excessive amplitude
build-up, but as previously mentioned, it was recently discovered that
the anchor bolts have to he retightened at frequent intervals.
My reply regarding cost of designing to a 0.2 - seismic factor was
included with the reply to Mr. Ludwigs comments.
Mr. Fred Ruud's statement that the U. S. Bureau of Reclamation
uses a seismic analogy method for some of their tall vessels is
particularly interesting in view of the fact that seismic factors of 0.25
to 0.30 were mentioned, and the author wishes to thank Mr. Ruud for
his interest and express regret that he did not have the opportunity to
prepare written discussion on this paper.

FEBRUARY 1959 / 91