Engineering Structures 101 (2015) 621–630

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Comparison of the shell design methods for cylindrical liquid storage

tanks
Eyas Azzuni, Sukru Guzey ⇑
Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907, USA

a r t i c l e i n f o a b s t r a c t

Article history: The three methods for determining the shell thickness of steel cylindrical liquid storage tanks designed in
Received 11 April 2015 conformance with API Standard 650, Welded Tanks for Oil Storage (API 650) are: (1) one-foot method
Revised 27 July 2015 (1FM), (2) variable-design-point method (VDM) and (3) linear analysis. We compared the shell designs
Accepted 31 July 2015
based on these three methods for different tank properties: diameter, height and allowable stress. For lin-
ear analysis, we developed a stiffness–flexibility matrix method based on thin shell theory that gives the
theoretical displacements and stresses at each shell course without any approximation or simplification.
Keywords:
Results show that shell designs using VDM may produce overstressed shell courses for some of the large
Storage tank
Shell design
steel liquid storage tanks when VDM is permissible to use. Linear analysis would give more accurate shell
API 650 designs for those cases.
Thin shell theory Ó 2015 Elsevier Ltd. All rights reserved.
One-foot method
Variable-design-point method

1. Introduction calculated using the circumferential stress at a point 0.3 m (1-ft)

API 650 is an industry standard used for the design and con-
struction of large cylindrical storage tanks for liquid products
[1–3]. API 650 storage tanks are vertical, cylindrical, closed- and
open-top welded tanks with uniformly supported flat bottom.
They are used to store petroleum, petroleum products, and other
liquid products [1].
Recently, considerable research effort has been devoted to the
analysis, design, and evaluation of the liquid storage tanks [4].
Much of the research conducted has focused on the buckling of
and wind effects on the storage tanks [5–8]. Some researchers
worked on dynamic effects related to earthquakes [9–11]. Chen
et al. worked on developing a simple method to calculate shell
stress [12].
A typical storage tank has a number of shell courses of uniform
plate thickness. The thickest course is at the bottom and each shell
course above is typically thinner than the previous one. See Fig. 1
for a typical storage tank shell cross-section.
There are three methods allowed by API 650 to determine the
required plate thickness of the shell. The first method is the
one-foot-method (1FM) which is based on the ‘‘membrane theory’’.
The required shell plate thickness for each shell course is
⇑ Corresponding author.
E-mail addresses: eazzuni@purdue.edu (E. Azzuni), guzey@purdue.edu
(S. Guzey).
above the lower horizontal weld seam of the shell course due to
hydrostatic pressure of the stored liquid. The reasoning behind this
assumption is that the tank bottom plates provide restraint to
reduce circumferential stress due to hydrostatic pressure at the
bottom 0.3 m (1-ft) of the lowest shell course. Similarly, a shell
course other than the lowest shell course, has generally thicker
shell plates below. The plate below provides some restraint at
the lower portion of the shell course in consideration. The 1FM is
used successfully for the majority of the tanks. However, the
designs based on the 1FM may become conservative and cost pro-
hibitive for larger diameter tanks. Therefore, API 650 limits the
applicability of this method to tanks up to 61 m (200-ft) in
diameter.
The second method to calculate the required shell plate thick-
ness is the variable-design-point method (VDM) that is also based
on the ‘‘membrane theory’’. The VDM was proposed by Zick and
McGrath in 1968 [13] and later adopted by API 650 as a refined
method to calculate the required shell plate thickness especially
for tanks more than 61 m (200-ft) in diameter. The VDM takes into
consideration the restraint provided by the tank bottom plates to
the first shell course and the restraint provided by each lower shell
course to the upper shell course. The VDM uses a variable distance
instead of fixed distance of 0.3 m (1-ft), as used in 1FM, above the
circumferential seam for each shell course to calculate the maxi-
mum stress due to hydrostatic pressure. The variable distance in
VDM is a function of the shell plate thickness above and below

http://dx.doi.org/10.1016/j.engstruct.2015.07.050
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
622 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630
Nomenclature
H distance from the maximum product level to bottom of
the shell course under consideration (m)
L ð500 DtÞ0:5 (mm)
h1 height of the bottom shell course (m)
D nominal tank diameter (m)
r nominal radius of the tank (m)
G the design specific gravity of the stored liquid
CA corrosion allowance (mm)
Sd allowable design condition stress (MPa)
St allowable hydrostatic test condition stress (MPa)
t thickness of the shell
ti thickness of a shell course
td design shell thickness (mm)
t 1d design shell thickness for the first shell course (mm)
t dx design shell thickness (mm)
tt hydrostatic test shell thickness (mm)
t 1t hydrostatic test shell thickness for the first shell course
(mm)
t tx hydrostatic test shell thickness (mm)
tu corroded thickness of the upper shell course; approxi-
mated using 1FM for the first iteration (mm)
tL thickness of the lower shell course (mm)
the seam. Most of the time designs based on VDM are more eco-
nomical compared with those based on the 1FM. However, for
some tank geometries the VDM may become unconservative and
the tank shell thicknesses designed in accordance with VDM may
be overstressed. Buzek showed that the restraint provided by the
tank bottom on the tank shell produces circumferential stresses
of sinusoidal nature varying with the distance from the tank bot-
tom [14]. For certain tank diameter and height proportions, this
sinusoidal varying restraining stress may add to the stress due to
the hydrostatic circumferential stress and the design based on
VDM may become unconservative. Therefore, API 650 limits the
applicability of the VDM for the tanks with L/H ratio less than
1000/6 in SI units (refer to the nomenclature for the definition of
these terms). For the storage tanks where the L/H ratio is more than
1000/6, tank shell thickness should be determined using linear
analysis.
The shell thickness calculation using linear analysis is the third
method given in API 650. In this approach the boundary conditions
for the analysis should be a plastic moment related to yielding of
the plate under the shell and fully restrained radial movement at
the bottom of the shell. API 650 does not describe a specific linear
analysis method. In this study we developed a new method using
thin shell theory to perform a linear analysis for the shell thickness
calculation. In this method we are using exact stiffness–flexibility
relations and exact shape functions originating from the so called
‘‘short shell’’ solution of the governing equations from the thin
elastic shell theory. Therefore, we do not have any approximations
or simplifications. The displacements, section forces and stresses
obtained from this method are exactly matching the theoretical
solution of thin shell theory. Overwhelming majority of texts
employ only a single course solution of shell cylinder without
extension to multiple shell courses with stepwise thicknesses.
One can find very few references dealing with multiple shell
courses in which only approximate solutions were obtained. In
our treatment, we present an attractive and easy to implement for-
mulation that renders analytical solution without any
approximation.
We shall investigate the efficiency and limitations of each
method described above. The efficiency is defined in terms of
t1 corroded thickness of the bottom shell course (mm)
t2 minimum design thickness of the second shell course
(mm)
t 2a corroded thickness of the second shell course (mm) as
calculated for the upper shell courses
w radial displacement of the cylindrical shell
x axial length coordinate of the cylindrical shell
E modulus of elasticity
Ds shell bending rigidity
p pressure
m Poisson’s ratio
b a parameter
C1, . . ., C4, integration constants
f(x) particular solution to the governing equation
Li length of a shell course
Q1, Q2 end shearing forces of a shell course
M1, M2 end bending moments of a shell course
w1, w2 end radial displacements of a shell course
h1, h2 end rotations of a shell course
tn
t2 H
BOTTOM
SHELL TANK BOTTOM
COURSE t1
D
Fig. 1. Typical tank shell cross-section.
minimum required shell thickness for each shell course and corre-
sponding total weight of steel for the shell plates. A decrease in the
weight of steel can be achieved by reducing the shell thickness,
which will lead to a decrease in cost. Achieving a smaller design
thickness is important for large diameter tanks because tank fabri-
cators in North America typically order a plate thickness with
0.01 in. (0.25 mm) increments as well as commercially available
1/16 in. (1.59 mm) increments directly from a steel mill for suffi-
ciently large weight of steel, about 20 tons. Another reason to
achieve a smaller thickness may be to comply with the maximum
thickness limit of 1 in. (25.4 mm) to avoid stress relieving require-
ment. Therefore, even a reduction of 0.01 in. (0.25 mm) in design
thickness would be significant for large diameter tanks. Our main
objective is to investigate the accuracy of 1FM and VDM and pos-
sibility of obtaining an economical result by using the linear anal-
ysis for the required shell thickness for storage tanks.
We shall first summarize the three design methods for the stor-
age tanks. Then we shall compare the shell designs based on the
classic 1FM with those based on VDM. Furthermore, we shall focus
on comparison of the VDM results with the theoretical solutions
obtained from thin shell theory. Finally, we shall discuss the results
and give conclusions.
 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630
2. Design methods
We conducted three different sets of studies of the available
storage tank shell design methods in API 650:
(1) Comparison of shell designs based on the classical 1FM with
those based on VDM for the storage tank sizes where designs
based on 1FM is permissible in API 650 (i.e. the tanks up to
61 m (200-ft) in diameter).
(2) Comparison of the shell designs based on VDM with those
based on linear analysis using thin shell theory for the larger
storage tanks up to 122 m (400-ft) in diameter.
(3) Study of the tank sizes where VDM is no longer permissible
and compare the VDM results with those of obtained using
thin shell theory.
The storage tanks are typically constructed using 2.4 m (8-ft) or
3 m (10-ft) high shell courses of uniform thickness. In this study
we used 2.4 m (8-ft) high shell courses because they are used more
often than 3 m (10-ft) high shell courses.
Material selection for a tank shell is an important task in the
design process. The tank manufacturer, usually responsible for
the tank design, selects a suitable material specification for each
shell course based on cost, availability and brittle-fracture consid-
erations. In this study, we used three different allowable stress rep-
resenting low strength, medium strength and high strength
material specifications permissible in API 650 tank designs.
There are two different allowable stresses used in tank design:
(1) design allowable stress and (2) hydrostatic test allowable
stress. Design allowable stress is used with the specific gravity of
the product that is going to be stored in the tank. The product
specific gravity does not usually exceed 1.0 for the petroleum prod-
ucts. The second allowable stress is that for the hydrostatic test
used for the hydrostatic test condition. The specific gravity for
the hydrostatic test is usually 1.0 to represent the specific gravity
of the test liquid. Hydrostatic test is a proof test before a tank is
placed in service. During this test, the tank is filled with water
and shell plates are stressed to the levels that the tank may not
see again during its entire service life considering the product
specific gravity is less than 1.0 in most cases. In order to be more
general, we used a liquid specific gravity of 1.0 and selected the
allowable stresses from the hydrostatic test allowable stresses.
In the following sections we shall summarize the three
shell-thickness design methods for storage tank design.
2.1. One-foot method
The 1FM can be used for tanks up to 61 m (200-ft) in diameter.
It has the advantage of being simple and easy to use with no iter-
ations needed. The 1FM design takes the stress 0.3 m (1-ft) above
the bottom of each shell course to determine its thickness. The
thickness required is the larger of the two thicknesses calculated
using Eqs. (1) and (2) from API 650:
4:9DðH  0:3ÞG
td ¼ þ CA ð1Þ
Sd
4:9DðH  0:3Þ
tt ¼ ð2Þ
St
Note that Eqs. (1) and (2) above are unit dependent and the
equations in this work are given in the SI units. The reader is
referred to API 650 for the equations in USC system of units. Eq.
(1) is used to calculate the design shell thickness resulting from
hydrostatic pressure of the liquid to be stored and desired corro-
sion allowance for the design life of the storage tank. Eq. (2) is used
623
to calculate the hydrostatic test shell thickness resulting from
hydrostatic test pressure of the test liquid. As stated earlier, to be
more general, we used Eq. (2), and corresponding hydrostatic test
allowable stress in our study.
2.2. Variable-design-point method
The VDM can be used for tanks with L/H ratio of up to 1000/6 in
SI units (refer to nomenclature for the definition of L and H terms).
In general, most storage tanks used in practice would have an L/H
ratio smaller than the limiting value of 1000/6. Therefore, VDM can
be used to determine the shell design thickness most of the time.
The L/H limit is present because the VDM was based on the
assumption that the base plastic moment has a small effect on
the circumferential stresses in the lower shell course. The effect
of base moment is of a sinusoidal nature: It may amplify the cir-
cumferential stress due to hydrostatic pressure for bigger tanks.
The assumption used in the VDM for the base moment only works
for tanks with L/H ratio less than 1000/6 [14].
The calculation of the shell courses thicknesses using VDM is
more elaborate than the calculation using the 1FM. The VDM does
not take a specific point for all the shell courses but rather finds the
point where the stress is highest in each shell course to calculate
the thickness. There are three categories of calculations: first shell
course, second shell course, and upper shell courses. In accordance
with API 650 [1] the following equations are used.
For the first shell course from the bottom the greater of the two
thicknesses obtained using Eqs. (3) and (4) is used as the shell
thickness.
sffiffiffiffiffiffiffi! 
0:0696D HG 4:9HDG
t1d ¼ 1:06  þ CA ð3Þ
H Sd Sd
sffiffiffiffi! 
0:0696D H 4:9HD
t1t ¼ 1:06  ð4Þ
H St St
Note that the equations above are unit dependent and the equa-
tions in this work are given in SI units. If the calculated first shell
course thickness using Eqs. (3) and (4) is more than the thickness
obtained using the traditional 1FM, the thickness calculated from
1FM may be used as the first shell course thickness.
Calculation of the shell thickness for the second and upper shell
courses requires an iterative process. In this process the shell thick-
ness converges to the required shell thickness in two or three iter-
ations. The calculation procedure for the second and upper shell
courses are given in Appendix A.
2.3. Linear analysis using thin shell theory
Linear analysis is the required method of API 650 to calculate
shell plate thickness when 1FM and VDM are not permissible.
This method should be used when L/H ratio is greater than the lim-
iting value of 1000/6 in SI units. For the storage tank geometries
that is constructed in practice, it is very difficult to reach and
exceed an L/H ratio of 1000/6. However, in this study we aimed
to examine the linear analysis when the shell design based on
VDM is still permissible and compare the results of the design
based on linear analysis with those of the design based on VDM.
In addition, we investigated the linear analysis and VDM when
the design based on VDM is not permissible, i.e. L/H ratio is more
than the limiting value of 1000/6.
The API 650 document does not specify how to perform a linear
analysis other than some instructions on how to choose the bound-
ary conditions at the base of the shell cylinder. We performed the
linear analysis using thin elastic shell theory that combines the
624 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630

bending deformations in addition to the membrane deformations. attached to a joint is needed for each joint. However, long shell
Because we are dealing with thin shells, the out-of-plane shear solution may not give accurate results when semi-infinite length
deformations are neglected. assumption cannot be justified [16].
Consider a circular cylindrical shell of finite length and subject Short shell solution gives more accurate results than the long
to axisymmetric, torsion-free loading. The governing equation of shell solution because it incorporates all the integration constants
radial displacement, w may be written the following way if the into the solution. However, this approach requires more involving
thickness of the shell, t, is constant [15–17]. treatment because each joint cannot be solved independently.
4   Simultaneous solution of entire tank joints equations needed.
d w Et p This problem may be calculated by approximate methods [16].
4
þ w¼ ð5Þ
dx Ds r 2 Ds In this work we will use short shell solution and instead of solv-
where E is the modulus of elasticity, r is the radius of shell, and p is ing the joint equations approximately we will solve them analyti-
the pressure. Shell bending rigidity Ds is defined as cally. We will cast the shell equations for a finite length cylinder for
each shell course in a matrix form using the approach suggested by
Et3 Calladine [15]. Then we will enforce continuity and equilibrium of
Ds ¼ ð6Þ
12ð1  m2 Þ each joint in similar way to the direct stiffness or finite element
method and solve the system of equations simultaneously.
where m is the Poisson’s ratio. Eq. (5) is very similar to the beam on
Now let us consider an axisymmetric circular cylindrical shell of
elastic foundation equation [18]. Now we will define a parameter b
length Li , constant thickness of t i subject to end shearing forces, Q 1
where
and Q 2 and bending moments M 1 and M 2 . Radial displacements
Et 3ð1  m2 Þ and rotations at the end of the cylinder is given as w1 ; w2 and
b4 ¼ ¼ ð7Þ
4Ds r 2
r2 t2 h1 ; h2 . See Fig. 2 for the geometry and sign convention for displace-
ment, rotation, bending moment and shearing forces. Although
Therefore, the governing equation can be written as
there are many ways to relate displacement, rotation, bending
4 moment and shearing forces at the ends of a short cylinder we will
d w p
4
þ 4b4 w ¼ ð8Þ use the following stiffness–flexibility matrix suggested by
dx Ds
Calladine [15].
The general solution of the above fourth order ordinary differ- 2 3
2 3  2Dbb3 a d c 2 3
ential equation of constant coefficients may be written as [16] w1 s b 2Ds b3 b Q1
6 7
6M 7 16 a bð2Ds bÞ  bc dð2Ds bÞ 76 7
w ¼ ebx ðC 1 cos bx þ C 2 sin bxÞ þ ebx ðC 3 cos bx þ C 4 sin bxÞ 6 17 6 b 76 h1 7
6 7¼ 6 76 7 ð11Þ
4 w2 5 2 6  d 3  bc b a 7
 b 5 Q2 5
4
þ f ðxÞ ð9Þ 4 2Ds b 2Ds b3
M2 c
dð2Ds bÞ a
bð2Ds bÞ h2
b b
where C 1 ; . . . ; C 4 are the four integration constants to be determined
from the boundary conditions and f ðxÞ is a particular solution. In where the coefficients a; b; c; d are defined by the following
our case membrane solution due to hydrostatic pressure of the functions
stored liquid will be the particular solution and integration con-
stants C 1 ; . . . ; C 4 above will be found from the combined membrane
a ¼ ðSinh 2n  sin 2nÞ=k
and bending solution. b ¼ ðSinh 2n þ sin 2nÞ=k
ð12Þ
Note that all the other variables rotation, h, bending moment, M c ¼ 2ðCosh n sin n  Sinh n cos nÞ=k
and shearing force, Q can be expressed in terms of radial displace- d ¼ 2ðCosh n sin n þ Sinh n cos nÞ=k
ment, w by the following equations
dw
h¼ ;
dx
2
d w
M ¼ Ds 2 ; ð10Þ
dx
3
dM d w
Q¼ ¼ Ds 3 :
dx dx
Cylindrical liquid storage tanks usually have shell courses of
several different thicknesses. So for each shell course we need to
solve the governing Eq. (8) and thus, find the four integration con-
stants and particular membrane solution appeared in Eq. (9). In
general there are two ways to approach the problem [16]. First
one is to consider each shell course as semi-infinite length shell
that one end of the shell does not see the effects of the loading
at the other end. This approach is commonly named as long shell
solution. The second approach is to consider each shell course as
finite length shell that one end of the shell does see the effects of
the loading at the other end. The second approach is commonly
named as short shell solution.
Long shell solution is relatively simple to obtain because the
semi-infinite length assumption eliminates the integration con-
stants C 1 and C 2 and remaining two integration constants may
be found by construction of continuity of radial displacement w
and rotation at each shell joint (weld seam). Each joint may be
treated independently and only the information from a shell course Fig. 2. The sign convention used to develop the thin shell solution.
 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630 625

where
n ¼ bLi
ð13Þ
k ¼ Cosh2n  cos 2n
Now we can form the stiffness–flexibility matrix for each shell
course using Eq. (11) and assemble system matrix similar to finite
element method and enforce continuity of displacement and equi-
librium of forces together with membrane solution. Solution of
system matrix will give us the shearing forces Q and rotation h
at each joint (weld seam). The boundary conditions of the bottom
of the tank were set to be pinned along the circumference of the
tank. The top of the tank was free to move. Plastic moment was
applied to the bottom of the tank simulating a bottom plate of
6.35 mm (0.25-in.) thick steel with a yield stress of 250 MPa
(36,000 psi) along the circumference of the bottom shell course.
Note that solution of the governing Eq. (8) was originally
defined in terms of radial displacement w. Therefore, we need to
recover w at any point along the shell to calculate stresses. While
there are many ways to recover the integration constants appeared
in Eq. (9), we would like to mention two of them here. First method
is given for beams on elastic foundations in Hetenyi’s book and
named as ‘‘method of initial conditions’’ [18]. This approach Fig. 3. Simply supported cylindrical shell with uniform pressure.
expresses displacement function w in terms of w0 ; h0 ; M 0 , and Q 0
E
quantities at the end x ¼ 0 of the shell course. The second method rhoop ¼ w ð17Þ
to obtain the integration constants is suggested by Gould’s book on r
shells [17]. Gould’s method constructs a 4  4 coefficient matrix Before we move to the comparison studies of each tank design
that relates the integration constants to the end bending moments methods we would like show validation of our linear analysis using
and shearing forces, M 0 , and Q 0 at both ends of a shell course at thin shell theory. As we mentioned earlier our method uses exact
x ¼ 0 and x ¼ Li . We used both of the recovery methods and they stiffness–flexibility matrix and exact shape functions to describe
gave identical results as expected. Here we would like to explain the radial displacement within the usual assumptions of thin shell
Hetenyi’s approach in detail. theory. To validate this we shall use a circular cylindrical shell of
Displacement function w at any point along the shell course length Li with a uniform internal pressure, p. For both ends of the
under consideration can be written as shell pinned–pinned condition the closed form solution of radial dis-
placement is given by Timoshenko by the following expression [16]
1 1
wðxÞ ¼ w0 F 1 ðbxÞ  h0 F 2 ðbxÞ  2 M0 F 3 ðbxÞ 
b b Ds pL4i 2 sin a Sinh a
w¼ 1 sin bx Sinh bx
1 64Ds a 4 cos 2a þ Cosh 2a
 Q 0 F 4 ðbxÞ ð14Þ 
b3 Ds 2 cos aCosh a
 cos bxCosh bx ð18Þ
cos 2a þ Cosh 2a
where
Where
F 1 ðbxÞ ¼ Cosh bx cos bx
F 2 ðbxÞ ¼ 12 ðCosh bx sin bx þ Sinh bx cos bxÞ bLi
ð15Þ
a¼
1 2
F 3 ðbxÞ ¼ 2
ðSinh bx sin bxÞ
F 4 ðbxÞ ¼ 1
ðCosh bx sin bx  Sinh bx cos bxÞ In Eq. (18) origin of coordinate x is taken at the middle of the
4
cylinder. We solved the above problem using our thin shell
Note that the displacement function w in Eq. (14) is the approach for different thicknesses and obtained identical results
homogenous solution or bending solution of displacement of the as the closed form solution given in Eq. (18). Fig. 3 shows the
governing equation. We need to add membrane solution associ- geometry and problem description of the problem. Fig. 4 shows
ated with the hydrostatic liquid pressure to find the general solu- the normalized radial displacement with respect to the membrane
tion for the displacement at each shell course. Radial displacement solution using thin shell approach and closed form solution for dif-
due to membrane solution at any point under hydrostatic liquid ferent thickness. Our thin shell solution matches the closed form
pressure can be expressed by the following equation solution exactly as expected.

pr 2
wm ¼ ð16Þ 3. Comparison studies
Eti
where hydrostatic pressure p is We performed three different comparison studies to investigate
the shell design methods.
p ¼ cðhi  xÞ

where c is the weight per unit volume of the stored liquid and hi is
the height of the liquid from the bottom joint (seam) of a shell
course under consideration.
Once the general radial displacements, including both bending
and membrane deformations, are combined the circumferential
stress of the tank shell at any point can be calculated simply by
the following expression.
3.1. Study 1: Comparison of 1FM and VDM
To compare the results of 1FM with those of VDM, a spread-
sheet was generated to compute the thicknesses of shell courses
for different tank diameter and height. This comparison study
was done using allowable stresses of 158.6 MPa (23,000 psi),
206.8 MPa (30,000 psi), and 236.5 MPa (34,300 psi). We started
626 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630

1.2
t = 0.01 t=0.1 t=1
1

0.8
w/w membrane

0.6

0.4

0.2

0
0 20 40 60 80 100
% x/L
Fig. 5b. Comparison of 1FM and VDM, ratio of volume of steel resulting from 1FM
Fig. 4. Normalized radial displacement in a pinned–pinned cylinder subject to
to VDM vs tank diameter for different tank heights in meters for the allowable
uniform internal pressure. The horizontal axis is the ratio of the height to the entire
stress of 206.8 MPa (30,000 psi).
length, and the vertical axis is the normalized radial displacement w with respect to
the membrane radial displacement.

Fig. 5a. Comparison of 1FM and VDM, ratio of volume of steel resulting from 1FM Fig. 5c. Comparison of 1FM and VDM, Ratio of Volume of Steel Resulting from 1FM
to VDM vs tank diameter for different tank heights in meters for the allowable to VDM vs Tank Diameter for Different Tank Heights in Meters for the Allowable
stress of 158.6 MPa (23,000 psi). Stress of 236.5 MPa (34,300 psi).

In total, 27 tanks were tested. Four different heights were con-

with a tank height of 7.3 m (24-ft) with three 2.4 m (8-ft) shell
sidered: 12.2 m (40-ft), 14.6 m (48-ft), 17.1 m (56-ft), and 19.5 m
courses and the diameter ranging from 3 m (10-ft) to 61 m
(64-ft). The diameters were selected from API 650 Annex K table
(200-ft). Then we increased the tank height with 2.4 m (8-ft) incre-
for an allowable stress of 158.6 MPa (23,000 psi). Each shell course
ments up to a tank height of 19.5 m (64-ft) with eight 2.4 m (8-ft)
was 2.4 m (8-ft) high, and the thickness of each course was also
shell courses with the diameter ranging from 3 m (10-ft) to 61 m
determined using the same table.
(200-ft).
We calculated the actual stresses using thin shell theory and the
stress profile was then obtained. The points where the tank was
3.2. Study 2: Validation of VDM using the analytical approach when overstressed were found.
VDM is permissible to use

API 650 Annex K provides tables to summarize the required
shell thickness based on VDM for different tank sizes and allowable
stresses. In this part of the study, the tank designs specified in
Annex K for an allowable stress of 158.6 MPa (23,000 psi) were
tested with the linear analysis using thin shell theory. The linear
analysis provided the theoretical stresses in each course of the
tank. With this study we may be able to see if the designs based
on VDM are actually overstressed or not.
3.3. Study 3: Validation of VDM using the analytical approach when
VDM is not permissible to use
This part of the study was performed to compare the shell
designs based on VDM and linear analysis using the thin shell the-
ory when the L/H ratio is more than 1000/6 in SI units (i.e. VDM
method is no longer permissible). We investigated if the limiting
L/H ratio of 1000/6 is a good limit for all diameters, heights and
 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630 627

20 20 20
D = 64.0 m D = 67.1 m D = 79.2 m
15 15 15

Height (m)
Height (m)

Height (m)
10 10 10

5 5 5

0 0 0
0 50 100 150 0 50 100 150 0 50 100 150
Stress (MPa) Stress (MPa) Stress (MPa)

20 20 20
D = 85.3 m D = 91.4 m D = 97.5 m
15 15 15

Height (m)
Height (m)

Height (m)
10 10 10

5 5 5

0 0 0
0 50 100 150 0 50 100 150 0 50 100 150
Stress (MPa) Stress (MPa) Stress (MPa)

20 20 20
D = 103.6 m D = 109.7 m D = 115.8 m
15 15 15
Height (m)
Height (m)

Height (m)
10 10 10

5 5 5

0 0 0
0 50 100 150 0 50 100 150 0 50 100 150
Stress (MPa) Stress (MPa) Stress (MPa)

Fig. 6a. Circumferential stress profiles for different tank diameters for tank height of 12.2 m (40 ft) for allowable stress of 158.6 MPa (23,000 psi).

allowable stresses. This exercise was carried out by modeling tanks
of relatively smaller height and relatively large diameters using
thin shell theory with shell thicknesses calculated using the
VDM. The modeled tanks were 3.7 m (12-ft), 4.9 m (16-ft), and
6.1 m (20-ft) high, with each tank having two shell courses each
having the height of half the height of the tank. The tanks were
modeled over a range of diameters and with the thicknesses of
the shell courses calculated for the allowable stresses of
158.6 MPa (23,000 psi), 206.8 MPa (30,000 psi), and 236.5 MPa
(34,300 psi). The circumferential stress results from the analytical
solution were compared with the allowable stress.
4. Results and discussions
4.1. Study 1: Comparison of 1FM and VDM
Comparisons of the shell designs based on 1FM and those based
on VDM showed that the designs based on VDM led to more
economical designs for almost all the tank sizes. For the tank diam-
eters smaller than 15 m (50-ft) both 1FM and VDM produce similar
shell thicknesses. However, for tank diameters more than 15 m
(50-ft) design based on VDM are starting to be more economical.
Figs. 5a–5c show the results of the comparison for the allowable
stresses of 158.6 MPa (23,000 psi), 206.8 MPa (30,000 psi), and
236.5 MPa (34,300 psi), respectively. The vertical axis is the ratio
of the volume of the steel required by the 1FM to the volume of
the steel required by the VDM: V1FM/VVDM. If the ratio is higher
than 1, then the 1FM requires more volume of steel than the
VDM. If the ratio V1FM/VVDM was less than 1, then it requires less
steel. The horizontal axis represents the diameter of the tank.
There are six curves representing six different heights of tanks.
4.2. Study 2: Validation of VDM using the analytical approach when
VDM is permissible to use
The stress analysis for the API 650 tanks presented 11 out of the
27 tanks which have an overstress of more than 0.5% of the allow-
able stress.
Shown in Figs. 6a and 6b are the results of stress profiles for
tanks designed using VDM with an allowable stress of 158.6 MPa
(23,000 psi). The vertical axis represents the height from the bot-
tom of the tank in meters, and the horizontal axis represents the
circumferential stress at that height. The solid vertical line repre-
sents the allowable stress of 158.6 MPa (23,000 psi). Any point
has a stress greater than that in the tank represents an overstress
in the tank.
Fig. 6a shows the stress profile for 9 tanks with the height of
12.2 m (40-ft). Most of the overstress happens in the upper shell
courses. This is due to the upper courses approaching the L/H limit
of 1000/6. For example, the tank with the height of 12.2 m (40-ft)
and diameter of 115.8 m (380-ft) is facing an overstress of 1.4%.
The L/H ratio for the tank as a whole is 122, but if the top two
courses were considered by themselves as a separate tank the
L/H ratio becomes 186, which is greater than the 1000/6 limit.
Fig. 6b shows the stress profiles of 3 tanks with the height of
14.6 m (48-ft), 3 tanks with the height of 17.1 m (56-ft), and 3
tanks with the height of 19.5 m (64-ft). The overstress is more
prevalent in the bottom shell courses. This is due to the way
VDM is formulated. The first course’s design is formulated without
the consideration of the upper shell courses. With taller tanks, the
forces transferred from upper courses to the bottom course accu-
mulate and eventually increase the overall stress experienced by
the bottom shell course.
4.3. Study 3: Validation of VDM using the analytical approach when
VDM is not permissible to use
In this part tanks designed using VDM, when VDM is not per-
missible to use were analyzed using the linear analysis based on
thin shell theory. The study shows the ratio of maximum stress,
obtained from the solution for a tank designed using VDM in the
tank increases with the increase in the L/H ratio.
The results are presented in Figs. 7a–7c. The figures represent
three allowable stresses: 158.6 MPa (23,000 psi), 206.8 MPa
(30,000 psi), and 236.5 MPa (34,300 psi) respectively. The horizon-
tal axis shows the L/H ratio, and the vertical axis represents the
ratio of the maximum hoop stress divided by the allowable stress.
628 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630

20 20 20
D = 54.9 m D = 73.2 m D = 79.2 m
H = 14.6 m
15 15 15

Height (m)

Height (m)

Height (m)
10 10 10

5 5 5

0 0 0
0 50 100 150 0 50 100 150 0 50 100 150
Stress (MPa) Stress (MPa) Stress (MPa)

20 20 20
D = 67.1 m D = 73.2 m D = 75.3 m
H = 17.1 m

15 15 15
Height (m)

Height (m)

Height (m)
10 10 10

5 5 5

0 0 0
0 50 100 150 0 50 100 150 0 50 100 150
Stress (MPa) Stress (MPa) Stress (MPa)

D = 48.8 m D = 54.9 m D = 64.4 m

20 20 20
H = 19.5 m
Height (m)

Height (m)

Height (m)
15 15 15
10 10 10
5 5 5
0 0 0
0 50 100 150 0 50 100 150 0 50 100 150
Stress (MPa) Stress (MPa) Stress (MPa)

Fig. 6b. Circumferential stress profiles for different tank diameters and heights for allowable stress of 158.6 MPa (23,000 psi).

1.1 1.1

1.05 1.05

1 1
exact/ allowbale

exact/ allowbale

0.95 0.95

0.9 L/H=1000/6 0.9 L/H=1000/6

0.85 0.85

0.8 0.8
160 180 200 220 240 160 180 200 220 240
L/H L/H
H = 3.7 m H = 4.9 m H = 6.1 m H = 3.7 m H = 4.9 m H = 6.1 m
Fig. 7a. The ratio of maximum circumferential stress obtained from linear analysis
Fig. 7b. The ratio of maximum circumferential stress obtained from linear analysis
to allowable stress of 158.6 MPa (23,000 psi) vs. L/H (the shell thickness is
to allowable stress of 206.8 MPa (30,000 psi) vs. L/H (the shell thickness is
determined using VDM).
determined using VDM).

The points of interest are those at which the plots cross with unity 5. Conclusions
in the vertical axis.
Based on the results of our study, The L/H limit ratio of 1000/6 We developed a stiffness–flexibility method based on thin shell
was found to be reasonably conservative and safe to use. The low- theory that gives the theoretical displacements and stresses at
est L/H ratio found using our calculations was for the 6.1 m (20-ft) each shell course without any approximation or simplification.
high tank, with a diameter of 152.4 m (500-ft). The L/H ratio at Based on the results of our studies, shell designs based on VDM
which the allowable stress is equal to the maximum experienced may become up to 4% more economical than those of the 1FM: The
stress was established through interpolation. The L/H value found larger the tank size either in terms of the diameter or the height,
was 183. This value represents a tank with a diameter between the greater the advantage of the VDM. For the tank diameters
152.4 m (500-ft) and 167.6 m (550-ft) with a height of 6.1 m smaller than 15 m (50-ft) the designs based on 1FM and VDM
(20-ft). are very close. Therefore, for the tank diameters smaller than
 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630 629

sffiffiffiffiffiffiffi! 
1.1
0:0696D HG 4:9HDG
t1d ¼ 1:06  þ CA ðA1Þ
H Sd Sd
1.05
sffiffiffiffi! 
0:0696D H 4:9HD
t1t ¼ 1:06  ðA2Þ
1 H St St
exact/ allowbale

Note that the equations above are unit dependent and the equa-
0.95 tions in this work are given in SI units. If the calculated first shell
course thickness using Eqs. (A1) and (A2) is more than the thick-
ness obtained using the traditional 1FM, the thickness calculated
0.9 L/H=1000/6
from 1FM may be used as the first shell course thickness.
The shell thickness for the upper shell courses are calculated the
0.85 following way:
The location of the maximum hoop stress is found first using
the lowest of the values
0.8
160 180 200 220 240
x1 ¼ 0:61ðrt u Þ0:5 þ 320CH
L/H
H = 3.7 m H = 4.9 m H = 6.1 m x2 ¼ 1000CH

Fig. 7c. The ratio of maximum circumferential stress obtained from linear analysis
to allowable stress of 236.5 MPa (34,300 psi) vs. L/H (the shell thickness is x3 ¼ 1:22ðrt u Þ0:5
determined using VDM).
With C ¼ ðK 0:5 ðK  1ÞÞ=ð1 þ K 1:5 Þ and K ¼ tL =t u
The thickness of the course is then found using the greater of:
15 m (50-ft) there is no need to use VDM. Instead 1FM should be x
4:9DðH  1000 ÞG
used. tdx ¼ þ CA
Sd
Comparing shell designs based on linear analysis with those
based on VDM, it is found that the linear analysis does not lead x
4:9DðH  1000 Þ
to more economical designs than VDM most of the time. ttx ¼
St
However, results show that shell designs using VDM may produce
overstressed shell courses for some of the large steel liquid storage Second course thickness:
tanks when VDM is permissible to use. Linear analysis would give
more accurate shell designs for those cases. Considerations should h1
6 1:375 t2 ¼ t1
ðrt1 Þ0:5
be given to use linear analysis for larger thanks even if VDM is per-  
h1 h1
missible to use. 1:375 6 < 2:625 t2 ¼ t 2a þ ðt1  t2a Þ 2:1 
ðrt 1 Þ0:5 1:25ðrt1 Þ0:5
If L/H ratio is larger than 1000/6 in SI units VDM shall not be 2:625 6 h1 t2 ¼ t 2a
ðrt 1 Þ0:5
used based on the rules of API 650. Our linear analysis results also
support this limitation on the use on VDM.
Our studies have shown that to obtain economical results for
shell thickness depending on the diameter and L/H ratio one should
use:

References
D 6 15 m ð50 ftÞ D > 15 m ð50 ftÞ and L
H > 1000=6
L=H 6 1000=6 [1] API Standard 650. Welded tanks for oil storage. 12th ed. Washington,
One-foot Variable-design-point Linear DC: American Petroleum Institute; 2013.
[2] DiGrado BD, Thorp GA. The aboveground steel storage tank handbook. Van
method method and linear analysis
Nostrand Reinhold; 1995.
analysis [3] Myers PE. Aboveground storage tanks. McGraw-Hill; 1997.
[4] Zingoni A. Liquid-containment shells of revolution: a review of recent studies
on strength, stability and dynamics. Thin-Walled Struct 2015;87:102–14.
http://dx.doi.org/10.1016/j.tws.2014.10.016.
[5] Jaca RC, Godoy LA, Flores FG, Croll JG. A reduced stiffness approach for the
Acknowledgments buckling of open cylindrical tanks under wind loads. Thin-Walled Struct
2007;45(9):727–36. http://dx.doi.org/10.1016/j.tws.2007.07.001.
[6] Uematsu Y, Koo C, Yasunaga J. Design wind force coefficients for open-topped
Research by the first author was supported by the Lynn oil storage tanks focusing on the wind-induced buckling. J Wind Eng Indust
Fellowship and Purdue University. Research by the second author Aerodyn 2014;130:16–29. http://dx.doi.org/10.1016/j.jweia.2014.03.015.
[7] Chen L, Rotter JM, Doerich C. Buckling of cylindrical shells with stepwise
was supported by Purdue University.
variable wall thickness under uniform external pressure. Eng Struct
2011;33(12):3570–8. http://dx.doi.org/10.1016/j.engstruct.2011.07.021.
[8] Zhao Y, Lin Y, Shen YB. Wind loads on large cylindrical open-topped tanks in
Appendix A group. Thin-Walled Struct 2014;78:108–20. http://dx.doi.org/10.1016/
j.tws.2014.01.002.
For completeness the VDM calculation procedure are given here [9] Nachtigall I, Gebbeken N, Urrutia-Galicia JL. On the analysis of vertical circular
cylindrical tanks under earthquake excitation at its base. Eng Struct
in SI system of units. The reader is referred to the API 650 for the
2003;25(2):201–13. http://dx.doi.org/10.1016/S0141-0296(02)00135-9.
further details of the VDM and equations in USC system of units [10] Virella JC, Godoy LA, Suárez LE. Fundamental modes of tank-liquid systems
[1]. In addition, Annex K of API 650 provides example calculations under horizontal motions. Eng Struct 2006;28(10):1450–61. http://dx.doi.org/
for VDM. 10.1016/j.engstruct.2005.12.016.
[11] Matsui T. Sloshing in a cylindrical liquid storage tank with a floating roof
For the first shell course from the bottom the greater of the two under seismic excitation. J Pressure Vessel Technol 2007;129(4):557–66.
thicknesses obtained using following equations. http://dx.doi.org/10.1115/1.2767333.
630 E. Azzuni, S. Guzey / Engineering Structures 101 (2015) 621–630

[12] Chen ZP, Duan YY, Shen JM, Jiang JL. A simplified method for calculating the
stress of a large storage tank wall. Proc Inst Mech Eng, Part E: J Process Mech
Eng 2007;221(3):119–27. http://dx.doi.org/10.1243/09544089JPME125.
[13] Zick LP, McGrath RV. Design of large diameter cylindrical shells. Proc Division
Refining, Am Petrol Inst New York 1968;48:1114–40.
[14] Buzek J. Hoop forces in cylindrical liquid storage tanks. Correspondence-API
Subcommittee on Pressure Vessels & Tanks; 1979.
[15] Calladine CR. Theory of shell structures. Cambridge University Press; 1983.
[16] Timoshenko S, Woinowsky-Krieger S. Theory of plates and shells. 2nd
ed. McGraw-Hill; 1959.
[17] Gould PL. Analysis of shells and plates. Prentice Hall; 1999.
[18] Hetenyi M. Beams on elastic foundation; theory with applications in the fields
of civil and mechanical engineering. University of Michigan Press; 1946.