International Journal of Pressure Vessels and Piping 77 (2000) 511±517

www.elsevier.com/locate/ijpvp

Comparison of design methods for a tank-bottom annular plate and

concrete ringwall
T.Y. Wu, G.R. Liu
Department of Mechanical and Production Engineering, The National University of Singapore, 10 Kent Ridge Crescent, Singapore, Singapore 119260
Received 7 March 2000; revised 27 June 2000; accepted 3 August 2000

Abstract
This paper studies three design methods for a bottom annular plate of large storage tanks and the design of the concrete ring wall. It is
shown that the annular-plate width, thickness, and the width that will project beyond the outer surface of the shell can be calculated
analytically. The design of the annular bottom plate and the concrete ringwall is discussed and the need for an accurate representation
for the interaction between the bottom plate and the foundation is emphasised. q 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Storage tanks; Tank-bottom; Tank foundation; Concrete ringwall; Annular plate

1. Introduction ®rst shell course in API 650. The third is a minimum width
that projects beyond the outside edge of the weld attaching
Risks of storage tank failure and environmental contam- the bottom to the shell plate or beyond the outer surface of
ination are just two of the problems facing oil storage. As a the shell plates. It is 25 mm (1 in.) in API 650 and 50 mm in
result, more attention should be given to tank design and BS2654 and SH3046-92. The one-foot method or variable-
operation standards in addition to stringent safety controls, design-point method can be used to calculate the shell thick-
since the majority of oil and its products are stored in large ness. However, no code or standard in any of the countries
capacity tanks. Many large above ground storage tanks have listed uses equations to calculate the stresses of the bottom
been built with vertical cylindrical shells and ¯at bottoms annular plate to determine the above-mentioned three
that rest directly on simply prepared subgrades and are dimensions. This paper sets out a discussion and recommen-
supported on a ringwall beneath the tank shells. Large, verti- dation for the design equations of the bottom annular plates.
cal, cylindrical, single-, and ¯at-bottomed, unanchored, The designer of the concrete ringwall must ®rst obtain the
welded steel storage tanks under hydrostatic loads are data on superimposed loadings to be used for the ringwall
considered, as shown in Fig. 1. Design calculation methods design. These data are not provided in the current standards.
for the tank-bottom annular plate are discussed. This plate is The forces applied on the ringwall by the tank body are
normally designed empirically, there being no recognized obtained if the ¯exion mechanism of the annular plate is
code, standard or common practice for design. clearly known.
The way in which the bottom plate and the lowest tier of
the tank shell interrelate is important in relation to the
design of the bottom annular plate and the concrete ring 2. Three design methods for the tank-bottom annular
wall. The tank design standards for various countries [1± plate
5] only stipulate three dimensions for the annular plate. The
The American Petroleum Institute published a stress
®rst is a minimum nominal width of the annular plates, such
calculation method for the annular plate in 1968 [6].
as 1.8 m in Chinese Standard SH3046-92, 1.8 m (72 in.) in
Later, the Academy of Sciences of China (ASC) produced
API 650, or 500 mm in British Standard BS2654. The
an alternative approach in 1979 [7]. Recently, Wu [8]
second is a minimum nominal thickness, which is deter-
proposed a more accurate method. They all considered a
mined by the nominal plate thickness of the ®rst shell
vertical cylindrical tank with a ¯at bottom that rests directly
course, and also related to the hydrostatic test stress in the
on simply prepared soils and is supported on a concrete
ringwall.
E-mail address: [email protected] (T.Y. Wu). The stress calculation of the tank-bottom annular plate
0308-0161/00/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0308-016 1(00)00055-7
512 T.Y. Wu, G.R. Liu / International Journal of Pressure Vessels and Piping 77 (2000) 511±517

Nomenclature
A1 ; A2 ; A3 ; A4 coef®cients
C distance from shell plate centre to outer edge of bottom annular plate
D shell diameter
Db Et13 =‰12…1 2 y 2 †Š
Ds Ets3 =‰12…1 2 y 2 †Š
E steel elasticity modulus, 206 GPa
G tank shell weight per length of circumference
H tank water-®lling height
Kb elastic coef®cient of the tank's ¯exible foundation
Ks Ets =R2
L the design width of the annular plate
M0 bending moment at the tank shell-bottom junction
M2 bending moment of annular plate at one speci®c point
P tank-bottom water pressure, P ˆ H g
Q0 shear force at the tank shell-bottom
q2 supporting force per square metre in ASC method
R shell radius
R1 supporting force of annular plate at bottom plate periphery
R2 shear force of annular plate at one speci®c point
S distance for which the annular plate lifts clear of the foundation in API and Wu methods
ts shell wall thickness of the lowest tier
t1 bottom annular plate thickness
t2 bottom inner-plate thickness
w horizontal
p4  displacement of the shell-bottom junction
bb p4 K b =4Db

bs Ks =4Ds
ub rotation angle of the bottom annular plate at the shell-bottom junction
us rotation angle of the shell plate at the shell-bottom junction
g speci®c weight of water, 9.81 kN/m 3
ss minimum yield stress of the bottom annular plate
s max theoretical maximum bending stress of the bottom annular plate
s 20 annular-plate bending stress at the point 20 mm from the shell's inner wall inward of the tank centre
y Poisson's ratio, 0.3

requires a proper description of the interaction between the This simpli®cation is accurate from an engineering view-
bottom and the foundation, as shown in Fig. 1. The bottom point [9]. The compatibility conditions at the tank shell-
of a storage tank is a large steel membrane constrained by bottom junction are the continuities of the respective displa-
the base and the shell. As a product is added to the tank, the cement and rotation angle. The three methods are simply
weight of the product presses the bottom inner plates into presented below.
complete contact with the foundation. Therefore, the inner
bottom plates can be assumed as non-stressed because of 2.1. The API method [6]
uniform support from the foundation. Only a minimum
nominal thickness, exclusive of any corrosion allowance, As shown in Fig. 3, a small part of the annular plate lifts
is speci®ed for the inner bottom plates in many codes or clear of the foundation for some distance S. The bottom
standards. Since the concrete ringwall cannot ¯ex down- plate ¯exural bending is of nonlinear nature with a changing
ward and the shell applies a large bending moment to the boundary condition, since a different hydrostatic load will
bottom plate at the shell-bottom junction, the only possible cause a changing uplift distance. The outer edge of the
movement of the bottom annular plate is to ¯ex upward, i.e. annular plate is simply supported with a zero vertical displa-
to lift it off the concrete ringwall. The uplift hypothesis for cement. The compatibility conditions are used at the shell-
the bottom annular plate is reasonable and adopted by all the bottom attachment, where the shell horizontal displacement
three calculation methods mentioned above. To simplify the is zero. The boundary conditions at the inner support point
calculation, the ®rst shell plate tier is represented as a long are zero for both vertical displacement and rotation angle,
half cylinder under hydrostatic pressure, as shown in Fig. 2. assuming the ringwall has suf®cient width to support the
 T.Y. Wu, G.R. Liu / International Journal of Pressure Vessels and Piping 77 (2000) 511±517 513

Centre of the Therefore, the stresses of the annular plate can be calculated
ringwall and shell analytically. It should be noted that the relationship between
Tank shell M0 and ub is nonlinear rather than linear as taken in Ref. [6].

Annular plate Inner plate

2.2. The ASC method [7]

The width of the concrete ringwall for many large tanks is

relatively narrow, compared with the distance S obtained
Ground
level using the API method. Therefore, the inner support point
Compacted in Fig. 3 cannot satisfy its boundary conditions Ð zero
clean sand vertical displacement and zero rotation angle. The inner
point falls on the elastic foundation adjacent to the ringwall
Coarse
gravel or in most cases. As shown in Fig. 4, the ASC method consid-
crushed ers the effect of the inner ¯exible foundation. It assumes the
stone Compacted same distance, which falls on the ¯exible foundation and
refill
bears a triangular reaction force, as the uplift distance.
However, it does not consider the section that projects
beyond the outer surface of the shell plates. The annular
plate is simply supported at the point just beneath the
Fig. 1. The tank shell-bottom junction and foundation with a concrete shell and thus has a zero vertical displacement therein.
ringwall. Another boundary or compatibility condition at the shell-
bottom attachment is that the ®rst shell plate and the annular
annular plate. The calculation formulae are as follows: plate have the same rotation angle therein. The rotation
angle at the most inner support point is zero, and the vertical
‰GC 1 0:25P…S2 2 C 2 †Š…C 2 2 S2 † displacement is q2 =Kb . The calculation formulae are as
M0 ˆ …1†
…S2 2 3C 2 † follows:
  !
‰0:5P…S2 2 C2 † 1 GC 1 M0 Š 11 ts 3 3 1 2b2s P
R2 ˆ …2† PS 1 2g
S 240 t1 4b4s 2bs 1 …1 2 y†t s =…Rt2 †
M0 ˆ   ;
1 1 17S ts 3
g…H bs 2 1† M0 2 2
us ˆ 1 ; …3† 2bs 1 …1 2 y†t s =…Rt2 † bs 40 t1
Ks 2 bs D s
…5†
" #
1 R2 …S 2 C†2 P…S 2 C†3  
ub ˆ 2 : …4† 1 47 960Db
Db 2 6 M0 ˆ 2 PS2 ; …6†
…1920Db =Kb S4 † 2 49 6 Kb S4
Assuming a value of S, then M0 ; R2 , us and ub are calculated  
from the above equations, respectively. When a certain S 12 2M0
q2 ˆ P1 2 : …7†
results in the same us and ub ; this S corresponds to this 5 S
combination of conditions, along with M0 ; R2 , us and ub :
The solution procedures are the same as above. A certain
x value of S results in an identical M0 in Eqs. (5) and (6).

2.3. The Wu method [8]

In view of the special use of the concrete ringwall, Wu [8]

proposed a new method to account for the effect of the
elastic foundation. As shown in Fig. 5, Wu's method also
(H-x)γ considers an uplift part with a length of S, as in the API
method. However, the inside part of the annular plate
toward the tank centre is considered as a semi-in®nite
plate ¯exibly supported by the foundation. The reaction
force of the subgrade is linear with the annular plate vertical
y displacement, while the elastic coef®cient of the subgrade
Q0 adjacent to the ringwall is Kb. The shell horizontal displace-
M0
ment is zero at the shell-bottom junction. The calculation
Fig. 2. Shell forces. formulae have been obtained as follows:
514 T.Y. Wu, G.R. Liu / International Journal of Pressure Vessels and Piping 77 (2000) 511±517

Tank wall centre the annular plate can be obtained according to Fig. 5.
G P=Hγ M ˆ R1 x …x , C†; …12a†
M0
M ˆ R2 …S 2 x† 2 0:5P…S 2 x†2 2 M2 …C , x , S†;
…12b†
R1 R2
C
S
M ˆ 2e2bb x {M2 ‰sin…bb x† 1 cos…bb x†Š 1 R2 sin…bb x†=bb }

…x . S†: (12c)
Fig. 3. A force distribution of the API model for the tank-bottom plate.

0:25P…2A1 A4 2 A2 A3 † 1 GC…A2 C 2 2 A1 †
M0 ˆ ; …8†
A1 2 3A2 C 2 3. Discussion

" 3.1. Theoretical comparison of the three methods

1 R2 P
ub ˆ …S 2 C†2 2 …S 2 C†3
Db 2 6 With the exception of tanks founded on solid rock, hard-
# core, or similar sub-bases, some settlement is bound to take
1 place. Large, and perhaps even moderate, irregularities in
2M2 …S 2 C† 2 …R2 1 2bb M2 † ; …9†
2b2b settlement may lead to serious distortion of the bottom plate
and other important elements of the tank. Due to the rela-
0:5PA4 2 GC 2 M0 tively small width of the concrete ringwall for many large
M2 ˆ ; …10† tanks, the inner support point in Fig. 3 falls on the inner
A2
elastic foundation adjacent to the ringwall. Therefore, the
0:5P…S2 2 C2 † 1 GC 1 M2 1 M0 inner point cannot satisfy its boundary conditions Ð zero
R2 ˆ : …11† vertical displacement and zero rotation angle. The API
S
method is not suitable for thin ringwall cases. However,
where many crushed stone or gravel ringwalls have suf®cient
3S width to support the annular plate and meet the required
A1 ˆ 3S2 1 bb S3 1 ; boundary conditions.
bb
The ASC calculation is simpli®ed in terms of the effects
of the ¯exible foundation and does not consider the annular
A2 ˆ 1 1 bb S; plate's protruding length. In the ASC formulae, the bottom
plate's horizontal displacement is considered as [7]
…1 2 y†RQ 0
2S3 6S wˆ : …13†
4
A3 ˆ C 2 S 2 4
1 3; Et2
bb bb
This equation means that the bottom plate is regarded as a
circular plate with the bottom inner-plate thickness and with
S a load of Q0 acting at its periphery. This equation does not
A 4 ˆ C 2 2 S2 2 : consider the friction force applied by the elastic foundation.
bb
Due to the hydrostatic pressure, the bottom inner plate may
The shell rotation angle is calculated using Eq. (3). The not have a horizontal displacement due to the friction force.
solution procedure is similar to the above. A certain value Indeed, the formulae for the API and Wu methods employ
of S results in the same us and ub . The bending moment of zero horizontal displacement therein. However, there is

Tank wall centre

P=Hγ

M0

q2
R1

S/2 S/2

Fig. 4. A force distribution of the ASC model for the tank-bottom plate.
 T.Y. Wu, G.R. Liu / International Journal of Pressure Vessels and Piping 77 (2000) 511±517 515

Tank wall centre

x P=Hγ
G P=Hγ
M2 M2
M0

x
R2 R2 Ù ∞
R1
C
S

Fig. 5. A force distribution of the Wu model for the tank-bottom plate.

little in¯uence on the ®nal results whether this displacement of the 100 000-m 3 tank and is representative of the bending
is taken as zero or given by Eq. (13), as veri®ed in Refs. stresses using the Wu method. The other two methods have
[8,9]. a similar distribution. If the API method is used, there is no
If the ringwall or subgrade has suf®cient width to support bending stress at the distance S and beyond. The ASC
the annular plate in such a way that the annular plate can be method does not consider the annular-plate part projecting
completely supported by a rigid ¯at surface, the elastic beyond the shell surface. From Fig. 6, it is clearly seen that
coef®cient Kb can be taken as large, for example 100 the annular-plate stresses reach a maximum value at the
times the normal value. Theoretically, the API method can inner point of the shell-bottom junction and then become
be obtained using the Wu method when Kb tends to in®nity. low and die out rapidly. The bending stresses at a distance of
When a ¯at-bottom tank is designed without anchoring 20 mm to the inner surface of the shell wall were measured.
the shell to a counterbalancing weight, the bottom is They are 520 MPa for the 50 000 m 3 tank [7] and 430 MPa
designed to carry all the weight and pressure forces distrib- for the 20 000 m 3 tank. The 100 000 m 3 tank value is
uted on the bottom and to transfer the uplift forces from the 390.7 MPa, taken as an average of 416, 403.5, 389 MPa
sidewall through the bottom plates. This case, as required in measured at one position and 387.5, 375.5, and 370.1 MPa
section 3.11.3 of API standard 620, is also in the discussion measured at another position [10]. Using the raw data in
scope, if the shell is cylindrical. Table 1, we obtained the theoretical results shown in
Table 2, where the errors are calculated using the annular-
3.2. Theoretical calculations and experimental data plates' theoretical and experimental bending stresses at a
analysis distance of 20 mm to the inner surface of the shell. The
maximum annular plate bending stresses are between
To calculate the stresses of the tank-bottom annular plate yield and twice yield stress for the 50 000 and 20 000 m 3
using either the ASC method or the Wu method, the elastic tanks.
coef®cient of the ¯exible foundation adjacent to the ringwall It is clearly shown from Table 2 and Fig. 6 that, for the
is required. The value Kb ˆ 39:24MN=m3 is long established 100 000 m 3 tank, the Wu method predicts more accurately
in engineering practice in China and is further discussed the maximum annular plate stresses than the other two
below. In China, three tanks of representative volumes methods. This is probably because the whole settlement of
20 000, 50 000, and 100 000 m 3 were strain-gauged during the subgrade and the ringwall is small, only a few centi-
the water-®lling test, when they were ®rst constructed. This metres. Therefore, the ¯exible foundation has a good
paper examines these ®eld results to discuss the three support to the bottom plate, and the empirical value Kb
proposed methods. applies to the calculations. The empirical Kb is not measured
Fig. 6 shows the annular plate bending stress distribution adjacent to the ringwall, but is an average value

Fig. 6. Bending stresses of the bottom annular plate for 100 000 m 3 tank.
516 T.Y. Wu, G.R. Liu / International Journal of Pressure Vessels and Piping 77 (2000) 511±517

Table 1
Data for three test storage tanks

Nominal volume (km 3) D (m) H (m) C (mm) L (m) ts (mm) t1 (mm) t2 (mm) G (kN/m) s s (Mpa)

20 40 15 60 0.8 24 9 6 20 240
50 60 18.04 60 0.8 32 12 9 26.3 350
100 80 20.22 116.25 2.4 32.5 21 12 29.7 500

Table 2
Theoretical calculation results for three test storage tanks

Tank (km 3) Method C (mm) S (mm) M0 (kN) R1 (kN/m) R1 =G M2 (kN) R2 (kN/m) s max (MPa) s 20 (MPa) Error (%)

20 Wu 60 195.1 28.318 58.76 2.94 0.899 218.88 355.0 268.7 237.5

Wu 75.58 197.8 28.987 57.85 2.89 1.088 219.86 341.2 257.7 /
API 60 404.1 27.410 56.93 2.85 / 13.71 295.9 213.9 250.2
ASC / 605.6 26.541 48.61 2.43 / / 484.5 380.4 11.5

50 Wu 60 293.3 216.57 91.49 3.48 0.675 224.01 461.7 368.6 229.1

Wu 113.8 320.9 220.41 88.81 3.38 1.141 225.95 429.3 340.3 /
API 60 554.1 215.33 90.02 3.42 / 23.52 413.9 323.1 237.9
ASC / 778.9 213.59 75.95 2.89 / / 566.2 463.9 10.8
100 Wu 116.25 521.1 248.91 147.4 4.96 0.407 237.35 432.4 376.1 23.73
Wu 188.22 522.8 254.01 138.9 4.68 2.422 241.86 378.8 327.1 /
API 116.25 928.6 247.04 147.1 4.95 / 43.71 407.3 351.2 210.1
ASC / 1234. 239.93 136.7 4.60 / / 549.5 483.0 23.8

accumulated over a considerable area and in various the ringwall would not support the annular plate as required.
subgrade Kb measurements. Annular plate stresses for the The elastic coef®cient Kb would have been considerably
20 000 and 50 000 m 3 tanks are discussed in the section reduced to cause larger stresses in the annular plate. There-
below. fore, it is a gross mistake or misconception that the concrete

3.3. Excessive differential settlement G

It is necessary to have the elastic coef®cient of the inner

R1
¯exible foundation to apply the ASC and Wu methods. P=Hγ
Average load-bearing capacity or elastic coef®cient over a Concrete
considerable or whole area is usually used in the engineer- ring
ing design, but it may be deceptive. Its value near the ring- wall
wall may still be misleading. The reason is that these values Bottom
are obtained through directly testing the foundation. In this annular
plate
way, we have ®rst assumed that the ringwall and its inner
foundation have the same or uniform settlement. However,
this is unlikely to be true, especially in large settlement
cases.
The annular-plate stresses obtained using the Wu method Lateral
are larger than those using the API method, since a weak pressure
support is expected from the ¯exible foundation. As shown
in Table 2, the experimental annular plate stresses for the
20 000 and 50 000 m 3 tanks are much larger than those
predicted by the Wu and API methods. This is most
probably caused by the excessive differential settlement
local to the inner wall of the ringwall, because the ®eld
settlement measurements show that the inner elastic founda-
Bottom pressure
tion has much more settlement than the ringwall and the
total settlement of the whole foundation is more than one
metre. In this case, the inner elastic foundation adjacent to Fig. 7. Typical soil pressure distributions on the concrete ringwall.
 T.Y. Wu, G.R. Liu / International Journal of Pressure Vessels and Piping 77 (2000) 511±517 517

ringwall should be constructed as strong as possible, stresses would have been smaller as predicted by the Wu
because this causes excessive differential settlement, method.
makes the annular-plate seemingly hanging, and results in 2. The annular-plate's bending stresses have a very local
even much larger annular-plate stresses. characteristic. For example, they are very small beyond
the width 500 mm for the 100 000 m 3 tank.
3.4. Concrete ringwall design 3. For the Wu and API methods, the wider the width that
will project beyond the outer surface of the shell is, the
It is a long-standing practice that the ringwall centre coin-
smaller is the maximum stress in the annular plates.
cides with the tank shell centre. However, the point where
4. The point where the annular plate applies loadings to the
the annular plate applies loadings to the ringwall is the most
concrete ringwall is the peripheral point of the annular
peripheral point of the annular plates, and it is not the tank
plate, and it is not the tank shell centre. Therefore, it is
shell centre, as shown in Fig. 7. The soil pressure distribu-
not justi®ed that the concrete ringwall centre conforms to
tions on the concrete ringwall were measured [11]. It is
the tank shell centre.
typical of Fig. 7 to re¯ect this pressure distribution. It is
5. The ringwall loading and its application point should be
clearly shown that the ringwall bottom pressure at the
considered according to the API or Wu method.
outside is larger than that at the inside. This implies that
the annular plate applies load to the ringwall from the most
peripheral point of the annular plates. This implicitly proves
that the annular plate's protruding length is at work and it is References
improper not to consider this width in the ASC method. Due
to a very narrow width of the ringwall, this force eccentri- [1] API Standard 650: Welded steel tanks for oil storage. The tenth
city will produce considerable torsion in the ringwall. edition, November 1998, American Petroleum Institute, Washington,
Therefore, it is not justi®ed that the ringwall centre DC.
[2] API Standard 653: Tank inspection, repair, alteration, and reconstruc-
conforms to the tank shell centre.
tion. The second edition, December 1995, American Petroleum Insti-
The present design load of the concrete ring wall is taken tute, Washington, DC.
as the total weight of the tank shell and top, and it is G ˆ [3] API Standard 620: Design and construction of large, welded, low-
29:724 kN=m in the 100 000 m 3 tank. The actual force pressure storage tanks. The ninth edition, February 1996, American
applied to the concrete ring wall is R1 ˆ 149:8 kN=m, Petroleum Institute, Washington, DC.
[4] BS 2654: Speci®cation for manufacture of vertical steel welded non-
which is about ®ve times the present design load G. It is
refrigerated storage tanks with butt-welded shells for the petroleum
apparent that the ringwall loading and its application point industry. The third edition, 1989.
should be considered in their real circumstances. [5] China Standard SH3046-92: Petro-chemical design speci®cation for
vertical cylindrical steel welded storage tanks. Petro-Chemical Indus-
try Press, Beijing 1992 (in Chinese).
4. Conclusion [6] Denham JB, Russell J, Wills CMR. How to design a 600 000-Bbl.
tank. Hydrocarbon Processing 1968;47(5):137±42.
The design of the annular-plate width, thickness, and the [7] Li G-C. Stress analysis of the stepped shell wall and bottom plate
width that will project beyond the outer surface of the tank of large cylindrical storage tanks. Mechanics and Practice
1979;1(4):38±41.
shell is discussed, along with the design of the concrete [8] Wu TY. More accurate method devised for tank-bottom annular plate
ringwall. A few remarks are listed as a conclusion to this design. Oil Gas J 1996;94(21):81±3.
work: [9] Wu TY. A new stress analysis method for petroleum storage tanks
and its calculation veri®cation. Petro-Chemical Equipment
1. The excessive differential settlement local to the inner 1997;26(5):15±19 (in Chinese).
wall of the ringwall should be avoided in order to reduce [10] Chen D-F, Li X-T. Stress gauging and analysis of 10 £ 10 4 m 3 ¯oat-
ing-roof tanks. Oil Gas Storage Transp 1988;7(6):29±36 (in Chinese).
the stresses in the bottom annular plates. If the excessive [11] Wu TY. Stress analysis of storage tanks with their bottom annular
differential settlements in the 20 000 and 50 000 m 3 plates and center plates lap-welded. Oil Gas Storage Transp
tanks had not been that large, the annular plate bending 1997;16(8):13±18 (in Chinese).