You are on page 1of 5

Proceedings of the 9th Biennial ASME Conference on Engineering Systems Design and Analysis

ESDA2008
July 7-9, 2008, Haifa, Israel

ESDA2008-59272
SEISMIC ANALYSIS OF A LARGE OIL TANK WITH FLOATING ROOF
Alexander V. Kultsep
Alexey M. Berkovsky
CKTI-Vibroseism Co. Ltd., St. Petersburg, Russian Federation
Phone: +7 812 3278599, Fax: +7 812 3278599
E-mail: akultsep@cvs.spb.su

ABSTRACT
Seismic qualification of the large oil tanks requires
consideration of a lot of specific failure modes. One of them is
the failure induced by dynamic behaviour of the floating roofs
or pontoons: a collision between floating steel roof and tank
wall during an earthquake can lead to the post crash fire with
severe system fault. Seismic behaviour of a 50000 m3 tank with
floating pontoon has been investigated in a numeric study.
Seismic safety limits of the considered tank including floating
roof movement are presented. A validation study using
numerical experiments of the tank with floating pontoon has
been performed in order to verify the analytical approach. An
influence of the tank anchorage on the tank seismic behaviour
and specific failure modes connected with tank bottom uplift
has been also investigated.
1

INTRODUCTION

Design of the big Oil Reservoirs in Russia is mostly


regulated by the National Building Codes and Standards.
Established several decades ago these Standards still provide a
reliable and safe design of structures for Normal Operation
Conditions. However, fast development of Russian Oil Industry
in recent years has challenged new requirements for Seismic
Qualification of these structures. Existing in Russian
engineering practice normative documents sometime could not
provide all sided guidance for Seismic Qualification. Due to
this reason, for consideration of some specific Reservoirs
modes of failure, it is necessary to use an accumulated
domestic and international experience of seismic design.
Special care should be given to the checking reservoirs against
overturning and failure during the seismic event. The beginning
of the overturning of the cylindrical free supported on the base
reservoirs is connected with a specific mode of deformation uplifting of the part of the bottom above the base. In this case,
dangerous stresses in the connection joint between tank's
bottom and wall can lead to the structural failure. To prevent
this phenomena in the regions with high seismicity an anchor

fastenings of the reservoir wall to the basement are required by


normative documents. Nevertheless, for conditions of moderate
seismic loading, it proves to be possible to exclude the anchor
fastenings, allowing the uplift of the bottom above the base. In
this case it is possible to remove from the reservoir wall the
anchor supports, that create stress concentration, and on the
other hand somewhat reduce the overturning moment,
transferred to the base of reservoir.
Another problem of the oil tanks is the seismic interaction of
the floating roofs or pontoons with tank's inside structure.
There are no clear requirements regarding floating roofs
available in domestic or international rules. However, a
collision between floating structure and reservoirs wall can
produce sparks, followed by flammable vapor ignition resulting
in severe tank fire damages as observed after the Kocaeli
earthquake and other seismic events [1,2,3].
The presented positions and the proposed calculation procedure
are demonstrated based on the example of a standard reservoir
with the volume of 50000 m3.
NOMENCLATURE
FZ
vertical reaction;
horizontal reaction;
Fh

friction coefficient
overturning moment about X axis
MXX
overturning moment about Y axis
MYY
pontoon horizontal movement
dh
pontoon vertical movement
dv
pontoon to reservoir wall relative movement
dr
w
seismic wave height of the product
p
angle of plastic hinge rotation
I
moment of inertia for
NOC
Normal Operation Conditions
DBE
Design Basis Earthquake
ZPA
Zero Period Acceleration

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Copyright 2008 by ASME

METHOD OF ANALYSIS

2.1 General information


The finite element method is used for calculation of
reservoirs response. The tank structure and floating roof is
modeled using shell and beam elements. The outer ring of
reservoirs bottom is connected with basement using set of nonlinear spring elements that simulate contact boundary
conditions during bottom's uplift. For modeling of overall
basement (soil) stiffness a conventional spring elements were
also included in the model. Potential formulation is used for
modeling of liquid sloshing effects [9]. Since the potential
formulation has restricted possibilities to consider floating body
dynamics, an elastic foundation layer was added to the pontoon
elements in order to model floating body buoyancy.
Assumption of the small structure displacements considered in
the liquid potential formulation is obviously not valid in the
gap between the pontoon and the reservoir wall because if the
gap is going to close, the fluid pressure will be calculated for
non-deformed model geometry. However, it is assumed that the
calculated gap pressure is lower compared to the real pressures.
Therefore the calculated movement of the floating roof should
be larger, i.e. more conservative.
Because of the nonlinear nature of system (liquid with the
free surface and the possibility of separation of bottom from
base) dynamic calculation was carried out by the direct implicit
integration method. Overall damping in system is taken into
account with the use of Rayleigh dissipation matrix:
[C]= [M]+ [K],
where [M] is the mass matrix; [K] is the stiffness matrix.

of static equilibrium. The wave heights calculated with both


methods have quite good correspondence (Figure 2). Similar
results were obtained for the floater vertical motion.

Figure 1 Test tank with floater


However, for the horizontal floater motion the results
obtained with potential fluid formulation (SOLVIA) show
significantly larger floater motions (Figure 3). Therefore it can
be concluded that the using of potential fluid formulation for
the tank with floating roof will produce more conservative
results for the roof movement under seismic load compared to
the methods with improved fluid-structure interaction
procedure.
0.2

SOLVIA
LS-DYNA

0.2
0.1
wave height (m)

The used coefficients and provide relative damping


about 2% for the main structural eigenfrequencies in the range
from 3.5 Hz (tank breathing frequency) up to 10 Hz (horizontal
and turnover vibration). Different damping coefficients were
used for fluid elements in order to get 0.5% relative damping
for the main sloshing mode (0.1 Hz).
Calculations were performed with finite element program
SOLVIA developed by SOLVIA Engineering AB, Sweden [9].

0.1
0.0
0

-0.1
-0.1
-0.2
-0.2
time (s)

Figure 2 Calculated wave heights (m)


0.2
SOLVIA

0.2

LS-DYNA

0.1
floater move (m)

2.2 Calculation method validation


The most unclear point in the chosen calculation procedure
is the lack of the reliably fluid-structure interaction (FSI). In
order to verify it and "feel" the margins of such approach a
numerical test has been performed. Dynamic response of a
benchmark tank (Figure 1) with the length of 2 m and the
depth of 1 m with a floater was investigated using two different
approaches: potential fluid formulation realized in the program
SOLVIA and Arbitrary Lagrange Euler fluid formulation using
the program LS-DYNA of Livermore Software Technology
Corporation (LSTC), USA [10].
The tank was exposed to static gravity load and to the
horizontal acceleration of 0.1 g applied step-wise after reaching

0.1
0.0
0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

-0.1
-0.1
-0.2
-0.2
time (s)

Figure 3 Floater horizontal motions (m)

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Copyright 2008 by ASME

C Z = bo E (1 + A10 / A )

CALCULATION MODEL
Calculated reservoir has followed data:

Reservoir diameter
60.7
m
Wall shell thickness (variable along height) from 26 to 13 mm
44274 t
Total liquid mass (density 900kg/m3)
Reservoir steel structure mass
765
t
Mass of domical roof
66
t
Pontoon mass
296 t

[N/m3]

where =5.00E+07 Pa is the ground elasticity module, is the


base area (A=200 m2 for the basement large than 200 m2), A10
= 10 m2, bo=1.5 (large-grained soil).
Coefficients, which characterize the stiffness of the
foundation for other base movements directions are C
(rocking), Cx (horizontal shift). They can be determined from
the formulas: CX=0.7 CZ , C=2.0 CZ [8]. Stiffnesss of the
whole soil base are presented in table 1.

Vertical stiffness
Shift stiffness
Rocking stiffness

Table 1 Basement stiffness


KZ=2.66E+11 N/m
KZ=ACZ
KX=1.86E+11 N/m
KX=ACX
K=1.22E+14 N/m
K=IC

Seismic load was applied simultaneously in three


orthogonal directions using synthesis accelerograms. Two
different sets of accelerograms that represent different
earthquake sources were used with relation between vertical
and horizontal components of 2/3 (table 2).
spings
Figure 4. Calculation model of the reservoir
The whole calculation model contains outer shell and
bottom modeled with 7289 shell elements (Figure 4), pontoon
modeled with 3934 solid elastic elements and naphtha
(product) modeled with 91936 fluid elements (Figure 5).
Nonlinear spring elements are also included in order to model
contact conditions on the bottom outer ring. Linear springs
simulate base soil response.

Table 2 Seismic load data (vertical direction)


parameter
Seismic
Seismic
load 1
load 2
ZPA, g
0.2
0.2
peak acceleration in Z direction
0.64
0.66
(for 5% damping), g
frequency range for peak
2.6-7.1
7.1-18
acceleration, Hz
earthquake duration, s
8.5
3.5

CALCULATION RESULTS

4.1 Basement reactions


The maximum values of the dynamic base reactions are
given in table 3. In the case if the anchoring of reservoir is
absent, fixation against the slippage on the base is achieved
only due to the friction between the reservoir bottom and the
base. Absence of the sliding motion is determined by the
relationship:
Pontoon elements

Fluid elements
Figure 5 Calculation model details
The calculation of the ground base stiffness was carried
out on the basis of the coefficient of the elastic uniform
compression, determined according to the formula [8]:

FZ > Fh
For the reactions, given in table 3 and assuming friction
coefficient be equal =0.3, the condition indicated above is
satisfied for whole duration of seismic simulation.
As shown in the table 3, basement reactions decrease if no
wall anchorage is installed. The reason for it is obviously lower
dynamic stiffness of an unanchored tank compared to the
anchored tank.

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Copyright 2008 by ASME

Table 3 Extreme basement reactions


Force component
Seismic load 1
Seismic load 2
without
with
without
with
anchors
anchors
anchors
anchors
MXX, MN*m
1772
2185
1133
1292
MYY, MN*m
1981
2504
1349
1453
2140
3264
1508
1604
(MYY2+MXX2),
MN*m
max FZ, MN
508
507
539
533
min FZ, MN
347
351
359
333
max. FX, MN
90.6
115
61.7
67.3
max. FY , MN
78.8
100
63.0
68.1
4.2 Free surface motion and structure deflections
As shown in table 4 the maximum dynamic deflections and
movements increase for the case of unanchored tank.
Especially relative motion of the pontoon to the tank shell dr
will arise up to 16% for an unanchored tank. The calculated
free surface deflection (seismic wave height w) calculated for a
tank without pontoon is up to 30% higher compared to the
calculated maximum pontoon vertical motion, i.e. a pontoon
can reduce the free surface waves up to 30%.
The maximum calculated bottom uplift 49 mm is
connected with specific shell deformation due to seismic load
(Figure 6). Because partial uplift of the bottom above the base
is considered as possible, it is necessary to examine the
permissibility of the stress state under these conditions.
Table 4 Extreme deflections and movements (mm)
Seismic load 1
Seismic load 2
without
with
without
with
anchors
anchors
anchors
anchors
dh,
82
80
24
21
d v,
230
196
71
70
wave height
257
73
w
Bottom
49
0
15
0
uplift u
Shell
62
20
19
11
deflection
dr
116
100
26
23

The bottom uplift can be accompanied by the appearance


of a plastic hinge in the margin bottom plate, in the shell-tobottom joint. In compliance with design practice the
permissible value of the angle of plastic hinge rotation in the
joint is established. This angle can be calculated in accordance
with [4] as follows:

p = p* M p / p / Db ;
p* =

10 + 7 2 *3 2 + 2 *
l
l t
6
2
l
l* =
(2 + 2 ) M p / p

where Db = Et b / 12(1 v ) is the cylindrical stiffness of the


3

bottom, E and are the Young modulus and the Poisson ratio
of the reservoirs material, tb is the bottom thickness, Mp is the
bottom plastic moment, l is the width of the bottom uplift zone
which can be calculated from the following equations set in
two unknown l and 0:

6 Db
2 Db
pl 2

+
= 0;
0
u
12
l
l2
6 Db
4 Db
pl 2
t Dt 0 2 u +
0
= 0;
l
12
l
where tt is the thickness of the reservoirs shell lower course, Dt
is the cylindrical stiffness of this shell, u is the bottom uplift,

t = (12(1 2 ) / tt2 Dt2 )1 / 4 .

A calculation performed according presented method


results for u = 49 mm in the value of p = 1.5, which is less
than the permissible value of [p]=15 according [5].
4.3 Shell stresses
Equivalent Tresca stresses in the middle surface of sheet
are considered as deign stresses for the reservoirs shell.
Absence of residual deformations of the shell after the seismic
action is the criterion of strength. In this case the local plastic
deformations due to bending moments in the shell are allowed
for example in the zone of the shell-to-bottom joint. The
performed calculations show that the greatest equivalent
stresses appear in the middle of the shell height, in the zone
where thickness decreases from 18 mm to 14 mm. High
stresses in this region are connected with the large
circumferential stretching stresses in the shell as a result of the
hydrodynamic pressure action under seismic load (Figure 7).

Bottom uplift

Figure 6. Reservoirs shell deflection due to seismic


load (magnified).

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Copyright 2008 by ASME

validate proposed computation scheme for the tanks with


floating roofs.
The other finding from this evaluation is that tank's
anchorage required by the design rules in order to prevent tank
overturning can be avoided based on analysis results performed
in accordance with described scheme. It was shown that
absence of tank's anchorage can simultaneously reduce
basement moments and reservoirs shell stresses.

Figure 7 Distribution of maximum equivalent dynamic


stresses (Tresca)
As shown in table 5 shell stresses are higher for an anchored
tank.
Table 5 Maximum equivalent stress in the shell (MPa)
Operation
Seismic load 1
Seismic load 2
condition
without
with
without
with
anchors
anchors
anchors
anchors
NOC
173
173
173
173
NOC+DBE
358
443
314
313
Other criteria for the shell stresses are elastic and elasticplastic buckling criteria (elephant-foot buckling). These criteria
were checked according to the references [6, 7] and have met
design requirements.
5

CONCLUSION REMARKS

The performed analyses demonstrate that potential fluid


formulation can be successfully used for seismic evaluation of
big oil tanks with floating roofs. However, use of this approach
for solving of FSI problem leads to the conservative results for
assessment of horizontal pontoon movement. At the same time,
this conservatism could be considered as acceptable in case of
low or moderate seismic input when even conservatively
produced results give a positive answer for seismic
qualification (criteria is possibility of collision between floating
roof or pontoon and tank's inside wall). From the other hand,
solution based on the potential formulation has significantly
less computational costs. Let say, analysis of 80000 elements
carried out with LS-DYNA (Arbitrary Lagrange Euler fluid
formulation) takes ~ 70 hours. For comparison: same problem
with use of "potential" algorithm can be solved on the same
computer for 6 minutes only. It looks like a good engineering
compromise between "complexity" of the problem and
available tools and budget for its solving. However, additional
computations for different tank dimensions and laboratory
seismic tests on tank models should be performed in order to

6 REFERENCES
1. Report on damage to industrial facilities in the 1999
KOCAELI earthquake, Turkey, K. Suzuki, Journal of
Earthquake Engineering, Vol. 6, No. 2 (2002), 275-296
2. ZAMA SHINSAKU Damage and Failure of Oil Storage
Tanks due to the 1999 Kocaeli Earthquake in Turkey and ChiChi Earthquake in Taiwan - Journal of High Pressure Institute
of Japan; vol: 41, issue: 2 , page: 79-86, 2003
3. Failure knowledge database:
http://shippai.jst.go.jp/en/Detail?fn=0&id=CC1300013;
http://shippai.jst.go.jp/en/Detail?fn=0&id=CC1200074,
http://shippai.jst.go.jp/en/Detail?fn=0&id=CB1012035
4. K.Ishida, N.Kobayashi, An effective method of analyzing
rocking motion for unanchored cylindrical tanks including
uplift Fluid-Structure Dynamics PVP Vol. 9
5. BS EN 1998-4:2006 Eurocode 8, Part 4, Design of
structures for earthquake resistance. Silos, tanks and pipelines,
European Standard 1998
6. Welded steel tanks for oil storage API standard 650, tenth
edition, Add. 1,2; Nov. 2001
7. Jp.Touret, Methods of seismic calculation for tanks 18tn
European regional earthquake engineering seminar, Lyon, Sept.
1995
8. A.N. Birbraer, Seismic analysis of structures SanktPetersburg, Nauka, 1998
9. SOLVIA Finite element system. Users Manual, Report SE
03-1: SOLVIA Engineering AB, Vasteras, Sweden
10. LS-DYNA Users Manual, Version 970: Livermore Software
Technology Corporation, Livermore, California, April 2003

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Copyright 2008 by ASME

You might also like