seismic isolation of cylindrical liquid storage tanks -...

8
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 Porto, Portugal, 30 June - 2 July 2014 A. Cunha, E. Caetano, P. Ribeiro, G. Müller (eds.) ISSN: 2311-9020; ISBN: 978-972-752-165-4 3145 ABSTRACT: Seismic excited liquid filled tanks are subjected to extreme loading due to hydrodynamic pressures, which can lead to nonlinear stability failure of the thin-walled cylindrical tanks, as it is known from past earthquakes. The overall seismic behavior of tanks is, however, quite complex, since the dynamic interaction effects between tank shell and liquid must be considered. A significant reduction of the seismically induced loads can be obtained by the application of base isolation systems, which have to be designed carefully with respect to the modified hydrodynamic behavior of the tank in interaction with the liquid. Firstly, this paper presents the state of the art of tank design for anchored tanks with a fix connection to a rigid foundation with special focus on the practicability of the available design rules. Then a highly sophisticated fluid-structure interaction model is introduced, which has to be applied for a realistic simulation of the overall dynamic behavior of an isolated tank. KEY WORDS: liquid storage tank; impulsive vibration mode; fluid structure interaction model; base isolation; LS-DYNA 1 INTRODUCTION Steel tanks are preferably designed as cylindrical shells, because they are able to carry the hydrostatic pressure from the liquid filling by activating membrane stresses with a minimum of material. In combination with the high strength of steel this leads to thin-walled constructions, which are highly vulnerable to stability failures caused by seismic induced axial and shear forces. However, an earthquake- resistant design of rigid supported tanks especially for high seismic hazard levels requires unrealistic und uneconomic wall thicknesses. Compared to increasing the wall thickness an earthquake protection system can be a much more cost- effective alternative. Especially a base isolation with elastomeric bearings offers advantages in terms of an earthquake-friendly tank design. The calculation capabilities of base-isolated, liquid-filled tanks are quite limited because of the complex interaction of the isolation and the combined modes of vibrations of tank and fluid. For rigidly supported tanks different calculation methods are available for describing the vibration behavior and the seismic induced load components in a more or less simplified way. These generally accepted methods are either quite simple [1] or very accurate but complex [2] – but valid only for anchored tanks. To capture the hydrodynamic loading of isolated tanks, a complete modeling of the fluid-structure interaction including the behavior of the seismic isolation is necessary. Firstly, this contribution presents a comprehensive and feasible calculation method for anchored, rigidly supported liquid storage tanks and then introduces the modelling of a complex fluid-structure-interaction model, which is used to calculate the seismic behavior of isolated tank structures. 2 LOAD CALCULATION OF ANCHORED LIQUID STORAGE TANKS As a result of seismic excitation hydrodynamic pressure components, produced by the movement of the fluid, appear and have to be superimposed with the hydrostatic pressure. These seismic loads acting on wall and bottom of a cylindrical tank (Figure 1) can be divided into the following components: - the convective load component; the fluid vibration in the rigid tank (sloshing), - the impulsive rigid load component; caused by the inertia of the liquid, if the rigid tank moves together with the foundation, - the impulsive flexible load component; representing the combined vibration of the flexible tank shell with the liquid. Since the oscillation periods of those individual seismically induced pressure components are far apart, each mode of vibration can be determined individually. Figure 1. Cylindrical tank. 2.1 Convective pressure component Figure 2 shows the mode of vibration and the pressure distribution corresponding to the convective pressure component. The pressure distribution is defined as: Seismic isolation of cylindrical liquid storage tanks Julia Rosin 1 , Christoph Butenweg 1 1 Chair of Structural Analysis and Dynamics, RWTH Aachen University, Mies-van-der-Rohe-Str. 1, 52072 Aachen, Germany email: [email protected], [email protected]

Upload: trinhtuyen

Post on 05-Apr-2018

248 views

Category:

Documents


8 download

TRANSCRIPT

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 Porto, Portugal, 30 June - 2 July 2014

A. Cunha, E. Caetano, P. Ribeiro, G. Müller (eds.) ISSN: 2311-9020; ISBN: 978-972-752-165-4

3145

ABSTRACT: Seismic excited liquid filled tanks are subjected to extreme loading due to hydrodynamic pressures, which can lead to nonlinear stability failure of the thin-walled cylindrical tanks, as it is known from past earthquakes. The overall seismic behavior of tanks is, however, quite complex, since the dynamic interaction effects between tank shell and liquid must be considered. A significant reduction of the seismically induced loads can be obtained by the application of base isolation systems, which have to be designed carefully with respect to the modified hydrodynamic behavior of the tank in interaction with the liquid. Firstly, this paper presents the state of the art of tank design for anchored tanks with a fix connection to a rigid foundation with special focus on the practicability of the available design rules. Then a highly sophisticated fluid-structure interaction model is introduced, which has to be applied for a realistic simulation of the overall dynamic behavior of an isolated tank.

KEY WORDS: liquid storage tank; impulsive vibration mode; fluid structure interaction model; base isolation; LS-DYNA

1 INTRODUCTION Steel tanks are preferably designed as cylindrical shells, because they are able to carry the hydrostatic pressure from the liquid filling by activating membrane stresses with a minimum of material. In combination with the high strength of steel this leads to thin-walled constructions, which are highly vulnerable to stability failures caused by seismic induced axial and shear forces. However, an earthquake-resistant design of rigid supported tanks especially for high seismic hazard levels requires unrealistic und uneconomic wall thicknesses. Compared to increasing the wall thickness an earthquake protection system can be a much more cost-effective alternative. Especially a base isolation with elastomeric bearings offers advantages in terms of an earthquake-friendly tank design. The calculation capabilities of base-isolated, liquid-filled tanks are quite limited because of the complex interaction of the isolation and the combined modes of vibrations of tank and fluid. For rigidly supported tanks different calculation methods are available for describing the vibration behavior and the seismic induced load components in a more or less simplified way. These generally accepted methods are either quite simple [1] or very accurate but complex [2] – but valid only for anchored tanks. To capture the hydrodynamic loading of isolated tanks, a complete modeling of the fluid-structure interaction including the behavior of the seismic isolation is necessary. Firstly, this contribution presents a comprehensive and feasible calculation method for anchored, rigidly supported liquid storage tanks and then introduces the modelling of a complex fluid-structure-interaction model, which is used to calculate the seismic behavior of isolated tank structures.

2 LOAD CALCULATION OF ANCHORED LIQUID STORAGE TANKS

As a result of seismic excitation hydrodynamic pressure components, produced by the movement of the fluid, appear

and have to be superimposed with the hydrostatic pressure. These seismic loads acting on wall and bottom of a cylindrical tank (Figure 1) can be divided into the following components:

- the convective load component; the fluid vibration in the rigid tank (sloshing),

- the impulsive rigid load component; caused by the inertia of the liquid, if the rigid tank moves together with the foundation,

- the impulsive flexible load component; representing the combined vibration of the flexible tank shell with the liquid.

Since the oscillation periods of those individual seismically induced pressure components are far apart, each mode of vibration can be determined individually.

Figure 1. Cylindrical tank.

2.1 Convective pressure component Figure 2 shows the mode of vibration and the pressure distribution corresponding to the convective pressure component. The pressure distribution is defined as:

Seismic isolation of cylindrical liquid storage tanks

Julia Rosin1, Christoph Butenweg1

1Chair of Structural Analysis and Dynamics, RWTH Aachen University, Mies-van-der-Rohe-Str. 1, 52072 Aachen, Germanyemail: [email protected], [email protected]

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

3146

pk ξ, ζ, θ, t =2·R·ρL

(λn2-1)

J1(λn · ξ)

J1(λn)cosh(λn· γ · ζ)

cosh(λn· γ)

n=1

cos(θ) akn t · Γkn

(1)

with: pk convective pressure component due to horizontal

excitation n summation index; number of considered sloshing

modes (here: n = 1) R inner tank radius ρL liquid density J1 first order Bessel function:

J1 λn· ξ = -1 k

k! · Γ(1 + k + 1)·λn · ξ

2

2k+1∞

k=0

λn null derivation of Bessel function: λ1 = 1,841, λ2 = 5,331, λ3 = 8,536

ξ dimensionless radius: ξ = r/R ζ dimensionless height: ζ = z/H θ angle of circumference γ tank slenderness: γ = H/R akn(t) horizontal acceleration-time history as a result of an

equivalent single-degree-of-freedom system with a period Tkn for the nth-eigenmode of the sloshing wave. By using the response spectrum method the spectral accelerations corresponding to the natural periods Tkn should be calculated based on the elastic response spectrum.

Γkn participation factor for the convective pressure component for the nth-eigenmode.

Figure 2. Convective pressure component – mode of vibration

and pressure distribution.

Taking into account the first sloshing eigenmode (n = 1) and the pressure distribution of the tank shell (ξ = 1), equation (1) can be simplified to:

pk ξ = 1, ζ, θ, t = R · ρL 0,837 · cosh(1,841· γ · ζ)

cosh(1,841· γ)cos(θ) ak1 t · Γk1

(2)

The natural period Tkn for the nth-eigenmode of the sloshing wave is calculated with

Tkn= 2πg · λn· tanh (λn· γ)

R

(3)

2.2 Impulsive rigid pressure component due to horizontal seismic excitation

Figure 3 shows the mode of vibration and the pressure distribution corresponding to the rigid impulsive pressure component. The pressure distribution is given by the expression:

pis,h ξ, ζ, θ, t = 2 · R· γ· ρL· -1 n

νn2

I1νnγ · ξ

I νnγ

n=0

cos νn· ζ cos(θ) ais,h(t) · Γis,h

(4)

with: pis,h rigid impulsive pressure component due to horizontal

excitation νn νn= 2n + 1

I1 modified first order Bessel function:

I1νnγ

· ξ = J1 i·νnγ · ξ

in=∑ 1

k! · Γ(1 + k + 1)∞k=0 ·

νnγ · ξ

2

2k+1

I1’ Derivation of the modified Bessel function regarding to

[3]:

I νnγ

·ξ = I0νnγ

·ξ -I1

νnγ ·ξ

νnγ · ξ

I =∑ 1k! · Γ(0 + k + 1)

·νnγ ·ξ

2

2k+0∞k=0 -

∑ 1k! · Γ(1+k+1) · 

νnγ · ξ

2

2k+1∞k=0

νnγ · ξ

ais,h(t) horizontal acceleration-time history. By using the response spectrum method ais,h(t) should be replaced by the spectral acceleration Sa corresponding to T = 0 s.

Γis,h participation factor for the rigid impulsive pressure component: Γis,h = 1.0, because the rigid tank is moving together with the foundation.

Figure 3. Impulsive rigid pressure component – mode of

vibration and pressure distribution.

Taking into account the pressure distribution of the tank shell (ξ = 1), equation (4) is simplified to:

The corresponding natural period is T = 0.

2.3 Impulsive flexible pressure component due to horizontal seismic excitation

Figure 4 shows the mode of vibration and the pressure distribution corresponding to the flexible impulsive pressure component.

The flexible impulsive pressure component is calculated in an iterative procedure using the added-mass-model according to [3], Annex A. Within the framework of the procedure the tank wall is loaded with iterative calculated additional mass

pis,h ξ = 1, ζ, θ, t = R·ρL2 · γ · -1 n

νn2

I1νnγ

I νnγ

cos νn·ζ∞

n=0

cos(θ) ais,h(t)·Γis,h

(5)

portgive

pif,h

n=0

withpif,h

f(ζ)

aif,h(

Γif,h

Fi

Tpres

withpif,h

s(ζ)

Tshel

T

Tif,

withWL

= R

tions of the en by the expr

h ξ, ζ, θ, t =

2R ρL

I1νγ

νnγ · I0

h: flexible horizontal

deflectionoscillation

(t) horizontalequivalentthe respacceleratioTif,h,1 shou

h participaticomponen

igure 4. Impulvib

The participatssure compon

Γif

h: (ζ) pressure f

componen) wall thick

Taking into acll (ξ = 1), equ

The first natura

f,h,1 = 2 · F γ

h: fluid w

R·ρL 2· νnγ

n=0

activated fluiression:

νnγ · ξνnγ

cos ν

impulsive pl excitation

n curve of then of the tank-fl accelerationt single-degreponse spectron corresponduld be used. ion factor for nt due to horiz

lsive flexible pbration and pre

tion factor Γnent is calculat

f,h=pif,h ζ d1

0

f(ζ) · pif,h ζ10

function of tnt as a functionkness of the tan

ccount the pration (6) can b

al period Tif,h,1

WLπ · g · E · s(ζ = 1/3

weight: WL = π

pif,h ξ = 1, ζ

I1νnγ

nγ I νn

γ·cos

cos(θ) aif,h

id. The press

νn· ζ f ζ · c1

0

cos(θ)

pressure com

e first (anti-syfluid combinatn-time historyee-of-freedomrum analysding to the f

the flexible izontal excitati

pressure compessure distribu

Γif,h for the fted as follows

ζ dζ, s(ζ) = co

the flexible in of the fillingnk

ressure distribbe simplified

1 is calculated

3)= 2 · R · F γ

π · R2· H · ρL·

ζ, θ, t =

νn·ζ f ζ · c1

0

t · Γif,h

Pr

sure distributi

os νn· ζ dζ

aif,h t · Γif,h

mponent due

ymmetric) motion

y as a result om system. By sis the spfirst natural p

impulsive preon

ponent – modution.

flexible impus:

onst.

impulsive preg height

bution of the to:

d as follows:

γ H ·ρLE · s(ζ = 1/3)

g

cos νn·ζ dζ

roceedings of the

ion is

(6)

e to

de of

of an using ectral

period

essure

e of

ulsive

(7)

essure

tank

(9)

F(y)s(ζ =

Thdepewordjoint[3], Ato thshap"drydeter

Hobecapressmath

Ththatby ucorreand m

Ththe p‐ t‐ t‐ t‐ aHo

smaldeterdeter

is suθ=0.the t

Thcomptabulthe t

Foalso

p

withpjCj(ζ,Γj(γ)

(8)

9th International

correcti= 1/3) wall thihe function cuends on the imds the aforemt bending formAnnex A thehis the resonape, is added to" tank modrmined (Figur

Fi

owever, evenause the calcusure functionhematical softwhough, with cthe bending f

using a paramectly to the namaterial confi

he bending forparameters a, bthe tank slendthe Poisson’s rthe ratio fluid a changing waowever, the ll. By using rmine the flexrmine the pres

pif,h ξ,ζ,θ,t

uitable. Cif,h(ξ, Figure 6 shoank shell (ξ =

he participatioponent Γif,h, wlated as well, ank mass is in

or simplicity, tprovided in a

pj ξ 1,ζ,θ =

: pressur

,γ) factor o) particip

tabulate

l Conference on S

ion factor: F γickness of the urve of f(ζ) inmpact of the

mentioned prem must be corr

added-mass cating fluid, aco the tank wael, the morere 5, see also [

igure 5. Iterati

this iterativeulation is basen (8), whichware tool withomprehensiveform of any ta

meterized sineatural frequenigurations [5],

f ζ =a sin π2

rm of the comb, c, d – depenerness (γ = H/ratio ν, mass to tank m

all thickness alast named the sine fun

xible impulsivssure for pract

t =R ρL cos

ζ) correspondws the variati 1) for differe

on factor for twhich is specassuming tha

nsignificant. the formulatiostandardized,

R ρL Cj(ζ,γ) c

e component jof pressure compation factor ed in [6]

Structural Dynam

γ = 0,157 · γ2

tank at 1/3 filn (8) is generliquid onto t

essure functiorectly determiconcept is proctivated with all density. The accurate b[5], [6]).

ion process [5

e process is red on a comph requires ah a finite eleme parameter stank geometrye wave, whichncies for all co, [6]: π2

ζ-b c+d

mbined vibrationds on /R),

mass and along the tank

influences anction (10) e pressure wittical use the e

θ Γif,h aif,h t

ds to the normion of the factent tank slendethe flexible imcified in equat the tank wa

ons of all pres, tabulated for

cos θ Γj γ

j = {k; is,h; ifmponent j, tab

of pressur

mics, EURODYN

2 + γ + 1,49 lling height rally unknowthe tank, in oon pif.h. Thusined iterativelyoposed. Accor

the first benhen with thebending form

5].

rather impracplex mathemaa coupling oment program.tudies it is sh

y can be descrh can be mapommon geom

on – and with

height. are comparatiit is possiblethout iterationquation

Cif,h(ξ,ζ)

malized pressutor Cif,h(ξ=1,ζernesses. mpulsive pres

uation (7), canall is constant

ssure functionsrm [6]:

Sa,j Tj

f,h} bulated in [6]re componen

2014

3147

wn. It other s the y. In

rding nding

new m is

ctical atical of a

hown ribed pped

metric

(10)

h this

ively e to

n. To

(11)

ure at ) for

ssure n be t and

s are

(12)

nt j,

Proc

314

Sa,j(

F

Inpj cthe strea th

2.4 FuraccoandhoriactiimpflexcorrTheheigthe for be f

F

Athe struseis

3 Bucstor

ceedings of the 9t

8

(Tj) spectraperiod

Figure 6. Norm

n the followinan be appliedtank structure

ess distributionhree dimension

Vertical srthermore the ount, which l

d correspondinizontal seismivated by thepulsive flexibxibility of thresponding pre calculation ght and radiushorizontal vithe pressure

found.

Figure 7. Verti

As this paper base isolatio

ucture due tosmic action is

STABILITYckling inducerage tanks and

th International C

al acceleratiod of the pressu

malized flexib

ng those seismd as equivalene, which allown by performinnal model of t

eismic excitatvertical seismleads to two ng pressure d

mic action, the rigid-body ble pressure he tank shelressure distribof the press

s is simplifiedibration modefunctions as w

ical seismic acpressure d

discusses the on mainly affo horizontal s

disregarded in

Y FAILURE Oed damage is d should be ta

Conference on Stru

n correspondure component

ble impulsive p

mic induced prnt static loads ws an accurateng a finite elethe tank.

tion mic excitation

additional mdistributions. he impulsive

motion of component i

ll. The vibrbutions are shsure distributid analogues toes. In [6] the well as the tab

ction: Modes distributions.

behavior of ffects the behseismic excitn this contribu

OF ANCHORespecially pr

aken under ser

uctural Dynamics

ding to the nat 

pressure Cifh [

essure compoon the dry sh

e calculation oement analysis

must be takenmodes of vibra

In analogue rigid pressu

the tank andis caused by

ration modes hown in Figuion over the o the calculati

relevant formbulated value

of vibration a

isolated tankhavior of the ation, the veution.

RED TANKS rofound for lrious consider

s, EURODYN 20

atural

[1].

onents hell of of the s with

n into ations

to a ure is d the y the

and ure 7.

tank ion of mulas es can

and

s and tank

ertical

liquid ration

for tsubsand tankthe rand verticausbuckoccuexcitseismpressthe (p =whicbuckconscomp(p =of thwall press

Fidi

Geand circudangstresmomand tensicompby thsheadirec

F

Acat di

a)

a)

014

the design of equently buckenvironmentas can be obseresult of the vminimum pr

ical seismic exe different str

kling” is a formurs when the mtation coexistsmic excitationsure due to hominimum prpstat + ph - pv)

ch can lead tokling”. Finalsidered on theponents are pstat - ph - pv),

he tank to arethickness bu

sure can be ob

gure 8. Stabilamond shaped

enerally, theshydrostatic

umferential agerous for stabss at the bottomment arises (F

hydrodynamiile stress in thpressive streshe sloshing or stress occuction (Figure 9

igure 9. Qualiseismic exci

circum

ccording to thifferent periph

xy

f tank structurkling can leadal hazard. Somerved in Figuvarious combiressure compxcitation – deresses endangm of plastic bmaximum pres with the man (p = pstat + porizontal seismressure due t, the stabilizi

o axial bucklinly, for the e opposite sisubtracted fr which can le

eas with negauckling due tbserved in the

ity failure: a) d buckling [8]

internal pr

e different cpressure co

and shear stbility failure: m of the tank

Figure 9a). Thic pressure re lower tank w

ss arises in thof the liquid urs at the sid9c).

itative stress ditation in x-dimferential stre

is, the decisivheral angles

y

b)

b)

res, as insuffd to significanme typical bu

ure 8. These dinations betweonents due t

escribed in [2]gering stabilitbuckling at thessure due to h

aximum pressuph + pv). Whe

mic excitation to vertical sng internal prng known as

same seismide of the tanfrom the hydead especiallyative pressure.to circumferense areas.

elephant foot

], c) buckling ressure[8]. ombinations

ome along wtresses, whichFirstly, high a

k resulting fromhe combinatiesults in higwall area, whi

he top area of (Figure 9b).

des parallel t

distribution durection: a) axiess, c) shear s

ve stresses for θ: axial com

x

ficient design nt loss of propuckling modesdamage formseen the maximto horizontal ] and [6] –, wty. “Elephant e tank bottomhorizontal seisure due to veren the maximis combined

eismic excitaressure decrea“diamond sha

mic combinank, both presdrostatic presy in the upper . In case of sntial compres

buckling[7], due to negativ

of hydrodynawith high ah are extremaxial compresm the overturon of hydros

gh circumfereile circumferef the tank indu

In addition, o the earthqu

ue to horizontaial stress, b) tress.

the design apmpressive stres

y

c)

c)

and perty s for s are mum

and which

foot m and

smic rtical mum with ation ases, aped ation ssure ssure part mall ssive

b) ve

amic axial, mely ssive rning static ntial ntial uced high uake

al

ppear ss at

x

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

3149

θ = 0°, circumferential compressive stress at θ = 180° and shear stress at θ = 90°.

3.1 Earthquake protection by base isolation To prevent an earthquake damage of an arbitrary building structure two different possibilities exist: One the one hand a stiffening of the structure results in higher resistance to the earthquake impact and on the other hand a decoupling of structure and underground results in reduced seismic loads acting on the structure. For a tank structure the stiffening could be render possible by an increase of the shell thickness, which especially for high seismic hazard levels requires unrealistic und uneconomic wall thicknesses. Compared to increasing the wall thickness an earthquake protection system can be a much more cost-effective alternative. Especially a base isolation with elastomeric bearings offers advantages in terms of an earthquake-friendly tank design.

A base isolation is aimed at a decoupling of the building and the ground motion. Elastomeric bearings are a widely used base isolation and can optionally be installed with or without reinforcement, often in combination with a lead core [9]. It is recommended to use reinforced bearings, since elastomers are subjected to high deformations up to 25% under vertical loads. These deformations cause lateral strains, by which unwanted rocking motion in vertical direction in case of seismic excitation can occur. Reinforced bearings can be considered as quasi-rigid in the vertical direction, so they are suitable to transfer vertical loads. Under cyclic loading, elastomers behave almost like springs. They have – depending on the material properties – a certain stiffness, which causes a reset of the bearing and thus of the entire system after the release. Through the use of high-damping elastomers (addition of oils, resins, extra fine carbon black and other fillers), or a lead core, the damping capacity of the bearings can be increased significantly. In case of a distortion of 100%, high damping elastomers have damping rates from 0.1 to 0.2 while normal elastomers are about 0.04 to 0.06 [9]. The use of elastomeric bearings as earthquake protection systems for tanks has already been realized, e.g. in [10]. So far, to capture the hydrodynamic loading due to seismic excitation simple mass oscillators with simplistic approaches have been used: In this case of calculation the fluid was assumed as rigid, so that the interaction between the tank wall and fluid was neglected. In addition, the determination of the stress distribution in the tank wall is not possible.

4 FLUID STRUCTURE INTERACTION MODEL The software LS-DYNA [11] is used for the simulation of the fluid-structure interaction of a liquid-filled tank. The software provides an explicit solver, which offers advantages especially for the solution of dynamic contact problems. In addition, LS-DYNA provides materials for the modeling of fluids and standardized contact formulations to represent the interaction of the tank shell and the fluid during a seismic excitation. Details of the following material, element and contact formulations can be found at the LS-DYNA manuals [12].

4.1 Material formulation Basically both, elastic and plastic approaches are applicable for the tank shell. Since the focus is set on the reduction of the seismic loading by applying a base isolation, an elastic

behavior of the tank shell is assumed (MAT_ELASTIC). For the base plate the concrete material model MAT_CSCM is used, which is applied with default settings. The fluid is performed by a linear (elastic) material formulation (MAT_ELASTIC_FLUID), which idealizes the fluid as incompressible and friction-free. Here, only the fluid density as well as the bulk modulus of the fluid is needed as input for the calculation. The change in the hydrostatic pressure fraction p is described by the bulk modulus K and the strain rate εii in the main directions of the material:

p = -K εii (13)

The shear modulus is set to zero. The deviatoric stress component Sik

n+1 is given by:

Sikn+1 = VC ΔL C ρL εik

' (14)

Herein, VC is a tensor viscosity coefficient, ΔL is the characteristic element length, C is the fluid bulk sound speed (C = K/ρL), ρL is the fluid density, and εik

' is the deviatoric strain rate. The chosen values for water are given in Table 1.

Table 1. Material properties for the fluid formulation.

Symbol Property Water value K Bulk modulus 2100 [N/mm²] ρL Density 1000 [kg/m³] C Speed of sound 1450 [m/s] VC Tensor viscosity coefficient 0.1

4.2 Element formulation The tank wall and the tank bottom are idealized by Belytschko-Lin-Tsay shell elements with reduced integration, which are characterized by a high efficiency in terms of computing power required for explicit analysis [13]. To avoid hourglassing, which is particulary a problem by application of reduced integrated elements, LS-DYNA offers an hourglass control. The foundation plate is idealized by 8-node solid elements and the fluid is represented by an Arbitrary-Lagrangian-Eulerian finite element formulation (ALE). When using the ALE formulation extra volume elements are generated within the region of the freeboard up to the top edge of the tank wall, so the fluid surface can move freely (sloshing). Also, the ALE mesh must enclose the Lagrange mesh. For this reason a series of elements at the top and bottom of the tank, below the base plate and outside of the tank wall are generated. The elements are assigned to the vacuum material MAT_VACUUM, which has no physical meaning, but merely represents a region within the ALE mesh in which the fluid can move. Due to the very moderate deformation of the actual FE-mesh by using the ALE element formulation, no hourglass problems for the fluid result.

4.3 Contact formulation The interaction between the tank shell and the fluid represents an important aspect of modeling. If a contact of the two parts appears, compression stresses are transferred while the transfer of tensile and shear stresses is disregarded. LS-DYNA provides essentially two different contact formulations for coupling the tank shell (Lagrange) and the fluid (ALE) with

Proc

3150

eachCONformthe exaformresutank

4.4 Theelemsuppelemdambaseexcdireuplimeain F

F

4.5 In agravseconexoscithe Thiusin

Tfluia rivertdisr

5

5.1 The(tenthic(comtank17 (stif

ceedings of the 9t

0

h other: NSTRAINEDmulations are latter formul

mple a separamulation allowult of the hydk shell.

Base isolae base isolatioments, whichporting nodements exhibitmping represene isolation iitation, where

ection fixed sifting of the asures. The FFigure 10.

Figure 10. Bas

Sequence a first step thvity. The loadond to avoid a

xt half second illations subsisupporting n

s time historyng a baseline c

The base isolad-tank-structuigid behaviortical directioregarded in the

RESULTS INTERACT

Example oe following cnsile strenghckness, firmlympression strk is illustratedhigh dampin

ffness 825 kN

th International C

ALE_FSI_PD_LAGRANG

well suited folation offers mate specificatiows the evaluadrostatic and

ation on is idealize

h are integrates and the nt a corresponnting the behais acting in eas in verticalsupports are tank is preveSI-model with

se isolated flui

of loading he system is ld is linearly an excessive othe system is

ide. Then the nodes as a hoy is artificial gcorrection. ated support oure due to horr of the elastoon. The vee calculations

OF THTION MODEL

of calculationcalculations arht fyk = 240 Ny anchored to rength fck = 50d in Figure 11ng elastomer

N/m, damping

Conference on Stru

PROTECTIONGE_IN_SOLIDor fluid-structumore configuon of dampination of the chydrodynami

ed by linear sted in the mnodes of thending equivaavior of the bthe direction

l and horizontapplied. It is

ented by meah seismic isol

id-structure in

loaded with tapplied withi

oscillation of ts unloaded, soseismic excit

orizontal velogenerated from

only affects thrizontal seismomeric bearinertical seisms.

HE FLUIDL

re carried ouN/mm²) witha reinforced c

0 N/mm²). Th. The base pla

ric bearings, g rate 15%) ar

uctural Dynamics

N (AFP) D (CLIS). ure interaction

uration optiong. Furthermor

contact forcesc pressures o

spring and damodel betweene base plate. alent stiffnessbase isolation.n of the setally perpendis assumed thans of construlation is illust

nteraction mod

the acceleratioin a period othe system. Foo that the resutation is appliocity time hism a code spec

he behavior oic excitation, ngs is assum

mic excitation

D STRUCT

ut for a steel h constant concrete base

he geometry oate is supporte

which propre taken from

s, EURODYN 20

and Both

n, but s, for re the s as a on the

amper n the

The s and The ismic icular

hat an uctive trated

del.

on of f one or the ulting ied to story. ctrum

of the since ed in n is

TURE

tank wall plate

of the ed on erties [14].

The elastverti(ρL =

Fbear

Figuseismvelocmodthe comprate dampand in th

F

Th0.26modprincrelevstrucadvashiftacceincre(Figu

5.2Figupress

014

calculation mtomeric bearinical direction= 1,0 t/m³).

Figure 11. Sysrings (left), lo

ure 12 shows mic hazard zocity time histel is artificialresponse speponent (Figur– according tping of 2.5%0.5% for the

he code.

Figure 12. Acc

he sloshing fHz; the freqe according ciple is 9.1 Hzvant is, whiccture. With anantages: Becaut in the lowlerations is eases, which ure 12).

Validationure 13 showsure over the

model considerngs, whereas n. The fillin

stem of the isoocation of the

baseplate

the acceleratone, which is tory used for lly generated ectrum methore 2 to Figureto the Europe

% for the impuconvective vi

celeration respdampin

frequency of quency of theto an analysiz, so that the mch conducts n isolated suppuse of the sof

wer frequencyobtained. Fu

also reduce

ws the distrib tank height

rs the horizonta rigid behav

ng is assume

olated tank wielastomeric bee (right).

tion responsetaken for thethe calculatifrom this sp

od every ind 4) gets an in

ean standard Eulsive flexiblibration mode

ponse spectrumng rates.

the rigidly se impulsive is basing on maximum speto a high loport there are ft isolation syy range withurthermore, thes the spect

bution of thand the rang

tal stiffness ovior is assumeed to be w

ith elastomericearings under

e spectrum ofe calculation. ion with the ectrum. By udividual pres

ndividual damEurocode 8 [3le vibration me is recommen

m for differen

supported tanflexible vibrathe added-m

ectral acceleraoad on the two consider

ystem a frequeh lower spehe damping tral accelerat

he hydrodynage. The maxim

f the ed in water

c the

f the The FSI-

using ssure

mping 3]. A mode nded

nt

nk is ation

mass-ation tank

rable ency

ectral rate

tions

amic mum

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

3151

value relating to the range occurs at the peripheral angle of θ = 0°.

Figure 13. FSI-model: Qualitative hydrodynamic pressure

distribution over tank height and range.

To validate the fluid-structure interaction model the hydrodynamic pressure and the stress results of a rigidly supported tank are compared to those from an analysis with equivalent static pressure distributions (ESP) presented in chapter 2. For the analysis with equivalent static pressure distributions the seismic induced pressure components are calculated separately for each vibration mode using the response spectrum method. Afterwards they are superimposed to the resulting hydrodynamic pressure using the SRSS-rule. Finally the resulting hydrodynamic pressure is applied as an equivalent static load to the dry tank wall to calculate the stress distribution of the tank shell. For this calculation the hydrodynamic pressure is combined with dead load and hydrostatic pressure.

The models are validated by means of parametric studies with different tank geometries. The varied calculation parameters are the tank slenderness (H/R-ratio), the wall slenderness (R/t-ratio) and the seismic hazard level (different subsoil conditions according to Eurocode 8 [4]. The investigated H/R and thickness ranges are 0.3 ≤ H/R ≤ 3 and 500 ≤ R/t ≤ 1000, which contains the calculation example presented in chapter 5.1 (H/R = 2 and R/t = 833).

Figure 14. Hydrodynamic pressure distribution for two

different tank geometries.

Figure 14 shows the hydrodynamic pressure distribution due to horizontal seismic action at the peripheral angle of 0° over the wetted tank height for two selected tank geometries: the geometry of the calculation example and, in addition, a very squat geometry. According to [3] damping values of 2% for the tank shell and 0.5% for the fluid are applied for the

rigidly-supported FSI-model. By applying these damping values the numerical simulation results of the rigidly-supported tank considering fluid-structure interaction effects show a very good agreement with results according to the method with equivalent static pressure distribution (ESP).

5.3 Base-isolated support The calculations of the base isolated tank are carried out for an elastic behavior of the tank itself and a damping of the fluid of 0.5%. Figure 15 shows the resulting displacement and the acceleration time history of two different nodes: the one is a supporting node, which means a node directly affected from the seismic excitation. So the displacement and the acceleration are exactly the same than the ground movement raised by the earthquake. As the ground movement is filtered by the isolation system the tank structure itself moves different. The acceleration affecting a node of the tank bottom is reduced significantly. Of course the reduced acceleration is at the expense of an increase of the displacement. The maximum horizontal displacement of the isolated tank structure is about 3 cm.

Figure 15. Resulting displacement and acceleration time history of the isolated tank above and under the isolation

support.

Figure 16 shows a significant decrease of the resulting hydrodynamic pressure, which results from the frequency shift and the increasing damping rate caused by the isolated support.

Figure 16. Hydrodynamic pressure distribution for the isolated

tank in comparison to a rigid support.

Figure 17 to Figure 19 show the circumferential (Figure 17), axial (Figure 18) and shear stress distribution (Figure 19) over

ρ

Proc

3152

the the circshea

Fi

F

F

FdistbetwdistGentensExcpresstabsectthe calcdecthe

ceedings of the 9t

2

tank height. Adesign ap

cumferential sar stresses (θ =

igure 17. Circ

Figure 18. Ax

Figure 19. Sh

or the rigidltribution, the ween the atributions annerally, the hosile stresses dcept the upperssures shows bility problemtion. The axia

results accculations of rease of axialcircumferent

th International C

As mentionedpear at dif

stresses (θ = 1= 90°).

cumferential sthei

xial stress distr

ear stress distr

ly supportedresulting str

analysis withnd the calcuoop stresses ofdue to the hydr edge of the local compre

ms of the thinal and shear strcording to the base isol and shear sttial stresses a

Conference on Stru

d before, the dfferent perip80°), axial str

tress distributight.

ribution over

ribution over

d tank, just resses show ah static equlation with f the tank waldrostatic prestank wall wi

ssion stressesn steel sheet resses complythe simplifi

olated tank shtresses, while are negligible

uctural Dynamics

ecisive stressepheral angleresses (θ = 0°

tion over the ta

the tank heigh

the tank heigh

like the prea good agree

quivalent prethe FSI-m

ll are dominatssure of the liith low hydro which can lein the upper

y qualitativelyed method. how a signifthe differenc

because they

s, EURODYN 20

es for es θ: °) and

ank

ht.

ht.

essure ement essure model.

ed by iquid.

ostatic ead to r tank y with

The ficant ces of y are

dompresscircu

6The tanklead of tapplipaperigidvaluebaseLS-Daccoagreeimplcan circuAlthveryslenddampuse sseism

REF [1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

014

minated by thsure. But espumferential co

CONCLUSIseismic exci

s activates hto uneconom

the seismic ication of bas

er introduced dly supported, es. After this -isolated tank

DYNA [11] aount. The resement for a rilementation obe reduced s

umferential anough the obtapromising fu

derness ratioping and stifsimplified calmic behavior o

FERENCES

Housner, G. W.,Seismological SoFischer, F.D., RDesign of AnchThree-Dimension371, 1991 DIN EN 1998-4:Erdbeben – Teil Institut für NormDIN EN 1998-1:Erdbeben – Teil Hochbauten, DeuHoltschoppen, BVereinfachtes BInteraktionsschwseismischer BelaMeskouris, K., Hund Erdbeben, VEl-Zeiny, A.: Unanchored LiCalifornia, IrvineShih, C.-F.: FailuDissertation, CalPetersen, C., Be2005: ErdbebenKuhlmann, U. (eBachmann, H., Erdbebensanieru2000. LS-DYNA, Liv2013. LS-DYNA, KeyR6.1.0, LivermorHallquist, J. OTechnology CorpBaumann, T., BöTanks, Structural

he tensile strpecially in th

ompressive str

ONS itation of rig

hydrodynamic mic wall thickn

induced loadse isolations wfirstly a simanchored tana highly soph

ks was presentand takes the sults of bothigid support af a base isolasignificantly

nd shear stressained results wurther investigos and isolatffness propertculation methof isolated tan

, The dynamic beociety of America

Rammerstorfer, Fhored and Unannal Earthquake

2006: Eurocode 4: Silos, Tankba

mung, Berlin, 20062010: Eurocode 1: Grundlagen, E

utsches Institut füB., Cornelissen Berechnungsverf

wingung bei flüssastung, Baustatik Hinzen, K.-G., B

Vieweg und TeubnNonlinear Tim

iquid Storage e, 2000. ure of liquid storaifornia Institute o

eutler, H. Braun,nschutzsysteme ds.), Ernst und So

Wenk, T., Scung, Schweizer In

vermore Softwar

word User’s Mare Software Tech

O., LS-DYNA Thporation, 2006. öhler, J., Seismic l Engineering Inte

resses due tohe upper tanresses are elim

gidly supportepressure co

nesses. A signds can be awith elastome

mplified calculnks basing on thisticated calcted. This modfluid-structur

h models shoat the tank botated support tand with thi

ses endangerinwith the deve

gations are netion systemsties. In Addithods are needenk structures.

ehavior of water a, 53(2), 381-387.G., Scharf, K., chored, Liquid Excitation. Struc

8: Auslegung vauwerke und Roh6. 8: Auslegung vErdbebeneinwirkür Normung, Berl

P., Butenweg, fahren zur Bsigkeitsgefüllten Baupraxis 11, InnButenweg, C., Mner Verlag, 3. Au

me-Dependent SeTanks. Disserta

age tanks due to of Technology, Pa, C., Mangering,für den Hoch

ohn, 864-929, 20chwächen statt ngenieur und Arc

re Technology

anual Volume I hnology CorporatiTheory Manual,

Design of Base-Iernational, 11(2),

o the hydrosnk wall area

minated.

ed liquid stomponents, wnificant reducachieved by ric bearings. lation methodtabulated presculation modedel is realized re interaction ow a satisfacttom. Throughthe induced ls the high a

ng stability. eloped modelseded for diffe

s with nonlition, for praced to calculate

tanks, Bulletin o7, 1963.

Earthquake ResStorage Tanks uctural Dynamics

von Bauwerken ghrleitungen, Deut

von Bauwerken gkungen und Regelin, 2010.

C., Mes-kourisBerücksichtigung

Tankbauwerken nsbruck, 2011

Mistler, M., Bauwuflage, 2011. eismic Responsation, University

earthquake excitaasadena, 1981. , I., Stahlbaukalh- und Brücke005.

Verstärken beichitekt, 118(3), 1

Corporation (LS

and II, Versionion, 2012. Livermore Sof

Isolated LNG-Sto, 139-144, 2001.

static the

orage which ction

the This

d for ssure el for with into

ctory h the oads

axial,

s are erent inear ctical e the

of the

sistant under 317-

gegen tsches

gegen ln für

s, K., der

unter

werke

se of y of

ation.

ender nbau,

i der 18-23,

STC),

n 971

ftware

orage-