You are on page 1of 6

Simple and practical analysis of effect of soil liquefaction on response of

structure-pile system
Kotaro Kojima, Kohei Fujita & Izuru Takewaki
Dept. of Architecture and Architectural Eng., Kyoto University, Japan

ABSTRACT: The time-history response of a structure-pile system during soil liquefaction is highly compli-
cated and several analytical methods have been proposed through the comparison with the experimental
works. However the analytical methods with higher accuracy often require large computational loads and
are not necessarily preferred in the actual design practice. On the other hand, while the response spectrum
method is not accurate compared to the aforementioned methods, it can provide useful design guidelines in
the preliminary stage for structure-pile systems under soil liquefaction with acceptable accuracy. In this pa-
per, a response spectrum method for a structure-pile-soil system is used where the effect of soil liquefaction is
taken into account by introducing the so-called p-multiplier method. It is shown that, while in the case of
inner partial liquefaction with a non-liquefied layer at the top, the demand on the pile moment is large due to
the inertial effect of that non-liquefied layer at the top, in the case of overall liquefaction near the ground sur-
face, the demand is smaller than the case of inner partial liquefaction.

1 INTRODUCTION 1975)) have been used in practice. In the static


method, a seismic deformation mode of the surface
Since the 1964 Niigata earthquake (Idriss and Bou- ground is assumed for evaluating the kinematic ef-
langer 2008), soil liquefaction is one of the major fect which is input to the soil-pile system and an as-
subjects in the field of earthquake geotechnical en- sumed inertial force is applied to the soil-pile system
gineering. The 2011 Tohoku (Japan) earthquake also for evaluating the inertial effect. However the stat-
caused a devastating damage to buildings and hous- ic method has a difficulty in its accuracy and the fi-
es due to soil liquefaction (AIJ 2012). While pile nite element method has some problems, e.g. huge
foundations are effective for settlement of buildings computational resources, complicated input data
during and after liquefaction, its resistance in lateral- preparation.
ly spreading ground during soil liquefaction is not The time-history response of a structure-pile sys-
necessarily clear. Actually during the 1995 Hy- tem during soil liquefaction is highly complicated
ogoken-Nanbu (Kobe) earthquake, many reports on and several analytical methods have been proposed
damaged piles were provided (AIJ 1995, Japanese through the accuracy verification based on the com-
Geotechnical Society 1996, Takewaki 1998). parison with the experimental works. However the
In the evaluation of pile bending moments, both analytical methods with higher accuracy often re-
the effect of the forced displacement of a free-field quire large computational loads and are not neces-
ground and the effect of the inertial force from a su- sarily preferred in the actual design practice. On
perstructure have to be considered in an appropriate the other hand, although the response spectrum
manner (for example Gazetas & Dobry 1984, method is not accurate compared to the aforemen-
Kavvadas & Gazetas 1993, Miyamoto et al. 1996, tioned methods, it may be able to provide useful de-
Mylonakis et al. 1997, Murono & Nishimura 2000, sign guidelines in the preliminary stage for struc-
Tokimatsu et al. 2005, Kishida & Takewaki 2010, ture-pile systems under soil liquefaction with
Ashford et al. 2011). However these two effects acceptable accuracy.
have different characteristics and it seems difficult In this paper, the ground motion is defined at the
to include these in a simple way keeping a reasona- engineering bedrock (see Figure 1) and a single-
ble accuracy. This is because the analysis of the input response spectrum method with well-designed
overall system is usually required in evaluating the free-field-pile interaction is proposed (Figure 1(a)).
latter effect. Conventionally two methods (static The previously proposed response spectrum method
method and finite element method (Lysmer et al. (Kishida & Takewaki 2010, Nakamura et al. 1996,
Takewaki 2004, Kojima et al. 2013) in terms of GL- 0
sand SPT value=2 Ground A
complex modal quantities for a structure-pile-soil 5
sand SPT value=3
system is used where the effect of soil liquefaction is 10
clay SPT value=4
taken into account by introducing the so-called p- 15

Depth (m)
multiplier method (Ashford et al. 2011). This re- 20 clay SPT value=2

sponse spectrum method is based on the result for 25


SPT value=22
sand
superstructures (Igusa et al. 1984, Yang et al. 1990). 30

The p-multiplier method introduces a reduced coef- 35


clay SPT value=3
SPT value
ficient for the p-y relation. Two liquefaction mod- 40
clay Engineering Bedrock =44
0 100 200 300 400 500
els are investigated. It is shown that, in the case of shear wave velocity (m/s)

inner partial liquefaction with a non-liquefied layer Figure 4. Soil profile of ground model A and SPT values
at the top, the demand on the pile moment is large

0
due to the inertial effect of that non-liquefied layer

stiffness reduction ratio G/G


1.2 0.4

at the top and, in the case of overall liquefaction 0.35

damping ratio
1 stiffness
of clay damping ratio 0.3
near the ground surface, the demand is smaller than 0.8 of sand
0.25
stiffness
the case of inner partial liquefaction. The reliabil- 0.6 of sand 0.2
damping ratio
ity and accuracy of the single-input response spec- 0.4 of clay
0.15
0.1
trum method with the help of the p-multiplier meth- 0.2
0.05

od are investigated through the comparison with the 0


10
-6
10
-5
10
-4
10
-3
10
-2 -1
10
0

result by the time-history response analysis includ- shear strain

ing the multi-input from the free-field ground and Figure 5. Dependence of shear moduli and damping ratios on
the nonlinear restoring-force characteristics around the strain level
the pile.
2 RESPONSE SPECTRUM METHOD IN TERMS
distributed
springs OF COMPLEX MODAL QUANTITIES
S1 = k x1 + i cx 1
free-field finite
ground element
linear
free-field Consider two models (model 1 and model 2) as
disp. ground
z1 shown in Figure 2. Model 1 shows an inner partial
VS1 cubic VS1
liquefaction model with a non-liquefied layer at the

disp. u g1 ( z1 , t ) u p1 ( z1 , t )

E p , I p , mp
top and model 2 indicates an overall liquefaction
VS 2 E p , I p , mp
VS 2 model near the ground surface.

engineering bedrock
The mean value of the maximum bending mo-
engineering bedrock
E F ment and shear force at the pile head may be evalu-
2E
(a) (b) ated by
Figure 1. Two approaches to using acceleration response spec-
{Z s(i ) ss(ij ) Z s( j ) + 2Z s(i ) sc(ij ) Zc( j ) + Zc(i ) cc(ij ) Zc( j ) }
n n
tra at engineering bedrock surface, (a) FEM, (b) Winkler-type M max =
spring). i =1 j =1

{Ys(i ) ss(ij )Ys( j ) + 2Ys(i ) sc(ij )Yc( j ) + Yc(i ) cc(ij )Yc( j ) }


n n
Model 1 Model 2
Qmax =
i =1 j =1

liquefaction liquefaction
where
Liquefiable

() ()
Z s( ) = EI S Ds Re[ ( ) (i ) ] , Z c( ) = EI S Dc Re[ ( ) (i ) ]
soil

i i i i i i

() ()
Ys( ) = EI S Ds Re[ ( ) (i ) ] , Yc( ) = EI S Dc Re[ ( ) (i ) ]
i i i i i i
Figure 2. Two cases of liquefaction and analytical model

In the above equations, EI is the bending stiffness


acceleration response spectrum(m/s )
2

10
damping ratio=0.05 of the pile and ( i ) is the curvature component at
8 the pile head in the i-th complex eigenmode. (i )
is the derivative of the curvature component with re-
6 spect to the pile axial coordinate at the pile head in
(i ) (i )
safety limit the i-th complex eigenmode. S Ds and S Dc are
level input
4 the sine and cosine spectra, respectively, (assumed
S A = 5.12 / T
here to be equal to the displacement response spec-
trum) and ( i ) is the i-th complex participation fac-
2 damage limit
level input
S A = 1.024/T
0
0.16 0.64 tor. This method has been developed in the refer-
0 0.5 1 1.5
natural period(s)
2 2.5 3 ences (Kishida & Takewaki 2010, Nakamura et al.
1996, Takewaki 2004, Kojima et al. 2013) based on
Figure 3. Acceleration response spectrum
the accomplishments for superstructures (Igusa et al.
1984, Yang et al. 1990). The accuracy and reliabil- Lysmers analogue velocity (= Vs here), the Win-
ity of this method have been confirmed in the refer- kler-type stiffness of the interaction spring (strain-
ences (Tassoulas & Kausel 1983, Kausel 2000). dependent equivalent one), the fundamental natural
The design earthquake ground motion defined in circular frequency of the superstructure. The radia-
the Japanese seismic-resistant design code (2000) is tion damping ratio is zero in 1 G and is linear
employed. The acceleration response spectrum is in G 1 2G . The final damping ratio at each
shown in Figure 3 (Safety-limit level). soil layer is evaluated at the fundamental natural
frequency of the superstructure as a frequency-
3 GROUND MODEL independent one. These data are used in the com-
putation of the kinematic effect, the inertial effect
A surface ground model, referred to as ground and the total response. From Figures 6(a), (b), it
model A (rather soft ground) is considered which in- can be seen that the present response spectrum
cludes a liquefiable sand layer. However, the soil method has a reasonable accuracy compared with
is treated in this section as the non-liquefaction the SHAKE program.
model. The soil profile is shown in Figure 4. The The accuracy of the response spectrum method
SPT values in the soil layers for ground model A are has been demonstrated (Kishida and Takewaki
also shown for reference. The mass density of sur- 2010). A Winkler-type continuum model has been
face soil layers is assumed to be 1.8 103 ( kg / m3 ) used as another model. Another comparison has
and Poissons ratio is 0.45. The ground model A been made with an actual record and the present
consists of six soil layers on the engineering bedrock method has been proven to be accurate enough when
(the mass density is assumed to be 2.0 103 taking into account the strain-amplitude nonlinearity
(kg / m3 ) ). In this ground model, each soil layer of soil deposit (Takewaki 2005).
has been divided into sub-layers of thickness of It will be shown in the next section that, by the
1(m). As a result, the ground model A consists of use of the response spectrum method, the distribu-
38 sub-layers (engineering bedrock is one layer). tions of bending moments in piles in the analysis of
To evaluate the strain-dependent nonlinearity of the kinematic and inertial interactions can be ob-
the ground, an equivalent linearization method tained efficiently.
(Takewaki 2004, Schnabel et al. 1972, Ching & Gla- 0 0

ser 2001) has been used. The strain-dependent 5 5

nonlinear relations are shown in Figure 5 for clay 10 10 Ground A


safety limit level/
and sand which are taken from the Japanese seismic- 15 depth (m) 15
depth (m)

damping ratio
Ground A (spectrum method)
resistant design code (June 2000). The effective 20
safety limit level/
20 safety limit level/
damping ratio
strain equal to 0.65 (the maximum strain) is as- 25 stiffness ratio
(spectrum method)
25 (SHAKE)
safety limit level/
sumed and several iterations are conducted for con- 30
stiffness ratio
30
(SHAKE)
vergence. The derived stiffness reduction ratios 35 35

and equivalent damping ratios by the response spec- 40


0 0.2 0.4 0.6 0.8 1 40
0 0.1 0.2 0.3 0.4 0.5
G/G0
trum method (Kishida & Takewaki 2010, Takewaki damping ratio
Figure 6. Convergent shear modulus ratio to the initial one and
2004) and the SHAKE program (Schnabel et al. damping ratio (safety limit level): response spectrum method
1972) for safety-limit level motion are shown in and SHAKE for one simulated motion (Takewaki 2004)
Figures 6(a), (b), respectively. The fundamental
natural period TG (= 2 / G ) of the ground model Table 1. p-multipliers to account for liquefaction
(Ashford et al. 2011)
A is 1.6(s). In case of using the SHAKE program,
ground-motion time histories are necessary and ten SPT Value p-multiplier
simulated ground motions have been used (Kishida <8 0.0 to 0.1
& Takewaki 2010, Takewaki 2004). 8-16 0.05 to 0.2
The damping ratios derived here are substituted 16-24 0.1 to 0.3
into the term of the hysteretic damping and the >24 0.2 to 0.5
following radiation damping is added to those ones
(Gazetas & Dobry 1984, Kavvadas & Gazetas
1993). 4 BUILDING MODEL ON GROUND MODEL
INCLUDING A LIQUEFIABLE SOIL LAYER
5

1
2k d V Vc 4 1d 4
cx = x 1 s s +
V V
1 + (1 2G ) A two-story building model is considered. The
G k x
s s floor mass of the building for a single pile is

40 103(kg) and the corresponding foundation mass
In this equation, cx , d , s , Vs , Vc , k x , 1 are is 120 103(kg). A cast-in-place reinforced con-
the damping coefficient of the interaction dashpot, crete pile is used and its pile diameter is 1.5(m).
the pile diameter, the soil mass density, the soil The Youngs modulus of concrete is 2.1 1010
shear velocity (strain-dependent equivalent one), the (N/m2) and the concrete mass density is 2.4 103
(kg/m3). The fundamental natural period TB of 5 COMPARISON WITH TIME-HISTORY RE-
the model with a fixed base is 0.3(s). SPONSE ANALYSIS
A coefficient 1/9 in the p-multiplier method has
been employed based on the work (Table 1: Ashford Since it seems important to investigate the accuracy
et al. 2011). The coefficient has been multiplied on and reliability of the present response spectrum
the equivalent shear modulus of the liquefiable soil method, a time-history response analysis has been
(see Figure 6). The coefficient 1/9 is a little bit out conducted and compared using the multi-input mod-
of recommendation (recommendation is 0.0-0.10), el (see Figure 1(b)) where the interaction spring is
but it seems reasonable judging from the complicat- modeled by the Hardin-Drnevich model (Hardin &
ed phenomena during soil liquefaction. Drnevich 1972) and the Masing hysteretic rule.
Figure 7 shows the pile bending moments after The coefficient 1/9 has been multiplied on the shear
liquefaction for models 1 and 2. Two reduced co- modulus of the liquefiable soil (from GL-4(m) to
efficients (1/9 and 1/100) in the p-multiplier method GL-10(m)). In the multi-input model, the input
have been employed (Ashford et al. 2011) and the ground motions at many depth levels have been gen-
case of 1/9 is shown in Figure 7. It has been demon- erated by using SHAKE program in consideration of
strated that the combination of the response spec- this coefficient and the interaction spring has been
trum method with the p-multiplier method for evalu- modeled by taking this coefficient into account.
ating the deterioration of soil resistance in liquefied Figure 7(a) shows the comparison of the pile
layers is a simple and effective method. It has also bending moments between the present response
been shown that, in the case of inner partial liquefac- spectrum method and a multi-input model including
tion with a non-liquefied layer at the top, the de- nonlinear horizontal interaction springs placed at
mand on the pile moment is large due to the inertial every pile node (every 1m) (Liquefaction model 1,
effect of that non-liquefied layer at the top and, in the coefficient of the p-multiplier method=1/9). It
the case of overall liquefaction near the ground sur- can be observed that, while the present response
face, the demand is smaller than the case of inner spectrum method provides somewhat different re-
partial liquefaction. This may result from the effect sponse distributions of pile stresses due to the en-
of action of the liquefied soil layer near the ground semble-mean characteristic of the response spectrum
surface on the pile. The softer ground stiffness method, the overall properties can be predicted with-
near the ground surface seems to weaken the action in an acceptable accuracy. Especially the bending
on the pile by reducing the dynamic effect of the moment at the pile top exhibits a good correspond-
liquefied soil in Liquefaction model 2. ence.
Figure 8 presents the pile bending moments for In Figures 7 and 8, the pile was assumed to
the non-liquefaction model. It can be found that behave elastically. In order to make the pile
both liquefaction models exhibit larger bending response more realistic, the elastic-plastic behavior
moments than the non-liquefaction model. of the pile is considered. The concrete strength is
0 0
21MPa and the stiffness-degrading trilinear model is
5 5
used.
kinematic
10 10 total
inertial
15 15
depth (m)
depth (m)

Response Response Spectrum Method


Spectrum
Method 20
20
Nonlinear
Time-history 25
25
Response
Analysis 30
30
2-story (Ground A) 35
35 2-story (Ground A)
7 7 7
40 7 7 7
0 1 10 2 10 3 10 0 1 10 2 10 3 10
bending moment (Nm) bending moment (Nm)

(a) (b)
Figure 7. Pile bending moment after liquefaction (left: model
1, right: model 2; reduced coefficient=1/9)
0
(a) (b)
5
Figure 9. (a) Force-deformation relation of the interaction
kinematic spring at the ground surface level, (b) Bending moment-
10 total
inertial rotation relation at the pile head
15
depth (m)

20
Figure 9(a) shows the force-deformation relation
25
of the interaction spring at the ground surface level
30
and Figure 9(b) presents the bending moment-
35
2-story (Ground A) rotation relation of the pile at the pile head.
40
0 1 107 2 107
bending moment (Nm)

Figure 8. Pile bending moment (non-liquefaction, response


spectrum method)
Figures 12(a) and (b) show the maximum
displacements of the pile and soil for the elastic pile
response and elastic-plastic pile response when the
soil is not liquefied.

(a)

Figure 13. Maximum displacements of pile and soil (Soil:


Liquefaction)
Figures 13(a) and (b) present the maximum
displacements of the pile and soil for the elastic pile
(b) response and elastic-plastic pile response when the
Figure 10. Time-histories of the force and deformation of the
soil is liquefied. It can be observed that the pile
interaction spring at the ground surface level
displacement is influenced greatly by the inclusion
Figures 10(a) and (b) illustrate the time-histories of the plasticity in the pile.
of the force and deformation of the interaction
spring at the ground surface level .

Figure 14. Time-histories of the pile-head displacement and the


(a)
free-field ground surface displacement

Figure 14 shows the time-histories of the pile-


head displacement and the free-field ground surface
displacement. A residual deformation can be
observed in the pile.

(b)
Figure 11. Time-histories of the bending moment and rotation
at the pile head
Figure 11(a) and (b) show the time-histories of
the bending moment and rotation at the pile head.

Figure 15. Bending moment-rotation relation at GL-10m

Figure 15 illustrates the bending moment-rotation


relation at GL-10m. It can be observed that the
pile rotation demand at GL-10m is not large
comapared to that at the pile head (Figure 9(b)).
This is because the pile head rotation is fixed.
Figure 12. Maximum displacements of pile and soil (Soil:
Nonliquefaction)
6 CONCLUSIONS Japanese Geotechnical Society 1996. Special Issue of Soils and
Foundations.
Kausel E. 2000. The thin-layer method in seismology and
The conclusions may be summarized as follows: earthquake engineering, in Wave Motion in Earthquake En-
(1) A response spectrum method in terms of com- gineering, WIT Press, 193-213.
plex modal quantities is a useful method for Kavvadas M & Gazetas G. 1993. Kinematic seismic response
evaluating the overall seismic responses of a and bending of free-head piles in layered soil. Geotechnique
building-pile-soil model including a liquefiable 43(2): 20722.
soil layer to a ground motion defined at the en- Kishida, A. & Takewaki, I. 2010. Response spectrum method
for kinematic soil-pile interaction analysis, Adv. in Struct.
gineering bedrock surface. Eng., 13(1), 181-197.
(2) The combination of the response spectrum meth- Kojima, K, Fujita, K. & Takewaki, I. 2014. Unified analysis of
od with the p-multiplier method for evaluating kinematic and inertial earthquake pile responses via single-
the deterioration of soil resistance in liquefied input response spectrum method, Soil Dynamics and Earth-
layers leads to a simple and effective method. quake Engineering, 63: 36-55.
Lysmer J, Udaka T, Tsai C-F & Seed HB. 1975. FLUSH: A
The reliability and accuracy of the p-multiplier computer program for rapproximate3-D analysis of soil-
method have been investigated through the com- structure interaction problems, Earthquake Engineering Re-
parison with experimental results in the previous search Center Report/No. EERC 75-30, Univ. of California,
research. The reliability and accuracy of the Berkeley.
single-input response spectrum method used in Miyamoto Y, Fukuoka A, Adachi N & Sako Y. 1996. Pile re-
sponse induced by inertial and kinematic interaction in liq-
the present paper with the help of the p- uefied soil deposit, J. Structural Construction Engineering,
multiplier method have been verified through the AIJ, 482: 53-62 (in Japanese).
comparison with the result by the time-history Murono Y. & Nishimura A. 2000. Evaluation of seismic force
response analysis including the multi-input from of pile foundation induced by inertial and kinematic interac-
the free-field ground and the nonlinear restoring- tion, Proc. 12th World Conference on Earthquake Engi-
force characteristics around the pile. neering, Paper No.1496.
Mylonakis G., Nikolaou A. & Gazetas G. 1997. Soil-pile-
(3) While in the case of inner partial liquefaction bridge seismic interaction: Kinematic and inertial effects.
with a non-liquefied layer at the top, the demand Part I; Soft soil, Earthquake Engineering and Structural
on the pile moment is large due to the inertial ef- Dynamics 26: 337-359.
fect of that non-liquefied layer at the top, in the Nakamura T, Takewaki I & Asaoka Y. 1996. Sequential stiff-
case of overall liquefaction near the ground sur- ness design for seismic drift ranges of a shear building-pile-
soil system, Earthquake Engineering and Structural Dy-
face, the demand is smaller than the case of inner namics 25(12): 405-420.
partial liquefaction. Newmark NM & Hall WJ. 1982. Earthquake Spectra and De-
sign, Earthquake Eng. Research Institute, Berkeley, CA.
Schnabel PB, Lysmer J & Seed HB. 1972. SHAKE: A com-
ACKNOWLEDGEMENTS puter program for earthquake response analysis of horizon-
tally layered sites, A computer program distributed by NI-
Part of the present work is supported by the Grant- SEE/ Computer Applications, Berkeley.
in-Aid for Scientific Research of Japan Society for Takewaki I. 1998. Remarkable response amplification of build-
the Promotion of Science (No.24246095, ing frames due to resonance with the surface ground, Soil
Dynamics and Earthquake Engineering 17(4): 211-218.
No.25630233). This support is greatly appreciated. Takewaki, I. 2004. Response spectrum method for nonlinear
surface ground analysis, Adv. in Struct. Eng., 7(6), 503-514
Takewaki I. 2005. Frequency domain analysis of earthquake
REFERENCES input energy to structure-pile systems, Engineering Struc-
tures 27(4): 549-563.
AIJ (ed) 2012. Preliminary reconnaissance report of the 2011 Tassoulas JL & Kausel E. 1983. Elements for the numerical
Tohoku-Chiho Taiheiyo-Oki earthquake, Springer. analysis of wave motion in layered strata. Int. J. Numerical
Ashford, S, Boulanger, R & Brandenberg, S. 2011. Recom- Methods in Eng. 19(7): 10051032.
mended design practice for pile foundations in laterally Tokimatsu K., Suzuki H. & Sato M. 2005. Effects of inertial
spreading ground, PEER Report, No.PEER 2011/04. and kinematic interaction on seismic behavior of pile with
Ching J-Y, & Glaser SD. 2001. 1D time-domain solution for embedded foundation, Soil Dynamics and Earthquake En-
seismic ground motion prediction. J. Geotechnical Geoenvi- gineering 25: 753-762.
ronmental Engineering, ASCE 127(1): 36-47. Yang JN, Sarkani S & Long FX. 1990. Response spectrum ap-
Gazetas G. & Dobry R. 1984. Horizontal response of piles in proach for seismic analysis of nonclassically damped struc-
layered soils, J. Geotech. Eng., ASCE 110(1): 20-40. tures, Engineering Structures 12: 173-184.
Hardin, B.O. & Drnevich, V.P. 1972. Shear modulus and
damping in soils: Design equations and curves. J. Soil
Mech. Found. Div. ASCE; 98(SM7): 667-692.
Idriss, IM & Boulanger RW. 2008. Soil liquefaction during
earthquakes, EERI.
Igusa T, Der Kiureghian A & Sackman J 1984. Modal decom-
position method for stationary response of non-classically
damped systems, Earthquake Engineering and Structural
Dynamics 12(1): 121-136.

You might also like