Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

PAPERS

Characteristics of Local Site Effects on Seismic Motion


- Non-linearity of Soil and Geological Irregularity -
Yoshitaka MURONO Akihiko NISHIMURA
Engineer Manager
Foundation & Geotechnical Engineering G., Structure Technology Development Div.
Structure Technology Development Div.

To determine seismic ground motion for aseismic design, it is necessary to take site
effects into account, such as non-linearity of soil and geological irregularities. In this paper,
we discuss two aspects of site effects. One is the estimation of nonlinear earthquake re-
sponse of soil deposits, and the other the estimation of seismic motion on the ground that
has topographical and geological irregularities. We propose a new stress-strain model
that is applicable to earthquake response analyses of soil deposits under large earthquakes.
Next, we develop a practical method to evaluate amplification characteristics of ground
motion on the irregularly layered ground.

Keywords : site effects, seismic motion, non-linearity, geological irregularity

1. Introduction ered soil deposit during large earthquakes, it is necessary


for a shear stress-strain model to satisfy the following
Earthquake waves travel from a fault to the ground conditions; the model i) can describe the stress-strain re-
surface through the seismic bedrock and engineering seis- lationship (τ~γ) of various geomaterials ranging from
mic base layer. Finally, the earthquake motion propagates soft clay to hard rock over a wide pre-failure strain range,
vertically in the surface ground and reaches at the ground ii) has a minimum possible number of parameters, each
surface (see Fig.1). Since surface ground is soft, it may having clear physical meaning, iii) can express the damp-
exhibit nonlinear behavior caused by the vertically propa- ing-strain relationship (h ~γ ) over a wide strain range
gating S-wave during large earthquakes. A surface wave obtained from laboratory tests, iv) can reflect the concept
may also propagate at the site, which is influenced by the of failure strength and v) can easily be applicable to seis-
irregularity of base layer. These characteristics are so- mic design.
called ‘site effects’ in the field of earthquake engineering. A number of models have been proposed in the past,
To determine seismic ground motion for designing struc- including Hyperbolic model1),2), Ramberg-Osgood model 3)
tures, it is important to adequately estimate these site and others4),5),6). However, these models don’t satisfy all
effects. conditions mentioned above. In this paper, we propose a
We discuss the following two points in this paper. One new stress-strain model that satisfactorily meets these
is a share stress-strain model for nonlinear analysis of conditions.
horizontal layered soil deposit. The other is a simple
method to evaluate the influence of geological irregulari- 2. 1 Skeleton curve
ties.
The nonlinear model of shear stress-strain model con-
Surface wave Local site effects sists of two factors, skeleton curve and hysteretic curve.
Body wave We use GHE (General Hyperbolic Equation) model as
a skeleton curve. This model was originally proposed by
Tatsuoka and Shibuya7) and is described,
Surface ground
γ
Seismic bedrock Engineerig seismic base layer
τ = γr y= x
τf 1 + 1 γ or 1 + x ・・・・・・(1)
Path effects C1 C2 γ r C1 C2

where x is the normalized shear strain defined as x= γ/


γr and y is the normalized shear stress defined as y= τ/
Source effects
τf. τf is the failure stress and γr is the reference strain
Fig. 1 Schematic figure showing wave propagation from defined as
fault to ground surface
τf
γr = ・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・(2)
2. Shear strain-stress model for nonlinear analysis Gmax
of horizontal layered deposit
Two parameters C 1 and C 2 are defined as
When we perform nonlinear analysis of horizontal lay-

QR of RTRI, Vol. 40, No. 3, Oct. ’99 139


as follows (see Fig. 2).
1) We obtain C1(0)=1.0, for a particular condition that dy/
C1(0) + C1(∞) C1(0) – C1(∞) π
C1(x) = + ⋅ cos dx=0 at x=0.
2 2 α/x + 1
2) C2(0) is the intercept at the y-axis of the linear rela-
・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・(3a)
tion starting at the origin (y/x=1, y=0), fitted to the
C (0) + C2(∞) C2(0) – C2(∞) π initial part of the observed relation (experimental
C2(x) = 2 + ⋅ cos
2 2 β /x + 1 data).
・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・(3b) 3) The coefficient of C 1(x= ∞ ) and C 2(x= ∞ ) are the in-
tercepts at the y/x-axis and y-axis, respectively, of the
These equations have six parameters C 1(0), C 2(0), linear relation fitted to the observed stress-strain re-
C 1(∞), C2(∞), α and β. Tatsuoka and Shibuya originally lation at large strains.
applied this model to investigate the static behavior of 4) At the point ‘A’ in Fig2, the diagonal implying x=1 in-
soil7). In this study, we apply this model to investigate tersects with the observed relation line.
the dynamic behavior of soil, and determine these param- 5) We draw a line tangent to the observed relation at
eters from dynamic deformation characteristics such as point ‘A’. The intersects of this line with the y/x-axis
the G/G max ~γ relationship obtained from laboratory and y-axis are C1(x=1), C2(x=1), respectively.
tests. The values of these parameters can be determined 6) We obtain the value of α and β by substituting these
values of C 1(x=1) and C2(x=1) into the equation (3).
For example, the value of y(x=1) is 1.0 and 0.5 for the
linear relation and the Hyperbolic relation, respectively.
The results of fitting this model to the test data
(Kasumigaura sand) are shown in Fig.2 and Fig.3. Fig-
ure 2 shows the y~y/x relationship, and Fig.3 shows the
G/G max ~γ, τ~γ relationship compared with the labo-
ratory test data. We recognize that this model agrees with
the test data for very small to large strains.

2. 2 Hysteretic loop

We use Masing’s 2-nd rule as a hysteretic loop with


some improvements. When Masing’s rule is applied to the

Fig. 2 Determination of parameters for the GHE model


(Kasumigaura sand)

(a) G/Gmax ~γ/ γr relationship


(a) Method to modify Masing’s rule

(b) τ/ τf ~γ/ γr relationship (b) Damping ratio


Fig. 3 The results of fitting GHE model to test data Fig. 4 Schematic figure of revision method of Masing’
Masing’ss rule

140 QR of RTRI, Vol. 40, No. 3, Oct. ’99


equation (1), the hysteretic loop equation is given by Fig.4(b)).
This new skeleton curve satisfies the following two
γ ± γa conditions.
τ ± τa = Gmax ⋅
1 + 1 γ ± γa ・・・・・・・・・・・・(4a) 1) skeleton curve should pass through the turning point
C1 C2 2γ r A( γa, τa).
C (0) + C1(∞) C1(0) – C1(∞) π 2) damping constant h calculated from the histeretic
C1 = 1 + ⋅ cos
2 2 α{(γ ± γ a)/2γ r}+1 curve should safety h ~γ relationship of laboratory
tests.
C2(0) + C2(∞) C2(0) – C2(∞) π
C2 = + ⋅ cos
2 2 β {(γ ± γ a)/2γ r}+1
・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・・(4b)

The damping ratio h, which is calculated by the area


surrounded by the hysteretic loop (eq. (4)), is overesti-
mated in the region of large strains. In order to avoid this
problem, we assume a temporary skeleton curve when the
strain changes its direction (see Fig.4(a) root O→A→B).
When Gmax is small, the area surrounded by the hyster-
etic curve becomes small, and also when G max is large, it
becomes large. Therefore, we regarded G max of hysteretic
curve (eq.(4)) as a parameter that depends on the strain
level in this study, and decided G max so that the damping
ratio h calculated from the histeretic curve satisfies the
h ~γ relationship obtained from laboratory tests (see Fig. 5 Stress-Strain curves obtained from various model

Fig. 6 Comparison of experimental maximum acceleration response and displacement response with simulated values

QR of RTRI, Vol. 40, No. 3, Oct. ’99 141


2. 3 Shaking table test motion is amplified to cause severer damage. It is diffi-
cult to propose a method to precisely estimate such ground
We carried out a shaking table test with a large-scale motion amplification, because so many factors are in-
laminar box filled with dry conditioned sand in order to volved in various ways. Nevertheless, a method to evalu-
verify the accuracy of the proposed model. The size of ate the ground motion amplification to be considered in
laminar box is 12×6×3.5m. Acceleration of soil and de- seismic design is necessary, even if it does not provide
formation of box are measured. We simulated the experi- perfectly correct estimation.
mental result by using the following five models, ① Hy- In this chapter, we suggest a method to estimate the
perbolic model (Case1), ② Hyperbolic model (Case 2), ③ amplification of earthquake motion on irregular grounds
RO model, ④ GHE model + normal Masing’s rule and ⑤ with reasonable accuracy.
the proposed model described in Section 2.2 (GHE mode
+ Modified Masing’s rule). When we use Hyperbolic 3. 1 Conditions for analysis
model, there are two methods to define the reference
strain γr. In case1, we define γr as the strain when G/ Since ground motion amplification is affected by vari-
G max becomes 0.5. In case2, we calculated γr from eq(2). ous factors including the scale of irregularity and input
Stress-strain curves are shown in Fig.5, and peak re- motion characteristics, it is almost impossible to analyze
sponse values are shown in Fig.6. the phenomenon by taking all factors into account. There-
The Hyperbolic model (Case1) underestimates accel-
eration for large inputs because the maximum shear
stress τ f is underestimated as seen in Fig.5.
The Hyperbolic model (Case2) shows larger peak ac-
celeration and smaller peak displacement for 100 ~ 400
(gal) inputs. But it shows better agreement in peak ac-
celeration for the G0C input. Because the Hyperbolic
model (Case2) overestimates the shear stress in small to
middle strain range, and estimates the shear stress ad-
equately at the large strain level.
The Ramberg-Osgood and GHE + Masing models show
acceleration more accurately because these models fit the
τ~γ relationship adequately. But these models under-
estimate the displacement about 10 ~ 20% for large ac-
celeration inputs such as the G0C wave, because the
damping ratio h is overestimated in the region of large
strains.
On the other hand, the proposed model can satisfac-
torily simulate both peak acceleration and displacement.
We can say that, in order to estimate the nonlinear
behavior of ground motion in the surface ground, it is nec-
essary to evaluate both the τ~γ relationship and the h Fig. 8 Response lateral acceleration obtained from 2D-
~γ relationship. FEM analysis

3. Effects of geological irregularity

Through an investigation of damage caused by earth-


quakes, it has been found that, on a ground which has
geological or topographical irregularities, seismic ground

Fig. 7 Ground models for 1D and 2D analyses Fig. 9 Time history of wave obtained from (2D)-(1D)
Time

142 QR of RTRI, Vol. 40, No. 3, Oct. ’99


fore, the grounds with rather simple irregularities are frequency domain).
considered in this study,.
G(ω, x) = F(ω) +α ⋅ F(ω) ⋅e– iω∆t
3. 1. 1 Ground models ・・(5)
Two-dimensional (2D) and one-dimensional (1D) = F(ω) ⋅ {1 + α ⋅ e– iω∆t} = F(ω) ⋅ η(ω)
analyses are considered in this study (See Fig.7). For 2D
analyses, we constructed ground models by changing the where G( ω , x) is the fourier spectra of seismic mo-
following three parameters: inclination of hard layer (θ), tion at the ground surface; F(ω) is the fourier spectra of
thickness of soft layer (H), and impedance ratio ( κ). In direct SH wave f(t); △t is a time delay; and α is the am-
1D analyses, all conditions such as soil properties and plitude ratio of f(t) and f ’(t).
thicknesses of soft and hard layers are set equal to those
of the corresponding 2D models as shown in the Fig.7. 3. 2. 2 Amplitude ratio α
The amplitude ratio α is defined as
3. 1. 2 Mechanism of seismic motion amplification due to
irregularity f ′(t)
α= ・・・・・・・・・・・・・・・・・・・・・・・・・・・・(6)
Fig.8 shows the computed acceleration response for a f (t)
SH incident Ricker wavelet having a predominant period
of 0.8 sec, which is the same as the fundamental resonance
period of the layer. As the soft layer is thick, accelera-
tion is amplified. Especially, its amplitude becomes 2.4
times that of incident motion near the edge of the basin
(x=40m). The duration time of response becomes longer
at places remote from the edge of basin.
These characteristics are due to ground irregularities.
To separate the effects purely due to ground irregulari-
ties, we deduct the 1D response time histories from the
2D response time histories (See Fig.9). We can see a lat-
erally propagating wave which is generated at the edge
of the basin. Since the phase velocity of this wave is nearly
equal to that of Raleigh wave, the Raleigh wave is a ma-
jor component of this laterally propagating wave. From
these results, it may be said the seismic wave on the (a) Influence of inclination θ
ground having irregularities consists of the direct propa-
gating SH wave and the laterally propagating Raleigh
wave.

3. 2 The evaluation of irregularity

3. 2. 1 Method to estimate ground motion


We suppose the following three assumptions.
1) The seismic response on a ground having irregulari-
ties has two component waves. One is a direct propa-
gating wave f(t) and the other a laterally propagat-
ing wave f ’(t) (Raleigh wave) which arrives with some
time delay.
2) The laterally propagating wave arrives after ∆ t sec (b) Influence of thickness H
of the arrival of the direct wave.
3) As for the laterally propagating wave, the amplitude
is α times that of SH wave.
Under this assumptions, the seismic motion on a
ground having irregularities is described as follows (in the

x = 0(m)
+ direction

Soft layer

Hard layer

(c) Influence of impedance ratio κ


Fig. 10 Distance x from the edge of basin Fig. 11 Amplification ratio α
11

QR of RTRI, Vol. 40, No. 3, Oct. ’99 143


Although α, f(t) and f ’(t) are functions of x, which is (3) κ : Impedance ratio
the distance from the edge shown in Fig.10, it is not writ- The impedance ratio κ plays a most important role
ten explicitly. We investigate how each factor affects α to determine the amplitude ratio α as shown in Fig.11(c).
by earthquake response analyses of ground models shown From these results, we propose the following equation
in Section 3.1.1. to determine the amplitude ratio α .
(1) θ : Inclination of hard layer
The inclination θ affects the amplitude ratio α sig-
α = 0.3 × exp – 7.0 × 1
κ × exp( – 0.44 X) ・・・・(7)
nificantly as shown in Fig.11(a). θ
(2) H: Thickness of soft layer
Thickness of soft layer does not affect the amplitude 3. 2. 3 Time delay △ t
ratio α , when we take the normalized distance X = x/H The time delay △t is determined as shown in Fig.12
on the cross axis as shown in Fig.11(b). and is described as

L
∆t = t1 + t2 + t3 – t4 = H + H + x – H ・・・・・・(8)
Vb C1 C2(ω) Vs

The definitions of H, LH and x are shown in Fig.12. Vb


and Vs are the shear velocity of the base and soft layer,
respectively. C1 is the traveling velocity of laterally propa-
gating wave in the region where the hard layer slants.
C2( ω ) is the phase velocity of Raleigh wave.

3. 3 Applicability of the proposed method


Fig. 12 Determination of time delay ∆ t
We investigate the applicability of the proposed
method by comparing test results with these of 2D-FEM
dynamic analysis. Figure 13 shows the fourier amplitude
of G(ω) compared with that obtained by 2D-FEM analy-
sis.
Figure 14 shows the time history of acceleration com-
pared with that obtained by 2D-FEM analysis. These re-
sults are compatible. We conclude that the proposed
method effectively determines the seismic motion ampli-
fication for seismic design.

4. Conclusion

We discussed the following two points in this paper.


One is a stress-strain model for analysis of horizontal lay-
ered soil deposit. The other is a simple method to evalu-
Fig. 13 Comparison of fourier amplitudes of 2D-FEM ate the influence of geological irregularities.
analysis with that of the proposed method (x = We developed a non-linear shear stress-strain model
20m) that is applicable to analyses of soil deposits during large
earthquakes. This model fits dynamic deformation char-
acteristics obtained from laboratory tests, such as G/Gmax
~γ ,h ~γ relationships over a wide strain range. Fur-
thermore, this model reflects failure limit stress τf. We
carried out a shaking table test to verify the effectiveness
of this model. From a comparison of analytical results
with the experimental results, it was proved that this
model was very effective for seismic design.
We also propose a practical method to estimate the
amplification characteristics of seismic motion on irregu-
lar grounds which have spatial variation of surface layer
thickness. This method is based on the general idea in
which vertically propagating SH wave and the horizon-
tally spread wave interfere with each other. The proposed
method gives good estimates for the frequency range 1~
3 Hz.
Fig. 14 Comparison of time histories of 2D-FEM analysis
with that of the proposed method (x = 0, 20, 40 m)

144 QR of RTRI, Vol. 40, No. 3, Oct. ’99


References 5) Wakai, A., Ugai, K., Li, Q., Matsuo, O. and Shimazu,
T. : “Dynamic elasto-plastic analyses of the sliding dis-
1) Konder, R.L. : “Hyperbolic Stress-strain Response : placement during earthquake”, Proc. Int. Sym. on De-
Cohesive Soils”, Proc. ASCE, SM1, pp.115-143, 1963. formation and Progressive Failure in Geomechanics,
2) Haridin, B.O. and Drnevich, V.P. : “Shear modulus and pp.635-640, 1997.
damping in soils : design equations and curves”, Jour- 6) Yoshida, N., Tsujino, S. and Ishihara, K.: “Stress-
nal of SMFD, Proc., ASCE, Vol.98, No.SM7, pp.667- strain Model for Nonlinear Analysis of Horizontally
692, 1972. Layered Deposits”, Summaries of the Technical Pa-
3) Jennings, P.C. : “Periodic Response of a General Yield- pers of Annual Meeting of AIJ, pp.1639-1640, 1990.
ing Structure”, Proc. ASCE, EM2, pp.131-163, 1964. 7) Tatsuoka, F. and Shibuya, S. : “Deformation charac-
4) H. Hayashi and T. Sugawara : “Modeling of the non- teristics of soils and rocks form field and laboratory
linear shear stress-strain behavior of soils”, The 8th tests”, Theme Lecture 1, Proc. of Ninth Asian Regional
Japan Earthquake Engineering Symposium, pp.777- Conference on Soil Mechanics and Foundation Engi-
782, 1990. neering, Vol.2, pp.101-170, 1992.

QR of RTRI, Vol. 40, No. 3, Oct. ’99 145

You might also like