0033

DYNAMIC FE ANALYSIS FOR STABILITY OF EMBANKMENT ON SILTY CLAY

DURING EARTHQUAKES
K YASUHARA1 , S MURAKAMI2 And O MATSUO 3

SUMMARY
Based on a series of dynamic centrifuge tests (1:50 scale) on embankments founded on clay the
following was made clear : (i) a dynamic load with = 250 gal, for example, induced large
deformation and settlement in both embankment and clay under a gravity field of 50g; (ii)
deformation and instability were minimized by installing the cement treated layer up to the bottom
of ground. Despite these findings, however, it is still uncertain whether embankment should be
safer than a clay foundation or which embankments or soil layers would give rise to instability
during strong motion earthquakes. This paper presents the results of dynamic FE method applied
to above problem. Increasing strength of embankment material and cement treated clay leads to
increased overall stability of the embankment and foundation together. A design chart has been
produced in order to decide which combination between embankment and cement treated clay
foundation would tend to be more stable during earthquakes such that engineers can decide
whether it is necessary to strengthen the embankment or foundation layer

INTRODUCTION
It has generally been said that fine-grained soils during earthquakes is more stable than sandy soils. However,
some of the previously reported large earthquakes such as in Alaska (1964) or Mexico (1985) [Mendoza and
Auvenet, 1988] as well as the recent earthquakes such as in Sagueney (1987) [Lefebrre et al, 1992] Loma Prieta
(1989) [Boulanger et al, 1989] or Northridge (1994) [Holtzer et al, 1999] fine-grained soils including from
highly plastic Mexico city clay to lean clay or low plasticity silt have suffered from damages in surface ground
or structures on ground. The authors have concluded that fine-grained soils in these case records must play a
role, to a greater or lesser degree, in causing damages during and after earthquakes [Yasuhara et al, 1998].

Recently in Japan, the high-graded embankments for the large rivers near the urban areas have sometimes been
constructed or have been under construction on soft cohesive soils. For preventing serious damages from these
river dykes on cohesive soils, a careful attention is required from seismic stability even if fine-grained soils were
more stable than sandy soils because most of the large rivers pass through the urbanized area of the large cities.
The present paper introduces the design chart for monitoring the situation of river embankments from stability to
instability and performing the earthworks for banking of materials. The method proposed is constructed by
plotting both the settlement ratios of ground and embankments based on the results from dynamic analysis using
a modified FE code, SADAP which was developed by PWRI (Public Works research Institute), Ministry of
Construction, Japan. From the design chart, in particular, an attempt has been made to estimate the effects of
cement stabilization on increasing the stability of embankment and ground during earthquakes.

1
2
3

Department of Urban and Civil Engineering, Ibaraki University, Hitachi, Japan, email:[email protected]
Department of Urban and Civil Engineering, Ibaraki University, Hitachi, Japan, email:[email protected]
Soil Dynamics Division,Public Works Research Institute, Ministry og Construction, Japan

2. DYNAMIC-CENTRIFUGE TESTS ON EMBAKMENT UNDERLAIN BY SILTY CLAY

2.1 OutlineofCentrifugeTests
Aseriesofcentrifugemodeltestsforinvestigatingthe behaviourofembankmentsonfine-grainedsoilsandthe effect of
cement treated layeragainstits eformation and instability were carried out at Soil Dynamic Research Centre,
Public Work Research Institute, Ministry of Construction. The details of the experiments were herefore
described in the previous paper [Matsuo et al., 1997]. The modeled embankment on clay as illustrated in Fig. 1
was installed in the box-type soil tank with 89cm width, 30cm height and 20cm length. Three

Figure 1: Cross Section of 2-D Dynamic Centrifuge Model Tests (Matsuo et al, 1997)
Kinds of the models were adopted : the model embankment with cement treated layer (CTL) up to full
foundation depth ; (cases of M3 and M2) CTL up to certain depth and the model without CTL (case of M1). The
representative index properties are s = 2.662, wn = 51%, Ip = 17). The stabilising layer was formed by mixing
cement with clay soil. To enhance the strength of cement-stabilized column, the model with the stabilised layer
was left for 7days before starting the centrifuge test. The unconfined compression strength of silt for ground at
this stage was 90 kPa on the average. After consolidating the modeled silt ground with and without the
stabilising layer, dynamic loads were applied with and without sinusoidal wave and 60Hz to the modeled ground
by stepping 4 stages from 100 gal to 350 gal. The number of load cycles was 20 for each step. The centrifuge
acceleration rate was 50G during consolidation through dynamic loading.
2.2 Summary of the Results
Fig. 2 shows a set of typical illustration of deformation manner at the embankment and ground system observed
at the final stage of dynamic loading with 350 gal. The following was pointed out from Fig. 2 :
(i)

a dynamic load with = 250 gal, for example, induced large deformation and settlement in clay under a
gravity field of 50 gal. In addition, failure and large deformations were also observed in the embankments;

0033

(ii)

deformation and instability were minimized by installing the cement stabilising layer up to the bottom of
the clay foundation ground ;

(iii) cement stabilized columns with partial depth did not have significant decrease in the deformation and
instability of the clay foundation and embankment.
(iv) overall configuration of embankment and ground, particularly in the case without cement stabilising
column seems to belong to lateral flow deformation rather than circular failure.

Figure 2: A Typical Example of Deformation Obtained in Dynamic Centrifuge Tests (Matsuo et al, 1997)
(CTL: Cement Treated layer)
3.

ANALYTICAL PROCEDURE

The computer code "SADAP" used in the present paper was developed by PWRI (Public Works research
Institute), Ministry of Construction, Japan. This code incorporates the two models: (1) Hardin-Drnevich model
[Hardin and Drnevich, 1972] for hyperbolic stress-strain relations during cyclic loading, (2) Ishihara-Towhata
model [Ishihara and Towhata, 1980] for pore pressure generation and dissipation. It is characterized that it is
capable of considering the degradation of stiffness step by step which is caused by generation of excess pore
pressures during cyclic loading. The following is the procedure of computation using SADAP:
1)

in the static analysis where displacement of ground due to self-weight of embankment is determined,
construct the element stiffness matrix [k] for each element using the isotropic elastic stress-strain matrix [D].

2)

determine the whole stiffness matrix [K] by superimposing [k], and calculate the displacement matrix {u}
by solving:

{f } = [K ]{u}
3)

(1)

on the other hand, in the dynamic response analysis, displacement during earthquakes is evaluated by
integrating directly the motion equation:

[M ]{u&&} + [C ]{u& } + [K ]{u} = {Q(t)}

(2)

0033

4)

for every time step using the Newmark- method. In the equation, [M] is mass matrix, [C] is damping
matrix, [K] is stiffness matrix, and {Q(t)} is external force matrix.

5)

predict the excess pore pressure corresponding to the calculated shear stress determined through the
displacement and stress calculated using Eq. (2). The magnitude of excess pore pressure is reflected to
evaluation of degradation in stiffness of soils [Yasuhara and Hyde, 1997]. This has been given by :

Gi , cy
Gi , NC

1 C

1
ln( OCRq )

(3)
OCRq where C and are experimental constants [Yasuhara and Hyde, 1997] and OCR
q

is given by:

OCRq =

1
1 u / p' c

(4)

This process is repeated until a prescribed time period is reached. In this way, deformation behavior during
earthquakes can be predicted from time to time.
3.1 Parameter Determination
The parameters for index and mechanical properties of soils constituting embankment, ground and cementstabilized body are summarized in Table 1. Strength parameters c' and ' were determined from undrained
triaxial compression tests on each soil specimen. The initial shear modulus Gi was determined by extrapolating G
~ curves obtained from dynamic deformation triaxial tests. To determine B'p and B'u related for generation of
excess pore pressures during earthquakes, undrained cyclic triaxial tests were carried out.

Figure 3: Various Models of Embankment on Cement Treated Soft Soil

Table 1 : Input Data for Finite Element Analysis

Sand
Clay
Emb
CTL

Gi
(tf/m2)

t
(tf/m3
)

c
(tf/m2)

(deg)

K0

n1

Bp

Bu

10000
406.4
400
11180

1.75
1.75
1.70
1.57

30.0
0.56
1.5
18.17

40.0
32.79
10.5
19.39

.357
.408
.45
.617

.263
.290
.3
.381

.263
.290
.4
.381

1000
812.7
800
22360

0.6
.497
0.5
0.5

1
1
1
1

1.1
.75
0
0.1

.2
.05
0
.02

40
45
0
45

0033

4. RESULTS FROM NUMERICAL ANALYSIS

The model grounds embankments used for numerical analysis are shown in Fig. 3. The present study particularly
aims at explaining the effects of cement-treated layers. In addition, the effect of the sinusoidal (regular) and
impact (irregular) waves were considered in the analysis. The acceleration rate for each case was 350gal and
400gal, respectively. Fig. 4 illustrates a set of the typical results of FE analysis to compare the seismic-induced
deformation pattern of embankment founded on fine-grained soils with and without cement-treated layers. Fig.
4b shows the results corresponding to models in Fig. 3a. It demonstrates that installation of cement-treated layers
prevents from large settlements and deformation of

Figure 4: Deformation Pattern of Example Problem

both embankment and foundation together. In order to understand the mechanism and the effects of cementtreated layers on improvement for increasing stability of embankment on grounds during earthquakes, an attempt
was made to construct a chart for execution management. For this purpose, the results from numerical analyses
are shown in Fig. 5 in which the plots are given as both the settlement ratios Sg /Hg (Sg : settlement of ground and
Hg : height of ground layer) and Se/He (Se : settlement of embankment and He : height of embankment) in
longitudinal and horizontal axes. Figs. 5a and 5b correspond to the cases of sinusoidal and impact waves,
respectively, as shown in Fig. 6a and Fig. 6b which are induced by earthquakes. The straight lines drawn in both
figures indicate the critical value of settlement (20cm in this case) for high-graded river embankment based on
the design manual tentatively regulated by the Ministry of Construction, Japan (1998). In other words, the
embankment of soft soils would be safer if the plot were within the triangular space surrounded by the
longitudinal axis, horizontal axis and this critical line. From Fig. 5, the following trends have been observed:
1) the wider the cement-treated layers are, the more marked the decrease in settlement of embankment is.
2) the deeper cement-treated layers are more effective for diminishing the settlement of grounds.
Therefore, configuration of cement-stabilisation of ground by the case <1> is the most effective among

0033

Figure 5: Plots of Settlement Ratio for Embankment and Ground Obtained from Dynamic FEM

Figure 6: Sinusoidal and Impact Waves Used in FEM

the cases shown in Fig. 3 from decreasing the settlement and increasing the stability of river embankment on soft
grounds. In other words, cement stabilized columns which were not long enough to reach the bottom of the clay
foundation layer did not decrease the deformation and instability of the clay foundation and embankment. The
above-stated behavior is in very good agreement with that observed in dynamic centrifuge model tests reported
earlier.

0033

5. CONCLUSIONS

1) Behaviour of embankment founded on fine-grained soils during simulated earthquake loading can be
predicted by the dynamic analysis model which is capable of considering the cyclic-induced degradation in
stiffness of soils.
2) The computer code modified from SADAP which was developed at PWRI is useful for explaining the
effectiveness of installing cement-treated layers in embankment and foundation on increasing stability and
decreasing settlements during earthquakes.

REFERENCES
Boulanger, R. W., Meyers, M. W., Mejia, L. H. and Idriss, I. M. (1989), Behavior of a fine-grained soil during
the Loma Prieta earthquake, Can. Geotech. J, 35, pp146-158.
Holtzer, T. L., Bennett, M. J., Ponti, D. J. and Tinsley III, J. C. (1999), Liquefaction and soil failure during
1994 Northridge earthquake, J. Geotechnical & Geoenvironmental Eng., ASCE, 125, 6, pp438-452.
Ishihara, K. and Towhata, I. (1980), One-dimensional soil response analysis during earthquakes based on
effective stress method, J. of the Faculty of Engineering, The University of Tokyo, (B), XXXV, 4, pp655-700.
Lefebrre, L., Lebouf, D., Horynch, P. and Tanguay, L. (1992), Slope failures associated with the 1998
Sanguenay earthquake, Quebec, Canada, Canadian Geotechnical Journal, 29, pp117-130.
Matsuo, O., Shimazu, T. and Tamoto, S. (1997), Investigation on seismic stability of embankment founded on
cohesive soils , J. Civil Eng. Techniques, 39, 9, pp56-61,(in Japanese).
Mendoza, M. J. and Auvenet, G. (1988), The Mexico earthquake of September 19, 1985 - Behavior of building
foundations in Mexico city, Earthquake Spectra, 44, pp753-770.
Public Work Research Institute (1981), Manual for SADAP, Ministry of Construction, Japan (in Japanese).
Stark, T. D., Contreras, I. A. (1998), Fourth avenue landslide during 1964 Alaskan earthquake, J. Geotech. &
Geoenviron. Eng.,Proc. ASCE, 124,2, pp99 109.
Yasuhara, K. and Hyde, A.F. L. (1997), Method for estimating postcyclic undrained secant modulus of clays,
J. of Geotechnical and Geoenvironmental Eng., ASCE, 123, 3, pp204-211.
Yasuhara, K., Tokita, K., Matsuo, O. and Nasu, M. (1998), Chapter 3, Case histories on dynamic properties of
cohesive soils , "Dynamic Properties of Cohesive Soils", "Tsuchi-to-Kiso", J. of JGS, 46, 7, pp57-62, (in
Japanese).

0033