6th International Conference on Earthquake Geotechnical Engineering 1-4 November 2015 Christchurch, New Zealand

Numerical Simulation of Centrifuge Model Tests on Embankment with a New Liquefaction Countermeasure by Ground Improvement considering

Constraint Effect

Y. Okochi1, S. Sreng2, M. Matsumoto3, H. Miki4, M. Tsuda5 , H. Ito6

ABSTRACT This paper presents the results of three dimensional numerical simulations of centrifuge model

tests using a new liquefaction countermeasure for embankment foundation conducted by the authors. Two kinds of centrifuge model tests were conducted in 70g of centrifugal acceleration in order to investigate the basic behavior of embankment during liquefaction with and without countermeasure. A newly developed countermeasure which is considering the constraint effect was adopted in the experiment. Numerical simulations using fully coupled effective stress 3D FEM were carried out to evaluate the experimental results and applicability of the analysis program to solve the practical problem. Conventional countermeasure TOFT method was also analyzed. As a result, analyses showed reasonable agreement with the experiments, therefore it is suggested that the analysis method have a potential applicability to the practical design.

Introduction

As one of the reliable countermeasures for the liquefaction, lattice shaped ground improve method (TOFT method) has been adopted in Japan. Lattice shaped wall type stabilized bodies are constructed in this method and regarding the depth, whole liquefaction layer is solidified generally. For the spacing of the lattice, 1/2~1/3 of the thickness of liquefaction layer has been proposed in conventional design method to reduce the excess pore pressure inside the improvement effectively (Taya et. al. 2008). As a result, the improvement ratio is around 50%. However, if the deformation control of the ground is the main purpose, larger spacing can be adopted for this purpose by considering the confining effect of the stabilized bodies against the unimproved ground, which was statically examined by Miki et. al. 2011. This new ground improve method considering confining effect is called CGI. With this method, wider spacing of the lattice can be applied and liquefaction layer beneath the stabilized bodies can be left. As a result, the cost of countermeasure can be reduced significantly. A series of centrifugal model tests (half section models) and their 3D FE analysis have been conducted to investigate the effect of the liquefaction countermeasure for embankment foundation using a CGI method (Sreng et al. 2014). By comparing the conventional TOFT method and CGI, it was found that CGI method

1 CEO, NOM Co., Ltd., Tokyo, Japan, nom-yo@jcom.home.ne.jp 2 Researcher, Research and Development Center, Nippon Koei Co., Ltd., Tsukuba, Japan, sokkheang-sr@n-koei.jp 3 CTO, NOM Co., Ltd., Tokyo, Japan, ms-matsu@jcom.home.ne.jp 4 Director, Miki Geo-Technology Research Institute Co., Ltd., Tsukuba, Japan, miki-egri@nifty.com 5 Manager, Nippon Koei Co., Ltd., Tokyo, Japan, a4074@n-koei.co.jp 6 Manager, Kato construction Co. Ltd., Tokyo, Japan, hirokuni.ito@kato-kensetu.co.jp

was effectively controlled the ground deformation and excess pore pressure despite of enlarged spacing of the lattice. Then, two cases of centrifuge model tests of full section model were carried out in order to avoid the boundary effect of the half section model. The first case was without countermeasure and the other case was with CGI method. Numerical simulations using fully coupled effective stress 3D FEM were carried out to evaluate the experimental results and applicability of the analysis program to solve the practical problem. Conventional countermeasure TOFT method was also analyzed.

Centrifuge Model Tests Centrifuge mode tests were carried out under centrifugal force field of 70g. The centrifugal model test system used in this study belongs to the Research and Development center of Nippon Koei Co. Ltd., Japan. In the model tests, the input earthquake motion of sine wave (30 waves) with frequency 2.0Hz and maximum acceleration 350gal (prototype scale) was applied. Toyoura sand and Silica sand #3 were used in the tests. The main properties of these sands are summarized in Table 1. Toyoura sand was used for the liquefaction layer and Silica sand #3 was used for the foundation layer.

Table 1. Physical properties of test materials

Physical properties Toyoura sand Silica sand #3 Maximum grain size, Dmax (mm) 0.425 2 Mean grain size, D50 (mm) 0.169 1.46 Uniformity coefficient, Uc 1.44 1.47 Coefficient of curvature, Uc’ 0.974 0.975 Minimum dry density, ρdmin (g/cm3) 1.343 1.490 Maximum dry density, ρdmax (g/cm3) 1.649 1.673 Density of soil particles, ρs (g/cm3) 2.641 2.636

The test cases carried out are summarized in Table 2.

Table 2. Test cases for full model (70g)

Case Liquefaction countermeasure

Case 1 Not adopted

Case 2 CGI method Figures 1(a), (b) present the outline of the centrifuge model tests in model scale. Depth of liquefaction layer corresponds to 10m of prototype scale. The model for stabilized bodies were made by mixing sand with cement then cured one week in the form aiming the target unconfined compression strength (qu) 1000kN/m2. Model ground was prepared in a rigid soil container with dimensions of 1000×490×200 mm by an air pluviation method and resulting relative density of the sand was 60%. The embankment

was made by compacting the mixed material of Toyoura sand 80% and kaolin clay 20% into an embankment mold with its wet density 1.9g/cm3 then frozen before mounted on the sand layer.

(a) Without Countermeasure (Case 1) (b) CGI method (Case 2)

Figure 1. Centrifuge model (full model)

Numerical Simulation In order to simulate the test results numerically, 3D version of LIQCA (Oka et al. 2006) was used. LIQCA is a fully coupled FEM for liquefaction analysis. Three cases carried out are listed in Table 3.

Table 3. Analytical cases

Case Liquefaction countermeasure Simulated centrifuge test

Case A-1 Not adopted Case 1

Case A-2 CGI method Case 2

Case A-3 TOFT method -

The model meshes used are illustrated in Figure 2-4. Eight node solid element with one integration point was used. Half portion was modelled because of their symmetry. Regarding the total volume of stabilized body, CGI method is about 63% of TOFT method.

20

490

1000Unit:mm

Unit:mm

1:1.8

200

28.614.3

150

142.9

142.9(10m)

312.9312.9

42.942.9

64.3(4.5m)

ρt=1.9g/cm3

1/70 model

374.3Silica sand #3 Dr=95%

Toyoura sand Dr=60%

Elevation

Plan

AccelerometerPore pressure meter

Laser disp. gaugeEmbankment

PPT2PPT3PPT4

ACC1ACC2ACC3

Disp3Disp2ACC10 ACC11

ACC13

20

490

1000

142.9

Unit:mm

Unit:mm

1:1.8

200

28.614.3

ACC13

150

114.3

142.9(10m)

312.9312.9

42.942.9

64.3(4.5m)

ρt=1.9g/cm3

1/70 model

374.3

105.7

14.3

105.7 90.5105.7

Silica sand #3 Dr=95%

Toyoura sand Dr=60%

Elevation

Plan

AccelerometerPore pressure meter

Laser disp. gaugeEmbankment

PPT2PPT3PPT4

ACC1ACC2ACC3

Disp3Disp2

Stabilized body

ACC10 ACC11

Figure 2 Mesh of no countermeasure (no. of nodes:7938, no. of elements:6448)

Figure 3 Mesh of CGI method (no. of nodes:7938, No. of elements:6448

Figure 4 Mesh of TOFT method (no. of nodes:8154, no. of elements:6624) The constitutive models used for each material are summarized in Table 4.

Case A-1

Case A-2

Case A-3

Liquefaction layer

Unimproved liquefaction layer

Unimproved liquefaction layer

Stabilized body

Stabilized body

Table 4. Constitutive model

Material Constitutive model

Embankment Mohr-coulomb

Toyoura sand (dry) Mohr-coulomb

Toyoura sand (saturated) Cyclic elasto-plastic (for liquefaction)

Stabilized body Mohr-coulomb

Silica sand #3 Ramberg-osgood Input parameters for cyclic elasto-plastic model was determined by simulation utility for laboratory test. The parameters used in the analysis are listed in Table 5. Please refer Oka et al. 2006 for the details of the parameters.

Table 5. Input parameter for Toyoura sand.

ρ e0 λ κ G0/σ’m0 k/γw M*m M*

f Kf 1.8 0.754 9.1×10-3 5.2×10-4 1323.6 1×10-3 0.77 1.12 2×106 B*

0 B*1 Cf Cd OCR* D*

0 n γP*r γE*

r 4089 54.5 0.0 2000 1.0 0.6 5.1 0.002 0.012

The parameters for other soils were assumed as general values used in conventional design. As the stabilized body is solidified by cement, its friction angle can be negligible for the stress state of the experiment, only cohesion is specified. Currently, as the program can’t cut off the tensile strength of the material with large cohesion, the tensile strength of the stabilized body might be overestimated because it is regarded about 10-20% of the compressive strength in general. Considering this situation, cohesion of the stabilized body was assumed quarter of qu.

Analysis Results Deformed meshes with a contour of vertical displacement are shown in Figure 5. Apparently deformation is controlled in the cases with countermeasure. It must be noted that hourglass mode in many elements were observed in both countermeasure cases perhaps due to usage of one integration point element.

CaseA-1

CaseA-2

CaseA-3

Figure 5 Deformed mesh Comparison of final vertical deformation distribution between experiment and analysis is shown in Figure 6. FEM can simulate Case 1 (without countermeasure) quite reasonably. However, the effect of countermeasure is slightly overestimated by FEM. Provably, tensile strength of the stabilized body is still too high.

DISPLACEMENT DZ (m)

0.500 0.384 0.227 0.091

-0.046 -0.182 -0.318 -0.455 -0.591 -0.727 -0.864 -1.000

DISPLACEMENT DZ (m)

0.500 0.384 0.227 0.091

-0.046 -0.182 -0.318 -0.455 -0.591 -0.727 -0.864 -1.000

DISPLACEMENT DZ (m)

0.500 0.384 0.227 0.091

-0.046 -0.182 -0.318 -0.455 -0.591 -0.727 -0.864 -1.000

Case 1

CaseA-1

Case 2

CaseA-2

Figure 6 Comparison of final vertical deformation distribution between experiment and analysis The result of CaseA-3 is shown in Figure 7. Vertical deformation is controlled better by TOFT method than CGI method in the analysis. However, the effect might be also overestimated.

CaseA-3

Figure 7 Final vertical deformation distribution

DISPLACEMENT DZ (m)

DISPLACEMENT DZ (m)

DISPLACEMENT DZ (m)

0.500 0.384 0.227 0.091

-0.046 -0.182 -0.318 -0.455 -0.591 -0.727 -0.864 -1.000

0.500 0.384 0.227 0.091

-0.046 -0.182 -0.318 -0.455 -0.591 -0.727 -0.864 -1.000

0.500 0.384 0.227 0.091

-0.046 -0.182 -0.318 -0.455 -0.591 -0.727 -0.864 -1.000

Comparison between experiment and analysis of vertical displacement with time is shown in Figure 8-9. Regarding Case1, FEM can simulate quite reasonably. Vertical displacement at the center of embankment is too underestimated in Case 2, although it is well simulated at the shoulder.

Figure 8 Vertical displacement ~ time (Disp1) Figure 9 Vertical displacement ~ time (Disp2)

Comparison of excess pore pressure at the middle part under the center of embankment is shown in Figure 10. Although peak value of the excess pore pressure in experiment is much higher than FEM, average values during shaking (about 10 seconds) are around 60 kN/m2. As FEM results are about 55kN/m2, mean value of excess pore pressure during shaking is similar.

Figure 10 Excess pore pressure ~ time (CH25) Comparison of acceleration at the center of embankment surface is shown in Figure 11-12. Although the tendency that rapid decrease after liquefaction is similar, FEM certainly overestimated absolute value.

-80

-60

-40

-20

0

0 20 40

Settl

emen

t(cm

)

Elapsed time (s)

Case 1 CaseA-1Case 2 CaseA-2

Embankment center

-80

-60

-40

-20

0

0 20 40Se

ttlem

ent(c

m)

Elapsed time (s)

Case 1 CaseA-1Case 2 CaseA-2

Embankment shoulder

0

25

50

75

100

0 10 20 30

Exce

ss p

ore

pres

sure

(kN

/m2 )

Elapsed time (s)

Case 1 CaseA-1Case 2 CaseA-2

CH25:Middle under Embankment center

Figure 11 Acceleration ~ time (CH19) Figure 12 Acceleration ~ time (CH19)

Figure 13 Settlement ~ time Figure 14 Settlement~ time

Figure 15 Settlement~ time

-600

-400

-200

0

200

400

600

0 10 20 30

Acc

eler

atio

n(ga

l)

Elapsed time (s)

Case 1

CaseA-1

Embankment center

-600

-400

-200

0

200

400

600

0 10 20 30

Acc

eler

atio

n(ga

l)

Elapsed time (s)

Case 2

CaseA-2

Embankment center

-30

-20

-10

0

0 20 40

Settl

emen

t(cm

)

Elapsed time (s)

CaseA-2 CaseA-3

Embankment center

-30

-20

-10

0

0 20 40

Settl

emen

t(cm

)

Elapsed time (s)

CaseA-2 CaseA-3

Embankment shoulder

-600

-400

-200

0

200

400

600

0 10 20 30

Acc

eler

atio

n(ga

l)

Elapsed time (s)

CaseA-2

CaseA-3

Embankment center

Comparison between CGI and TOFT method is shown in Figure 13-15. TOFT method shows better effect regarding vertical settlement partly because of the greater volume of stabilized body. However, some damage such as differential settlement might occur for embankment because of the settlement of the intermediate unimproved part of foundation ground. Occurrence of differential settlement at the top surface of embankment was not clear because of hourglass mode. In addition, acceleration of embankment of TOFT case is certainly greater than CGI case. It also can be a cause of damage for the embankment. Centrifuge test is expected for TOFT case.

Conclusions Numerical simulations using fully coupled effective stress 3D FEM were carried out to evaluate the experimental results and applicability of the analysis program to solve the practical problem. As a results, following conclusions are obtained.

1) FEM can simulate quite reasonably regarding the deformation especially for non-countermeasure case. It can be used for the practical design.

2) FEM overestimated the deformation control effect of countermeasure perhaps due to much higher tensile strength of stabilized body.

3) TOFT method shows better effect regarding deformation control than CGI because the volume of stabilized body is certainly greater. However some possibility of damage for embankment might be anticipated.

It is expected that carrying out centrifuge test for TOFT method. Also, it is expected to introduce tension cut off material for stabilized body to carry out better simulation.

References Miki H, Okochi Y, Makino M. Basic study for a new design concept of ground improve method evaluating constraint effect by the stabilized bodies. Proc. of GEOTEC Hanoi 2011, Hanoi;: 107-112.

Oka F, Kodaka T, Kimoto S, Kim Y-S, Yamasaki N. A multi-phase coupled FE analysis using an elasto-viscoplastic model for unsaturated soil. Geomechanics II, Geotechnical Special Publication, ASCE, Proc. 2nd US-Japan workshop on Geomechanics Ambrose 2006;: 124-131.

Okochi Y, Miki H, Sreng S, Kobayashi K, Makino M. Centrifugal model tests on the effect of a liquefaction countermeasure using enlarged spacing of the lattice shaped ground improvement method. Proc. of the 5th International Geotechnical Symposium-Incheon 2013;: 521-526

Sreng S, Kobayashi K, Tanaka H, Okochi Y, Miki H, Makino M. Dynamic centrifuge model tests of embankment with a new liquefaction countermeasure by ground improvement considering constraint effect. Proc. of the 5th Asia Conference on Earthquake Engineering 2014; 223

Taya Y, Uchida A, Yoshizawa M, Onimaru S, Yamashita K, Tsukuni S. Simple method for determining lattice intervals in grid-form ground improvement. Japanese Geotechnical Journal 2008; 3(3): 203-212.