See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/292345003

Identification of dynamic soil properties through shaking table tests on a

large saturated sand specimen in a laminar shear box

Article  in  Soil Dynamics and Earthquake Engineering · April 2016

DOI: 10.1016/j.soildyn.2016.01.007

CITATIONS READS

7 215

3 authors, including:

Chi-Chin Tsai Jiunn-Shyang Chiou

National Chung Hsing University National Taiwan University
35 PUBLICATIONS   166 CITATIONS    31 PUBLICATIONS   190 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Geotechnical reconnaissance View project

Site effect and site response analysis of vertical ground motion View project

All content following this page was uploaded by Chi-Chin Tsai on 17 May 2017.

The user has requested enhancement of the downloaded file.

Identification of dynamic soil properties through shaking table tests on a large

saturated sand specimen in a laminar shear box

Chi-Chin Tsai1, Wei-Chun Lin2, and Jiunn‐Shyang Chiou3

ABSTRACT

Many laminar shear boxes have recently been developed into sliding-frame containers

that can reproduce 1D ground-response boundary conditions. The measured responses of such

large specimens can be utilized to back-calculate soil properties. This study investigates how the

boundary effect in large specimens affects the identified soil properties through shaking table

tests on a soil-filled large laminar box conducted at the National Center for Research on

Earthquake Engineering in Taiwan. The tested soil-box system is unique because only 80% of

the container is filled with soil. This system can be regarded as a two-layer system: an empty top

and soil-filled bottom. The dynamic properties of this two-layer system are identified through

various approaches, including theoretical solutions of wave propagation, free vibration, and

nonparametric stress-strain analyses. Therefore, the coupling effect of the box and soil can be

evaluated. Results show that, compared with the two-layer system considering the influence of

the box, the conventional approach with a single-layer system slightly underestimates shear wave

velocity but obtains the same damping ratio of the soil layer. In addition, the identified modulus

reduction and damping curves in the two-layer system are consistent with those obtained in a

1
Associate Professor, Department of Civil Engineering, National Chung Hsing University, Taiwan,
tsaicc@nchu.edu.tw
2
Research Assistant, Department of Civil Engineering, National Chung Hsing University, Taiwan.
3
Assistant Professor, Department of Civil Engineering, National Taiwan University, Taiwan.
laboratory test on a small specimen. Furthermore, based on detailed acceleration measurements

along different depths of soil, a piecewise profile of shear wave velocity is built. The identified

shear wave velocity increases with depth, which is not uniform and differs from the constant

velocity typically assumed for the specimen.

Introduction

Soil in a level ground of infinite extent under earthquake shaking is usually modeled as a soil

element undergoing a simple shear loading condition. Small soil specimens are generally tested

in laboratories (e.g., using triaxial compression apparatuses and simple shear devices) under

regular or irregular dynamic loads to study soil behavior, such as stress-strain relationship and

liquefaction. Stress conditions and deformations in soil elements in these test types are

significantly affected by boundary conditions because of the size of the specimens. Moreover,

loading conditions generally do not reflect real-field situations because of the limitations of

loading devices.

An increasing number of downhole arrays are currently available for measuring motions on

ground surface and within a soil profile. These arrays provide valuable data in understanding in

situ soil behavior under earthquake shaking. Different approaches, including nonparametric

stress-strain analysis ([1] and [2]) and parametric ([3] and [4]) inverse analysis schemes, have

been developed to identify dynamic soil behavior through downhole measurement. However,

learning soil behavior from field measurements is an inherently inverse problem that can be

challenging to solve. The extraction of soil properties also depends largely on the availability and

spacing of measurements in a downhole array. Furthermore, the uncertainty of in situ conditions

is uncontrollable.

2
As an intermediate condition between small-specimen laboratory tests and less controllable

downhole measurements, large soil specimens are placed on shaking tables [5, 6] or centrifuges

[7]; thus, soil behavior under realistic seismic loading conditions can be observed and analyzed.

Several laminar shear boxes (e.g., [6, 13]) have recently been developed into sliding-frame

containers to reproduce 1D ground-response boundary conditions. Dietz and Wood [6] evaluated

dynamic soil properties by shaking table test on a large soil-filled laminar box. Three different

excitation motions: random [9], pulse [10] and sinusoidal [11] were employed to evaluate the

strain-dependent shear stiffness and damping. Afacan et al. [7] constructed centrifuge models in

a laminar container to study the site response of soft-clay deposits over a wide strain range.

Dense sensor arrays were used for the back-calculation of modulus-reduction and damping

values. Mercado et al. [4] identified shear wave velocities and their reduction against strain

through an optimization analysis of the centrifuge test of a laminar container. Experimental data

from large specimens complemented the laboratory geotechnical investigation technique.

However, the boundary effect can also be an issue that affects how soil properties are identified,

as in the case of small specimens. Lee et al. [12] investigated the boundary effects of a laminar

container on the seismic response acquired from accelerometers and pore pressure transducers at

various depths and distances from the end walls. The results of the analysis revealed minimal

boundary effects on the seismic responses, which confirmed the finding of the previous study [7]

[13] that measurements on the container agree with that within the specimen. Therefore, a

laminar container may be used effectively to simulate 1D shear wave propagation in shaking

table tests and the measurements on the container are commonly adopted for analysis.

3
Nevertheless, although boundaries do not affect responses and measurements, they can still

influence the identified properties. Such an issue has not yet been discussed.

In this study, the boundary effect on how dynamic properties are identified is investigated

through a series of shaking table tests on a large soil-filled laminar box. The tests are conducted

at the National Center for Research on Earthquake Engineering (NCREE) in Taiwan [8]. Unlike

the common test condition that the container is fully filled with soil, only 80% of the container is

filled with soil in these tests. Therefore, not only the shear wave propagation characteristics of

sand can be identified, but also the coupling effect of the box and soil can at the same time be

evaluated. The coupling effect of the box and soil is assessed by regarding the box and soil as a

two-layer system with an empty top and soil-filled bottom. The shear wave velocity (Vs),

damping ratio (D), and their variation against strain are identified through various approaches,

including theoretical solutions of wave propagation, free vibrations, and nonparametric stress-

strain method. The boundary effect (i.e., the laminar box) on the identified value is discussed.

Laminar Shear Box Test

The laminar shear box developed at NCREE is composed of 15 layers of sliding frames, as

shown schematically in Figure 1 [13]. The size of the soil specimen in the box is 1880 mm ×

1880 mm × 1300 mm. These 15 layers of frames (80 mm each) are separately supported on the

surrounding rigid steel walls, one above the other, with a vertical gap of 20 mm between adjacent

layers. Therefore, each layer of frames can move independently at different depths and directions,

thereby mimicking 1D ground-response boundary conditions. The mass of each frame layer is

approximately 8% of the mass of a 100 mm layer of soil enclosed by an inner frame.

4
Instrumentation

Transducers for acceleration measurements (i.e., Ay1 to Ay15 on each frame layer) were placed

at various depths on the outside rigid walls (Figure 2); piezoresistive accelerometers (e.g.

PCB4Y) for acceleration measurements were also placed at different locations within the soil.

Several comparisons were conducted to show that the accelerations on the frames are consistent

with those within soils prior to liquefaction [13]. Afacan et al. [7] also pointed out that the

measurement on the outer frames can represent the response within soils. Thus, the accelerations

on the laminar box frame were directly utilized to analyze the induced motion under shaking.

Sample preparation

Fine silica sand from Vietnam was used in the experiments. The basic properties of this sand are

given in Table 1. In this test, the wet sedimentation method was used for specimen preparation.

Only 80% of the container was filled with soil. Figure 2(a) shows that Ay3 is located on top of

the soil. The P-wave velocity of the specimen was measured horizontally across the specimen at

different depths to confirm the saturation of the soil specimen. The measured P-wave velocities

were between 1500 and 1700 m/s at various depths, except on the sand surface probably because

of the minor unevenness of the surface and trapping of air bubbles. The range of the measured P-

wave velocities showed that the soil specimen prepared through the wet sedimentation method

was fully saturated [8].

Selected test cases for analysis

A series of shaking table tests was conducted on the sand specimen in a biaxial laminar shear

box starting from August 2002 [8]. The purpose of test was to investigate the liquefaction

5
behavior of sand. In this study, we selected 15 cases in the test sequence conducted in October

2004 for analysis. These cases were selected because no liquefaction was observed. The excess

pore water pressure during shaking was very small. The input motions of Cases 19, 35, and 54

were scaled acceleration recorded at She-tou seismograph stations during the Chi-chi earthquake,

and the input motions of the others were sinusoidal acceleration with different amplitudes and

frequencies (Table 2). The relative density, Dr (%), of each test was estimated based on the

accumulated settlement of the previous tests and initial relative density, also shown in Table 2.

The Dr of the specimen in the later test was higher because of the densification caused by

shaking after each test.

Analysis Approach

Three methods are collectively utilized to identify the dynamic properties of Vietnam sand (Vs

or D) in a large shear box. The influence of very small excess pore water pressure generated in

the tests on the soil properties is ignored. Each method is described below.

Wave propagation method

The first approach to identifying system properties is based on the theory of wave propagation

using transfer function (TF). TF is the ratio of the measurement (e.g., acceleration time history)

at any two locations in the frequency domain. This ratio is theoretically a function of the Vs and

D of soil. Thus, system properties can be identified by adjusting Vs and D until the theoretical

TF that best matches the experimental TF is obtained. Considering the soil in the shear box as a

uniform soil layer, the system can be modeled as a two-layer system that includes the empty top

6
part of the box as one layer and the soil-filled bottom part of the box as another layer, as shown

in Figure 2(b).

TF of the Two-Layer System

The theoretical TFs of the two-layer system between the boundaries can be derived by solving

the wave equation through strain compatibility and stress equilibrium at these boundaries.

The three TFs of the top of the frame (Top) to the top of the soil (Mid), Top to bottom of the soil

(Bottom), and Top to Bottom are obtained as follows:

Top 1 (1)
TF
Mid
Mid 2 (2)
TF
Bottom 1 1
1 cos 1 cos
2 2
Top 1 (3)
TF
Bottom

where

α (4)

1 ;i=1,2 (5)
k ; i=1,2 (6)

Vs1 and Vs2 are the Vs of Layers 1 and 2, respectively, representing the Vs of the frame and that

of the frame with soil combination, respectively; 1 and 2 are the Ds of Layers 1 and 2,

representing the Ds of the frame and the frame with soil combination, respectively; 1 and 2 are

the densities of Layers 1 and 2, respectively. The ratio of 1/2 is approximately 8%, given the

specific gravity Gs of frame = 2.70 (aluminum) and the sizes of the frame and specimen. Given a

special case of Vs1 = 0 (i.e., a single-layer system), Eq. (2) yields the following:

7
 1 (7)
TF

which is the same as the TF of a single-layer system [14]. The natural frequency of the two-

layer system that is a function of the properties of the two layers can be expressed as follows [15]:

2 ∙ 2 ∙ (8)
tan tan α

Procedure of determining the parameters

The properties of the system can be identified by adjusting the Vs and damping of the two layers

until the theoretical TF that best matches the experimental TF is obtained. The procedures of

determining the parameters of the two-layer system are described as follows:

1. Vs1 is adjusted to match the frequency of the first peak of the theoretical TF (Top/Mid)

(Eq. (1)) with the frequency of the experimental TF (Ay1/Ay3) because the peak location

of this TF mainly depends on Vs1.

2. Vs2 is adjusted to match the frequency of the first peak of the theoretical TF (Mid/Bottom)

(Eq. (2)) with the frequency of the experimental TF (Ay3/Aybn).

3. 2 and 1 are adjusted simultaneously to match the magnitudes of the first peak of the

theoretical TFs of Top/Mid, Mid/Bottom, and Top/Bottom (Eqs. (1) to (3)) with the

experimental TFs of Ay1/Ay3, Ay3/ Aybn, and Ay1/Aybn, respectively.

Figure 3 shows two examples of the wave propagation method using the data of Cases 1 and 19.

Through the best match of the theoretical and experimental TFs by the above procedure, the

identified Vs1 and Vs2 in Case 1 are 12.5 and 62 m/s, respectively, and are labeled as Profile 1 in

Figure 4. The identified 1 and 2 are 4% and 9%, respectively. Vs2 and 2 represent the

8
combined properties of the soil and frame. Obtaining soil properties requires further calculation,

as elucidated later.

Free-vibration method

Vibration decreases in amplitude and eventually disappears when a system is set into a state of

free vibration. A reduction in the amplitude of vibration is caused by internal damping within the

mass of the material. Therefore, D can be estimated by approximating the free-vibration response

of the acceleration time history at the end of the input motion as follows [16]:

(9)

where fn is the natural frequency of the system, which can be identified in the Fourier spectrum

of the free vibration. Therefore, in addition to obtaining the damping, Vs can be calculated as

4 (10)

where h is the height or thickness of the system. The Vs and damping identified by this approach

reflect the overall system behavior and are neither of the soil nor of the frame. Further

investigations are required to identify the Vs and damping of the soil and frame. This issue is

elucidated by observing actual records.

Figure 6 shows the free-vibration responses of Ay1 (on the top of the frame), Ay3 (on the top of

the soil), PCB4Y (within the soil) in Case 1 after 10.9 s. The corresponding Fourier spectra are

shown in Figure 7. A comparison of Figure 7 indicates that all free-vibration records show the

natural frequency (i.e., the first peak) around 10 Hz, but Ay1 shows an additional peak at 15 Hz

that does not appear in Ay3 and PCB4Y. The peak of Ay1 at 15 Hz is due to the wave traveling

back and forth within the first layer (i.e., the empty frame). Therefore, the Vs of the frame can be

determined as 12 m/s based on Eq. (10) given fn = 15 Hz and h = 0.2 m. The Vs of Layer 2 is 60
9
m/s based on Eq. (8) given fn = 10 Hz (according to either Ay3 or PCB4Y), h1 = 0.2 m, h2 = 1.3

m, and Vs1 = 12 m/s. The Vs identified based on the measurements on the frame and within the

soil are identical.

The free-vibration response on the frame (Ay3) and within the soil (PCB4Y) are similar although

the elevations of two measurements are different. However, it behaves differently before and

after 11.2 s. Ay1, Ay3, and PCB4Y exhibit clear decay before 11.2 s, but the decay of the

response becomes insignificant after 11.2 s. This observation can be explained by the two-layer

system, in which the stiffness and damping of Layer 2 are higher than those of Layer 1. Ay1 still

vibrates after 11.2 s, whereas Ay3 or PCB4Y (which represents Layer 2) almost stops after 11.2

s because of high damping. The residual vibration is due to the interaction with the remaining

free vibration of Layer 1 at the top. Therefore, approximating the decay by Eq. (9), it is obtained

that 1% damping of 1 primarily on the basis of Ay1 decay after 11.2 s, whereas it is obtained

that 5% damping of 2 primarily on the basis of the measurement Ay3 or PCB4Y before 11.2 s.

The damping identified based on the measurements on the frame and within the soil are identical.

Nonparametric stress–strain approach

The nonparametric stress–strain approach calculates the stress and strain time histories using

acceleration measurements. The stiffness and damping can be determined based on the stress–

strain curve [1]. This study, however, only uses the obtained strain history and its variation

against depth (i.e., strain profile) to discuss the uniformity of the specimen. The average shear

strain at depth z can be computed by assuming 1D wave propagation conditions (i.e., = ∂u/∂z):

(11)
γ t
∆
10
where u(t) is the displacement that can be obtained from the double integration of the

acceleration time history. The numerator in Eq. (11) represents the differential horizontal

displacement between the two adjacent accelerometers (i.e., Ay4 and Ay5), and the denominator

represents the vertical distance between these accelerometers (i.e., 100 mm). The acceleration

time history is filtered and baseline-corrected prior to integration to the displacement time

history. The acceleration time series is high-pass filtered in the frequency domain using a

Butterworth filter to remove low-frequency noises; the selection of the corner frequency follows

the protocols described by [17] and is intended to apply the smallest possible amount of filtering

while achieving realistic velocity and displacement.

Results and discussion

Vs and damping

The Vs and damping of Layers 1 and 2 identified by the wave propagation and free-vibration

methods are summarized in Table 3 and Table 4. Given that Vs2 and 2 are the properties that

combine the frame and soil, obtaining the Vs and of soil requires further calculation. Assuming

that the frame and soil system act as a composite material during seismic shearing, the Vs of soil

(Vs,soil) can be back-calculated based on the shear force equilibrium and strain compatibility of

the composite material:

, , (12)

where Vs,frame is equal to Vs1, Atotal is the section area (3.7636 m2), Asoil is the area inside the

frame (3.5344 m2), and Aframe is the section area of the frame (0.2256 m2). Taking Case 1 as an

11
example, the Vs of soil is obtained as 63.9 m/s given that Vs1 (frame) = 12.5 m/s and Vs2 (frame

+ soil) = 62 m/s. The damping of soil is obtained through the conservation of energy loss:

, , (13)

where ,frame is equal to 1. Taking Case 1 as an example, the  of soil is obtained as 9% given

that  (frame) = 4% and  (frame + soil) = 9%. The identified soil properties are also

summarized in Table 3 and Table 4.

Figure 8 compares the Vs and damping obtained from the wave propagation and free-vibration

methods. The identified Vs of the frame is similar, primarily between the range of 12 and 13m/s.

The identified Vs of the soil is widely distributed between the range of 45 and 76 m/s. The Vs

obtained from the two methods agree with each other and do not show any bias, which indicates

that the obtained number is reliable. By contrast, the identified damping shows much scattering.

The identified damping of the frame is between the range of 1% and 6%, and the identified

damping of the soil is between the range of 2% and 10%. The results significantly depend on the

method. The damping identified by the wave propagation method is approximately 50% larger

than that of the free-vibration method. The difference may result from how well the fitting of the

theoretical and experimental TFs are achieved and how well the decay in the free vibration is

approximated by Eq. (9).

The overall results indicate that the box produces 3% to 4% damping, which also leads to the

energy loss of the entire system. However, given its low stiffness and light mass compared with

the soil specimen, the produced energy loss is minimal. Consequently, the identified soil D is

essentially the same as that of Layer 2, which indicates that the frame contribution can be

neglected. By contrast, the presence of the frame reduces the Vs of the system. The identified Vs
12
can thus be underestimated (average of 3% error) if the frame contribution is ignored. Figure 3(b)

also provides a comparison of the theoretical TF of a single-layer system by Eq. (7) with Vs = 62

m/s and  = 9%. The single- and two-layer theoretical TFs are similar and both match the

experimental TF, thereby indicating that the effect of the top frame portion is also insignificant in

terms of TF. Therefore, the conventional approach assuming a single-layer system that includes

the frame effect can also be used to rationally estimate dynamic soil properties (only slightly

underestimates the Vs of soil).

Variation in the Vs profile

The specimen in a sand box is usually assumed to be uniform (i.e., Vs = constant) because it is

prepared through the wet sedimentation method using the same material. However, previous

studies [18] [19] have shown that the Vs of sand can increase with depth because of an increase

in confining pressure. In this section, whether or not the large sand specimen also presents a

pressure-dependent property is investigated. A depth-dependent Vs profile is first developed

based on the previous empirical model, and its TF is compared with the experimental TF. The

shear modulus of sand at a small strain Gmax can be empirically modeled as follows [20]:

2.17 e . (14)
G A∙ ∙ ′
1 e

wherev’ is the effective vertical stress, e is the void ratio, and A is the calibrated model

constant. In this study, e is estimated based on the basic information in Table 1 and Dr in Table 2.

The constant A is back-calculated given that the average Vs of the pressure-dependent Vs profile

is identical to the constant Vs profile identified previously in Case 1. Case 1 is selected to obtain

A because its input motion is the weakest, and the resulting soil response is closest to the small
13
strain condition. Taking Case 1 as an example, the resulting pressure-dependent Vs profile

(Profile 2) is shown in Figure 4, and the theoretical TF of the pressure-dependent Vs profile is

shown in Figure 3(b) for comparison. The theoretical TF of the pressure-dependent Vs profile is

obtained from a frequency-domain analysis using the site-response analysis program DEESPOIL

[21] because the analytical solution for the pressure-dependent Vs profile, such as Eqs. (1) to (3),

is complex [22] and not commonly used. Although the Vs profile differs with depth, the

corresponding TF is still similar to that of uniform soil and does not show a better match with the

experimental TF. Therefore, determining the soil profile cannot rely on the TF between the top

and bottom of the soil specimen.

The theoretical TFs between two close locations should be compared with the corresponding

experimental TFs that provide more information/constraint to further evaluate the

aforementioned issue. Thus, the TFs of Ay3/Ay7, Ay7/Ay11, and Ay11/Aybn are analyzed.

Figure 5 shows that the TFs of the uniform (Profile 1) and pressure-dependent profiles (Profile 2)

do not agree with the experimental TF for all the three pairs. The TF of the pressure-dependent

profile does not show much improvement in terms of matching the experimental TF, which

indicates that the Vs assumed as the pressure-dependent profile by Eq. (14) cannot represent the

real condition. Therefore, it is attempted to identify a step-wise Vs profile to ideally match the

theoretical TF with the experimental TF. Figure 4 shows an optimized Vs profile (Profile 3), and

the corresponding TF is shown in Figure 5. These results indicate that Vs is not constant but

increases with depth. Given a low Vs between Ay3 and Ay7 in the optimized profile, the first

peak of the theoretical TF can align with that of the experimental TF of Ay3/Ay7 at 20 Hz,

which cannot be reproduced in the TFs that use the other profiles. Figure 4 also shows the tip

14
resistance profile of the cone penetration test (CPT) for comparison. CPT was performed by mini

cone with diameter of 17.8mm to check the uniformity of specimen. The trend of the optimized

profile agrees well with the measured tip resistance profile of CPT, which confirms the variation

of Vs profile with depth. A similar trend is found in the other cases.

Strain history and the maximum strain profile

The maximum strain profile along the depth obtained based on the calculated strain histories is

shown in Figure 9. The maximum strain in the soil specimen generally increases with depth,

which is consistent with the trend of the analytical solution of the linear frequency-domain

analysis subjected to the same motion. The identified maximum strain profile, however, exhibits

several “kicks” at different depths, thereby indicating that the soil specimen is not uniform with

depth. The trend of analytical strain profile using Profile 3 is more consistent with that of

experimentally identified profile than those using Profiles 1 and 2, which confirms the increasing

behavior of the Vs profile, as discussed previously. However, the identified strain is higher than

the analytical one. The identified strain is calculated based on the displacement history that

integrates baseline-corrected and filtered-acceleration time histories. Thus, certain errors may be

induced through the process. In addition, the precision of the acceleration time history is not

sufficient to accurately calculate the strain given a close space between two measurements (100

mm). Therefore, the analyzed strain profile may be more representative than the identified strain

profile.

Nonlinear curve

As mentioned earlier, the tests are conducted in a sequence. The specimen is subjected to a

stronger motion in the later tests. On this basis, the Vs of the specimen should be lower because

15
more soil nonlinearity is encountered. However, the identified Vs is mostly the same or even

higher than those in the later test. After each test, the specimen is densified as a result of shaking;

thus, e decreases, or Dr increases. Theoretically, the Vs of the specimen should increase in the

subsequent test according to Eq. (14) and thereby compensate for the reduction because of the

strong shaking.

The following analysis is performed and a nonlinear curve that shows the variation in G against

strain is constructed to further investigate the aforementioned issue and to separate the factors of

the motion intensity and density of the specimen on the identified Vs. First, an empirical Vs at a

small strain condition is calculated based on Eq. (14). The identified Vs of Case 1 is used to

calibrate the constant A because Case 1 is subjected to the weakest motion. Once constant A is

determined, Vs at a small strain condition for the other cases can be calculated empirically given

Dr (i.e., e). Therefore, the Vs reduction or shear modulus reduction can be obtained by the ratio

between the identified Vs and the empirical Vs. Second, the average strain encountered during

each test is estimated by the linear site-response analysis, in which the identified Vs, representing

an equivalent Vs, is used. The reason for using the analytical strain instead of the identified strain

is that the latter may not be accurate, as discussed earlier. Lastly, based on the identified modulus

reduction and the corresponding strain of each test, nonlinear behavior can be found, as shown in

Figure 10a. The modulus reduction in Case 1 is set to match the Seed and Idriss curve [18] at the

corresponding strain, and those of the other cases are adjusted accordingly. The reason for such a

process is that Gmax at a small strain (i.e., 10−4%) is not available; thus, only the relative

nonlinearity is compared. The identified modulus reduction by both wave propagation and free

vibration methods generally agrees with the experimental curve but presents a more noticeable

16
nonlinearity. The discrepancy between them may be caused by several reasons, such as the

different soil types and testing devices. The identified damping is also plotted against the

corresponding strain to compare the damping curve of Seed and Idriss in Figure 10b. The

identified damping of the soil agrees with the damping curve of Seed and Idriss. The results

confirm that the soil does undergo certain nonlinearity and explain why the identified damping is

high.

Conclusions

This study investigates the boundary effect on how soil properties are identified by conducting

shaking table tests on a soil-filled large laminar box. Because only 80% of the container is filled

with soil,the system is regarded as a two-layer system with an empty top and soil-filled bottom.

The dynamic properties of the soil, box, and soil with box, such as damping and Vs, are

successfully identified through wave propagation and free-vibration methods.

The identified Vs values of the frame are mainly 12.0 m/s while those of soil are mostly 60 m/s

to 70 m/s depending on the density and the level of shaking. The identified Vs of the soil can be

underestimated (i.e., at an average of 3% error) without considering the influence of the box. On

the other hand, the identified Ds of frame are mainly 3% to 4% and those of soil are mostly 5%

to 10 %. The identified soil D is essentially the same as that of Layer 2, which indicates that the

contribution of energy loss by the frame can be neglected because of its light weight and low

stiffness. Based on the results of this study, the conventional approach assuming a single-layer

system that includes the effect of the frame can also be used to reasonably estimate dynamic soil

properties, but it slightly underestimates the Vs of the soil.

17
The wave propagation method by the TF only between the surface response of soil and the

bottom excitation or the free vibration method by only using the acceleration measurement of

soil regards the soil specimen as a single layer. Therefore, the Vs identified by either way is a

single equivalent value for the whole specimen (i.e. the specimen is assumed to be uniform). In

this 1g shaking table testing, more detailed acceleration measurements along different depths of

soil are available and therefore a piecewise Vs profile with depth can be built. The results based

on both wave propagation method and nonparametric stress-strain method show that the large

specimen is not uniform. In addition, the Vs of soil increases with depth. The observed

increasing trend, nevertheless, differs from that estimated by the pressure-dependent empirical

correlation. The difference may be due to the non-uniformity of specimen (i.e. the void ratio

varies with depth). The identified modulus reduction and damping curves show a similar trend to

those in a previous experimental test on a small specimen. However, the identified modulus

reduction presents a more notable nonlinearity. The discrepancy between them may be caused by

several reasons, such as the different soil types and testing devices.

Acknowledgements
The authors would like to thank the National Center for Research on Earthquake Engineering,

National Applied Research Laboratories of Taiwan for providing the shaking table test data and

the associated test reports.

References

[1] M. Zeghal, A.-W. Elgamal, H.T. Tang, J.C. Stepp, Lotung downhole array II: evaluation of soil

nonlinear properties, Journal of geotechnical engineering, 124 (1995) 363-377.

[2] C.-C. Tsai, Y.M.A. Hashash, Learning of Dynamic Soil Behavior from Downhole Arrays, Journal of

18
 Geotechnical and Geoenvironmental Engineering, 135 (2009) 745-757.

[3] A. Elgamal, T. Lai, Z. Yang, L. He, Dynamic soil properties, seismic downhole arrays and

applications in practice, in: 4th International Conference on Recent Advances in Geotechnical

Earthquake Engineering and Soil Dynamics, San Diego, California, USA, 2001.

[4] V. Mercado, W. El-Sekelly, Zeghal, M., T. Abdoun, Identification of soil dynamic properties through

an optimization analysis of arching pressures in ideal elastic soil, Computers and Geotechnics, 65

(2015) 175-186.

[5] A. Aghaei Araei, I. Towhata, Impact and cyclic shaking on loose sand properties in laminar box using

gap sensors, Soil Dynamics and Earthquake Engineering, 66 (2014) 401-414.

[6] M. Dietz, M. Wood, Shaking table evaluation of dynamic soil properties, in: 4th International

Conference on Earthquake Geotechnical Engineering Thessaloniki, Greece, 2007.

[7] K.B. Afacan, S.J. Brandenberg, J.P. Stewart, Centrifuge Modeling Studies of Site Response in Soft

Clay over Wide Strain Range, Journal of Geotechnical and Geoenvironmental Engineering, 140

(2014) 04013003.

[8] T.S. Ueng, C.H. Chen, L.H. Peng, W.C. Li, Large-Scale Shear Box Soil Liquefaction Testing on

Shaking Table (III)—Preparation of Large Sand Specimen and Preliminary Shaking Table Test, in,

National Center for Research on Earthquake Engineering, Taiwan, 2006.

[9] A.R. Dar, Development of a flexible shear stack for shaking table testing of geotechnical problems, in,

University of Bristol, 1993.

[10] K.L. Fishman, J.B. Mander, R. Richards Jr., Laboratory study of seismic free-field response of sand,

Soil Dynamics and Earthquake Engineering, 14 (1995) 33-43.

[11] A.J. Brennan, N.I. Thusyanthan, S.P.G. Madabhushi, Evaluation of shear modulus and damping in

dynamic shaking table tests, Journal of Geotechnical and Geoenvironmental Engineering, 131 (2005)

1488-1496.

[12] C.-J. Lee, Y.-C. We, Y.-C. Kuo, Boundary effects of a laminar container in centrifuge shaking table
19
 tests, Soil Dynamics and Earthquake Engineering, 34 (2012) 37-51.

[13] T.S. Ueng, M.H. Wang, M.H. Chen, C.H. Chen, L.H. Peng, A Large Biaxial Shear Box for Shaking

Table Test on Saturated Sand, Geotechnical Testing Journal, 29 (2006) 1-8.

[14] S.L. Kramer, Geotechnical earthquake engineering, Prentice Hall, Upper Saddle River, N.J., 1996.

[15] C.H. Chen, L.Y. Chen, Dynamic response of two-layers soils subjected to vertically propagating

shear wave, Journal of the Chinise Institute of Civil and Hydraulic Engineering, 7 (1995) 263-271.

[16] A.K. Chopra, Dynamics of structures : theory and applications to earthquake engineering, Prentice

Hall, Englewood Cliffs, N.J., 1995.

[17] D.M. Boore, J.J. Bommer, Processing of strong-motion accelerograms: Needs, options and

consequences, Soil. Dyn. Earthq. Eng. Comput., 25 (2005) 93-115.

[18] H.B. Seed, I.M. Idriss, Soil moduli and damping factors for dynamic response analyses, College of

Engineering University of California Berkeley., Berkeley, 1970.

[19] T. Kawaguchi, H. Tanaka, Formulation of Gmax from reconstituted clayey soil its application to

measured Gmax in the field, Soils and Foundations, 48 (2008) 821-831.

[20] K. Ishihara, Soil behavior in earthquake geotechniques, Clrendon Press, Oxford, 1996.

[21] Y.M.A. Hashash, D.R. Groholski, C.A. Phillips, D. Park, M. Musgrove, DEEPSOIL 5.1, User

Manual and Tutorial, (2012) 107.

[22] H. Afra, A. Pecker, Calculation of free field response spectrum of a non homogeneous soil deposit

from bed rock response spectrum, Soil Dynamcs and Earthquqke Engineering, 22 (2002) 157-165.

20
Table 1 Basic properties of the specimen [8]
Original Da Nang,
Vietnam
Shape Angular
Color White
Gs 2.65
0.17
0.26
1.59
0.596
0.879

21
 Table 2 Selected cases for analysis
Frequency Recording rate e1 Gmax2
Case Amax (g) Duration (s) Dr (%)
(Hz) (time/s) (MPa)
01 2 0.03 10 1000 20.37 0.853 77.61

05 2 0.05 10 250 21.61 0.849 78.20

07 2 0.05 30 250 22.3 0.847 78.52

08 1 0.05 5 250 22.73 0.846 78.73

10 1 0.05 20 250 22.97 0.845 78.82

13 4 0.05 10 250 23.62 0.843 79.15

14 4 0.05 20 250 23.86 0.842 79.27

Chi-chi earthquake (PGA scales down 0.803 85.92

19 250 37.34
to 12.5%)
26 1 0.075 10 250 52.51 0.759 93.92

29 4 0.075 5 250 52.64 0.758 94.00

Chi-chi earthquake (PGA scales down 0.691 107.36
36 250 75.74
to 18.75%)
44 1 0.1 20 250 88.44 0.654 115.34

47 4 0.1 20 250 88.9 0.652 115.64

Chi-chi earthquake (PGA scales down 0.625 137.49
54 250 98.25
to 25%)
57 2 0.15 10 250 98.74 0.624 137.86
1. Estimated based on Dr and emax and emin in Table 1
2. Calculated according to Eq. (14)

22
 Table 3 Summary of the identified Vs by different methods
Method 1 (Wave propagation) Method 2 (Free vibration)
Case Vs1 Vs2 Frame Vs Soil Vs Vs1 Vs2 Frame Vs Soil Vs
(m/s) (m/s) (m/s) (m/s) (m/s) (m/s) (m/s) (m/s)
01 12.5 62.0 12.5 63.9 12.9 59.7 12.9 61.5
05 12.5 60.0 12.5 61.8 12.8 60.7 12.8 62.5
07 12.5 58.0 12.5 59.7 12.9 61.0 12.9 62.9
08 15.5 48.0 15.5 49.3 12.9 51.0 12.9 52.5
10 12.5 60.0 12.5 61.8 12.8 59.9 12.8 61.7
13 12.0 59.0 12.0 60.8 13.0 60.8 13.0 62.7
14 12.5 60.0 12.5 61.8 12.6 60.7 12.6 62.6
19 12.5 62.0 12.5 63.9 12.6 63.3 12.6 65.2
26 23.0 65.5 23.0 67.3 14.0 66.1 14.0 68.0
29 15.0 60.0 15.0 61.8 12.5 64.2 12.5 66.2
36 12.5 67.0 12.5 69.0 12.7 69.6 12.7 71.8
44 22.5 28.0 22.5 28.1 12.5 45.7 12.5 47.0
47 23.0 52.0 23.0 53.2 12.3 48.0 12.3 49.4
54 23.0 72.0 23.0 74.0 19.7 76.9 19.7 79.1
57 23.0 67.0 23.0 68.8 20.7 73.5 20.7 75.6

23
Table 4 Summary of the identified damping by different methods
Method 1 (Wave propagation) Method 2 (Free vibration)
Case Frame Frame 
1 (%) 2 (%) Soil  (%) 1 (%) 2 (%) Soil  (%)
 (%) (%)
01 4.0 9.0 4.0 9.0 1.0 5.0 1.0 5.0
05 5.0 9.0 5.0 9.0 0.5 2.0 0.5 2.0
07 5.0 10.0 5.0 10.0 1.5 2.0 1.5 2.0
08 3.0 5.0 3.0 5.0 2.0 5.0 2.0 5.0
10 5.0 9.0 5.0 9.0 1.0 3.0 1.0 3.0
13 4.0 8.0 4.0 8.0 3.0 5.0 3.0 5.0
14 5.0 10.0 5.0 10.0 2.5 5.0 2.5 5.0
19 5.0 10.0 5.0 10.0 - - - -
26 3.0 5.0 3.0 5.0 3.0 5.0 3.0 5.0
29 3.0 8.0 3.0 8.0 3.0 8.0 3.0 8.0
36 5.0 10.0 5.0 10.0 - - - -
44 5.0 9.0 5.0 9.2 2.0 8.0 2.0 8.0
47 6.0 10.0 6.0 10.1 2.5 6.0 2.5 6.0
54 6.0 10.0 6.0 10.0 - - - -
57 2.0 6.0 2.0 6.0 3.0 6.0 3.0 6.0

24
Figure 1 Schematics of the biaxial laminar shear box [13]

25
(a) (b)
Ay1 (Top)
Frame Ay1
Layer1: Vs1, h1,  (Mid)
Ay3 Ay3
Ay4
Sand
100 mm

1500 mm
Ay7

1300 mm
PCB4Y Layer2: Vs2, h2, 
Ay11
500 mm

(Bottom)
Aybn Aybn
Rigid base Rigid 
Figure 2. (a) Sketch of the laminar shear box (b) Equivalent two-layer system

26
TF
TF
TF

Figure 3. Comparison of the theoretical and experimental TFs using Case 1 data (left) and Case 19 data
(right)

27
 Vs(m/s)
0 20 40 60 80 100
0

0.4

0.8

1.2

1.6
0 1 2 3 4 5
Tip resistance(MPa)

Figure 4. Vs profile along the depth: (a) Case 1 and (b) Case 19

28
20

16

12

0
0 10 20 30 40 50
Frequency(Hz)
8

0
0 10 20 30 40 50
Frequency(Hz)
8

0
0 10 20 30 40 50
Frequency(Hz)

Figure 5. Comparison of the theoretical TF that assumes different Vs profiles and experimental
TF using Case 1 data

29
 Acceleration (g)
Acceleration (g)
Acceleration (g)

Figure 6. Free-vibration responses of Ay1 and Ay3 in Case 1

30
 Fourier Amplitude
Fourier Amplitude
Fourier Amplitude

Figure 7. Fourier spectrum of the free-vibration responses of Ay1 and Ay3 in Case 1

31
Figure 8. Comparison of the identified Vs and damping by Methods 1 and 2

32
 (a)case01 Strain(%) (b)case19 Strain(%)
0 0.02 0.04 0.06 0.08 0 0.04 0.08 0.12 0.16 0.2

0 0

Sand
Sand

0.4 0.4

0.8 0.8

1.2 1.2

1.6 1.6
Figure 9. Maximum strain profile in Cases 1 and 19

33
 1

0.8

0.6

0.4

Seed & Idriss (1970)

0.2
Free vibration method
Wave propagation method
0

0.0001 0.001 0.01 0.1 1 10

Strain (%)
30

20

10

0.0001 0.001 0.01 0.1 1 10

Strain (%)

Figure 10. Identified nonlinear curve

34

View publication stats