Keywords

1 Introduction

LEAP-ASIA-2019 project is a sequel to the LEAP-GWU-2015 project (Kutter et al., 2018; Manzari et al., 2018; Zeghal et al., 2018) and the LEAP-UCD-2017 project (Kutter et al., 2019; Manzari et al., 2019a, b) that investigated the repeatability and reproducibility of centrifuge tests, the sensitivity of the experimental results to variation of testing parameters and conditions, and the performance and validity of constitutive models and numerical modeling tools in predicting the observed response. The goals of LEAP-ASIA-2019 are to (1) validate the “generalized scaling law” for centrifuge modeling (Iai et al., 2005) on the seismic behavior of a liquefiable sloping ground and (2) fill the gaps of the LEAP-UCD-2017’s data with the aim to identify trends in the experimental results (in terms of a combination of Dr and PGA) and to build an experimental database for numerical modelers.

The LEAP-ASIA-2019 project involved nine numerical simulation teams from different academic institutions and geotechnical companies from around the world; they participated in the modeling of some of the centrifuge model experiments performed at several research institutions. The simulation exercise consisted of the calibration of constitutive model parameters, Type-B predictions, or Type-C simulations. In the first phase (i.e., model calibration), the numerical simulation teams were provided with a series of hollow cylinder torsional shear tests and direct simple shear tests to calibrate their constitutive models. Ueda et al. (2023) present an overview of the results of the first phase. In the second phase of the simulation exercise, the numerical simulation teams performed Type-C or Type-B simulations using their constitutive models, which were calibrated in the first phase, with or without iterative adjustment of the model parameters. This chapter presents a summary of key aspects of the second phase and their comparisons with the experimental data obtained from centrifuge model tests on the seismic behavior of a uniform-density, 20-m-long, and 5-degree sandy slope (Tobita et al., 2022, 2023).

2 LEAP-2019 Centrifuge Experiments

As described in the Introduction, one of the objectives of the LEAP-ASIA-2019 project is to fill the gaps in LEAP-UCD-2017’s experimental data. To this end, the LEAP-ASIA-2019 centrifuge experiments were designed to study the lateral spreading of a uniform-density, 20-m-long, 4-m-deep at center, and 5-degree sloping liquefiable deposit, similar to the LEAP-GWU-2015 and LEAP-UCD-2017 projects (Kutter et al., 2018, 2019). As illustrated in Figs. 3.1 and 3.2, the sloping ground was created with Ottawa F-65 sand in a rigid container. Three arrays of accelerometers and pore pressure transducers are placed in the central section and at 3.5 m away from the left- and right-side walls. In the vertical direction, the sensors were placed 1.0 m apart. Tables 3.1 and 3.2 show the specified locations of the accelerometers and pore pressure transducers, respectively. Also, 3D printed surface markers were placed on the ground surface to measure the surface horizontal displacements under seismic loading. Figure 3.3 illustrates the surface marker locations for displacement measurement (top view).

Fig. 3.1
2 schematic diagrams of the sand in a rigid container. 3 arrays of accelerometers and pore pressure transducers are placed in the central section. They have a water table to cover the sand during spinning and an approximately circular shape so that the soil surface has a constant radius on each cross-section.

Baseline schematic for LEAP-ASIA-2019 experiment for shaking parallel to the axis of the centrifuge (e.g., RPI, ZJU, NCU). L* in the figure corresponds to 1/μη

Full size image
Fig. 3.2
2 schematic diagrams present the slope of the sand in a rigid container. They indicate an approximate log spiral shape so that the surface has an angle of 5 degrees from the normal to the G field, positions of sensors, and a water table to cover the sand during spinning.

Baseline schematic for LEAP-ASIA-2019 experiment for shaking in the circumferential direction of the centrifuge (e.g., UCD, KyU). L* in the figure corresponds to 1/μη

Full size image
Table 3.1 Positions of the accelerometers in LEAP-ASIA-2019 experiments
Full size table
Table 3.2 Positions of the pore water pressure sensors in LEAP-ASIA-2019 experiments
Full size table
Fig. 3.3
A schematic of the surface marker locations for displacement measurement. It indicates the transverse array and the longitudinal array. The width is labeled W and the 2-D plane axes are Y r e f and X r e f.

Surface marker locations for displacement measurement (top view)

Full size image

As part of the LEAP-ASIA-2019 project, 10 institutes performed 24 centrifuge model tests in total (Tobita et al., 2022, 2023; Madabhushi et al., 2023; Stone et al., 2023; Okamura & Sjafruddin, 2023; Escoffier et al., 2023; Manandhar et al., 2023; Vargas et al., 2023; Huang & Hung, 2023; Korre et al., 2023; Ma et al., 2023). Since another aim of the LEAP-ASIA-2019 project is to validate the “generalized scaling law” for centrifuge modeling (Iai et al., 2005) as noted in Introduction, 11 model tests applied the conventional centrifuge scaling law (hereafter called “Model A”), while 13 model tests did the “generalized scaling law” (hereafter called “Model B”). Tables 3.3 and 3.4 show a summary of the centrifuge experiments “Model A” and “Model B,” respectively, selected for the LEAP-ASIA-2019 Type-B or Type-C simulations. The table lists the main characteristics of each experiment: the reported achieved soil density, relative density, combinations of the virtual 1G scaling factor μ and centrifuge scaling factor η, input acceleration levels (i.e., PGAeff (Kutter et al., 2018, 2019)), and observed horizontal displacements (average value) at the ground surface after shaking.

Table 3.3 Summary of centrifuge experiments, Model A in LEAP-ASIA-2019
Full size table
Table 3.4 Summary of centrifuge experiments, Model B in LEAP-ASIA-2019
Full size table

3 Type-B/Type-C Numerical Simulations

The second phase of the simulation exercise consisted of Type-B or Type-C numerical simulations of centrifuge model tests that were conducted as part of the LEAP-ASIA-2019 centrifuge modeling campaign. Table 3.5 shows the numerical simulation teams who participated in the Type-B or Type-C simulation exercises and submitted the report on their simulation results. The constitutive model and the analysis platform used by each numerical simulation team are also listed in the table as well as the simulation type (i.e., Type-B or Type-C). It should be noted that Simulations 1–3 and Simulation 8 can be classified as between Type-B and Type-C simulations (but close to Type-B); this is because the model parameters for dynamic deformation characteristics and dilatancy remain the same as those determined from the element simulations for the laboratory tests, but only the hydraulic conductivity was adjusted. Mode-detailed information of each constitutive model and the numerical simulation techniques used by each simulation team are provided in separate papers (Tanaka et al., 2023; Hyodo & Ichii, 2023; Fasano et al., 2023; Qiu & Elgamal, 2023; Elbadawy & Zhou, 2023; Reyes et al., 2023; Wang et al., 2023).

Table 3.5 Numerical simulation teams
Full size table

The main steps of the second phase were as follows:

  1. 1.

    Results of three sets (KyU_A_A2_1, RPI_A_A1_1, and UCD_A_A2_1 in Table 3.3) of centrifuge tests, known as Model A tests in LEAP-ASIA-2019, were provided to the numerical simulation teams. Some of the tests were performed as part of the LEAP-UCD-2017 project. The numerical simulation teams were able to refer to detailed information of the experimental conditions such as the achieved base excitation and the centrifuge specimen (e.g., the density of the prepared specimen, as-built geometry, location of the sensors). For the Type-C simulations, the measured pore water pressures, accelerations (the sensor positions are shown in Tables 3.1 and 3.2), and displacements (the surface marker locations are shown in Fig. 3.3) could also be used to fine-tune the parameters of their constitutive models and the key parameters of their numerical simulation platform (damping ratio, permeability, etc.).

  2. 2.

    The achieved base excitations, the density of the soil specimen, and the locations of the sensors in two sets (KyU_A_B2_1 and RPI_A_B1_1 in Table 3.4) of centrifuge tests conducted as Model B tests for LEAP-ASIA-2019 were provided (including the complete set of the experimental results) to the numerical modeling teams. They were asked to simulate the results of these tests in terms of time histories of excess pore pressure, accelerations, and lateral displacements at selected locations. The Model B tests had the same prototype and model sizes as the corresponding Model A tests, but the centrifugal acceleration (η) was scaled in accordance with the generalized scaling law (Iai et al., 2005). Since the centrifugal accelerations in the Model B tests at a few facilities were much lower than that in the Model A tests, the model parameters might have needed to be fine-tuned to take into account the low confining stress effect on liquefaction strength. Such issues (revision of the parameters and their effects on liquefaction strength curves) should have been clearly documented in the reports submitted by the numerical simulation teams.

  3. 3.

    (Optional) If time allowed, the numerical simulation teams were welcome to perform Type-C simulations for the other centrifuge experiments in Tables 3.3 and 3.4 and submit the simulation results. Test cases highlighted in green were not mandatory but highly recommended.

The timeline of the second phase of the numerical simulations was as follows:

  1. 1.

    All the necessary data regarding Model A and Model B tests were provided to the numerical modelers on January 18, 2019.

  2. 2.

    The numerical simulation teams were requested to submit the results of their simulations for the required test cases (highlighted in yellow in Tables 3.3 and 3.4) by February 8, 2019.

4 Summary of Type-B/Type-C Simulation Results for Models A and B

As shown in Table 3.5, six simulation teams submitted the results (i.e., time histories of predicted accelerations, excess pore water pressures, and displacements at selected locations) of the Type-B numerical simulations, including the Type-B+ simulations with the adjustment of the hydraulic conductivity, for the selected LEAP-ASIA-2019 centrifuge tests. Also, three simulation teams submitted the results of the Type-C numerical simulations. Due to space limitation, only a subset of these data for mandatory test cases highlighted in yellow in Tables 3.3 and 3.4 is presented herein. In the following, selected time histories of accelerations, excess pore pressures, and lateral displacements are compared to provide representative samples of the performance of each simulation in comparison with the experimental data.

Figures 3.4, 3.5, 3.6, 3.7, 3.8, 3.9, 3.10, 3.11 and 3.12 illustrate the comparison of the numerical simulation results with the experiments for the three sets of Model A tests in Table 3.3 (i.e., KyU_A_A2_1, RPI_A_A1_1, and UCD_A_A2_1). For KyU_A_A2_1, the following observations are noted from a comparison in Figs. 3.4, 3.5 and 3.6:

  • The acceleration responses at deep sensor locations (e.g., AH1) do not differ significantly among simulations. However, the amplitude and waveform shape differ to some extent among simulations near the ground surface (e.g., AH4), where soil nonlinearity associated with the excess pore pressure increase may be strong.

  • In addition to the peak value of excess pore pressure, the experimental dynamic amplitudes during shaking and dissipation processes after shaking are challenging to fully capture at some sensor locations (e.g., P1 and P2), even in the Type-C simulations.

  • Although the amount of the residual displacements varies among simulations to some extent, the experimental tendency of accumulating displacement associated with lateral spreading in one direction during shaking is adequately simulated in a majority of the numerical simulations.

Fig. 3.4
10 line graphs each in 6 parts plot acceleration A H 1, A H 2, A H 3, A H 4, A H 6, and A H 9 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend has high fluctuations between 5 and 17 seconds and becomes 0 in the rest times.figure 4

Comparison of the measured and computed acceleration time histories for KyU_A_A2_1 test. (a) AH1, (b) AH2, (c) AH3, (d) AH4, (e) AH6, (f) AH9

Full size image
Fig. 3.5
10 line graphs each in 6 parts plot excess pore water pressures P 1, P 2, P 3, P 4, P 6, and P 8 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend rises initially, fluctuates heavily for a few seconds, and then drops.figure 5

Comparison of the measured and computed time histories of excess pore water pressures for KyU_A_A2_1 test. (a) P1, (b) P2, (c) P3, (d) P4, (e) P6, (f) P8

Full size image
Fig. 3.6
10 line graphs each in 3 parts plot a horizontal displacement at markers 2 to 2, 2 to 3, and 2 to 5 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the line follows a logarithmic increasing trend with slight fluctuations, except in simulations 4 and 6.

Comparison of the measured and computed time histories of ground surface lateral displacements for KyU_A_A2_1 test. (a) Marker 2–2, (b) Marker 2–3, (c) Marker 2–5

Full size image
Fig. 3.7
10 line graphs each in 6 parts plot acceleration A H 1, A H 2, A H 3, A H 4, A H 6, and A H 10 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend has high fluctuations between 5 and 17 seconds and becomes 0 in the rest times, except simulation 7 in B.figure 7

Comparison of the measured and computed acceleration time histories for RPI_A_A1_1 test. (a) AH1, (b) AH2, (c) AH3, (d) AH4, (e) AH6, (f) AH10

Full size image
Fig. 3.8
10 line graphs each in 6 parts plot excess pore water pressures P 1, P 2, P 4, P 6, P 7, and P 8 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In most graphs, the trend rises initially, fluctuates heavily for a few seconds, and then drops.figure 8

Comparison of the measured and computed time histories of excess pore water pressures for RPI_A_A1_1 test. (a) P1, (b) P2, (c) P4, (d) P6, (e) P7, (f) P8

Full size image
Fig. 3.9
10 line graphs each in 3 parts plot a horizontal displacement at markers 2 to 2, 2 to 3, and 2 to 5 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. All have a logarithmic increasing trend with slight fluctuations, except in measured and simulations 4 and 6.

Comparison of the measured and computed time histories of ground surface lateral displacements for RPI_A_A1_1 test. (a) Marker 2–2, (b) Marker 2–3, (c) Marker 2–5

Full size image
Fig. 3.10
10 line graphs each in 6 parts plot acceleration A H 1, A H 2, A H 3, A H 4, A H 6, and A H 9 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend has high fluctuations between 5 and 17 seconds and becomes 0 in the rest time. 2 graphs are empty in E and F.figure 10

Comparison of the measured and computed acceleration time histories for UCD_A_A2_1 test. (a) AH1, (b) AH2, (c) AH3, (d) AH4, (e) AH6, (f) AH9

Full size image
Fig. 3.11
10 line graphs each in 6 parts plot excess pore water pressures P 1, P 2, P 3, P 4, P 7, and P 8 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In most graphs, the trend rises initially, fluctuates heavily for a few seconds, and then drops.figure 11

Comparison of the measured and computed time histories of excess pore water pressures for UCD_A_A2_1 test. (a) P1, (b) P2, (c) P3, (d) P4, (e) P7, (f) P8

Full size image
Fig. 3.12
10 line graphs each in 3 parts plot a horizontal displacement at markers 2 to 2, 2 to 3, and 2 to 4 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. All have a logarithmic increasing trend with slight fluctuations, except in simulation 4 which is stable.

Comparison of the measured and computed time histories of ground surface lateral displacements for UCD_A_A2_1 test. (a) Marker 2–2, (b) Marker 2–3, (c) Marker 2–4

Full size image

A close examination of Figs. 3.7, 3.8 and 3.9 shows the following trend for RPI_A_A1_1:

  • As in the case of KyU_A_A2_1, most of the numerical simulations can reproduce the experimental acceleration responses at deep sensor locations (e.g., AH1). However, the simulated accelerations near the ground surface (e.g., AH4) are different among simulations; Simulations 1–3, 5, and 8 are capable of capturing well the observed spike waveform due to positive dilatancy.

  • Compared to the variation in the simulated excess pore pressure for KyU_A_A2_1, the simulated variation for RPI_A_A1_1 is not large; Simulations 2, 5, and 8 can reasonably simulate the observed pore pressure response, including the peak value, dynamic amplitude, and dissipation phase.

  • The variation in the simulated horizontal displacements for RPI_A_A1_1 looks smaller than that for KyU_A_A2_1; the observed displacement waveform is well simulated, particularly in Simulations 1 and 3, including the residual value as well as the dynamic amplitude.

For UCD_A_A2_1, Figs. 3.10, 3.11 and 3.12 demonstrate the following trend:

  • As in the previously mentioned two cases, a majority of the numerical simulations can adequately reproduce the observed acceleration responses at deep sensor locations (e.g., AH1). When it comes to the accelerations near the ground surface (e.g., AH4), most of the numerical simulations show spike waveforms due to positive dilatancy; in particular, Simulations 1–3, 5, and 8 can capture well the experimental result.

  • The variation in the simulated excess pore pressure for UCD_A_A2_1 looks similar as that for RPI_A_A1_1; Simulations 1–3 have a larger dynamic amplitude, which may be caused by strong positive dilatancy, while the amplitude is small for Simulations 5–7.

  • A majority of the numerical simulations can adequately capture the observed horizontal displacement waveform; the variation in the simulated residual displacements is not considerable.

Figures 3.13, 3.14, 3.15, 3.16, 3.17 and 3.18 compare the numerical simulation results with the corresponding experiments for the two sets of Model B tests in Table 3.4 (i.e., KyU_A_B2_1 and RPI_A_B1_1). As in the case of the numerical simulations for Model A tests, the variation in the simulated responses (i.e., accelerations, excess pore pressures, and horizontal displacements) exists among simulations. However, the comparison demonstrates that a majority of the numerical simulations are capable of simulating well the experimental trends; in particular, the Type-C simulations are found to capture the measured responses more accurately by adjusting the model parameters.

Fig. 3.13
10 line graphs each in 6 parts plot acceleration A H 1, A H 2, A H 3, A H 4, A H 6, and A H 9 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend resembles a damping signal.figure 13

Comparison of the measured and computed acceleration time histories for KyU_A_B2_1 test. (a) AH1, (b) AH2, (c) AH3, (d) AH4, (e) AH6, (f) AH9

Full size image
Fig. 3.14
10 line graphs each in 6 parts plot excess pore water pressures P 1, P 2, P 3, P 4, P 6, and P 8 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend rises initially, fluctuates heavily for a few seconds, and then drops.figure 14

Comparison of the measured and computed time histories of excess pore water pressures for KyU_A_B2_1 test. (a) P1, (b) P2, (c) P3, (d) P4, (e) P6, (f) P8

Full size image
Fig. 3.15
10 line graphs each in 3 parts plot a horizontal displacement at markers 2 to 2, 2 to 3, and 2 to 4 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. All have a logarithmic increasing trend with slight fluctuations, except measured and simulation 6 in A.

Comparison of the measured and computed time histories of ground surface lateral displacements for KyU_A_B2_1 test. (a) Marker 2–2, (b) Marker 2–3, (c) Marker 2–4

Full size image
Fig. 3.16
10 line graphs each in 6 parts plot acceleration A H 1, A H 2, A H 3, A H 4, A H 6, and A H 9 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend has high fluctuations between 5 and 17 seconds and becomes 0 in the rest times.figure 16

Comparison of the measured and computed acceleration time histories for RPI_A_B1_1 test. (a) AH1, (b) AH2, (c) AH3, (d) AH4, (e) AH6, (f) AH9

Full size image
Fig. 3.17
10 line graphs each in 6 parts plot excess pore water pressures P 1, P 2, P 3, P 4, P 6, and P 8 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. In all, the trend rises initially, fluctuates heavily for a few seconds, and then drops.figure 17

Comparison of the measured and computed time histories of excess pore water pressures for RPI_A_B1_1 test. (a) P1, (b) P2, (c) P3, (d) P4, (e) P6, (f) P8

Full size image
Fig. 3.18
10 line graphs each in 3 parts plot a horizontal displacement at markers 2 to 2, 2 to 3, and 2 to 5 versus time in seconds. The ten graphs are for the measured and simulations from 1 to 9. All have a logarithmic increasing trend with slight fluctuations, except simulations 4 and 6.

Comparison of the measured and computed time histories of ground surface lateral displacements for RPI_A_B1_1 test. (a) Marker 2–2, (b) Marker 2–3, (c) Marker 2–5

Full size image

5 Comparison of Numerical Simulation Performance in Terms of Horizontal Displacement

In order to further assess the quality of numerical simulations’ fit to the centrifuge test results and their performance in terms of horizontal displacement due to lateral spreading, Fig. 3.19 compares the relationship between Dr, PGAeff, and horizontal displacement (i.e., residual value) in the form of a balloon plot; the balloon size represents the amount of measured and simulated horizontal displacements. In the figure, the displacement (i.e., balloon size) should be smaller as Dr increases, whereas the displacement should be larger as PGAeff increases. The experimental results are generally in line with this trend, but some results do not follow this trend; this may be due to the variability of the centrifuge experiments. If such experimental variability could not be known prior to the Type-B simulations (e.g., Simulations 3, 4, and 8), it would be difficult to reproduce the observed results of such experiments accurately. On the other hand, the Type-C simulations can reproduce the experimental results, as shown in Simulations 5–7, because the model parameters are adjusted to match the experimental results. Thus, it is considered quite challenging to capture all measured responses perfectly by taking into account the experimental variability, but accumulating high-quality experimental results as a database can be essential to improve constitutive models and analytical platforms further.

Fig. 3.19
Nine bubble plots of P G A e f f versus D r in percentage. They have bubbles of the Med U times 2 100, 200, and 300 based on experiments and simulations 1 to 9. In all, the Med U times 2 300 for the experiment has the maximum value, except in simulation 9, where the Med U times 2 400 is high.figure 19

Summary of the measured and computed lateral displacements at the center of ground surface. (a) Simulation 1, (b) Simulation 2, (c) Simulation 3, (d) Simulation 4, (e) Simulation 5, (f) Simulation 6, (g) Simulation 7, (h) Simulation 8, (i) Simulation 9

Full size image

6 Conclusions

In the LEAP-ASIA-2019 prediction campaign, nine numerical simulation teams submitted Type-B or Type-C simulations on the seismic behavior of a uniform-density, 20-m-long, and 5-degree sandy slope; this chapter presented an overview of the simulation results (i.e., time histories of response accelerations, excess pore water pressures, and lateral displacements at the ground surface) and their comparison with the results of a selected set of centrifuge model tests. The comparison demonstrated that a majority of Type-B and Type-C numerical simulations were capable of simulating well the experimental trends observed in the centrifuge tests; in particular, Type-C simulations were found to capture the measured responses more accurately by adjusting the model parameters. Although it is quite challenging to capture all measured responses perfectly, the simulation exercises indicated that the numerical simulations could be further improved by accumulating high-quality experimental results as a database.