Evaluation of liquefaction induced lateral displacements using genetic programming

doi:10.1016/j.compgeo.2006.05.001

Computers and Geotechnics

Volume 33, Issues 4–5, June–July 2006, Pages 222-233

https://doi.org/10.1016/j.compgeo.2006.05.001 Get rights and content

Abstract

Determination of liquefaction induced lateral displacements during earthquake is a complex geotechnical engineering problem due to the complex and heterogeneous nature of the soils and the participation of a large number of factors involved. In this paper, a new approach is presented, based on genetic programming (GP), for determination of liquefaction induced lateral spreading. The GP models are trained and validated using a database of SPT-based case histories. Separate models are presented to estimate lateral displacements for free face and for gently sloping ground conditions. It is shown that the GP models are able to learn, with a very high accuracy, the complex relationship between lateral spreading and its contributing factors in the form of a function. The attained function can then be used to generalize the learning to predict liquefaction induced lateral spreading for new cases not used in the construction of the model. The results of the developed GP models are compared with those of a commonly used multi linear regression (MLR) model and the advantages of the proposed GP model over the conventional method are highlighted.

Introduction

During an earthquake, significant damage can result due to the ground movement and instability of the soil in the area affected by internal seismic waves. One of the common phenomena resulting from earthquake is liquefaction in loose saturated sands. Liquefaction is known as one of the major causes of ground movement and failure related to earthquake. It is believed to occur when the pore pressure approaches the confining pressure under earthquake loading where, as a result of a rapid and dramatic loss of soil strength, it can initiate movement of large blocks of soil with amplitudes ranging from a few centimeters to 10 m or more, eventually causing extensive damage to buried utilities, lifeline networks and many other underground and surface civil engineering structures.

Liquefaction induced lateral displacement generally occurs on gentle slopes ranging from 0.3° to 3° [1] founded on loose sand with groundwater level fairly close to the ground surface; however, open faces like stream channels are also susceptible to lateral spreading.

Liquefaction occurrence and the resulting lateral spreading depend on the physical and mechanical characteristics of the soil layers in the site, the depth of the water table, the intensity and duration of the ground shaking, the distance from the source of the earthquake and the seismic attenuation properties of in situ soil. Because of the participation of a large number of factors, the determination of liquefaction-induced lateral displacement is a complex geotechnical engineering problem. Several researchers have attempted to model this phenomenon using different techniques, a summary of which is presented in the following section.

Section snippets

Current practice in estimation of liquefaction induced lateral displacement

Due to the complexity of the problem, several analytical, numerical and empirical methods have been developed to estimate liquefaction induced lateral spreading.

Genetic programming

Genetic programming (GP) is an evolutionary computing method that generates a structured representation of the data provided. It imitates the biological evolution of living organisms. The technique which was introduced in the early 90s by Koza [23] makes use of the principles of genetic algorithms (GAs) to manipulate and optimize a population of computer models composed of functions and terminals in order to find a model that best fits the problem. These functions and terminals represent the

Genetic programming approach for estimation of lateral spreading

In this paper a new approach is presented for the assessment of liquefaction induced lateral displacement using genetic programming. A database of SPT-based case histories, compiled by Youd and Bartlett [19] has been used in this study to find the relationship between liquefaction induced lateral displacement and its contributing factors. This database is a revised version of the original database which was used previously by Bartlett and Youd [17], [18]. The corrections and modifications which

GP model of lateral displacement for free face cases

In the GP procedure, initially a number of potential models are randomly generated. Each model is trained and tested using the training and testing cases respectively. The fitness of each model is evaluated by minimizing the error i.e., the difference between the predicted and measured lateral displacements. If the errors calculated for the models in the population do not meet the termination criteria the evolutionary process continues in order to create a new generation of models. The

GP model of lateral displacement for cases involving gently sloping ground

The database also includes 256 case histories of lateral displacement in gently sloping ground. These cases were used for the development of the second GP model for sloping ground conditions. Of these, 201 cases (78.5%) were used for training of the GP model and the remaining 55 cases (21.5%) were used for validation. After the analysis of a number of different alternative models, the following model was found to be the most appropriate model for gently sloping ground condition: ${Dh}_{c} = - 0.8 \frac{F_{15}}{M} +$

GP models for moderate lateral displacements

It is seen from Fig. 5, Fig. 6 that, in general, the GP models (Eqs. (5), (6)), predict lateral displacements with high accuracy, particularly for cases with displacements greater than 1.5 m. However, as shown in Figs. 7(a) and (b) (which are the enlarged views of the initial parts of Figs. 5(a) and (b) respectively), this approach does not provide the same level of accuracy in the estimation of lateral displacements for cases with moderate displacements of less than about 1.5 m. Youd et al. [19]

Sensitivity analysis

In order to examine the sensitivity of the models to data set size, the data sets for free face and gently sloping ground conditions were gradually reduced in size to 160, 120, 80 and 60 cases. For each case, the reduced data set was divided into training and validation subsets and GP models were constructed and validated. Table 6 shows the results of the sensitivity analysis for the GP models constructed with different number of cases. It is shown that the proposed GP methodology performs well

Results and discussion

A new approach, based on genetic programming, has been presented for prediction of liquefaction-induced lateral displacements during earthquake. Two separate models have been presented, one for whole range of displacements and one for moderate displacements of up to 1.5 m. Fig. 5, Fig. 6, Fig. 8, Fig. 9 show the prediction capabilities of the GP models as plots of measured versus predicted displacements. It is shown that the models are capable of learning, with a very high accuracy, the complex

Summary and conclusion

Determination of liquefaction induced lateral displacements during earthquake is a complex geotechnical engineering problem due to the heterogeneous nature of soils and the participation of a large number of factors involved.

In this paper, a new approach was presented, based on genetic programming, for determination of liquefaction induced lateral displacements due to earthquake. Two general GP models were trained and validated using a database of SPT-based case histories for free face and

Acknowledgements

This work used software developed by Ozario Giustlisi (Technical University of Bari) and Dragan Savic (University of Exeter).

References (25)

M. Hamada et al.
Study on permanent ground displacements induced by seismic liquefaction
Comput Geotech
(1987)
S. Yasuda et al.
The mechanism and simplified procedure for analysis of permanent ground displacement due to liquefaction
Soil Foundat
(1992)
I. Towhata et al.
Prediction of permanent displacement of liquefied ground by means of energy principle
Soils Foundat
(1992)
J. Wang et al.
A neural network model for liquefaction induced horizontal ground displacement
J Soil Dyn Earthquake Eng
(1999)
M.H. Baziar et al.
Evaluation of lateral spreading using artificial neural networks
J Soil Dyn Earthquake Eng
(2005)
S. Sette et al.
Genetic programming: principles and applications
Eng Appl Artificial Intell
(2001)
Committee on Earthquake Engineering. Liquefaction of soils during earthquakes, Committee on Engineering and Technical...
Rauch AF. EPOLLS: An empirical method for predicting surface displacements due to liquefaction-induced lateral...
W.H. Gu et al.
Progressive failure of Lower San Fernando Dam
J Geotech Eng
(1993)
W.H. Gu et al.
Postearthquake deformation analysis of Wildlife site
J Geotech Eng
(1994)

Yasuda S, Nagase H, Kiku H, Uchida Y. A simplified procedure for the analysis of the permanent ground displacement. In:...

Yasuda S, Nagase H, Kiku H, Uchida Y. Countermeasures against the permanent ground displacement due to liquefaction....

Cited by (138)

Predicting the small strain shear modulus of sands and sand-fines binary mixtures using machine learning algorithms
2024, Transportation Geotechnics
This study aims to develop several novel machine learning (ML) evolutionary algorithms for the prediction of small strain shear modulus ( $G_{\max}$ ) of clean sands and sand-fines binary mixtures. To this end, five key features of isotropic confining pressure ( $p$ ), void ratio ( $e$ ), uniformity coefficient ( $C_{u}$ ), particle shape descriptor ( $ρ$ ), and non-plastic fines content ( $FC$ ) are adopted as the inputs to artificial neural network (ANN) models as well as genetic programming (GP) algorithm so as to render the maximum shear modulus of granular soils as the output. Accordingly, a comprehensive dataset containing 1055 $G_{\max}$ data points is exploited to develop ML simulations. The validity of ML-based models in estimating the $G_{\max}$ of clean sands and sand-silt mixtures is rigorously examined through various statistical indices and measurement criteria. The results show that a novel ML model utilizing ANN-Levenberg Marquardt (LM) in conjunction with an evolutionary optimization method named Success History-based Adaptive Differential Evolution with Linear population size reduction (LSHADE) is capable of predicting $G_{\max}$ data with a very high precision rendering R² values of 0.9833, 0.9841, 0.9802, and 0.9835 for the whole, training, validation, and test datasets, respectively. Meanwhile, using the well-established GP algorithm, a new practical model is proposed to predict the $G_{\max}$ of clean sands and sand-fines mixtures containing non-cohesive silt inclusion with R² values of 0.9323, 0.9351, and 0.9312 for the whole, training, and test datasets, respectively. Finally, the proposed models of ANN-LSHADE-LM and GP are shown to be appreciably superior, in terms of accuracy, to all commonly used empirical correlations in the literature for $G_{\max}$ estimation.
Assessment of liquefaction-induced lateral spread using soft computing approaches
2023, Gondwana Research
Lateral spread is a common ground deformation induced by soil liquefaction, which has caused extensive damage to buildings and lifelines in past earthquakes. At present, several empirical methods for assessing lateral displacement have been proposed. Due to the complexity of the problem, it is difficult for traditional methods to predict the lateral spread displacement with high accuracy. In this study, a new database of 522 lateral spreading case histories was established based on the databases developed by Youd et al., Chu et al., and NGL (Next Generation Liquefaction). The confidence ellipse was used to evaluate the similarity between parameters of liquefiable soil layers for two site conditions (free face and gently sloping ground). Then, based on the new database, predictive models for assessing liquefaction-induced lateral displacement were developed by different soft computing approaches including Ridge Regression, Lasso & Lasso CV, Random Forest Regression (RF), eXtreme Gradient Boost (XGBoost), and Multivariate Adaptive Regression Splines (MARS). The MARS model showed good predictive performance (the R² values were 0.583 and 0.555 on testing set, respectively) and can generate explicit expressions. The results of SHAP investigation showed that log(T₁₅) is the most important feature variable in RF model, where T₁₅ is the thickness of liquefiable soil layer. Further analysis showed that thicker liquefiable soil layers, smaller epicentral distances, larger magnitudes, and unfavorable topographical conditions will increase the possibility of lateral displacement, while larger grain sizes and more fines contents will reduce the probability of lateral displacement.
Smart prediction of liquefaction-induced lateral spreading
2023, Journal of Rock Mechanics and Geotechnical Engineering
The prediction of liquefaction-induced lateral spreading/displacement (D_h) is a challenging task for civil/geotechnical engineers. In this study, a new approach is proposed to predict D_h using gene expression programming (GEP). Based on statistical reasoning, individual models were developed for two topographies: free-face and gently sloping ground. Along with a comparison with conventional approaches for predicting the D_h, four additional regression-based soft computing models, i.e. Gaussian process regression (GPR), relevance vector machine (RVM), sequential minimal optimization regression (SMOR), and M5-tree, were developed and compared with the GEP model. The results indicate that the GEP models predict D_h with less bias, as evidenced by the root mean square error (RMSE) and mean absolute error (MAE) for training (i.e. 1.092 and 0.815; and 0.643 and 0.526) and for testing (i.e. 0.89 and 0.705; and 0.773 and 0.573) in free-face and gently sloping ground topographies, respectively. The overall performance for the free-face topology was ranked as follows: GEP > RVM > M5-tree > GPR > SMOR, with a total score of 40, 32, 24, 15, and 10, respectively. For the gently sloping condition, the performance was ranked as follows: GEP > RVM > GPR > M5-tree > SMOR with a total score of 40, 32, 21, 19, and 8, respectively. Finally, the results of the sensitivity analysis showed that for both free-face and gently sloping ground, the liquefiable layer thickness (T₁₅) was the major parameter with percentage deterioration (%D) value of 99.15 and 90.72, respectively.
Prediction of lateral spreading displacement using conditional Generative Adversarial Network (cGAN)
2022, Soil Dynamics and Earthquake Engineering
Lateral spreading is the most pervasive type of earthquake-induced ground deformation, which can cause considerable damage to engineered structures and lifelines. There are several factors, such as soil properties and ground motion characteristics that affect the liquefaction induced lateral spread. This inherent complexity and nonlinear relationship between the variables make it difficult to predict lateral spread with high accuracy. There are several empirical and machine learning models developed to predict lateral spread. In this study, a conditional generative adversarial network (cGAN) is developed to predict the horizontal ground displacements. A ten-fold cross validation is used to assess the model performance. The average accuracy of the model for both free face and ground slope conditions are found to be 82% and 68%, respectively. Shapley additive explanations-based sensitivity analysis was carried out to identify the important parameters that influence the lateral displacement.
Use of soft computing techniques for tunneling optimization of tunnel boring machines
2021, Underground Space (China)
Citation Excerpt :
Jain et al. (2004) showed that the ANN model captured some physical process features in a rainfall-runoff ANN model. Some scholars used soft computing methods to establish simplified formulas for solving geotechnical engineering problems (Javadi, Rezania, & Mousavi, 2006; Johari, Habibagahi, & Ghahramani, 2006; Padmini, Ilamparuthi, & Sudheer, 2008; Samui, 2008). Kim et al. (2001) used neural networks to predict ground surface settlements during the construction of the Seoul subway.
Thanks to advances in tunnel boring machine (TBM) and monitoring, significant progress has been achieved in the application of soft computing techniques for the optimization of TBM tunneling and the reduction of disturbance related to tunneling in urban areas. Because experimental, analytical, and numerical methods have limitations in solving problems related to TBM tunneling, engineers can use soft computing techniques in analyzing the relationship between the target tunneling responses and influential design inputs parameters, including the geometrical, geological, and TBM operational factors. These techniques are useful in achieving robust and low-cost solutions. However, engineers face difficulties in making an optimal choice of the soft computing technique to solve the complex problems related to TBM tunneling. To help with this choice, this study presents state of the art about the use of soft computing techniques in TBM tunneling through practical applications. The study proposes recommendations for the optimal use of these techniques, in particular (i) the importance of preliminary analyses for the selection and reduction of input parameters, (ii) the necessity to complete insufficient data using laboratory tests and numerical modeling, (iii) the selection of reduced number of hidden layers and nodes to avoid overfitting, (iv) the use of recurrent neural networks to deal with time-series data, and (v) the association of soft computing methods with hybrid optimization techniques to reduce the risk of convergence to local minima.
Predicting resilient modulus of compacted subgrade soils under influences of freeze–thaw cycles and moisture using gene expression programming and artificial neural network approaches
2021, Transportation Geotechnics
This study aims at developing a gene expression programming (GEP) model and an artificial neural network (ANN) model to predict the resilient modulus (M_R) of compacted pavement subgrade soils based on their physical properties, external stress states, and environmental factors. A database of 2813 M_R measurements derived from 12 subgrade soils with different moisture-temperature histories was established for model development and validation. Influencing factors considered in this database include the weighted plasticity index (wPI), dry unit weight (γ_d), confining stress (σ_c), deviator stress (σ_d), moisture content (w), and the number of freeze–thaw cycles (N_FT). Sensitivity analysis was conducted to evaluate the importance of each factor. The order of influencing factors by decreasing importance was found to be wPI, γ_d, w, N_FT, σ_d, σ_c and the importance of the σ_c and σ_d is remarkably lower than other factors. This indicates that the M_R of compacted subgrade soils is strongly dependent on the soil type (wPI, γ_d) and is more sensitive to environmental factors (N_FT, w) than external stress states (σ_c, σ_d). The developed GEP and ANN models reasonably predicted the M_R in the database and achieved better performance compared to several existing empirical models.

View all citing articles on Scopus

View full text

Evaluation of liquefaction induced lateral displacements using genetic programming

Abstract

Introduction

Section snippets

Current practice in estimation of liquefaction induced lateral displacement

Genetic programming

Genetic programming approach for estimation of lateral spreading

GP model of lateral displacement for free face cases

GP model of lateral displacement for cases involving gently sloping ground

GP models for moderate lateral displacements

Sensitivity analysis

Results and discussion

Summary and conclusion

Acknowledgements

Comput Geotech

Soil Foundat

Soils Foundat

J Soil Dyn Earthquake Eng

J Soil Dyn Earthquake Eng

Eng Appl Artificial Intell

Progressive failure of Lower San Fernando Dam

J Geotech Eng

Postearthquake deformation analysis of Wildlife site

J Geotech Eng