Evaluation of liquefaction induced lateral displacements using genetic programming

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.     

基于贝叶斯网络的地震液化概率预测分析

基于解释结构模型和因果图法,选取12个具有代表性的定性和定量因素,在大量数据不完备的情况下提出了建立贝叶斯网络液化模型的方法。以2011年日本东北地区太平洋近海地震液化不完备数据为例,采用总体精度、ROC曲线下面积、准确率、召回率和F1值5项指标对模型进行综合评估,并与径向基神经网络模型进行对比。结果表明：贝叶斯网络液化模型的回判和预测效果都优于径向基神经网络模型,且对于数据缺失的样本的预测效果也较理想。此外,该模型对于不同土质的液化评估均有较好的适用性。分类不均衡和抽样偏差会对模型的学习和预测效果产生很大影响,建议应同时采用上述5项评估指标进行综合评估模型的优劣。     

Use of evolutionary computing for modelling some complex problems in geotechnical engineering

In this paper, the feasibility of using evolutionary computing for solving some complex problems in geotechnical engineering is investigated. The paper presents a relatively new technique, i.e. evolutionary polynomial regression (EPR), for modelling three practical applications in geotechnical engineering including the settlement of shallow foundations on cohesionless soils, pullout capacity of small ground anchors and ultimate bearing capacity of pile foundations. The prediction results from the proposed EPR models are compared with those obtained from artificial neural network (ANN) models previously developed by the author, as well as some of the most commonly available methods. The results indicate that the proposed EPR models agree well with (or better than) the ANN models and significantly outperform the other existing methods. The advantage of EPR technique over ANNs is that EPR generates transparent and well-structured models in the form of simple and easy-to-use hand calculation formulae that can be readily used by practising engineers.     

A Novel Methodology to Classify Soil Liquefaction Using Deep Learning

In this research, deep learning (DL) model is proposed to classify the soil reliability for liquefaction. The applicability of the DL model is tested in comparison with emotional backpropagation neural network (EmBP). The database encompassing cone penetration test of Chi–Chi earthquake. This study uses cone resistance (q_c) and peck ground acceleration as inputs for prediction of liquefaction susceptibility of soil. The performance of developed models has been assessed by using various parameters (receiver operating characteristic, sensitivity, specificity, Phi correlation coefficient, Precision–Recall F measure). The performance of DL is excellent. Consistent results obtained from the proposed deep learning model, compared to the EmBP, indicate the robustness of the methodology used in this study. In addition, both the developed model was also tested on global earthquake data. During validation on global data, both the models shows good results based on fitness parameters. The developed classification models a simple, but also efficient decision-making tool in engineering design to quantitatively assess the liquefaction potential. The finding of this paper can be further used to capture the relationship between soil and earthquake parameters.

     

Least square support vector machine and relevance vector machine for evaluating seismic liquefaction potential using SPT

The determination of liquefaction potential of soil is an imperative task in earthquake geotechnical engineering. The current research aims at proposing least square support vector machine (LSSVM) and relevance vector machine (RVM) as novel classification techniques for the determination of liquefaction potential of soil from actual standard penetration test (SPT) data. The LSSVM is a statistical learning method that has a self-contained basis of statistical learning theory and excellent learning performance. RVM is based on a Bayesian formulation. It can generalize well and provide inferences at low computational cost. Both models give probabilistic output. A comparative study has been also done between developed two models and artificial neural network model. The study shows that RVM is the best model for the prediction of liquefaction potential of soil is based on SPT data.     

基于自组织神经网络的砂土液化评价

在分析自组织特征映射(SOFM)神经网络基本学习算法的基础上,从提高算法收敛速度和性能出发,提出了一种改进算法:根据实际应用并结合专家经验确定初始连接权值;采用高斯函数作为拓扑邻域函数;将算法分为粗调整和细调整两个阶段,分别采用不同的学习率和邻域函数,然后采用改进后的SOFM算法对砂土液化进行评价。实例研究表明,应用SOFM神经网络评价砂土液化高效可行,为砂土液化评价提供了新方法。     

Prediction of cyclic resistance ratio for silty sands and its applications in the simplified liquefaction analysis

Liquefaction resistance of granular soils is commonly characterized by the cyclic resistance ratio (CRR) in the simplified shear stress procedure of liquefaction potential assessment. This parameter is commonly estimated by cyclic tests on reconstituted samples or empirical correlations between liquefied/non-liquefied case histories. The current study employs results of cyclic triaxial tests on reconstituted soil specimens and presents a predictive equation for cyclic resistance ratio (CRR) of clean and silty sands. The CRR equation is a function of relative density, effective mean confining pressure, non-plastic fines content, number of harmonic cycles for liquefaction onset, and some other basic soil properties. It is demonstrated that the developed relationship obtains reasonable accuracy in the prediction of laboratory-based CRR. Based on the developed CRR model, new relationships are then presented for the coefficient of effective overburden pressure (K_σ) and magnitude scaling factor (MSF), two important modification factors in the simplified shear stress procedure. These new modification factors are then compared with those recommended by previous researchers. Finally, the possible application of the proposed CRR model in field condition is shown for a specific case. This study provides a preliminary insight into the liquefaction resistance of silty sands prior to the complementary laboratory studies.     

Assessment and Prediction of Liquefaction Potential Using Different Artificial Neural Network Models: A Case Study

Soil liquefaction as a transformation of granular material from solid to liquid state is a type of ground failure commonly associated with moderate to large earthquakes and refers to the loss of strength in saturated, cohesionless soils due to the build-up of pore water pressures and reduction of the effective stress during dynamic loading. In this paper, assessment and prediction of liquefaction potential of soils subjected to earthquake using two different artificial neural network models based on mechanical and geotechnical related parameters (model A) and earthquake related parameters (model B) have been proposed. In model A the depth, unit weight, SPT-N value, shear wave velocity, soil type and fine contents and in model B the depth, stress reduction factor, cyclic stress ratio, cyclic resistance ratio, pore pressure, total and effective vertical stress were considered as network inputs. Among the numerous tested models, the 6-4-4-2-1 structure correspond to model A and 7-5-4-6-1 for model B due to minimum network root mean square errors were selected as optimized network architecture models in this study. The performance of the network models were controlled approved and evaluated using several statistical criteria, regression analysis as well as detailed comparison with known accepted procedures. The results represented that the model A satisfied almost all the employed criteria and showed better performance than model B. The sensitivity analysis in this study showed that depth, shear wave velocity and SPT-N value for model A and cyclic resistance ratio, cyclic stress ratio and effective vertical stress for model B are the three most effective parameters on liquefaction potential analysis. Moreover, the calculated absolute error for model A represented better performance than model B. The reasonable agreement of network output in comparison with the results from previously accepted methods indicate satisfactory network performance for prediction of liquefaction potential analysis.     

Application of the adaptive neuro-fuzzy inference system for prediction of soil liquefaction

The determination of liquefaction potential of soils induced by earthquake is a major concern and an essential criterion in the design process of the civil engineering structures. A purely empirical interpretation of the filed case histories relating to liquefaction potential is often not well constrained due to the complication associated with this problem. In this study, an integrated fuzzy neural network model, called Adaptive Neuro-Fuzzy Inference System (ANFIS), is developed for the assessment of liquefaction potential. The model is trained with large databases of liquefaction case histories. Nine parameters such as earthquake magnitude, the water table, the total vertical stress, the effective vertical stress, the depth, the peak acceleration at the ground surface, the cyclic stress ratio, the mean grain size, and the measured cone penetration test tip resistance were used as input parameters. The results revealed that the ANFIS model is a fairly promising approach for the prediction of the soil liquefaction potential and capable of representing the complex relationship between seismic properties of soils and their liquefaction potential.     

An evolutionary approach to modelling the thermomechanical behaviour of unsaturated soils

A new data‐mining approach is presented for modelling of the stress–strain and volume change behaviour of unsaturated soils considering temperature effects. The proposed approach is based on the evolutionary polynomial regression (EPR), which unlike some other data‐mining techniques, generates a transparent and structured representation of the behaviour of systems directly from raw experimental (or field) data. The proposed methodology can operate on large quantities of data in order to capture nonlinear and complex relationships between contributing variables. The developed models allow the user to gain a clear insight into the behaviour of the system. Unsaturated triaxial test data from the literature were used for development and verification of EPR models. The developed models were also used (in a coupled manner) to produce the entire stress path of triaxial tests. Comparison of the EPR model predictions with the experimental data revealed the robustness and capability of the proposed methodology in capturing and reproducing the constitutive thermomechanical behaviour of unsaturated soils. More importantly, the capability of the developed models in accurately generalizing the predictions to unseen data cases was illustrated. The results of a sensitivity analysis showed that the models developed from data are able to capture and represent the physical aspects of the unsaturated soil behaviour accurately. The merits and advantages of the proposed methodology are also discussed. Copyright © 2014 John Wiley & Sons, Ltd.     

二维场地液化势预测的神经网络方法

基于人工神经网络,提出了场地液化势预测模型。场地液化势的空间数据结构特征可由不同参数的自回归神经网络（GRNN）来模拟。该预测模型的一个重要参数spread可用地质统计学（Kriging）方法中的交叉验证技术来确定。研究表明,在最优spread参数条件下GRNN能够较好地映射场地液化势数据结构特征。用GRNN模型预测结果与经典的Kriging估计方法所得到的结果十分吻合。GRNN模型可以用于二维空间数据的预测及基于GIS决策系统。     

Predicting flow liquefaction,a constitutive model approach

In this paper, flow liquefaction criterion for contractive loose sands is analytically extracted based on the fundamental definition of flow liquefaction. In order to obtain the closed form of this criterion, Dafalias–Manzari constitutive model is employed; so the stress ratio at the onset of flow liquefaction is presented as a function of model parameters, state parameter and void ratio. Flow liquefaction line, as a graphical form of suggested criterion in stress space, shows that the peak points of undrained stress paths with same void ratios are not necessarily in a straight line. In order to validate the reliability of proposed flow liquefaction line to predict the onset of instability, it has been compared with the results of experimental tests performed on Toyoura, Ottawa and Leighton Buzzard sands. The verification results show that the present criterion can satisfactorily predict the onset of flow liquefaction in monotonic and cyclic undrained tests of saturated sands as well as the structural collapse in constant deviatoric stress tests of loose dry sands.     

利用模糊神经网络进行砂土液化势评判

利用模糊信息分析表达知识和人工神经网络在映射能力方面的优势, 选取应力比、震级、地面运动最大加速度、标贯击数、地下水位作为评价参数指标, 构造砂土液化势识别的模糊神经网络模型。验证和应用结果表明, 模糊神经网络模型可提供更高的映射能力, 是砂土液化势评价预测的有效手段。     

Probabilistic assessment of soil liquefaction considering spatial variability of CPT measurements

The conventional liquefaction potential assessment methods (also known as simplified methods) profoundly rely on empirical correlations based on observations from case histories. A probabilistic framework is developed to incorporate uncertainties in the earthquake ground motion prediction, the cyclic resistance prediction, and the cyclic demand prediction. The results of a probabilistic seismic hazard assessment, site response analyses, and liquefaction potential analyses are convolved to derive a relationship for the annual probability and return period of liquefaction. The random field spatial model is employed to quantify the spatial uncertainty associated with the in-situ measurements of geotechnical material.     

Developing constitutive models from EPR‐based self‐learning finite element analysis

A constitutive model that captures the material behavior under a wide range of loading conditions is essential for simulating complex boundary value problems. In recent years, some attempts have been made to develop constitutive models for finite element analysis using self‐learning simulation (SelfSim). Self‐learning simulation is an inverse analysis technique that extracts material behavior from some boundary measurements (eg, load and displacement). In the heart of the self‐learning framework is a neural network which is used to train and develop a constitutive model that represents the material behavior. It is generally known that neural networks suffer from a number of drawbacks. This paper utilizes evolutionary polynomial regression (EPR) in the framework of SelfSim within an automation process which is coded in Matlab environment. EPR is a hybrid data mining technique that uses a combination of a genetic algorithm and the least square method to search for mathematical equations to represent the behavior of a system. Two strategies of material modeling have been considered in the SelfSim‐based finite element analysis. These include a total stress‐strain strategy applied to analysis of a truss structure using synthetic measurement data and an incremental stress‐strain strategy applied to simulation of triaxial tests using experimental data. The results show that effective and accurate constitutive models can be developed from the proposed EPR‐based self‐learning finite element method. The EPR‐based self‐learning FEM can provide accurate predictions to engineering problems. The main advantages of using EPR over neural network are highlighted.     

http://www.sciencedirect.com/science/article/pii/S167498711100137X

This study presents promising variants of genetic programming (GP),namely linear genetic programming (LGP) and multi expression programming (MEP) to evaluate the liquefaction resistance of sandy soils....     

An intelligent multi-objective EPR technique with multi-step model selection for correlations of soil properties

Current multi-objective evolutionary polynomial regression (EPR) methodology has difficulties on decision-making of optimal EPR model. This paper proposes an intelligent multi-objective optimization-based EPR technique with multi-step automatic model selection procedure. A newly developed multi-objective differential evolution algorithm (MODE) is adopted to improve the optimization performance. The proposed EPR process is composed of two stages: (1) intelligent roughing model selection and (2) model delicacy identification. In the first stage, besides two objectives (model accuracy and model complexity), the model robustness measured by robustness ratio is considered as an additional objective in the multi-objective optimization. In the second stage, a new indicator named selection index is proposed and incorporated to find the optimal model. After intelligent roughing selection and delicacy identification, the optimal EPR model is obtained considering the combined effects of correlation coefficient, size of polynomial terms, number of involved variables, robustness ratio and monotonicity. To show the practicality of the proposed EPR technique, three illustrative cases helpful for geotechnical design are presented: (a) modelling of compressibility, (b) modelling of undrained shear strength and (c) modelling of hydraulic conductivity. For each case, a practical formula with better performance in comparison with various existing empirical equations is finally provided. All results demonstrate that the proposed intelligent MODE-based EPR technique is efficient and effective.

     

A Deep-Learning-Based Geological Parameterization for History Matching Complex Models

A new low-dimensional parameterization based on principal component analysis (PCA) and convolutional neural networks (CNN) is developed to represent complex geological models. The CNN–PCA method is inspired by recent developments in computer vision using deep learning. CNN–PCA can be viewed as a generalization of an existing optimization-based PCA (O-PCA) method. Both CNN–PCA and O-PCA entail post-processing a PCA model to better honor complex geological features. In CNN–PCA, rather than use a histogram-based regularization as in O-PCA, a new regularization involving a set of metrics for multipoint statistics is introduced. The metrics are based on summary statistics of the nonlinear filter responses of geological models to a pre-trained deep CNN. In addition, in the CNN–PCA formulation presented here, a convolutional neural network is trained as an explicit transform function that can post-process PCA models quickly. CNN–PCA is shown to provide both unconditional and conditional realizations that honor the geological features present in reference SGeMS geostatistical realizations for a binary channelized system. Flow statistics obtained through simulation of random CNN–PCA models closely match results for random SGeMS models for a demanding case in which O-PCA models lead to significant discrepancies. Results for history matching are also presented. In this assessment CNN–PCA is applied with derivative-free optimization, and a subspace randomized maximum likelihood method is used to provide multiple posterior models. Data assimilation and significant uncertainty reduction are achieved for existing wells, and physically reasonable predictions are also obtained for new wells. Finally, the CNN–PCA method is extended to a more complex nonstationary bimodal deltaic fan system, and is shown to provide high-quality realizations for this challenging example.

     

A rational method for development of limit state for liquefaction evaluation based on shear wave velocity measurements

This papers presents a new approach for developing a limit state for liquefaction evaluation based on field performance data. As an example to illustrate the new approach, a database that consists of, among many other features, in situ shear wave velocity measurements and field observations of liquefaction/non‐liquefaction in historic earthquakes is analysed. This database is first used to train a neural network to classify liquefaction/non‐liquefaction based on soil resistance parameters and load parameters. The successfully trained and tested neural network is then used to establish a limit state, a multiple dimension boundary that separates ‘zone’ of liquefaction from ‘zone’ of non‐liquefaction. The limit state yields cyclic resistance ratio for a given set of soil resistance parameters. Examination of all cases in the database show that the developed limit state has a high degree of accuracy in predicting the occurrence of liquefaction/non‐liquefaction. The developed neural network model can accurately predict the cyclic resistance ratio of soils. Copyright © 2000 John Wiley & Sons, Ltd.     

Modeling and Interpretation of Pressuremeter Test Results with Artificial Neural Networks

In this paper, three types of artificial neural network (ANN) are employed to prediction and interpretation of pressuremeter test results. First, multi layer perceptron neural network is used. Then, neuro-fuzzy network is employed and finally radial basis function is applied. All applied networks have shown favorable performance. Finally, different models have been compared and network with the most outstanding performance in two stages is determined. Contrary to conventional behavioral models, models based neural network do not demonstrate the effect of input parameters on output parameters. This research is response to this need through conducting sensitivity analysis on the optimal structure of proposed models.