Professor Jinsong Huang

A three-dimensional representation of the random distribution of fractures in rock masses, known as the discrete fracture network (DFN), is widely used to analyze the stability of jointed rock slopes. In this paper, a new framework for constructing three-dimensional DFN models in rock masses has been proposed to overcome the limitations of conventional circular or polygon-based models. The framework utilizes the universal elliptical disc (UED) model and integrates it with the discrete element method in 3DEC for the stability evaluation of jointed rock slopes. This paper starts by introducing the basic principles of the UED model. The procedures for constructing a three-dimensional DFN using the UED model is then outlined. In the present study, a case study of a rock slope in Zhejiang Province, China is used to demonstrate the implementation of the proposed framework. A comprehensive comparative study is conducted to investigate the impacts of several UED model parameters, including the ratio of major axis to minor axis and rotation angle, and discontinuity density, on the stability of rock slopes and compared to the conventional Baecher disc model. The results show that the framework can effectively integrate the UED model into 3DEC and provide a realistic representation of the three-dimensional DFN, leading to improved accuracy and efficiency in the stability evaluation of jointed rock slopes. The framework also shed light on the interactive effects of the UED model parameters and discontinuity density on the rock slope stability, providing a strong reference for using the UED model in constructing DFN models for rock slopes.

How to select non-landslide samples for landslide susceptibility prediction (LSP) modeling is an important uncertainty affecting the LSP results. To study the influence of different non-landslide sample selection methods on LSP modeling, five sampling methods were proposed (Randomly selected from the whole area, from the specific attribute area with a slope lower than 5°, from the area outside buffer zone which is 300 m from each landslide, selected by information value method, selected by Semi-supervised machine learning) with the same number of landslide grid units, and coupled with Random Forest (RF) to construct random selection-RF, low-slope RF, buffer-based RF, IV-RF, and semi-supervised RF models for LSP. Taking Nankang County of Jiangxi province as the study area, a total of 19 environmental factors such as elevation, slope, population density, and road density were acquired, and 233 landslide inventories were obtained. The landslide inventory was divided into 2598 grids as landslide samples to construct the input-output of the above-coupled model. Then, the prediction accuracy and the distribution characteristics of predicted landslide susceptibility indexes were used to analyze the LSP modeling uncertainty. To further solve the problem of unreasonable distribution of landslide susceptibility indexes predicted by the coupled model, a sample set with a 1:2 ratio of landslide to non-landslide was used for LSP, and the condition of the sample set with equal proportion was compared in semi-supervised RF. Results showed that: 1) The prediction accuracy of models such as low-slope RF, buffer-based RF, IV-RF, and semi-supervised RF was substantially better than that of the random selection-RF model, suggesting that accurate selection of non-landslide samples was critical for LSP. 2) The modeling performance of the semi-supervised RF was optimal, which predicted the distribution characteristics of landslide susceptibility indexes more accurately and reliably at landslide:non-landslide = 1:2 than at 1:1. It is necessary to explore the ratio of landslide to non-landslide samples in depth in future studies.

Fully supervised machine learning models are widely applied for landslide susceptibility prediction (LSP), mainly using landslide and non-landslide samples as output variables and related conditioning factors as input variables. However, there are many uncertain issues in LSP modelling; for example, known landslide samples may have errors, non-landslide samples randomly selected from the whole study area are not accurate, the ratio of landslide to non-landslide samples set as 1:1 is not consistent with the actual landslide distribution characteristics, it is unreasonable to assign samples labelled non-landslide a probability of 0, and it is difficult to achieve a comprehensive assessment of LSP performance. Based on a review of the literature, we innovatively propose a semi-supervised imbalanced theory to overcome these uncertain issues. First, based on landslide samples (occurrence probability assigned 1), randomly selected non-landslide samples (occurrence probability assigned 0), and slope units divided by the multi-scale segmentation method and related conditioning factors, a supervised machine learning model is constructed and used to predict the initial landslide susceptibility indexes (LSIs), which are then classified as very low, low, moderate, high and very high landslide susceptibility levels (LSLs). Second, the landslide samples with LSLs classified as very low are removed to reduce errors in landslides, and non-landslide samples are randomly selected from the low and very low LSL groups to ensure the accuracy of non-landslides. We refer to this type of sample selection as a semi-supervised learning strategy. Third, the sampling ratio of landslide to non-landslide samples is successively set to values from 1:1 to 1:200, the initial LSIs are assigned as the labels of the corresponding non-landslide samples, and the labels of landslide samples are still assigned the value 1. We call these processes as the imbalanced sampling strategy. Fourth, we use the labelled landslide and non-landslide samples to train and test the supervised machine learning again. Finally, the optimal ratio of landslide samples to non-landslide samples can be determined to obtain the final LSP results through comparisons of LSP accuracy and LSI distribution characteristics under different sampling ratios. Jiujiang City in Jiangxi Province of China is the study area. The results show that the ROC and prediction rate accuracies of semi-supervised imbalanced RF model gradually increase from 0.979 and 0.853 to 0.990 and 0.912, respectively, with the imbalanced ratios rise from 1:1 to 1:160. Then both accuracies tend to converge as the ratio rises from 160 to 200. Hence, the LSP results of the semi-supervised imbalanced theory are efficient when the ratio of landslides to non-landslides is1:160. We conclude that the proposed theory significantly improves the theoretical basis of LSP modelling.

Probabilistic inverse-analysis is an essential approach to infer statistical characteristics of uncertain soil parameters, making the slope reliability assessment closer to engineering reality. However, current probabilistic inverse analysis rarely integrates multi-source information, including monitored data, field observation information, and slope survival records. Conducting the probabilistic inverse-analysis of spatially varying soil parameters and slope reliability prediction under rainfalls by integrating the multi-source information is a challenging issue due to the involvement of thousands of random variables and the evaluation of high-dimensional likelihood functions. In this paper, a modified Bayesian updating with subset simulation (mBUS) method is combined with adaptive conditional sampling (aCS) algorithm to establish a framework for probabilistic inverse analysis of spatially variable soil parameters and reliability prediction of slopes. The effectiveness of this framework is validated using a highway slope as a case study. The research results show that the posterior statistical characteristics of soil parameters obtained by integrating multi-source information are in good agreement with field observation results. Additionally, the probability of slope failure under heavy rainfall on September 12, 2004 with the updated soil parameters is 23.1 %, which is in line with the actual slope instability. Within this framework, multi-source information can be fully utilized to address high-dimensional probabilistic inverse analysis problems.

Automatic and timely identification of mud pumping is important for the reliability and safety of railroads. The current mud pumping prediction model is based on monitoring the dynamic response of railway tracks. The essential geotechnical trigger factors such as the hydrological conditions are not well-considered in these prediction models, as that information is hard to be quantified, which unavoidably reduces the accuracy of the prediction. This paper proposes to utilize the Geographic Information System (GIS) to quantify the hydrological information along railway tracks. Through GIS analysis, the hydrological variables including elevation, near-river distance, rainfall, sink depth, and soil types are estimated and combined with in-service train monitoring data for model development. To deal with multi-attribute data, a dual-channel neural networks model is proposed to separately mine the characteristics in different attributes data for prediction. To further understand the prediction model, Shapely addictive explanations (SHAP) method is applied to estimate the importance of hydrological variables and reveal the possible relationships between the variables and the probability of mud pumping. The proposed approach is applied to a real-life case from the railway tracks in Australia to validate its effectiveness. The prediction results show that the proposed approach can predict mud pumping with balanced accuracy of 90.84%. The results confirm that integrating of GIS information and monitoring data can generate more accurate prediction and reduce the false prediction rate. Based on the explainable results, it is observed that rainfall is the most important hydrological variable that influences mud pumping occurrence. Apart from rainfall, groundwater-related variables show a greater impact on mud pumping occurrence than surface water-related variables. The explainable results also can help infrastructure managers to identify the most vulnerable sections in railway tracks, which facilitates targeted maintenance planning and track substructure design.

Geotechnical subsurface stratification based on sparse measurements presents a significant challenge. Learning from prior geological information, such as learning soil layer distribution patterns from stratification results (2D images) of adjacent data-rich projects, is an emerging approach to reducing uncertainties caused by data scarcity. Existing methods rely on pixel-based techniques that require training images to have identical soil types as the testing image. Additionally, pixel-based methods are prone to error accumulation. To address these issues, this study introduces a new framework that focuses solely on learning boundary information from training image rather than soil types. This eliminates the need for matching soil types between training and testing images. Tree-based model is proposed to learn boundary information from training images. Contrary to conventional tree-based models that use coordinates as input, this study employs a set of designed distance fields (boundary dictionary) to represent complex boundary patterns. A selection process is introduced to identify the most important distance fields from the training images. Using these selected distance fields for model input results in soil stratification that aligns well with both borehole data and training image boundaries. The proposed method's efficacy is validated through multiple simulated and real-world cases. The proposed method outperforms pixel-based methods in multiple cases, achieving up to a 25% improvement in accuracy. This method is robust and it yields consistent results for a variety of training images considered. Additionally, the proposed method also provides quantification of interpolation uncertainty through the Gini impurity method.

Due to the similarity of conditioning factors, the aggregation feature of landslides and the multi-temporal landslide inventory, the spatial and temporal effects of landslides need to be considered in landslide susceptibility prediction (LSP). The ignorance of this issue will result in some biases and time-invariance in landslide susceptibility. Hence, a novel framework has been proposed to update landslide susceptibility by simultaneously considering the spatial and temporal effects of landslides at the regional scale. In this framework, the landslide inventory of Chongyi County has been divided into pre- and fresh-landslide inventories. According to the LSP results predicted by the support vector machine (SVM) model using the slope unit-based conditioning factors and pre-landslide inventory, a normalized spatial distance index (NSDI) is calculated to quantitatively represent the spatial correlation between landslides and surrounding slope units to develop the SVM-NSDI model. Furthermore, the SVM-Updating model, which incorporates the LSP results of the SVM-NSDI model and fresh-landslide inventory, could be developed to update the LSP results. Subsequently, the confusion matrix, the area under the receiver operating characteristic curve (AUC) and frequency ratio (FR) accuracy are used to evaluate the prediction performance of the above LSP models. The F1-score values of the SVM, SVM-NSDI and SVM-Updating models are 0.776, 0.816 and 0.831, respectively. The AUC values are 0.869, 0.903 and 0.914 and the FR accuracies are 0.795, 0.853 and 0.873. It can be concluded that landslide susceptibility is a time-variant variable, which can be updated by considering the spatial correlation between landslides and surrounding slope units as well as the temporal effects of multi-temporal landslide inventory. This study provides a new framework to update landslide susceptibility over time and also provides more accurate LSP results for decision-makers.

Failure mechanism and reliability analysis of rainfall-induced slopes generally ignore the effects of field observation information, such as the observation that the slope keeps stable in natural conditions or after a historical rainfall event. In this paper, with an infinite slope model as an example, the BUS (Bayesian Updating with Subset simulation) method is adopted for the probabilistic back analysis of spatially variable hydraulic and shear strength parameters based on the field observation that the slope survived from a previous extreme rainfall event. The probabilities of slope failure under different rainfall durations are evaluated within the framework of Monte-Carlo simulation. The influence of ignoring/incorporating the field observation on the estimate of probability of slope failure is also investigated. The results indicate that the possibility of slope failing along the weak zones caused by the spatial variability of soil parameters can be effectively excluded through the probabilistic back analysis incorporating the field observation. Based on this, more realistic probability of slope failure induced by the rainfall can be produced. If the field observation that the slope survived from a previous extreme rainfall event is ignored, the probability of slope failure will be significantly overestimated, especially in the early stage of rainfall. The research outcomes provide a new perspective for interpreting the rainfall-induced slope failure mechanisms in the spatially variable soils.

Most literature related to landslide susceptibility prediction only considers a single type of landslide, such as colluvial landslide, rock fall or debris flow, rather than different landslide types, which greatly affects susceptibility prediction performance. To construct efficient susceptibility prediction considering different landslide types, Huichang County in China is taken as example. Firstly, 105 rock falls, 350 colluvial landslides and 11 related environmental factors are identified. Then four machine learning models, namely logistic regression, multi-layer perception, support vector machine and C5.0 decision tree are applied for susceptibility modeling of rock fall and colluvial landslide. Thirdly, three different landslide susceptibility prediction (LSP) models considering landslide types based on C5.0 decision tree with excellent performance are constructed to generate final landslide susceptibility: (i) united method, which combines all landslide types directly; (ii) probability statistical method, which couples analyses of susceptibility indices under different landslide types based on probability formula; and (iii) maximum comparison method, which selects the maximum susceptibility index through comparing the predicted susceptibility indices under different types of landslides. Finally, uncertainties of landslide susceptibility are assessed by prediction accuracy, mean value and standard deviation. It is concluded that LSP results of the three coupled models considering landslide types basically conform to the spatial occurrence patterns of landslides in Huichang County. The united method has the best susceptibility prediction performance, followed by the probability method and maximum susceptibility method. More cases are needed to verify this result in-depth. LSP considering different landslide types is superior to that taking only a single type of landslide into account.

The numerical simulation and slope stability prediction are the focus of slope disaster research. Recently, machine learning models are commonly used in the slope stability prediction. However, these machine learning models have some problems, such as poor nonlinear performance, local optimum and incomplete factors feature extraction. These issues can affect the accuracy of slope stability prediction. Therefore, a deep learning algorithm called Long short-term memory (LSTM) has been innovatively proposed to predict slope stability. Taking the Ganzhou City in China as the study area, the landslide inventory and their characteristics of geotechnical parameters, slope height and slope angle are analyzed. Based on these characteristics, typical soil slopes are constructed using the Geo-Studio software. Five control factors affecting slope stability, including slope height, slope angle, internal friction angle, cohesion and volumetric weight, are selected to form different slope and construct model input variables. Then, the limit equilibrium method is used to calculate the stability coefficients of these typical soil slopes under different control factors. Each slope stability coefficient and its corresponding control factors is a slope sample. As a result, a total of 2160 training samples and 450 testing samples are constructed. These sample sets are imported into LSTM for modelling and compared with the support vector machine (SVM), random forest (RF) and convolutional neural network (CNN). The results show that the LSTM overcomes the problem that the commonly used machine learning models have difficulty extracting global features. Furthermore, LSTM has a better prediction performance for slope stability compared to SVM, RF and CNN models.

Landslide susceptibility assessment can help us to effectively predict the spatial location of potential landslides, which is the basis of landslide hazard and risk assessment. Slope units are commonly employed to predict landslide susceptibility because they are extracted based on actual landforms and geomorphology with visible geological features. However, one of the key constraints limiting the applicability of slope units and the challenge in current research is how to efficiently and accurately extract slope units and take into account the heterogeneity of conditioning factors within slope units. The Chongyi County was selected as the case study. First, the aspect and shaded relief images were extracted as the initial fundamental data. The multi-scale segmentation (MSS) method was used to extract slope units and the optimal parameter combination including scale, shape weight and compactness weight was determined by combining the trial-and-error method with recorded landslide features. Then, a total of 15 conditioning factors such as elevation, slope and profile curvature were extracted based on slope units and were imported into the support vector machine (SVM) and logistic regression (LR) models to construct Slope¿SVM/LR models. Furthermore, the range and standard deviation values were used to represent the heterogeneity of conditioning factors within slope units to construct the Variant Slope¿SVM/LR models. Finally, the receiver operating characteristic (ROC) curves and frequency ratio (FR) accuracy were used to evaluate the predicted performance of landslide susceptibility models. The results show that: 1) when the parameters of scale, shape weight and compactness weight were set to 20, 0.8 and 0.8, respectively, slope units extracted by the MSS method in the study area were at their best. 2) The ROC accuracy of the Slope¿SVM, Variant slope¿SVM, Slope¿LR and Variant slope¿LR models was 0.812, 0.876, 0.818 and 0.839, respectively. The FR accuracy of those models was 0.780, 0.866, 0.792 and 0.865, respectively, indicating that the predicted accuracy of Variant slope¿SVM/LR models was better than that of Slope¿SVM/LR models. Therefore, it can be inferred that the MSS method is an effective method to accurately and automatically extract slope units, and the predicted performance of landslide susceptibility models can be significantly improved by considering the heterogeneity of conditioning factors within slope units.

In this study, an implicit finite difference program, 1DFDM, based on the fully coupled one dimensional elasto-viscoplastic (1D EVP) model combining radial and vertical drainage is developed to perform consolidation analysis of embankments constructed on soft soils. Class A prediction (prior prediction) is performed by using 1DFDM for the embankment constructed on soft clay in Ballina, Australia. However, the prediction deviates significantly from the measurements, and the model parameters need to be calibrated. Bayesian back analysis is thus combined with 1DFDM in this study to do Class C prediction. Three updating schemes (using monitoring settlement only, using monitoring pore water pressure only, and using both monitoring settlement and pore water pressure) are employed to shed light on the effects of the monitoring data types on settlement prediction. The results show that the long-term settlement and pore water pressure predictions could be further improved if both the monitoring settlement and pore water pressure data are incorporated. The coupling effects in the constitutive model can reduce the fluctuations in the updated optimum parameters. A stable and physically reasonable parameter set could be obtained if enough monitoring data is incorporated into the Bayesian back analysis.

Constructing three-dimensional discrete fracture network(DFN) is a commonly-used method to simulate the random distribution of discontinuities in the rock masses for jointed rock slopes. To overcome the shortcomings of traditional Baecher disc and polygon models in constructing three-dimensional DFN model of a slope,in this paper,a method for generating an universal elliptical disc(UED) model-based DFN in 3DEC software is proposed. Based on this,the stability of a jointed rock slope can be accurately and efficiently evaluated. At the same time,the basic principles of the UED model are briefly introduced,including planarization of polygon fractures,optimal elliptic disc model fitting and generation of spatial elliptical disc model,etc. The procedure of constructing the DFN model of jointed rock slopes based on the UED model in 3DEC software is presented. Finally,taking an open pit mine slope as the example,the influences of characteristic parameters(ratio of long axis to short axis and rotation angle) of elliptic disc model for discontinuity groups on slope stability are investigated,and the differences between the UED model and traditional Baecher disc model are compared. The results show that the proposed method can avoid tedious programming using fish language and inefficient internal modeling of 3DEC software,and quickly construct the three-dimensional DFN based on the UED model and conduct slope stability analysis in 3DEC software. It thereby provides an important technical means for efficiently evaluating the stability of three-dimensional complex jointed rock slopes. Additionally,the influences of the characteristic parameters of UED model on the slope stability are addressed. It is confirmed that it is of great necessity to construct the DFN of slope by using the UED model in engineering practice.

Reliable shear strength determination of large in situ discontinuities is still a challenge faced by the rock mechanics field. This is principally due to the limited availability of surface roughness and morphology information of in situ discontinuities and the unresolved management of the ¿scale effect¿ phenomenon. Recently, a stochastic approach for predicting the shear strength of large-scale discontinuities was established, encompassing random field theory, a semi-analytical shear strength model, and a stochastic analysis framework. A key aspect of the new approach is the application at field scale, thereby minimising or bypassing the scale effect. The approach has been validated at laboratory scale and an initial large-scale deterministic-based validation showed promising results. However, to date, no large-scale experimental-based validation has been undertaken. This paper presents the first rigorous application of the employed semi-analytical shear strength model and the stochastic approach on a 2¿m-by-2¿m discontinuity surface, with comparison of prediction to experimental shear strength data. The shear strength model was found to generally produce peak and residual predictions within a ± 10% relative error range, with good agreement between predicted and observed damage areas. It was observed that, applying the stochastic approach to seed traces with gradient statistics equivalent to that of the surface, produced predictions that closely resemble the experimental results. Whereas, predicting shear strength from different seed traces results in more variability of predictions, with many falling within ± 20% of the experimental data. The predictions of residual shear strength tended to be more accurate than peak shear strength.

In the Load and Resistance Factor Design (LRFD) of piles, several design codes recommend higher resistance factors if load tests are conducted. However, no information is provided on how these resistance factors are determined. In this paper, a probabilistic approach based on Bayes' theorem and the First Order Reliability Method (FORM) is proposed to calibrate resistance factors for different numbers of load tests and the corresponding test results. In addition, within-site variability, design methods, types of piles, and ground conditions can also be considered. The proposed approach applied to H-piles under axial load tests shows consistent results with current design codes. Results show that resistance factors are significantly increased even if only one positive test is observed among all the tests. For low variability sites, the differences of resistance factors between various design methods are significantly reduced if one or more tests are positive, while for high variability sites, the differences of resistance factors are only slightly decreased, indicating that design methods should be considered in the latter case. Most of the increase in resistance factors is achieved with a small number of tests. For ß-Method used in clay sites, 80% of the increase in resistance factors is achieved with two, four and five consecutive positive tests are observed for low, medium and high variability sites, respectively.

The selection of non-landslide samples has a great impact on the machine learning modelling for landslide susceptibility prediction (LSP). This study presents a novel framework for studying the uncertainty of non-landslide samples selection on the LSP results through the slope unit-based machine learning models. In this framework, the non-landslide samples are randomly selected from the non-landslide areas by multiple times (N = 1, 10, 100, 500, 1000, 5000) to construct LSP models and calculate N types of landslide susceptibility indexes (LSIs). Afterwards, the statistical analysis is used to represent the uncertainty of LSIs under each non-landslide selection. The maximum probability analysis (MPA) is applied to reduce the uncertainty of non-landslide samples selection in LSP, which calculates the probability of N types of LSIs falling into very high, high, moderate, low and very low landslide susceptibility levels and selects the optimal landslide susceptibility level with the highest probability for each slope unit. Chongyi County in China is selected as the example, slope unit-based logistic regression (LR) and support vector machine (SVM) models are constructed with 16 conditioning factors. The area under the receiver operating features curve (AUC) and frequency ratio (FR) accuracy are used to evaluate the LSP performance. Results show that the N types of LSIs in each slope unit exhibit a normal distribution rather than one constant value. The uncertainties of LSIs caused by non-landslide samples selection are well represented by statistical analysis. The AUC values of slope unit-based LR/SVM models range from 0.714/ 0.711 (N = 1) to 0.787/0.775 (N = 5000) and increase to 0.867/0.848, meanwhile, the FR accuracies range from 0.772/ 0.763 (N = 1) to 0.815/0.826 (N = 5000) and increase to 0.843/0.861 by the MPA method. It is concluded that some more scientific and accurate landslide susceptibility results are obtained by selecting non-landslide samples multiple times and using the MPA method.

Characterization of the spatial variability of rock mass parameters differs significantly from that of soil parameters due to complex structures inherently existing in the rock masses. At present, the spatial variability modeling of the mechanical parameters of highly fractured/weathered rock masses and those dominated by a single structural plane and associated slope reliability analysis have been extensively studied. However, few attempts have been made to depict the spatial variability of the mechanical parameters of wedge and anti-dip rock masses and conduct associated slope reliability analysis. In this paper, a method for spatial variability modeling of the mechanical parameters of structural planes and slope reliability analysis is proposed for the wedge and anti-dip rock masses. The varying ranges of the autocorrelation distances and scales of fluctuation of the mechanical parameters for different types of rock masses are systematically summarized. An interface program between the spatial variability modeling of the mechanical parameters and probabilistic slope stability analysis with FLAC3D software is developed. An explicit function relationship between the factor of safety of the slope and the random field vectors of structural plane parameters is constructed using the Back-Propagation(BP) neural network. Then the non-intrusive stochastic finite difference method is adopted to explore the influence of the spatial variability of the mechanical parameters of structural planes on the probability of slope failure. The results indicate that the autocorrelation distances and scales of fluctuation of the mechanical parameters are different among different types of rock masses. The single exponential autocorrelation function is the most widely used theoretical autocorrelation function. The direction simulation and direct Monte-Carlo simulation methods are adopted to validate the effectiveness of the proposed method. The proposed method can provide an effective means for the spatial variability characterization of the mechanical parameters of structural plans for the wedge and anti-dip rock masses in the slope reliability analysis. In addition, the probability of slope failure will be overestimated if the spatial variability of mechanical parameters is ignored, which can further result in conservative slope reinforcement design schemes.

In some studies on landslide susceptibility mapping (LSM), landslide boundary and spatial shape characteristics have been expressed in the form of points or circles in the landslide inventory instead of the accurate polygon form. Different expressions of landslide boundaries and spatial shapes may lead to substantial differences in the distribution of predicted landslide susceptibility indexes (LSIs); moreover, the presence of irregular landslide boundaries and spatial shapes introduces uncertainties into the LSM. To address this issue by accurately drawing polygonal boundaries based on LSM, the uncertainty patterns of LSM modelling under two different landslide boundaries and spatial shapes, such as landslide points and circles, are compared. Within the research area of Ruijin City in China, a total of 370 landslides with accurate boundary information are obtained, and 10 environmental factors, such as slope and lithology, are selected. Then, correlation analyses between the landslide boundary shapes and selected environmental factors are performed via the frequency ratio (FR) method. Next, a support vector machine (SVM) and random forest (RF) based on landslide points, circles and accurate landslide polygons are constructed as point-, circle- and polygon-based SVM and RF models, respectively, to address LSM. Finally, the prediction capabilities of the above models are compared by computing their statistical accuracy using receiver operating characteristic analysis, and the uncertainties of the predicted LSIs under the above models are discussed. The results show that using polygonal surfaces with a higher reliability and accuracy to express the landslide boundary and spatial shape can provide a markedly improved LSM accuracy, compared to those based on the points and circles. Moreover, a higher degree of uncertainty of LSM modelling is present in the expression of points because there are too few grid units acting as model input variables. Additionally, the expression of the landslide boundary as circles introduces errors in measurement and is not as accurate as the polygonal boundary in most LSM modelling cases. In addition, the results under different conditions show that the polygon-based models have a higher LSM accuracy, with lower mean values and larger standard deviations compared with the point- and circle-based models. Finally, the overall LSM accuracy of the RF is superior to that of the SVM, and similar patterns of landslide boundary and spatial shape affecting the LSM modelling are reflected in the SVM and RF models.

Geotechnical and geophysical testing data are conventionally considered as separated information or combined based on deterministic methods in site investigation programs, which causes loss of information and introduces additional uncertainties. This study aims to reduce the uncertainties and costs in inhomogeneous soil profile characterization and geotechnical analysis by quantitatively integrating these data. The intrinsic collocated co-kriging method (ICCK) is utilized to integrate Multi-channel analysis surface wave (MASW) and CPT data. This research shows the potential of using the MASW data to explore the horizontal scale of fluctuation (SOF) of the cone tip resistance (qc) field which is very difficult to get using the conventional methods due to the limited CPTs. The Markov model is used to avoid the tedious modeling of the cross-covariance relationship in the ICCK method. A series of synthetic case studies show that the combined soil profile is in good agreement with the ¿true¿ qc field with significantly reduced uncertainties. Based on estimating the uncertainties, the optimal distance between CPTs is suggested to be 1¿2 horizontal SOF. This framework is also applied to a real case in the Christchurch area, which involves integration of CPTs and MASW tests. Cross validation and edge detection methods are used to quantitatively compare the integration results, confirming that the proposed framework is fully applicable to field data and can provide realistic and reliable estimations of qc field.

This paper compares the performance of five popular machine learning methods, namely, particle swarm optimization¿extreme learning machine (PSO¿ELM), particle swarm optimization¿kernel extreme learning machine (PSO¿KELM), particle swarm optimization¿support vector machine (PSO¿SVM), particle swarm optimization¿least squares support vector machine (PSO¿LSSVM), and long short-term memory neural network (LSTM), in the prediction of reservoir landslide displacement. The Baishuihe, Shuping, and Baijiabao landslides in the Three Gorges reservoir area of China were used for case studies. Cumulative displacement was decomposed into trend displacement and periodic displacement by the Hodrick¿Prescott filter. The double exponential smoothing method and the five machine learning methods were used to predict the trend and periodic displacement, respectively. The five machine learning methods are compared in three aspects: highest single prediction accuracy, mean prediction accuracy, and prediction stability. The results show that no method performed the best for all three aspects in the three landslide cases. LSTM and PSO¿ELM achieved better single prediction accuracy, but worse mean prediction accuracy and stability. PSO¿KELM, PSO¿LSSVM, and PSO¿SVM always yielded consistent predictions with slight variations. On the whole, PSO¿KELM and PSO¿LSSVM are recommended for their superior mean prediction accuracy and prediction stability.

For accurate characterization of the statistical properties of orientations(i.e., dip angle and dip direction) of a rock mass discontinuity, a Copula method is proposed to model the discontinuity orientations accounting for the cross-correlation between the dip angle and the dip direction. Based on the optimal fittings of the marginal probability distributions and Copula functions(i.e., correlation structures) of the discontinuity orientations from the measurement data, a two-dimensional joint probability density function of the dip angle and the dip direction can be constructed. In the meantime, a visual comparison between the results obtained from the proposed method with those obtained from the traditional Fisher distribution and bivariate empirical distribution methods is conducted. The measured and simulated discontinuity orientations are compared on stereographic projection maps by using the stereographic projection method. Finally, four examples are investigated to illustrate the effectiveness of the proposed method. The results indicate that the traditional Fisher distribution and the bivariate empirical distribution methods cannot effectively characterize the cross-correlation between the discontinuity orientations, while the proposed method can be in a more flexible way to construct the joint probability density function of the discontinuity orientations that follow arbitrary marginal distributions and correlation structures based on a small amount of measurement data. In short, the proposed approach can better depict the cross-correlation between the dip angle and the dip direction, and circumvent the limitations of treating the dip angle and the dip direction as two independent variables in constructing the probability distribution models of the discontinuity orientations.

Heavy rainfall and engineering slope cutting are two key factors that trigger unstable red clay landslides with red clay soil as the sliding mass in the mountainous and hilly areas of South China. It is important to study the influence of engineering slope cuttings on changes in slope stability under heavy rainfall conditions. First, by summarising the main evolution and failure characteristics of landslides in Ganzhou City, Jiangxi Province, China, a general landslide physical model of red clay landslides with universal significance is constructed. Then, the rainfall characteristics of Ganzhou City are analyzed, and heavy rainfall occurring once in a period of 50¿years is applied to the general landslide physical model. Concurrently, the influences of different engineering slope cutting distances and angles on the changes in slope stability are explored. Finally, saturated and unsaturated infiltration theory and nonlinear finite element analysis are used to calculate the stability and pore water pressure changes in the landslides under the above-described conditions of heavy rainfall and engineering slope cutting. The results show that: (1) When there is no rainfall, the stability coefficient of the red clay slope rapidly decreases with increasing distance and/or angle of slope cutting; for a certain slope cutting angle, the stability coefficients of the landslide show a convex upward decrease with increasing slope cutting distance; for a certain slope cutting distance, the stability coefficients show a linear decrease with a gradually increasing slope cutting angle. (2) Under 5¿days of heavy rainfall reaching 210¿mm, the engineering slope cutting forms have increasing influence on stability reduction in a red clay slope. For a certain slope cutting distance, as the slope cutting angle increases, the slope stability coefficient shows a slow decrease. For a certain slope cutting angle, a greater slope cutting distance means a faster decrease in the slope stability coefficient. (3) The pore water pressure along the potential sliding surface of the red clay slope under heavy rainfall gradually increases, and there is a good inverse correspondence between the changes in the pore water pressure and the stability coefficient.

For linear conditioning factors such as rivers, roads, and geological faults, existing studies mainly use buffer analysis in Geographic Information System to obtain discrete variables such as distance to rivers and roads. These discrete variables have random fluctuations and are sensitive to the errors of point or line elements, leading to a decrease of landslide susceptibility prediction (LSP) accuracy. This study proposes continuous conditioning factors such as river density and road density to improve the suitability of the linear factors. Xunwu County in China is taken as an example; 337 historical landslides and 12 conditioning factors are acquired. First, the distance to rivers and roads and other 10 conditioning factors together constitute the original factors of LSP. Second, the distance to rivers and distance to roads are replaced by the road density and river density, respectively, to constitute the improved factors. Third, based on the support vector machine (SVM), logistic regression (LR), and random forest (RF), original factor- and improved factor-based SVM, LR, and RF models are constructed for comparisons. Finally, the LSP uncertainty is evaluated. Results show that (1) the improved factor-based models have higher LSP accuracies than original factor-based models, indicating that density factors are more feasible than linear factors with more explicit physical meaning; (2) landslide susceptibility indexes distribution features indicate that improved factor-based models reduce the uncertainty of LSP; (3) spatial density factors do not reduce the importance of conditioning factors in both original factor- and improved factor-based models.

This paper studies the potential of using the successive over-relaxation iteration method with polynomial preconditioner (P(m)-SOR) to solve variably saturated flow problems described by the linearized Richards¿ equation. The finite difference method is employed to numerically discretize and produce a system of linear equations. Generally, the traditional Picard method needs to re-evaluate the iterative matrix in each iteration, so it is time-consuming. And under unfavorable conditions such as infiltration into extremely dry soil, the Picard method suffers from numerical non-convergence. For linear iterative methods, the traditional Gauss-Seidel iteration method (GS) has a slow convergence rate, and it is difficult to determine the optimum value of the relaxation factor w in the successive over-relaxation iteration method (SOR). Thus, the approximate optimum value of w is obtained based on the minimum spectral radius of the iterative matrix, and the P(m)-SOR method is extended to model underground water flow in unsaturated soils. The improved method is verified using three test examples. Compared with conventional Picard iteration, GS and SOR methods, numerical results demonstrate that the P(m)-SOR has faster convergence rate, less computation cost, and good error stability. Besides, the results reveal that the convergence rate of the P(m)-SOR method is positively correlated with the parameter m. This method can serve as a reference for numerical simulation of unsaturated flow.

It is significant to improve the warning accuracy and spatial identification of rainfall-induced landslides. This study takes 156 typical rainfall-induced landslide events from 1980 to 2001 in Ningdu County Jiangxi Province, China as a case. Firstly, the time probability levels of different rainfall-induced landslides are calculated based on traditional EE-D (early effective rainfallrainfall duration) threshold. Then taking each time probability corresponding to each level critical rainfall threshold curve as dependent variable, and its early effective rainfall (early effective rainfall, EE) and rainfall duration (D) as independent variables, logistic regression is adopted to fitting nonlinear mapping relationship between probability of rainfall-induced landslides and EE and D to obtain continuous probability of landslides. Furthermore, prediction performance of landslide susceptibility between C5.0 decision tree and multilayer perceptron is compared. Finally, continuous probability of rainfall-induced landslides is coupled with landslide susceptibility to realize continuous landslide hazard warning. Results show follows: (1) logistic regression fitting equation of continuous probability rainfall-induced landslides is 1/P=1+e4.062+0.747 4×D-0.079 44×EE with R2 of 0.983. (2) most of 20 rainfall-induced landslides from 2002 to 2003 used for continuous probability critical rainfall threshold test fell in areas with continuous probability greater than 0.7, and only 4 of them fell in areas less than 0.7. (3) the C5.0 DT model has a better prediction performance than the multilayer perceptron. (4) the continuous probability hazard values of four rainfall-type landslides in the past five years are above 0.8, and the areas of high and very high warning zone are smaller than those of traditional landslide hazard warning. It is concluded that compared with the traditional hazard zoning method, the continuous landslide hazard warning method has higher warning accuracy and spatial identification, and the real time landslide hazard map carrying out spatial and time warning can be obtained through combination of landslide critical rainfall threshold and landslide susceptibility map.

Rainfall-induced landslide hazard warning, which refers to the prediction of the spatial-temporal probability of landslide occurrence in a certain area under the conditions of continuous rainfall processes, can be established based on landslide susceptibility mapping and critical rainfall threshold calculations. However, it is difficult to determine appropriate machine learning models for mapping landslide susceptibility. Additionally, it is significant to consider the influences of early effective rainfall on landslide instability in the critical rainfall threshold methods. Furthermore, the uncertainties of the critical rainfall threshold values generated by different calculation methods have not been well explored. To overcome these three drawbacks, first, frequency ratio analysis-based logistic regression (LR), support vector machine (SVM) and random forest (RF) models are adopted to predict landslide susceptibility for machine learning model comparison. Second, three different types of critical rainfall threshold methods, namely, cumulative effective rainfall-duration (EE-D), effective rainfall intensity-duration (EI-D) and cumulative effective rainfall-effective rainfall intensity (EE-EI) models, are proposed to calculate the temporal probabilities of landslide occurrence under rainfall conditions based on the concept of effective rainfall. The accuracies and uncertainties of these three critical rainfall threshold methods are discussed. Finally, the landslide susceptibility maps and the critical rainfall threshold values are coupled to predict the rainfall-induced landslide hazards. Xunwu County in China is selected as the study area, and several rainfall-induced landslides are used as the test samples of the proposed landslide hazard warning model. The results show that the RF model has remarkably higher susceptibility prediction accuracy than the SVM and LR models, and the prediction performance of the temporal probabilities of landslide occurrence using the EI-D values are higher than those of EE-D and EE-EI values. Furthermore, rainfall-induced landslide hazard warning is effectively implemented based on the coupling of the susceptibility map and EI-D model.

Spatial variability of soil properties was rarely taken into account directly in traditional slope stability analyses, rather some ¿average¿ or suitably ¿pessimistic¿ properties are assumed to act across the whole region of interest. In the last two decades, a large portion of published research papers on slope stability have tried to explicitly model the spatial variability of soil properties. In the first decade, research mainly focused on showing the importance of modeling the spatial variability directly in probabilistic slope stability analysis. In the last decade, a rapid development was observed including quantitative risk assessment of slope failure, improving computational efficiency, and directly using site investigation and field monitoring data. This review tries to summarize these advances in the hope that future research directions can be identified.

Existing studies relating to landslide susceptibility prediction (LSP) either do not pay enough attentions to the continuously updated landslide inventories or use batch learning methods for LSP, resulting in the insufficient use of the entire landslide inventory. To overcome this problem, the Incremental Learning theory combined with a Bayesian Network (ILBN) model is constructed for LSP. Wencheng County of China is taken as the study area, a landslide inventory from 1985 to 2019 and 10 conditioning factors are mapped and analyzed. Then, the LSP results of the ILBN model are compared with the batch learning-based multilayer perceptron (BL-MLP) and support vector machine (BL-SVM) models. Results show that the LSP accuracies of ILBN_0 (ILBN modeling of initial landslide inventory), ILBN_1 (the first Incremental Learning model), and ILBN_2 (the second Incremental Learning model) increase gradually with the AUC value of 0.807, 0.813, and 0.835, respectively. The LSM produced by the ILBN model is more consistent with the law of landslides distribution in the study area. The mean values of ILBN_0, ILBN_1, and ILBN_2 are 0.307, 0.287, and 0.245, and the standard deviations are 0.278, 0.281, and 0.308, respectively. Meanwhile, the characteristics of LSIs in Wencheng County are in line with the actual landslides distribution with the main controlling factors of lithology, elevation, and normalized difference building indexes determined by the weighted mean method. Furthermore, the LSP results of ILBN model are superior to those of the BL-MLP and BL-SVM models. It is concluded that the ILBN model can better address the long-term, continuous LSP using the new added landslide inventory.

Site investigation is an important step of geotechnical projects. Previous studies have investigated the benefits of undertaking site investigation for differing scopes by assuming the measurements obtained from site investigation tests are ¿true¿ measurements without measurement errors. However, measurement errors are inevitable in all types of site investigation testing methods which cannot be neglected. This paper attempts to quantify the effects of measurement errors on the optimal number of site investigation tests and optimal testing methods. Monte Carlo simulations, random field theory and Kriging fitting method are utilised to incorporate the uncertainties due to soil variability and measurement errors in the slope reliability analysis process. Results indicate that there are significant benefits from increasing the number of site investigation tests in terms of the risk reduction of slope design. By taking the cost of site investigation into account, the optimal number of site investigation tests can be identified by balancing the risk and the cost.

Timely detection and identification of rail breaks are crucial for safety and reliability of railway networks. This article proposes a new deep learning-based approach using the daily monitoring data from in-service trains. A time-series generative adversarial network (TimeGAN) is employed to mitigate the problem of data imbalance and preserve the temporal dynamics for generating synthetic rail breaks. A feature-level attention-based bidirectional recurrent neural network (AM-BRNN) is proposed to enhance feature extraction and capture two-direction dependencies in sequential data for accurate prediction. The proposed approach is implemented on a three-year dataset collected from a section of railroads (up to 350 km) in Australia. A real-life validation is carried out to evaluate the prediction performance of the proposed model, where historical data are used to train the model and future 'unseen' rail breaks along the whole track section are used for testing. The results show that the model can successfully predict nine out of 11 rail breaks three months ahead of time with a false prediction of nonbreak of 8.2%. Predicting rail breaks three months ahead of time will provide railroads enough time for maintenance planning. Given the prediction results, a Shapley additive explanations (SHAP) method is employed to perform a cause analysis for individual rail break. The results of cause analysis can assist railroads to plan appropriate maintenance to prevent rail breaks.

Bayesian estimation of spatially varying soil parameters is a challenging task in geotechnical engineering because a large number of random variables need to be learned. To address this challenge, three Bayesian methods are revisited, including Differential Evolution Adaptive Metropolis with sampling from past states [DREAM(zs)] method, Bayesian Updating with Structural reliability methods using Subset Simulation (BUS + SS), and modified BUS with Subset Simulation (mBUS + SS). The differences between the performances (i.e. convergences, computational accuracies, and efficiencies) of these three methods are not well understood. This study systematically investigates the differences of these three methods in the generation of random samples, convergence criterion, model evidence, and estimation of posterior probability of failure in slope reliability updating. Two slope examples are used for the comparative study. It is found that the BUS + SS method performs well not only in the low-dimensional Bayesian inverse problems but also in the high-dimensional Bayesian inverse problems of spatially varying soil parameters. The DREAM(zs) method is preferentially recommended to deal with the low-dimensional Bayesian inverse problems whereas the mBUS + SS method can well tackle the high-dimensional problems.

The rainfall-induced slope failure mechanism and reliability analyses rarely consider the spatial variability of hydraulic and shear strength parameters at the same time and ignore a fact that the slopes always keep stable under the natural condition. An infinite slope model is taken as an example to conduct probabilistic back analyses of spatially varying shear strength parameters using the observation information in advance. Then, a non-stationary random field model is established to simulate the spatial variability and non-stationary distribution feature of the hydraulic conductivity. The probabilities of slope failure and distributions of the critical slip surface under different rainfall durations are evaluated within the framework of Monte-Carlo simulation. Based on these, the rainfall-induced slope failure mechanisms considering the spatial variability of hydraulic and shear strength parameters simultaneously are investigated. The results indicate that the probability of slope failure evaluated based on the posterior information of shear strength parameters obtained from the probabilistic back analyses is reduced from 28.1% (prior) to 7.2%. It is found that the triggering factors for the slope instability are different for different rainfall stages. The rainfall-induced slope failure mechanism and probability of failure will be erroneously evaluated, especially at the initial stage of rainfall, if the field observation information is ignored.

The determination of mapping units, including grid, slope, unique condition, administrative division, and watershed units, is a very important modeling basis for landslide assessments. Among these mapping units, the slope unit has been paid a lot of attention because it can effectively reflect the physical relationships between landslides and the fundamental topographic elements especially in mountainous areas. Although some methods have been proposed for slope unit extraction, effectively and automatically extracting slope units remains a difficult and urgent problem that seriously restricts the use of slope units. To overcome this problem, the innovative multi-scale segmentation (MSS) method is proposed for extracting slope units. Thus, first, the terrain aspect and shaded relief images obtained from the digital elevation model with certain weights are used as the data sources of the MSS method. Second, the scale, shape, and compactness parameters of the MSS method are properly determined according to the improved trial-and-error method. Third, the initial slope units generated by the MSS method with appropriate parameters are automatically optimized through vector analysis in GIS. Finally, reasonable slope units are obtained and the extraction performance is discussed. The Chongyi County and Wanzhou District in China are selected as study areas. The conventional hydrological method is also adopted to extract slope units for qualitative and quantitative comparisons. It can be concluded that the MSS method can accurately and automatically extract the slope units for landslide assessments in hilly and mountainous areas and performs better than the hydrological method.

Timely detection and identification of substructure defects in railway track are crucial for the safety and reliability of railway networks. Instrumented in-service trains can provide daily data for assessing the track conditions. This study tries to develop a data-driven model for the prediction of mud pumping defects using daily in-service train data. The data-driven model is based on long short-term memory (LSTM) networks. Bayesian optimization method is used to select the optimal hyper-parameters in LSTM. Genetic algorithm (GA) method is used for feature selection. A four-year real-world dataset from a section of railway network in Australia is used to train and test the data-driven model. The t-distributed Stochastic Neighbor Embedding (t-SNE) method is used to examine the resulting LSTM networks. The results show that the proposed approach can be used to predict the mud pumping defects in advance leaving enough time for maintenance.

Due to the complexity and obvious random distribution features of joints and fractures in the rock slope, accurately simulating the random fracture network is crucial for slope stability evaluation. However, the current methods cannot effectively make full use of the measured data of structural outcrops. This paper adopts a Bayesian updating approach to optimize the probability distributions of geometric and shear strength parameters of structural planes and correct the random fracture network of rock mass by using field measured data. The field measured data of structural planes are characterized as sample distributions. Based on these, a non-intrusive stochastic finite element method is employed to conduct reliability analysis of jointed rock slopes considering the uncertainties of geometric parameters (e.g., dip and trace length) and shear strength parameters (e.g., friction angle and cohesion) of structural planes at the same time. Finally, a simplified slope model selected from the left bank of Xiaowan hydropower station is adopted to validate the effectiveness of the proposed method. The results indicate that the probability distributions of structural plane parameters inferred from the Bayesian updating approach agree well with the corresponding analytical solutions. The Bayesian updating approach can effectively reduce the estimation of uncertainties and optimize the probability distributions of the structural plane parameters by incorporating the field measured data. Furthermore, more realistic fracture network models and reliability analysis results of the slope can be obtained. When the posterior information of structural plane parameters is used to generate a fracture network model and conduct slope reliability analysis, the estimated posterior probability of slope failure will be greatly smaller than the prior probability of slope failure.

Soil-cement column is a geotechnical solution used for ground improvement in coastal areas. However, after long periods of exposure, the strength of these columns may decrease to below their designed safe bearing capacity, ultimately resulting in failure. In this paper, the effects of high sulphate concentrations (100%, 200%, 500% and 1000% that of seawater) on the durability of soil-cement samples were examined. In addition, the simple simulation model was applied to predict the deterioration depth and long-term strength of the soil-cement columns. The results show that the deterioration is more pronounced and occurs deeper in the presence of high sulphate concentrations. For instance, the strength of a 0.5 m diameter column exposed to 200% seawater will fall below the minimum design strength after 75 years. For higher sulphate environments (5 to 10 times that of normal seawater) the same column would never reach the minimum design strength requirement. Consequently, this has significant implications on soil-cement column when used to stabilise soils in high sulphate environments.

It is of great significance to explore the influences of the spatial resolution and the proportion of model training and testing dataset on the uncertainties of landslide susceptibility prediction(LSP). Taking the landslides in Ningdu County of Jiangxi Province as examples, the frequency ratios of various environmental factors under different spatial resolutions(15, 30, 60, 90 and 120 m) are firstly calculated. Then, the landslide and non-landslide samples are divided into different model training and testing datasets with the proportions of 9/1, 8/2, 7/3, 6/4 and 5/5, and the model input and output variables under 25 combined conditions are obtained. Furthermore, these input and output variables are imported into the Support Vector Machine(SVM) and Random Forest(RF) models to carry out LSP. Finally, the uncertainties of LSP modeling under the 25 combined conditions are discussed using the accuracy assessment as well as the distribution characteristics of landslide susceptibility indexes. The results show that the landslide susceptibility accuracy predicted by the RF model under the spatial resolution of 15 m and training and testing dataset proportion of 9: 1 is the highest, and that the more important environmental factors under each condition are elevation, slope and topographic relief, etc. With decreasing the spatial resolution and/or the proportion of training and testing dataset, the LSP accuracies of both SVM and RF models decrease gradually, and the mean values of landslide susceptibility indexes increase with a decrease of the corresponding standard deviation value. For all combined conditions, as the spatial resolutions and the proportions of training and testing dataset decrease, the LSP accuracies decrease gradually while the corresponding uncertainties increase. It is also indicated that the LSP accuracy of the RF model is better than that of the SVM model under various combined conditions, and that the influence of the spatial resolution on the RF model is significantly greater than that of the proportion of training and testing dataset while there is little difference between the effects of the two factors on the SVM model.

Machine learning algorithms are an important measure with which to perform landslide susceptibility assessments, but most studies use GIS-based classification methods to conduct susceptibility zonation. This study presents a machine learning approach based on the C5.0 decision tree (DT) model and the K-means cluster algorithm to produce a regional landslide susceptibility map. Yanchang County, a typical landslide-prone area located in northwestern China, was taken as the area of interest to introduce the proposed application procedure. A landslide inventory containing 82 landslides was prepared and subsequently randomly partitioned into two subsets: training data (70% landslide pixels) and validation data (30% landslide pixels). Fourteen landslide influencing factors were considered in the input dataset and were used to calculate the landslide occurrence probability based on the C5.0 decision tree model. Susceptibility zonation was implemented according to the cut-off values calculated by the K-means cluster algorithm. The validation results of the model performance analysis showed that the AUC (area under the receiver operating characteristic (ROC) curve) of the proposed model was the highest, reaching 0.88, compared with traditional models (support vector machine (SVM) = 0.85, Bayesian network (BN) = 0.81, frequency ratio (FR) = 0.75, weight of evidence (WOE) = 0.76). The landslide frequency ratio and frequency density of the high susceptibility zones were 6.76/km2 and 0.88/km2, respectively, which were much higher than those of the low susceptibility zones. The top 20% interval of landslide occurrence probability contained 89% of the historical landslides but only accounted for 10.3% of the total area. Our results indicate that the distribution of high susceptibility zones was more focused without containing more ¿stable¿ pixels. Therefore, the obtained susceptibility map is suitable for application to landslide risk management practices.

This paper aims to explore the influences of different attribute interval numbers (AINs) in the frequency ratio (FR) analysis of continuous environmental factors and the influences of different data-based models on the uncertainties of landslide susceptibility prediction (LSP). Taking Ningdu County of China as study area, 446 landslides and nine environmental factors are first acquired. Then the FR values of environmental factors under 6 different AINs (4, 6, 8, 12, 16 and 20) and 6 different data-based models (FR model, grey relational degree (GRD), logistic regression (LR), multilayer perceptron (MLP), C5.0 decision tree (C5.0 DT) and random forest (RF)) are set to 36 different conditions. Finally, the LSP results with uncertainties under all conditions are discussed. Results show that: 1) For a certain model, the LSP accuracy gradually increases with the AINs increasing from 4 to 8, and then the increase rate decreases until the accuracy is stable with the AINs increasing from 8 to 20; 2) For a certain AIN, the LSP accuracy of RF is higher than that of C5.0 DT, followed by the MLP, LR, FR and GRD; 3) The LSP accuracy is highest under an AIN of 20 and RF and is satisfied under an AIN of 8 and RF, while is the lowest under an AIN of 4 and GRD; 4) The landslide susceptibility indexes (LSIs) under AINs of 4, 6 and 12 are significantly different from the other AINs, and the LSIs calculated by the C5.0 DT and RF are significantly different compared to the other models; 5) The mean values and standard deviations of LSIs calculated by the MLP, C5.0 DT and RF models are relatively smaller and larger, respectively, than those of the other models, indicating that the LSIs calculated by these models are more consistent with the actual landslide distribution features.

The loss of effective stress due to increasing excess pore pressure that results in the upward migration of soil particles, that is, subgrade fluidization and mud pumping, has been a critical issue for railways over many years. Traditional methods such as experimental and analytical approaches can capture macroscopic quantities such as the hydraulic conductivity and critical hydraulic gradient, but they have many limitations when microscopic and localized behavior must be captured. This paper, therefore, presents a novel numerical approach where the microscopic properties of fluid and particles can be better captured when the soil is subjected to an increasing hydraulic gradient. While particle behavior is simulated using the discrete element method (DEM), the fluid dynamics can be described in greater detail using the lattice Boltzmann method (LBM). The mutual LBM-DEM interaction is carried out, so the particle and fluid variables are constantly updated. To validate this numerical method, laboratory testing on a selected subgrade soil is conducted. The results show that the numerical method can reasonably predict the coupled hydraulic and soil fluidization aspects, in relation to the experimental data. Microscopic properties such as the interstitial fluid flowing through the porous spaces of the soil are also captured well by the proposed fluid-particle coupling approach.

Estimating the shear strength of large in situ rock discontinuities is non-trivial because of the multiscale nature of roughness and the fact that only a very limited extent of discontinuity morphology is visible along traces. Recently, a novel stochastic method based on random field theory and Monte Carlo semi-analytical estimation of shear strength was proposed. The method was validated at laboratory scale and its application to one large natural surface showed that it has the potential to bypass the scale effect. However, a critical issue was reported by the authors of the study: it was found that the random field approach used could not generate the correct distribution of gradients on the simulated surfaces, which translates into an inaccurate prediction of shear strength. The authors had to manually adjust the input of the model to achieve a satisfactory prediction. This paper presents a new multiscale approach using random field theory, which now allows a rigorous generation of large synthetic 3D surfaces with controlled distribution of asperity heights and gradients from the profile of a 2D fracture trace (as might be visible in a rock face) referred to as a seed trace. Each seed trace is first decomposed into three daughter profiles corresponding to three levels of roughness. The statistics of each daughter profile form the input of the random field model at each scale level, allowing synthetic daughter surfaces to be created at each scale level. The synthetic daughter surfaces are then superimposed to obtain a composite rough surface comprising three levels of roughness. The approach was successfully validated with 25 input seed traces, coming from 5 different natural surfaces. This rigorous multiscale approach is essential to apply the stochastic method that was recently developed to predict the shear strength of large in situ discontinuities.

The landslide boundaries and their spatial shapes usually appear as irregular polygonal surfaces such as semi-circle and dustpan. However, literature review shows that inaccurate landslide point and buffer circle are commonly used as the landslide boundary for landslide susceptibility prediction(LSP), leading to some uncertainties in the LSP results. To study the effects of different landslide boundaries on the LSP modelling, 337 landslides and 10 types of environmental factors in Shangyou County of Jiangxi Province are taken as basic data. The correlations between environmental factors and landslides are built based on different landslide boundaries forms of point, buffer circle and accurate polygonal surface. Next, the multi-layer perceptron(MLP) and random forest(RF) are selected to build six kinds of LSP models, namely, Point, circle and polygon-based MLP and RF models. Finally, three methods are used to analyse the LSP uncertainties, including the ROC accuracy, difference significance analysis and distribution rules of landslide susceptibilityindexs. Results show that:(1) LSP uncertainties will increase under the landslide boundaries of landslide point or buffer circle, while the accuracy and reliability of LSP results will increase under the accurate landslide polygon boundaries. (2) The uncertainty rules of LSP obtained by MLP and RF models are consistent, however, the uncertainty of RF is lower than those of MLP.(3) The LSP results of point and buffer circle based models can also reflect the spatial distribution rules of landslide probability on the whole, and can be used as a substitute scheme in the absence of accurate landslide boundaries.

Piezocone penetration test (CPTU) and multichannel analysis of surface waves (MASW) are common methods for ground investigation. Integrating the data from the abovementioned two methods for accurate characterization of soil profiles is a critical and difficult issue. This paper proposes an approach for integrating the geotechnical and geophysical testing data based on Bayesian theory. The proposed approach is validated by a synthetic example, and is further applied to the Ballina site, Australia for integrating the site data and characterizing the profiles of soil properties. The results indicate that the updated random field after integrating the CPTU and MASW data provides an accurate and concise description of spatial distribution of the soil parameter throughout the two dimensional (2D) sub-region. When using the CPTU data only, the soil parameters at areas where CPTU is not performed can be inferred, and the local uncertainties of soil profiles near CPTU are reduced. When integrating the 1D CPTU and 2D MASW data, the characterized soil profile has good agreement with the "true" profile, and the uncertainties of soil profiles in the whole sub-region are further reduced. In the Ballina case, the uncertainties of 2D soil profiles are reduced when increasing the integrated site data, and the "weak" areas are effectively identified.

Stability analysis and design of soil slopes are a classic problem in geotechnical engineering. The current commonly-used deterministic analysis (e.g., single factor of safety) approach does not quantify the influences of various uncertainties in slope engineering, while the probabilistic analysis approach is time-consuming because it often requires performing multiple rounds of reliability analyses. A slope model reconstruction method that can well adapt to different slope angles is proposed. The reliability-based design of slope angles for spatially varying slopes based on a small amount of test data is carried out using the inverse first order reliability method. To validate the effectiveness of the proposed method, a representative sandy slope is taken as an example, to conduct the reliability-based design of slope angles. The results indicate that the proposed method can obtain a design scheme of slope angle based on a small amount of test data, which is well consistent with engineering practice. It thereby provides an effective tool for the reliability-based design of slope angles for spatially varying slopes. For the sandy slope in this study, an optimized slope angle that achieves various target probabilities of failure can be obtained after 4 or 5 iterative calculations. In contrast, the deterministic analysis method will obtain a biased design scheme since it cannot quantitatively account for the influences of multiple sources of uncertainties in the slope engineering. To yield a target probability of failure of 1×10-4 which is often acceptable for stability evaluation of slopes, the slope angle of the sandy slope designed using the proposed method should be smaller than 14.13°. By contrast, the slope angle designed using the deterministic analysis approach differs significantly from that designed using the proposed method.

Loess fill slopes are vulnerable to heavy rainfall because of water sensitivity and collapsibility of loess. Studies on the failure mode and mechanism of loess fill slopes are limited and incomplete. In this study, a laboratory flume test is carried out to simulate the failure mode of loess fill slope by monitoring and analyzing its soil hydrological and mechanical parameters including volumetric water content, pore water pressure and horizontal earth pressure. The results show that under continuous rainfall, loess fill slope fails in a backward retrogressive failure mode including gully erosion, partial failure, slope toe failure, central slope failure and top slope failure stages. The failure mechanism of each failure stage has been explained based on the variation of slope hydrological and mechanical conditions with the rainwater infiltration. The evolution processes of the fissures and its effects on the slope failure have been investigated. It is revealed that fissures play an important role in failing the slope by generating preferential flow. Finally, some engineering measures are recommended for the prevention of loess fill slope failure.

Development of large scale solar farms supported by large numbers of short piles has created new challenges for engineers to address. Solar arrays are highly flexible structures and the piles can be designed to move to enable more cost effective design. The structural reliability of the above-ground pile can be assessed and probabilities of failure for different section sizes calculated. Economic analysis incorporating capital cost and whole-of-life maintenance cost can be performed to work out whether adopting smaller section sizes provide the best cost outcome. Assessment of pile movements using Monte-Carlo calculations, unsaturated soil mechanics and updating material parameters with suction have been performed. The results show that soil movements are typically larger than pile movements and that soil can slip past the pile with no pile movement when the limiting conditions occur. The results also highlight that the largest soil and pile movements occur infrequently as a result of extreme wetting or drying conditions. Structural reliability analyses showed that correlating wind speed and direction results in a lower probability of failure than if wind load is considered to be uncorrelated with wind direction. The outcomes of the assessment were sensitive to the adopted probabilistic model for pile durability. The main limitation of the analyses is that there is limited information in the literature relating to the types of probability distributions and their input parameters. This adds uncertainty to the stochastic analysis.

Only the spatial variability of soil shear strength parameters is taken into account in most slope reliability analysis of embankment dams, whereas the spatial variation of unsaturated hydraulic model parameters is ignored.This paper develops a non-intrusive stochastic analysis method for evaluation of slope reliability of embankment dams, wherein the influence of the spatial variability of hydraulic model parameters on the slope reliability of embankment dam is incorporated.The phenomenon that the probability of slope failure decreases with increasing the coefficient of variation (COV) of hydraulic model parameter n is explained.The results indicate that the developed method has much higher computational efficiency in comparison with the direct Monte-Carlo simulation method.The probability of slope failure is positively correlated with the COVs of the cohesion, friction angle and saturated hydraulic conductivity.The variability of the cohesion affects the slope reliability the most, while that of the saturated hydraulic conductivity has the least effect on the slope reliability.In addition, the probability of slope failure almost keeps unchanged as the COV of hydraulic model parameter a varies.It is also interesting to note that the probability of slope failure decreases as the COV of hydraulic model parameter n increases.This is because more realization values of n that are close to the lower bound of 1.05 can be generated when a larger COV of n is used.As a result, a smaller hydraulic conductivity and a lower flow rate that passes through the dam body are induced.Consequently, the safer slope of embankment dam is achieved.

To obtain accurate values of the material parameters of tailings dams based on the limited data, a stochastic back analysis approach considering the uncertainties of the material parameters of tailings dams is proposed under the framework of Bayesian updating and finite element analysis. To improve the computational efficiency of back analysis, a polynomial chaos expansion is adopted to replace the implicit function between the displacements of tailing dams at the representative monitoring points and uncertain input parameters. A real tailings dam is taken as an example to demonstrate the effectiveness of the proposed approach for stochastic back analysis of parameters of multi-layered tailings materials based on the monitoring data of displacements. The results indicate that the proposed approach can effectively reduce the estimation in the uncertainties of the material parameters of tailings dams, accurately infer the probability distributions of the material parameters, and identify the influence degree of different material parameters (e.g., elastic modulus, Poisson's ratio) on the deformation of the tailings dams.

Accurate inference of the probability distributions of geotechnical parameters is a crucial step for reliability analysis and risk assessment in geotechnical engineering. At present, the probability distributions of geotechnical parameters are mainly inferred based on in-situ and/or laboratory test data. This paper aims to propose an adaptive Bayesian updating approach for the probability distribution inference of geotechnical parameters, in which a quantitative termination strategy for subset simulation is presented. Moreover, a framework for the inference and reliability analysis of the probability distributions of geotechnical parameters is constructed. The effectiveness of the proposed approach is verified by taking the landslide on No. 3 Freeway in Taiwan and a saturated clay slope as examples. Finally, the influence of the number of samples in each subset simulation level on the inference of probability distributions is addressed in this paper. The results indicate that, in comparison with the maximum likelihood and Markov chain Monte Carlo methods, the proposed approach is more efficient in calculation, simpler in programming, and can provide an effective way to solve the problem of probability distribution inference of geotechnical parameters at low acceptance probability levels. The number of random samples in each subset simulation level has certain influence on probability distribution inference. As the number of samples in each level increases, the posterior statistics of geotechnical parameters and threshold of subset simulation gradually converge. In addition, the rationality of the established quantitative termination strategy for subset simulation can be verified according to the variation of complementary cumulative distribution function with the subset simulation threshold.

Conventional planning of maintenance and renewal work for railway track is based on heuristics and simple scheduling. The railway industry is now collecting a large amount of data with the fast-paced development of sensor technologies. These data sets carry information about the conditions of various components in railway track. Since just before the beginning of the 21st century, data-driven models have been used in the predictive maintenance of railway track. This study presents a systematic literature review of data-driven models applied in the predictive maintenance of railway track. A taxonomy to classify the existing literature based on types of models and types of applications is provided. It is found that applying the deep learning methods, unsupervised methods, and ensemble methods are the new trends for predictive maintenance of railway track. Rail geometry irregularity, rail head defect, and missing rail components detection were the top three most commonly considered issues within the application of data-driven models. Prediction of rail breaks has received increasing attention in the last four years. Among these data-driven model applications, the collected data types are the most critical factors which affect selecting suitable models. Finally, this study discusses upcoming challenges in the predictive maintenance of railway track.

In order to ensure seepage analysis accuracy of tailings dam, deduce hydraulic conductivity probability distribution of tailings material and to reduce its uncertainty, sequential probabilistic back analysis method of material parameters based on Bayesian updating was proposed. Then, a surrogate model of water table and likelihood function were constructed. Finally, with Daheishan tailings dam taken as an example, sequential probabilistic back analysis of hydraulic conductivity of multi-layered tailings materials was conducted based on monitoring data of water tables. The results show that the proposed approach can effectively infer hydraulic conductivity and probability distributions as well as reduce their variation coefficients which is reduced by 18. 25% for soil layer closer to monitoring points. Realistic uncertainties of hydraulic conductivity and representation cannot be well deduced only from monitoring information of water levels, and it is necessary to further collect field information of multiple sources and incorporate it into probabilistic back analysis.

To study the uncertainties of a collapse susceptibility prediction (CSP) under the coupled conditions of different data-based models and different connection methods between collapses and environmental factors, An¿yuan County in China with 108 collapses is used as the study case, and 11 environmental factors are acquired by data analysis of Landsat TM 8 and high-resolution aerial images, using a hydrological and topographical spatial analysis of Digital Elevation Modeling in ArcGIS 10.2 software. Accordingly, 20 coupled conditions are proposed for CSP with five different connection methods (Probability Statistics (PSs), Frequency Ratio (FR), Information Value (IV), Index of Entropy (IOE) and Weight of Evidence (WOE)) and four data-based models (Analytic Hierarchy Process (AHP), Multiple Linear Regression (MLR), C5.0 Decision Tree (C5.0 DT) and Random Forest (RF)). Finally, the CSP uncertainties are assessed using the area under receiver operation curve (AUC), mean value, standard deviation and significance test, respectively. Results show that: (1) the WOE-based models have the highest AUC accuracy, lowest mean values and average rank, and a relatively large standard deviation; the mean values and average rank of all the FR-, IV-and IOE-based models are relatively large with low standard deviations; meanwhile, the AUC accuracies of FR-, IV-and IOE-based models are consistent but higher than those of the PS-based model. Hence, the WOE exhibits a greater spatial correlation performance than the other four methods. (2) Among all the data-based models, the RF model has the highest AUC accuracy, lowest mean value and mean rank, and a relatively large standard deviation. The CSP performance of the RF model is followed by the C5.0 DT, MLR and AHP models, respectively. (3) Under the coupled conditions, the WOE-RF model has the highest AUC accuracy, a relatively low mean value and average rank, and a high standard deviation. The PS-AHP model is opposite to the WOE-RF model. (4) In addition, the coupled models show slightly better CSP performances than those of the single data-based models not considering connect methods. The CSP performance of the other models falls somewhere in between. It is concluded that the WOE-RF is the most appropriate coupled condition for CSP than the other models.

The attribute interval numbers (AIN) in the frequency ratio analysis of continuous environmental factors and the landslide susceptibility model are two important uncertainties affecting the results of landslide susceptibility prediction (LSP). To study the effects of the two uncertain factors on the change rules of LSP, taking Shangyou County of Jiangxi Province, China, as study area, the AIN values of the continuous environmental factors are respectively set to be 4, 8, 12, 16 and 20. Meanwhile, five different data-based models (analytic hierarchy process (AHP), logistic regression (LR), BP neural network (BPNN), support vector machines (SVM) and random forests (RF)) are selected as LSP models. Hence, there are a total of 25 types of different calculation conditions for LSP. Finally, the accuracy and uncertainties of LSP are analyzed. The results show that: (1) For a certain model, the LSP accuracy gradually increases with the AIN value increasing from 4 to 8, then slowly increases to a stable level with AIN increasing from 8 to 20; (2) For a certain AIN, the LSP accuracy of the RF model is higher than SVM, followed by the BPNN, LR and AHP models; (3) Under all the 25 calculation conditions, the prediction accuracy of AIN=20 and RF model is the highest while that of AIN=4 and AHP model is the lowest, and the modeling efficiency and accuracy of AIN=8 and RF model are very high;(4) The landslide susceptibility indexes calculated by the higher AIN and more advanced machine learning models are more consistent with the actual distribution features of landslide probability and have relatively lower uncertainties. It can be concluded that an efficient and relatively accurate LSP model can be built under the condition of AIN value of 8 and RF model.

Among the machine learning models used for landslide susceptibility indexes calculation, the support vector machine (SVM) is commonly used; however, SVM is time-consuming. In addition, the non-landslide grid cells are selected randomly and/or subjectively, which may result in unreasonable training and validating data for the machine learning models. This study proposes the self-organizing-map (SOM) network-based extreme learning machine (ELM) model to calculate the landslide susceptibility indexes. Wanzhou district in Three Gorges Reservoir Area is selected as the study area. Nine environmental factors are chosen as input variables and 639 investigated landslides are used as recorded landslides. First, an initial landslide susceptibility map is produced using the SOM network, and the reasonable non-landslide grid cells are subsequently selected from the very low susceptible area. Next, the final landslide susceptibility map is produced using the ELM model based on the recorded landslides and reasonable non-landslide grid cells. The single ELM model which selects the non-landslide grid cells randomly, and the SOM network-based SVM model are used for comparisons. It is concluded that the SOM-ELM model possesses higher success and prediction rates than the single ELM and SOM-SVM models, and the ELM has a considerably higher prediction efficiency than the SVM.

How to efficiently assess the system reliability of rock slopes is still challenging. This is because when the probability of failure is low, a large number of deterministic slope stability analyses are required. Based on Subset simulation, this paper proposes an efficient approach for the system reliability analysis of rock slopes. The correlations among multiple potential failure modes are properly accounted for with the aid of the ¿max¿ and ¿min¿ functions. A benchmark rock slope and a real engineered rock slope with multiple correlated failure modes are used to demonstrate the effectiveness of the proposed approach.

It is important to monitor the displacement time series and to explore the failure mechanism of reservoir landslide for early warning. Traditionally, it is a challenge to monitor the landslide displacements real-timely and automatically. Globe Position System (GPS) is considered as the best real-time monitoring technology, however, the accuracies of the landslide displacements monitored by GPS are not assessed effectively. A web-based GPS system is developed to monitor the landslide displacements real-timely and automatically in this study. And the discrete wavelet transform (DWT) is proposed to assess the accuracy of the GPS monitoring displacements. Wangmiao landslide in Three Gorges Reservoir area in China is used as case study. The results show that the web-based GPS system has advantages of high precision, real-time, remote control and automation for landslide monitoring; the Root Mean Square Errors of the monitoring landslide displacements are less than 5 mm. Meanwhile, the results also show that a rapidly falling reservoir water level can trigger the reactivation of Wangmiao landslide. Heavy rainfall is also an important factor, but not a crucial component.

Spudcan foundations are often pushed into a spatially varying non-homogeneous seabed to provide bearing capacity for a mobile jack-up platform. The natural variability of soil properties coupled with the complexity of loading conditions make determining the bearing capacity of spudcan foundations a challenging problem. A random finite element method is established to investigate the bearing capacity of a spudcan foundation embedded in a spatially varying clayey seabed when subjected to vertical, horizontal and moment loadings. A criterion is proposed for determining the characteristic value of the shear strength for the random seabed. Results indicate that the spatial variability in the clayey seabed significantly reduces the bearing capacity of a spudcan foundation. This reduction is more significant in the vertical bearing capacity than in the horizontal and moment bearing capacities. The mean bearing capacity is smaller for the clay with larger coefficient of variation of undrained shear strength. A characteristic value of mean minus a standard deviation of the undrained shear strength is capable to ensure the probability of failure is not greater than 5%. This study provide an evaluation method for the spatial variability effect of a clayey seabed, paving the way for a cost-effective design of spudcan foundations.

Pile load tests are used to refine designs and for quality assurance. They can also be used to verify the reliability of piles and pile groups. Stochastic methods have previously been developed to verify the reliability of single piles. A general stochastic method to verify the reliability of pile groups is developed in this paper. The method can be used to assess the reliability of groups where pile tests have been conducted to the ultimate capacity, to below the ultimate capacity but exceeding specified capacity, and where pile tests fail to achieve the specified capacity. In the latter case, the method allows decisions to be made as to whether the reliability of the entire pile group is satisfactory or whether additional piles need to be installed.

The limit analysis and the finite element method are powerful tools for analysing the bearing capacity of foundations. Previous research mainly focused on the foundations in uniform soils. In realistic conditions, soil properties are always varying spatially due to complex physical, chemical, and biological process in earth evolution. This paper investigates the bearing capacity and failure mechanism of footings buried at various depths in clays with spatially variable distribution of undrained shear strength using the lower bound limit analysis, the upper bound limit analysis, and the finite element analysis. Results show that the bearing capacity increases with increasing buried depths in spatially random soils, which is the same as in the uniform soils. The bearing capacity factors calculated using the finite element method, the lower bound limit analysis, and the upper bound limit analysis for a footing in spatially varied soils are all smaller than the corresponding values in uniform soils. The majority of the bearing capacity factors obtained from the finite element method is bounded by those obtained from the lower bound and the upper bound limit analysis. The shear planes show a clearly unsymmetrical manner in spatially varied soils using the three methods, which is different from the symmetrical shear plane in uniform soils.

After a geotechnical design has been developed, it is common to monitor performance during construction using the observational method by Peck (published in 1969). The observational method is a process where data are collected and geotechnical models updated, allowing timely decisions to be made with respect to risk and opportunity by asset owners or contractors. The observational method is similar to the mathematical formulation for Bayesian updating of material parameters based on measurements. A proof of concept study has been performed to assess the potential for Bayesian updating to be combined with the observational method to allow timely and accurate decision-making during construction of embankments on soft soils. The method was able to converge to an accurate solution prior to 50% consolidation assuming small measurement errors. It is also demonstrated that confidence in the predicted settlement is relatively low at the prior ¿design¿ stage and rapidly increases with three or four measurements spaced over time during the posterior ¿construction¿ phase.

The mechanical properties of dry soil mix (DSM) columns can be highly variable. Variability can be accounted for in the construction specification for deterministic design and directly in reliability based design. Design methods and specifications to date adopt simplifications that do not take the variability of the columns fully into account. This paper uses both simple and advanced probabilistic methods to assess the performance/failure and system redundancy of dry soil mix columns. Reliability-based design methods and examples are given for the design of column strength and the adjustment of the column spacing to achieve a target probability of unacceptable performance or failure. An acceptance criteria chart is developed. The pull-out resistance tests on the DSM columns constructed for the Ballina Bypass motorway construction project in NSW Australia are compared to the chart to provide guidance with respect to acceptance criteria required to achieve the desired performance.

Rainfall induced landslides vary in depth and the deeper the landslide, the greater the damage it causes. This paper investigates, quantitatively, the risk of rainfall induced landslides by assessing the consequence of each failure. The influence of the spatial variability of the saturated hydraulic conductivity and the nature of triggering mechanisms on the risk of rainfall-induced landslides (for an infinite slope) are studied. It is shown that a critical spatial correlation length exists at which the risk is a maximum and the risk is higher when the failure occurs due to a generation of positive pore water pressure. © 2014 Elsevier B.V.

In the study of landslides, it is generally assumed that an impermeable boundary exists at a certain depth and failure occurs at this boundary. In reality this is not always the case and failures can occur at any depth. This paper aims to study the effect of boundary conditions on landslides, using a series of seepage and stability analyses performed over a range of rainfall intensities, and for different failure mechanisms, by studying the failure time and depths corresponding to fully drained, partially drained, and impermeable boundaries. It is shown that these conditions can significantly affect the occurrence and depth of rainfall-induced landslides. © 2014 Elsevier Ltd.

The observational method is one of the most successful processes in geotechnical engineering. Performance monitoring data are the most reliable information that engineers can use to predict future performance of geotechnical projects. This paper presents two examples where Bayesian statistical methods can be used for the prediction of future performance. The first example is to update the capacity of piles using load test results. The second example is to update embankment settlement predictions when field settlement monitoring data are available.

This paper investigates the soil properties, stratigraphy and spatial variability of the soils at the National Field Test Facility in Ballina based on extensive CPTU tests. The soil profile in this site consists of an alluvial crust over a relatively weaker layer of clay and underlain with a layer of sand and Pleistocene age stiff clay. The measured cone penetration resistance, sleeve friction and pore pressure for 26 CPTUs are presented along with the deployment of the CPTUs. The spatial variability in both vertical and horizontal direction of each layer of soils is explored based on the CPTU tests. An exponential autocorrelation function is found to best fit the autocorrelation coefficients. The scale of fluctuation in the vertical direction is 0.04 m in the alluvial crust layer, which is much smaller than that in the underlying clay layer, 0.15 m. The reason is that the clay was deposited under lower energy conditions compared to the more granular crust layers. The horizontal scale of fluctuation is 9.21 m in the alluvial crust layer and 4.92 m in the clay layer.

On the basis of new normative principles, criterions and methods such as Design specification for concrete gravity dams and Specifications for seismic design of hydraulic structures, we calculated the normal use ultimate strength and the carrying capacity ultimate strength of Dachaoshan dam according to the finite element method. The result indicates that the gravity dam is safe and credible; the dam's structure totally reaches the reliability level stated by criterions.

Two methods for estimating the ultimate bearing capacity of gravity dam are studied. The first method is to see whether or not the yield area across the dam. The second method called energy method is to define the ultimate state of the dam. Aiming at a real project, the example is given. It is found that the second method usually gets the lower result.

The paper presents a finite element model for strength analyses of a rotational composite shell under axial compression and internal pressure. The characteristics of stress distribution, stress locations and loads of failure are determined according to the model. The model is also developed to calculate the stability of the shell under axial compression, external pressure and combination of the both. Experiments are completed for studying the strength of the shell under axial compression and internal pressure and the stability of the shell under the combination of axial compression and external pressure.

A stability calculation method of laminated-composite dome structure is presented based on the modified-equivalent cylindrical shell model. The method is used to calculated the stabili ty of missile's dome under axial compression, external pressure and their combination respectively, and some experiments are completed for studying the stability of the dome under combined load. Finally, the relations between the dome's load of failure with characteristic size and laminated form are analyzed and discussed.

Since the charts ofTaylor, it has beenwell knownthat the location of the critical failure mechanism in a homogeneous undrained clay slope goes either deep (tangent to a firm base) or shallow (through the toe) depending on whether the slope angle is, respectively, less than or greater than about 53°. When slopes are made up of variable soils however, these expectations no longer hold true for all cases. In this paper, the influence of random soil strength and slope angle on the location of the critical failure mechanism and probability of failure is examined using the Random Finite Element Method (RFEM). It is found following Monte-Carlo simulation, that there exists a critical value of slope angle above which it would be unconservative to assume high spatial correlation length and below which it would be conservative to assume high spatial correlation length. For ß>48°, both correlation length and slope angle have no influence on the proportion of toe failures. For slope angle lying between 20° and 40°, slopes that have higher correlation length give lower proportion of toe failures. Research into the critical mechanism location forms part of a broader study of slope failure risk, in which the consequences of failure are assumed to be more serious in a deep failure, because a greater volume of soil is affected. © 2015 Taylor & Francis Group, London.

Bearing capacity and failure mechanism of a buried footing in uniformsoils have been simulated using limit analysis and finite element analysis in the past decades. In realistic conditions, soil properties always vary spatially. This dramatically affects the failure mechanism of a footing and, in turn, its bearing capacity. This paper illustrates an investigation into the failure mechanism and bearing capacity of a vertically and centrally loaded footing embedded in spatially variable clayey soils using random lower bound limit analysis, random upper bound limit analysis and random finite element analysis. The footing was embedded to 4 times its width. Monte Carlo simulation was performed for 400 realizations of random fields of undrained shear strength. The majority of the bearing capacity factors obtained from the finite element method is bounded by those obtained from the lower bound limit analysis and the upper bound limit analysis, but more close to the upper bound results. A full-flow failure mechanism is observed for the deeply embedded footing in spatially variable soil. The shear path of the footing shows an unsymmetrical pattern, which results from the spatially variable and unsymmetrical random field of soil shear strength. © 2015 Taylor & Francis Group, London.

Rainfall induced landslides can vary in depth and deeper the landslide, greater is the damage it causes. This paper investigates, quantitatively, the risk of rainfall induced landslides by assessing the consequence of each failure. The influence of the spatial variability of the saturated hydraulic conductivity on the risk of landslides is studied. It is shown that a critical spatial correlation length exists at which the risk is a maximum. © 2015 Taylor & Francis Group, London.

In deep drilling operation, borehole collapse due to insufficient drilling fluid pressure and borehole fracturing as a result of excessive drilling fluid pressure are two major modes of borehole failure. The latter may result in severe loss of drilling fluid into the formation, resulting in potential well control issues with influx of high pressure fluid or gas from adjacent formation layers. To prevent drilling fluid loss, the fluid pressure must not exceed an upper limit, otherwise the borehole will fail in tension or fracture. However, if the fluid pressure is too low, the borehole may collapse or fail in compression. Drilling fluid pressure window is therefore typically set by the upper limit of tensile failure and the lower limit of compressive failure. If the stress state and rock strength require a lower limit which is close to the upper limit, then the drilling fluid pressure design needs to stay within a very narrow window and consequently the feasibility of drilling may be questioned. Based on rigorous mechanics principles, this paper shows that it is possible to experience shear failure due to increasing drilling fluid pressure even before it reaches the upper limit of tensile failure. This may mean a more restricting drilling fluid pressure window in drilling weaker rocks. Implications of near wellbore shear failure are also briefly discussed in the context of water injection design which relies on injection at high pressure. © 2015 Taylor & Francis Group, London.

The drilling induced tensile fractures can be observed from borehole image logs but they must be differentiated from natural fractures. This paper focuses on the understanding of hydraulically or artificially induced tensile fractures while drilling. We present an inversion method using rigorous principles of mechanics, to determine the in-situ formation stress state from observed induced tensile fractures. Contrary to common practices where only vertical or near vertical wells can be analysed, the present method is applicable to wellbores of all orientations. For a geological rock formation and area where the in-situ stress regime can be assumed to be similar, all the relevant borehole image logs can be included to provide information to yield the most probable subsurface in-situ stress state. The proposed inversion method directly solves for the in-situ stress states given any single set of observed tensile fracture location and orientation. It provides not only an estimate for the minimum horizontal stress magnitude and direction, but also the maximum horizontal stress magnitude which is usually very difficult to pin down. The resultant equations are non-linear and a simple numerical scheme is adopted for the solution. Although published data on borehole images of fractures with corresponding in-situ stress information are scarce, two observed field data from published papers are chosen for comparison.

The mechanical properties of natural materials such as rocks and soils vary spatially. This randomness is usually modelled by random field theory so that the material properties can be specified at each point in space. When these point-wise material properties are mapped onto a finite element mesh, discretization errors are inevitable. In this study, the discretization errors are studied and suggestions for element sizes in relation with spatial correlation lengths are given. © (2014) Trans Tech Publications, Switzerland.

The homogenized stiffness of geomaterials that are highly variable at the micro-scale has long been of interest to geotechnical engineers. The purpose of this study is to investigate the influence of porosity and void size on the homogenized or effective properties of geomaterials. A Random Finite Element Method (RFEM) has been developed enabling the generation of spatially random voids of given porosity and size within a block of geomaterial. Following Monte-Carlo simulations, the mean and standard deviation of the effective property can be estimated leading to a probabilistic interpretation involving deformations. The probabilistic approach represents a rational methodology for guiding engineers in the risk management process. The influence of block size and the Representative Volume Elements (RVE) are discussed, in addition to the influence of anisotropy on the effective Young's modulus. © 2014 Taylor & Francis Group, London.

Analytical probabilistic analysis and Monte Carlo simulation based on elasto-plastic Finite Element Method (FEM) on dry soil mix columns are presented. It is shown that analytical method is over conservative because it ignores the supports from adjacent columns. Probabilistic FEM analysis can provide more accurate predictions, and thus lead to more economic designs. Probabilistic FEM analyses show that the effects of adjacent columns can be destructive when applied load is close to the strength. The reliability of the system of columns is analyzed by setting residual strength to zero. Results show that close spacing has more safety margin than loose spacing. © 2014 Taylor & Francis Group, London.

Computational limit analysis provides a fast and convenient means of evaluating the stability or bearing capacity of geostructures. It is based on numerical optimization techniques and the latest trend is to use robust conic programming algorithms. The shortcoming, however, is that the types of problems covered by conic programming are not very general. In practice, this means that only criteria containing linear and quadratic terms (such as Drucker-Prager) or those that involve linear terms in the principal stresses (such as Mohr-Coulomb) can be considered. In the present paper this shortcoming is addressed. The idea is to maintain an efficient and robust conic programming algorithm as the main solution engine. Nonlinear criteria are then handled by a dynamic linearization procedure that involves a sequence of standard conic programming solutions in an iterative scheme that turns out to converge relatively rapidly.

Program PES (Probabilistic Engineered Slopes) provides a repeatable methodology allowing the user to perform a slope stability analysis on a one-sided and two-sided sloping structure using a deterministic or probabilistic approach. Program PES, in contrast with other deterministic or probabilistic classical slope stability methodologies, is cable of seeking out the critical failure surface without assigning a pre-defined failure surface geometry. The probabilistic approach of program PES applies the Random Finite Element Method (RFEM) by Griffiths and Fenton (1993) taking into account the soil spatial variability and allowing the use of different random fields to characterize the spatial variation of any material type. The methodology is compared against the probabilistic approach proposed by the program SLOPE/W version 7.14 (Geostudio Group, 2007), and demonstrates its potential for predicting probability of failure (pf) in non-homogeneous soil structures characterized by phreatic conditions and potential post-earthquake liquefiable conditions. The pf results obtained by program PES have proved that underestimating the influence that the soil material variability has on the computation of pf will lead to lower results of probability and underestimate of the risk of slope instability. Program PES capabilities could be used by the engineering practice to prioritize intervention activities within a risk context. © 2011 American Society of Civil Engineers.

The paper describes probabilistic analyses performed as part of a large expansion to an existing cement manufacturing plant. A raft supported by drilled shafts was proposed for the project, but during installation, significant slurry and concrete loss began to occur indicating numerous voids existed in what was previously considered competent limestone bedrock. Since the possibility of voids, especially at the shaft tip, could serious reduce the shaft capacity, a probabilistic Monte Carlo 3D finite element simulation was proposed for the most heavily loaded raft foundation. The purpose of the simulation was to determine the probability of adverse performance, giving guidance as to whether any remedial measures (e.g., additional structural elements or thickened raft) might be required. © 2011 ASCE.

Reliability tools have been applied to slope stability analysis more than any other geotechnical application on account of the readily understood concept of "probability of failure" as an alternative or complement to the traditional "factor of safety". Probabilistic slope stability methods in the literature are reviewed. Particular attention is focused on the ability of the methods to correctly model spatially varying soil properties. A benchmark slope is reanalyzed and conclusions reached about their suitability for meaningful and conservative prediction of slope reliability. © 2010 ASCE.

The computer program Probabilistic Engineered Slopes (PES), coded in FORTRAN.95, provides a repeatable methodology, which allows the user to perform a slope stability analysis on a one- and two-sided sloping structure, using a deterministic or probabilistic approach. The program PES, in contrast with other deterministic or probabilistic classical slope stability methodologies, is capable of seeking out the critical failure surface without assigning a predefined failure surface geometry. The probabilistic approach of PES applies the Random Finite Element Method (RFEM) by Griffiths and Fenton (1993) [1], taking into account the soil spatial variability and allowing the use of different random fields to characterize the spatial variation of any material type. The methodology is compared against the probabilistic approach proposed with the program SLOPE/W, version 7.14 (Geostudio Group, 2007) [2], and demonstrates its potential for predicting probability of failure (pf) in nonhomogeneous soil structures for given phreatic conditions and potential postearthquake liquefiable conditions. The pf results obtained by program PES have proved that underestimating the influence that the soil material variability has on the computation of pf will lead to unconservative results of probability and underestimate of the risk of slope instability. The program PES has capabilities that could be used by the engineering practice to prioritize intervention activities within a risk context, test the stability conditions of dams during modification phases, and help estimate the probability of failure in cases involving postearthquake liquefaction.

The computer program Probabilistic Engineered Slopes (PES), coded in FORTRAN.95, provides a repeatable methodology, which allows the user to perform a slope stability analysis on a one- and two-sided sloping structure, using a deterministic or probabilistic approach. The program PES, in contrast with other deterministic or probabilistic classical slope stability methodologies, is capable of seeking out the critical failure surface without assigning a predefined failure surface geometry. The probabilistic approach of PES applies the Random Finite Element Method (RFEM) by Griffiths and Fenton (1993) [1], taking into account the soil spatial variability and allowing the use of different random fields to characterize the spatial variation of any material type. The methodology is compared against the probabilistic approach proposed with the program SLOPE/W, version 7.14 (Geostudio Group, 2007) [2], and demonstrates its potential for predicting probability of failure (pf) in nonhomogeneous soil structures for given phreatic conditions and potential postearthquake liquefiable conditions. The pf results obtained by program PES have proved that underestimating the influence that the soil material variability has on the computation of pf will lead to unconservative results of probability and underestimate of the risk of slope instability. The program PES has capabilities that could be used by the engineering practice to prioritize intervention activities within a risk context, test the stability conditions of dams during modification phases, and help estimate the probability of failure in cases involving postearthquake liquefaction.

The paper investigates the probability of failure of 2-d and 3-d slopes using the Random Finite Element Method (RFEM). RFEM combines elastoplasticity with random field theory in a Monte-Carlo framework. It is found that 2-d probabilistic analysis, by implicitly assuming perfect spatial correlation in the third direction, may underestimate the probability of failure of slopes. © 2009 IOS Press.

The influence of a spatially random coefficient of consolidation on one-dimensional uncoupled consolidation has been studied using the Random Finite Element Method. The results of parametric studies are presented, which describe the effect of the standard deviation and correlation length of the coefficient of consolidation on output statistics relating to the overall "effective" coefficient of consolidation. Three "effective" coefficient of consolidation are considered, namely harmonic mean, the log time method and the root time method. Copyright ASCE 2008.

The study investigates the role of spatially random soil on the stability of infinite slopes with application to landslides and other geohazards. The influence of the shear strength mean, standard deviation and spatial correlation length on the probability of failure is thoroughly investigated through parametric studies. The results show that the traditional "first order second moment" approach to this problem is inherently unconservative, due to its inability to allow the failure mechanism to "seek out" the critical depth below ground surface, which is frequently not at the base of the soil layer. Copyright ASCE 2008.

The paper presents probabilistic studies that demonstrate the influence of spatially random soil properties on the stability of shallow landslides using random fields Results indicate that traditional "first order" methods are inherently unconservative when applied to limit analysis problems unless they allow the failure mechanism to "seek out" the most critical location.