Quantitative Risk Assessment Approaches Based On The Simulated Landslide Evolution From Initiation To Post-Failure Large Deformation Of Soils

Landslides resulting from slope failure is a major geo-hazard in China,causing massive casualty and property loss each year.Quantitative risk assessment of landslides plays a significant role in mitigating landslide hazard and reducing its social-economic loss.Landslide risk is influenced by various uncertainties existing in geotechnical engineering,such as uncertainties in loading and spatial variability of soil properties.Risk assessment of landslides includes the assessment of slope failure probability(or slope reliability analysis)and landslide consequences.Slope reliability analysis theory and approaches have been rapidly developing in recent years.However,it remains a challenging issue to incorporate a large number of circular and/or non-circular potential slip surfaces into slope reliability analysis.Besides,most slope reliability methods cannot comprehensively assess landslide consequences.This may be attributed to that they use traditional slope stability analysis methods(e.g.,limit equilibrium methods and finite elements method)which face difficulties in simulating large deformation of soils during a landslide.This thesis aims for developing quantitative risk assessment approaches based on simulated evolution of landslides.Three key problems are summarized and addressed in this study,i.e.,simulation of the entire process of a landslide,slope reliability analysis considering a large number of circular and/or non-circular slip surfaces,quantitative risk assessment of landslides based on the simulated evolution of landslides.Major work and conclusions are summarized below.(1)Regarding the simulation of the entire process of a landslide,a pore water pressure approximation approach was developed to resolve the pressure oscillation problem caused by volumetric locking based on the Hu-Washizu variational principle.A hydro-mechanically coupled material point method program was developed and applied to a real-case landslide history induced by rainfall,i.e.,the Fei Tsui Road landslide occurred in Hong Kong on June13,1995.The entire process of the landslide was simulated and compared with field observations to verify the numerical model.The coupled material point method is able to reasonably simulate the evolution of rainfall induced landslides,from rainfall infiltration,advancement of wetting front,responses of pore water pressure,initiation of multiple failure modes and large deformation of soils.The incident includes with a minor failure and a major failure,and the major failure consists of a shallow failure and a deep failure.These failures collectively contribute to the failure consequence with large sliding volume and large run-out distance,which is more severe than typical small-scale rainfall-induced landslides in Hong Kong.For a slope under rainfall infiltration,perched water tends to form when a wetting front flows along a steep joint and encounters a relatively bedrock layer,and it deteriorates slope stability and possibly triggers a deep failure.Special attention should be paid to this geological condition.These results demonstrate the advantage of material point method in simulating the evolution of rainfall-induced landslides in comparison with traditional slope stability analysis methods and facilitate its further applications in landslide simulation.(2)With respect to slope reliability analysis considering a large number of slip surfaces,an adaptive Monte Carlo simulation(MCS)method was proposed for reliability analysis of large series systems with many correlated components.It is able to iteratively finds a small number of components and constitutes a subsystem that identifies all failure samples as the original system does,because failure of any component results in system failure and components are usually correlated.The equations for estimating system failure probability was validated.Various series system examples were used to verified its high performances in terms of convergence and computational efficiency.Its convergence is mainly affected by termination parameter,component correlation and nonlinearity of component performance functions.30-200 is the recommended range of the termination parameter.The computational efficiency of adaptive MCS is mainly affected by component number,component correlation,nonlinearity of performance functions,and termination parameter.Its computational efficiency improves when the component number and the component correlation increase,and becomes much higher than of direct MCS and subset simulation when the component number is large(e.g.,component number > 1000)or when the component correlation is strong.Adaptive MCS solves the challenging problem of reliability analysis of series systems with many correlated components.(3)Adaptive MCS was applied to slope system reliability analysis.It is able to consider a large,or even unlimited,number of circular and/or non-circular potential slip surfaces in slope system reliability analysis in an efficient manner.Its ability was verified by slope examples.Compared with direct MCS with brute-force exploration of failure samples after random samples are generated,the proposed method has a much higher computational efficiency.The assumption of slip surface shape is important to slope system failure probability for slopes with a weak layer.Assuming slip surfaces to be circular may overestimate slope system failure probability to the extent of one to two orders of magnitude,and result in unconservative slope design.There is a succinct slope subsystem comprising of a minimum,but sufficient,number of potential slip surfaces to represent the complete slope system given a set of direct MCS samples.The proposed method not only provides an easy and efficient tool for slope reliability analysis considering a large number of circular and/or non-circular slip surfaces,but also benefits for slope reinforcement design and hazard mitigation.(4)A quantitative risk assessment framework of landslides based on the simulated evolution of landslides was proposed.It is able to finish all necessary jobs by a single run,i.e.,slope reliability analysis,identification of elements at risk and vulnerability analysis.A random limit equilibrium and material point method was developed and adopted for slope reliability analysis.It takes advantage of both the computational efficiency of limit equilibrium methods and the ability of material point method in modeling large deformation of soils to reduce the total computational costs.The evolution of different landslide failure modes(i.e.,shallow,intermediate,deep and progressive failure modes)were investigated.In some cases,the landside failure mode can be non-circular in spatially variable soils.Limit equilibrium methods face difficulties in simulating the case when multiple failure modes may occur successively.The spatial variability of soil properties not only greatly affects the slope failure probability,as reported in the literature,but also determines the landslide failure mechanism and consequences.It is therefore of great importance to characterize site-specific spatial variability of soil properties for properly predicting the evolution of landslide failure,risk assessment and mitigation.Multiple features of landslide failure modes were quantified and used to for reliable vulnerability assessment and quantitative risk assessment of landslides.The proposed framework provides a new approach and new tool for quantitative risk assessment of landslides.