Reliability Analysis Of Slope Based On Random Field And Kriging Agent Model

The stability assessment of slope is an important step of landslide forecast and treatment.In slope engineering,on the one hand,there are many complex characteristics in slope,on the other hand,slope is affected by analytical model and method,so its stability shows uncertainty.With the development of the science and technology,there are more requirments for the precision and solution efficiency of slope stability analysis theory,but some traditional methods have limitations.To this end,aiming at the accuracy and solution efficiency of slope stability analysis theory,on the basis of the existing research achievement,the optimization algorithm of critical slip surface and slope reliability analysis method based on the random field model are studied in this article by theoretical analysis and numerical simulation.The main task is presented as follows:Firstly,in oder to the enhance the local search ability,the improved seeker optimization algorithm(ISOA)which combined chaos algorithm is proposed.Through functions optimization,ISOA has a promotion on accuracy,stability and efficiency.Combined with finite surface stress method,improved seeker optimization algorithm is applied to seach the critical slip surface.And the corresponding procedure is coding.Through two test examples of ACADS,it can be verified that ISOA is suitable for application in slopes.Secondly,based on the random field theory,according to the mean variance,correlation distance,correlation function and other statistical characteristics of geotechnical parameters,the random field model is established.And the random field is discreted by Karhunen-Loveve(KL)expansion method.The influence of different parameters on the discretization precision of random field is analyzed.Finally,on the basis of the random field model,the safety factor of slope is calculated by finite surface stress method.The agent model is established between the input that is independent standard random variable in the random field and the response that is factor of safety.Given the high dimension of sample points,slice inverse regressive(SIR)method is used to reduce the dimension.Then Kriging agent model is established between the sample points whose dimension is reduced and its response.With a large number of new samples,the Monte-Carlo method is used to obtain failure probability of the slope.And the influence of correlation distance,related coefficient and coefficient of variation on the failure probability of slope is analyzed.