Study On Stability Prediction Method Of Slopes With Bedrock Based On Neural Network

The geological conditions in Southwest China are complex,and the slopes with bedrock are widely distributed.Such slopes tend to have poor stability,which has an adverse impact on the construction of adjacent roads,railways,bridges and other infrastructure,and hinders China’s acceleration of the construction of the the Belt and Road in the new era and the realization of high-speed economic development.Taking appropriate methods to predict the stability of the slopes with bedrock can take timely measures to deal with the disaster and minimize the possible economic and property losses.In this paper,the neural network technology is used to establish the safety factor and stability grade prediction model of the slopes with bedrock.The safety factor prediction model aims to make a quantitative analysis of the slope stability when the known information is sufficient.The stability grade prediction model is mainly to make a rapid preliminary evaluation for the stability of the foundation covering slope in the field.The main research results are as follows:(1)The finite element limit analysis method can give the relatively accurate safety factor of the slopes with bedrock.Compared with the indoor shaking table model test,it is found that the numerical results can better show the deformation and failure mode of the slope.Because of the particularity of its structure,there are many types of stability influencing factors of the slopes with bedrock.The order of the importance of the factors affecting the safety factor of the slopes with bedrock is as follows:slope angle β,interface friction angle μ,internal friction angle φ,upper interface inclination angle θ1,slope height H,cohesion c,horizontal seismic coefficient k,soil weight γ,slope width L,lower interface inclination angle θ2.(2)In order to simplify the structure of the model,eight parameters which have a high degree of influence on the slope safety factor are taken as the network input,and the prediction accuracy is not much different from that of the model with all the parameters as input.The dimensionless input parameters have a positive effect on the network model,and the comprehensive performance of the model is better.Genetic algorithm(GA)and sparrow optimization algorithm(SSA)can further improve the prediction accuracy of network security factor.The two models have their own advantages.The GA-SSA model established by using the optimal weighted combination algorithm can combine the advantages of the two.(3)Based on the actual characteristics of the slopes with bedrock in Sichuan,the stability grades of the slope are divided into:stability,basic stability and instability,and the importance of the parameters is analyzed.The order of importance of each factor affecting the stability grade of the slopes with bedrock is as follows:roughness of foundation overburden interface,type of overlying soil,slope,inclination of upper foundation overburden interface,inclination of lower foundation overburden interface,thickness and height of accumulation body.Finally,the stability grade prediction model of the slopes with bedrock is established.GA-BP model can obtain the best classification results,then SSA-BP model,and finally the original BP model.(4)Combined with three specific engineering examples,the practical application effects of the safety factor prediction model and slope stability grade prediction model established in this paper are studied.The results show that the safety factors predicted by BP,GA-BP and SSA-BP models are basically consistent with the field measured safety factors,and the BP,GA-BP and SSA-BP stability grade prediction models also have high classification accuracy.The research results are of great significance to the actual landslide disaster prevention and control.