| Landslides are a very common and serious natural disaster.According to relevant statistics,there were more than 100,000 geological disasters in China between 2011 and 2020,of which 70,000 were landslides,accounting for 70% of the total.It is of great significance to analyze and research slope stability and reliability.Neural networks have been widely applied to the cross-section analysis and evaluation,Such as the network models and algorithms.On the premise of relevant research,sensitivity analysis of seven slope influencing factors is conducted in this paper,and then a GA-PSORBF prediction model on slope stability coefficients,reliability indicators and failure probabilities was established by combining genetic algorithm(GA)and particle swarm algorithm(PSO)through radial basis neural network(RBF).The main research elements and conclusions were as follows:(1)Based on the same single-layer homogeneous slope,the stability coefficient output by SLIDE software was used as a subject of study,and the sensitivity of the factors was analyzed by introducing univariate analysis,orthogonal experimental analysis and grey correlation degree analysis,which showed that among the seven selected influencing factors,the angle of internal friction φ was the most sensitive,and the performance of groundwater level h was the least sensitive.(2)In the process of the stability analysis,Consider the RBF neural network with good memory and strong self-learning ability,and to overcome the problem of unreasonable selection of its own parameters,it was optimized by embedding both GA and PSO,and then the GA-PSO-RBF prediction model on slope stability coefficients was established,and the samples generated by the orthogonal test were used to analyze the accuracy of the model.The accuracy of the model was analytically tested using the samples generated from the orthogonal test.The study shows that the model has advantages over the model under any single algorithm in terms of the adaptation curve,and has greatly improved in terms of prediction accuracy compared with the simple RBF prediction model.(3)The variation trend of soil parameter variation on the three indicators,including stability coefficient,reliability index and failure probability of slope,was analyzed.The study well-studied representation that in the process of variation coefficient changes,the first indicator did not tend to fluctuate much and remains almost unchanged,while the variation trend of slope reliability index and failure probability was much more obvious.(4)In view of the existence of coefficient of variation in soil parameters,an analytical method that combining with LHS and GA-PSO-RBF,and it can be used to predict the slope reliability index and failure probability,and the study of slope reliability prediction analysis was carried out for homogeneous single-layer,homogeneous three-layer and a practical engineering case successively.The research showed that the method proposed in this paper was more accurate than the related methods and can be used in the actual slope engineering stability problems,and it can lay a good foundation for the corresponding theoretical research. |
