Research On Early Warning Methods Of Landslide Based On Displacement Monitoring Data

Landslide disasters have strong suddenness,wide distribution,and high frequency.They are a very harmful and extremely common geological disaster.China has a vast territory and extremely complex geological conditions,which are deeply affected by landslide disasters.In the western and southwestern regions of China,mountainous areas account for as much as 85%of the area,and the steep mountain terrain formed by excessive wind and rain erosion also provides sufficient geological conditions for the formation of landslide disasters,leading to frequent occurrence of landslide disasters.This paper mainly uses two intelligent algorithms,BP and RBF,to carry out related research on landslide prediction.(1)By learning the relevant mechanisms and methods of BP neural network and applying it to landslide displacement prediction,several issues that need to be paid attention to in the prediction process are discussed.Introduced four improved BP network training algorithms,MOBP,CGBP,QNBP,LMBP.And the Gushuwu landslide displacement monitoring data was predicted,and a good prediction effect was obtained.(2)The related mechanism of RBF neural network and the method of applying it in landslide prediction are introduced,and the Gushuwu landslide is predicted by the RBF network landslide,and good prediction accuracy is obtained.Later,in order to further verify the prediction effects of the BP network and the RBF network,the Hongshibao landslide was predicted,and both networks can successfully predict it.Compared with BP network,RBF network has more general advantages.(3)An improved RBF landslide prediction method based on displacement increment sequence is proposed,and the displacement prediction of Gushuwu landslide and Hongshibao landslide is carried out,and good results are obtained.Then the results of the three methods were summarized and analyzed.The Gushuwu landslide has achieved the best results by the incremental method,while the Hongshibao landslide has not improved significantly.For the RBF network prediction,it has the characteristics of superior to the BP network.And for the RBF network prediction method modeled by the displacement increment sequence,compared with the direct RBF prediction,the prediction accuracy of the two is closely related to the original displacement sequence and the displacement increment sequence.Therefore,it is necessary to make a detailed comparison to determine in the actual application process.