Research On Early Warning Model Of Debris Flow In Railway Area Based On Big Data Analysis And Mining

Our country’s railway transportation is regarded as the main artery of the national economy.Nowadays,with the frequent occurrence of extreme weather,the frequency of debris flow disasters in railway areas has also increased,causing greater economic losses,so early warning is increasingly important to our country’s railway transportation.In this thesis,on the basis of studying the formation mechanism of debris flow,a breakthrough to the traditional monitoring facility and rainfall information warning model,research on early warning model of debris flow in railway area based on big data analysis and mining was carried out,and the following achievements was archived:1.The innovative introduction of In SAR technology to monitor the surface deformation of the railway area,and the use of big data analysis technology to study the influencing factors of debris flow.In addition,the law of the surface deformation-time curve of slope-type debris flow was studied by improving the Saito curve,and it was found that the occurrence of slope-type debris flow disasters was significantly related to the critical tangent angle.2.LSSVR algorithm combining sliding time window and hybrid kernel function is proposed,and it is combined with the grid cell division method to conduct data mining of debris flow disaster risk patterns and build a debris flow disaster risk prediction model.Experiments show that the accuracy of the prediction model is increased by 8.3%,and the false alarm rate is reduced by 4.5%.In addition,an early warning model of debris flow disaster risk is innovatively proposed,which combines the probability of occurrence of debris flow,the scale of disaster and the human settlement environment in the railway area.Experiments show that this early warning model performs better than other early warning models.3.On the basis of the debris flow risk prediction model and grid unit division,this thesis further solves the problem of predicting the spatial extent of debris flow disasters.Experiments show that this method has better prediction effects on small and medium debris flow disasters.4.Aiming at the prediction method of the surface deformation time series of debris flow,a combined algorithm of MV-PR-LSTM is proposed.This method innovatively uses the moving average method to decompose the surface deformation time series into internal factors and external factors.The regression algorithm predicts the time series of intrinsic factor items,and the LSTM neural network is applied to construct the time series prediction model of external factor items.Experiments show that the prediction method can effectively improve the accuracy of time prediction of debris flow disaster risk.5.An online early warning visualization system for debris flow disasters in railway areas is developed.The system realizes the visualization functions of debris flow disaster risk early warning,prediction of debris flow disasters in time and space,and dynamic update of multi-source monitoring flow data around monitoring sites.