Study On Hazard Assessment Method Of Landslide Along Railway

Landslide geological disasters along railway lines are characterized by strong sudden occurrence and great destructive power.Once geological disasters occur,ecological environment will be destroyed,people’s life and property safety will be damaged,and railway lines and driving safety will be seriously threatened.Under the organization of the railway Corporation,China has gradually established and improved the railway disaster prevention and safety monitoring system.At present,the system can monitor and warn of disasters such as wind,rain,snow,earthquake and foreign body encroachment,effectively protecting the safety of railway operation.However,the monitoring of landslide hazard along the railway is still in the research and test stage,and has not been popularized and applied in the field.Therefore,in order to monitor the landslide disaster along the railway and ensure the safety of railway operation,it is necessary to carry out the landslide risk assessment research along the railway.Based on different amounts of landslide data,this thesis selects different evaluation methods to provide schemes for landslide risk assessment and reduce the loss caused by landslide disasters.The details are as follows:(1)Through the actual investigation and analysis of landslide points in the study area,a total of 12 evaluation factors were selected from five aspects,such as topography,geological conditions,meteorology and hydrology,normalized vegetation cover index and human factors.The natural breakpoint method is used to grade each evaluation factor,and the area and landslide point in each grade area are analyzed statistically.In general,there is a certain correlation between the selected evaluation factors,which will make the model information redundant and lead to the reduction of the model accuracy.The collinearity analysis of factors is carried out by Pearson coefficient method,and the factors with less correlation are selected as the evaluation factors.(2)When the amount of landslide sample data is small,the improved mutation theory is used to conduct risk assessment and analysis of landslide data.Firstly,the relative importance of evaluation factors is calculated using entropy weight method,and then the data is standardized and normalized into the corresponding mutation model,and the total mutation result is finally obtained.Data normalization will lead to high and concentrated evaluation scores,and the original mutation results will be transformed by fitting function,so as to establish a new landslide risk judgment standard.The improved evaluation results are compared with the survey results,and the results show that the improved evaluation method is more accurate and can provide a basis for landslide hazard risk assessment.(3)When there is a large amount of landslide sample data,the evaluation model based on random forest is used to evaluate landslide risk.Firstly,three commonly used parameter optimization algorithms,namely whale algorithm,genetic algorithm and particle swarm algorithm,are selected to adjust the parameters of random forest respectively.Then,the selected evaluation factors were input into the unoptimized random forest and four evaluation models optimized by genetic algorithm,particle swarm algorithm and whale algorithm respectively.The accuracy of the four models was compared by the receiver characteristic curve.The results show that the random forest model optimized by the whale algorithm has the best evaluation effect,and the area under the subject characteristic curve is the largest(0.91).Therefore,the random forest model optimized by the whale method was used to evaluate the landslide risk in the study area.Arc GIS software was used to draw the landslide risk zoning map,and the natural discontinuous method was used to divide the risk zoning map into four danger levels: extremely high,high,medium and low.By comparing the hazard zoning map with the historical landslide disaster sites in the study area,about 77% of the landslides are in high and high risk areas.The evaluation results are in agreement with the actual results,indicating that this research method can provide a new way of thinking for landslide hazard risk assessment.(4)A landslide hazard warning system based on whale method is designed to optimize random forest model.By analyzing the historical landslide data,the system extracts the evaluation factor data and layers,and then uses the random forest model optimized by whale algorithm to classify and predict the landslide risk level.The system can realize the functions of visualization display of landslide data and release of landslide warning information.