Typical Landslide Monitoring Early Warning and Susceptibility Evaluation Based on BP Neural Network Model in Zhouzhi County

The method of in-situ monitoring and real-time early warning and forecasting is of great practical significance to effectively manage and reduce the risk and loss caused by landslide disaster. Therefore, this paper takes the typical landslide in Zhouzhi County, Xi'an as the research object, and a special group combined monitoring system was established by means of rainfall gauge, soil moisture gauge, dip gauge, crack gauge and GNSS monitoring equipment. Based on this, BP neural network model was used to evaluate the vulnerability of the landslide. The results show that the data collected on site can be sent to the monitoring data platform through 4G/5G, satellite communication and LORA AD hoc networking, which can effectively support the disaster early warning work. The monitoring platform can automatically analyze the monitoring data collected in real time, and can support the discrimination of various early warning models. When the preset conditions are triggered by changes in monitoring data, it can automatically send early warning information; Through the trial operation of the monitoring and warning system, the millimeter-level creep deformation of the landslide is captured. There is an obvious positive correlation between the change rule of the horizontal displacement of the landslide and the accumulated rainfall, and the curve shows the phased characteristics of decelerating creep, constant velocity creep and accelerating creep. The predicted value of landslide state based on BP neural network model fits well with the actual value. Through the implementation of three landslide monitoring and early warning in Zhouzhi County, Xi'an, the results can provide reference for similar landslide universal monitoring.