Research On Multi-Source Data Fusion Method For Train Operation Environment Based On Wireless Sensor Networks

With the increasing coverage of railways in China,the transportation environment for trains has become increasingly complex and ever-changing.The safety of railway transportation caused by natural disasters has become increasingly prominent.Natural disasters such as landslides and mudslides along the railway line seriously threaten the transportation safety and order of trains.This article takes a landslide along a certain railway line as the research object and uses the original time series data.Using a two-level data fusion model based on wireless sensor networks to predict the trend of landslide deformation in the future period,thereby improving the response ability to landslide disasters along the railway.The main research content of this article is as follows:(1)Construct a wireless sensor network topology structure and corresponding data fusion process for landslide monitoring along the railway line;Constructing a two-level data fusion model based on wireless sensor networks;Based on the obtained characteristics of landslide displacement sequence,the missing values,outlier and displacement coordinates are processed;This article introduces the monitoring equipment and basic layout used in landslide monitoring along the railway,including GNSS high-precision displacement monitoring station,rainfall monitoring station,and soil moisture content monitoring station.(2)For the X,Y,Z coordinate series of surface displacement collected by GNSS monitoring station,the missing values and outlier are processed,and the coordinates are converted into cumulative displacement in E,N,U directions.The soil moisture content at different depths is processed by adaptive weighted fusion method for data level fusion,providing high-quality data basis for subsequent feature level fusion.The results indicate that the adaptive weighted fusion algorithm can effectively fuse soil moisture content sequences at different depths.(3)When predicting the trend of landslide deformation in the future,most monitoring only considers a single variable of deformation,and the landslide is comprehensively affected by multiple factors during the incubation and development process.Therefore,feature level fusion is used to incorporate multi-source heterogeneous information such as other environmental factor variables(temperature,humidity,wind speed,soil moisture content,daily rainfall,cloud cover)into the fusion model,And based on m RMR feature selection,environmental factors with the greatest gain on displacement variables are selected to explore the effects of other environmental factors on landslide displacement deformation.By introducing Tent chaotic perturbation and Gaussian mutation,the optimization ability of Sparrow Search Algorithm SSA is improved.The improved Sparrow Search Algorithm ISSA is used to optimize the learning rate,maximum iteration times,and maximum depth of the XGBoost method.The experimental results show that the ISSA-XGBoost fusion model can effectively fuse multi-source heterogeneous monitoring data of landslides along the railway line.The environmental factor variables filtered through m RMR feature selection can better represent the future trend of landslide deformation in the future period than the unfiltered environmental factor variables,and the prediction accuracy is improved.