Analysis And Prediction Method For Monitoring Data Of Elevated Station Track System In High-Speed Railway

Due to the influence of topography,geomorphology,and the requirements for settlement,numerous welded turnouts are set on the bridges of high-speed railway.High-speed railway elevated station track system,consisted of the ballast-less track,welded turnout,as well as bridge,is difificult to be researched by conventional theoretical analysis methods.In order to promote the state-overhaul of the track,it’s necessary to build up a monitoring system to obtain the key indicators which should be monitored and predicted in real time.A large number of monitoring data of the elevated station in Beijing-Shanghai high-speed railway was paid more attention than ever before.The correlation of different monitoring indexes were analyzed,and prediction models were built.After that,visualized interface for early warning were designed for maintenance.The main work contents and achievement are as follows:(1)In order to get effective and reliable predict results,typical indicators of mechanical states directly related to the security of track system were extracted at the source.A series of finite element models were built to analyze the impact of the shape,material and installation method of the strain gage on the monitoring data.After that,a comparative experiment indicated this method is accurate and reliable.While,a correction method based on parallax was put forward to get the displacement of the switch rail converted.(2)Creditability of the monitoring data were validated by Kolmogorov Simonov test,periodic trend and spectral distribution were found in statistical method to describe the lagging of the track condition.Two types of outlier identification and correction methods includes Box-plot and Pauta criterion were contrasted and a fault detection mechanism for monitoring system was proposed.In addition,correlation between mechanics index include stress,displacement of the track system and rail temperature or air temperature were explored by regression models,especially their time series.The study may guiding significance for layout of transducers in field,reducing the number of sensors efifectively.(3)Regressive prediction models were built up in compared with the back-propagation and radial basis function artificial neural network prediction models to predict the mechanics behavior of the elevated station track system,include rail temperature,stress and displacement of the rail.Sample while the track system suffering from cold wave were used to verify the prediction model.According to the demand under different conditions,statistical results of prediction errors were used to evaluate different models.Comprehensive measures combined with BP neural network,RBF neural network and multiple regression models were proposed to forecast track state.(4)A mechanism combined with a series of static threshold and diagnostic approaches with distance or clustering was used to alarm and warning automatically.Static threshold roots in standards or statistical results,while diagnostic approaches based on distance and clustering can be used after the correlation weaken only.After that,a GUI providing forecast and warning information for maintenance was built.