Research On Fault Diagnosis Method For On-board Equipment Of Train Control System Based On BiLSTM-FWASVM Model

Chinese Train Control System is the core of safe operation of high-speed trains.As the key equipment of CTCS,On-board Equipment of Train Control System monitors the safe operation of the train by processing the ground information to generate and control the speed.During operation of the train in practice,on-board equipment of train control system in case of failure,light impact on traffic safety and transport efficiency,heavy accident safety.At present,after the fault of on-board equipment of train control system,it mainly depends on the work experience of field workers to judge,there is uncertainty and low efficiency of diagnosis.With CTCS and its related equipment becoming larger,more complex,network and intelligent,the professional quality of signal analysis personnel has higher requirements.Therefore,the realization of intelligent automatic fault diagnosis of on-board equipment of train control system can not only provide assistant decision-making for signal maintenance personnel,but also ensure the safe operation of trains and improve transport efficiency.In this thesis,the CTCS-300T on-board equipment of train control system suitable for CRH_380Band CRH_380BLEMU is taken as the research object,a fault diagnosis model for on-board equipment of train control system based on Bi-directional Long Short-Term Memory and Fire Works Algorithm optimization Support Vector Machine is designed.The thesis mainly consists of the following parts:（1）Firstly,the structure and function of the CTCS-300T train control on-board equipment are introduced,analyzes the characteristics of Application Event log,and summarizes the fault types,and lays a foundation for the fault diagnosis of train control on-board equipment in this paper.（2）Vectorized representation of on-board equipment of train control system text.In order to mine the useful information in AElog,it is necessary to clean the ATPCU document and keep the running state statement which can reflect the state of the device.Bidirectional Encoder Representations from Transformers model,which can mine semantic information,is used to represent the AElog as a text vector that can be recognized by the computer.（3）Fault diagnosis of on-board equipment based on Bi LSTM-FWASVM model.Bi LSTM realizes the automatic extraction of temporal features of text.Considering that softmax classifier in Bi LSTM only does the normalization operation according to the probability distribution,and has poor performance in multi-classification,so SVM,which has better classification performance,is introduced as the classifier.Finally,FWA is introduced to optimize the parameters of SVM,and intelligent fault diagnosis for on-board equipment of train control system is realized.Taking the AElog file of a railway bureau for two years as the data source,the fault diagnosis model of this thesis is compared with other diagnosis models,and all the evaluation indexes are improved.It shows that the model has a wide application prospect in the field of fault diagnosis for on-board equipment of train control system.