Research On Combined Diagnosis Model Of CNN And PSO-SVM For Fault Text Of On-board Equipment Of Train Control System

On-board equipment of train control system is a significant component of Chinese Train Control System,and rapid and precise location of its fault is a key to ensure safe train operation.On-board equipment generates a large amount of fault data during operation and maintenance,the description of the fault phenomenon is recorded in the form of unstructured text,and the fault diagnosis in the field work is mainly done manually,which may make the results arbitrary and inaccurate.Therefore,mining and analysis of fault text data is helpful for fault diagnosis research of on-board equipment of train control system.In this thesis,through the in-depth research and analysis of the fault text data of the most representative CTCS3-300 T on-board equipment in the CRH2 and CRH3 EMU trains,a fault diagnosis model for on-board equipment based on the combination of CNN and PSO-SVM is designed.The main research contents of the thesis are as follows:Firstly,the function and structure of on-board equipment of train control system are introduced in detail,the fault types are summarized,and the source of fault text data and its characteristics are analyzed.Secondly,the preprocessing work of fault text data of on-board equipment is completed by research on Chinese word segmentation technology.Then continuous bag-of-words model based on hierarchical softmax in Word2 vec is used to train word vectors,and each word after the word segmentation algorithm is converted into a low dimensional dense vector of the same dimension to realize the feature representation of fault text.Thirdly,for the combined diagnosis model,a CNN model for fault text classification of on-board equipment of train control system is designed first.The optimal hyperparameter combination is determined by training CNN,the trained CNN model is saved,and then it is used to perform feature extraction on the original input data.In view of the poor classification effect of the softmax classifier in CNN for small sample data,SVM with more stable classification effect is introduced to replace it.The feature vector extracted by CNN is used as the input of the SVM,and the PSO algorithm is introduced to optimize the parameters in SVM.The fault intelligent diagnosis of on-board equipment is realized by the construction of PSO-SVM classifier.According to the test analysis of the fault text data of on-board equipment recorded by a railway bureau and comparison with other models,the experimental results indicate that this thesis’ s model can obviously upgrade various evaluation indexes,among which precision rate and recall rate can reach 93.39% and 93.65%respectively,which prove that it can be used as an effective model for fault diagnosis of on-board equipment.Finally,on the basis of the designed model and experimental part,an auxiliary maintenance system of on-board equipment is constructed based on Py Qt5.Through the realization of the main function modules of the system,then the fault type identification of the fault text is completed and corresponding maintenance measures are provided,which can further prove the feasibility of the model in this thesis.The combined diagnosis model for on-board equipment and auxiliary maintenance system constructed in this thesis can effectively improve the efficiency and accuracy of fault diagnosis for on-board equipment,and realize the intelligent processing of fault text of on-board equipment on railway sites,which greatly improve complexity and inefficiency in manual text processing to a large extent.