| With the continuous acceleration of urbanization,subway,as a convenient,environmentally friendly and high-volume transportation mode,is the first choice for most citizens to commute.The safe operation of subway is becoming more and more important for citizens’ daily life.Signal system is the key system of urban rail transit,and its stable operation is closely related to the safe operation of trains.Therefore,ensuring the stable operation of signal system equipment is an important prerequisite for the realization of operation safety.If the signal system fails,the fault location can be quickly located,which is conducive to the orderly recovery of the faulty equipment and ensure the safe and stable operation of the train.At present,the text data generated in the process of system overhaul is not fully utilized,and the hidden information is not discovered and utilized.Therefore,this paper mainly takes the text data generated in the process of signal system maintenance as the research object,and realizes the fault diagnosis of signal system through text mining technology.Based on the analysis of the fault text data characteristics of the signal system,this paper builds a classification model conforming to the fault text characteristics.Firstly,text data is vectorized by word embedding,and a text classifier is constructed by using bidirectional long-and short-term neural network.Finally,a new fault diagnosis system for signal system is proposed.The research mainly includes the following contents:First of all,the automatic operation system components and equipment management methods,maintenance means are analyzed to understand the common equipment fault types and causes,equipment management deficiencies and signal system maintenance data content characteristics,laying a foundation for the construction of equipment fault diagnosis model.BiLSTM model was used to extract semantic data and generate text vector,and Adam optimizer was used to optimize BiLSTM model.At the same time,in the process of selecting the downstream task network structure,the performance of the models under different hidden layer structures is compared in the experiment,and the most suitable network structure proposed in this paper is obtained.In addition,in order to verify the excellent performance and effectiveness of this model,the results of BiLSTM model and other models are compared and analyzed experimentally.Experiments show that the average diagnostic accuracy of the Bil STM-based fault diagnosis model is 3 to 17 percentage points higher than that of other models.Finally,on the basis of the model established above,the construction of the fault diagnosis system of the signal system is completed,and the practical data is used to verify the feasibility of the system. |
