Deep Learning Based Diagnosis Method For Intelligent Computer-based Interlocking Maintenance System

Computer-based interlocking system(CBI)is a typical safety critical system for railway signaling,which is responsible for train route control and station operation safety protection.The current intelligent level of CBI operation and maintenance is low,which mainly relies on manual experience.This method is not able to meet the large-scale fault diagnosis requirements,and is likely to come across the problems of incomplete diagnosis and diagnosis errors caused by insufficient experience.The development of artificial intelligence technology and the massive data generated in the interlocking operation process bring opportunities to the intelligent operation and maintenance of the interlocking system.According to the characteristics of the interlocking system,a fault diagnosis method based on deep learning for intelligent operation and maintenance is proposed in this paper,and a CBI intelligent operation and maintenance support system is designed and implemented,which can effectively improve the automatic level of fault diagnosis.The main work of this thesis are as follows:(1)A tree data structure is proposed to describe the station topology.Considering that the interlocking logic depends on the station topology,a tree structure is proposed to describe the connected relations between signal equipments in station,so as to solve the problem that the interlocking data dimension is too large in the deep learning model.(2)A CBI logic fault diagnosis model based on tree structure is proposed.Five kinds of neural networks are established to model a large-scale actual station,and a comparative experiment is carried out.The experimental results show that the proposed tree structure model improves the accuracy of fault diagnosis by about 30%.(3)A deep learning model oriented to the interlocking temporal logic is proposed.The established tree-based neural networks are combined with the recurrent neural network and its variants respectively,and the fault diagnosis models for CBI control logic are formed.At the same time,the fault diagnosis models combined with temporal logic are compared with the models based only on tree structure,and the experimental results show that the accuracy of the model combined with temporal information is90.26% in the binary classification task and 88.00% in the multi-classification task,which proves its effectiveness.(4)Based on the proposed method,a CBI intelligent operation and maintenance support system based on deep learning is designed and implemented.The system takes the fault diagnosis model as the core and provides the visual interface and related functions of operation and maintenance for staffs,including station status display,interlocking fault diagnosis,equipment status query and evaluation.This paper focuses on the characteristics of interlocking system and integrates deep learning technology to provide a new idea and solution for fault diagnosis of interlocking logic,which has a certain reference significance for the CBI intelligent operation and maintenance.