Research On Automatic Fault Diagnosis System Of Subway On-board Signal Equipment

With the development of Chinese city,the urban mass transit in China has made great progress.Subway has become one of the main means of transportation for urban residents' daily travel.On-board signal equipment is one of core equipment of train,which determines the safety and efficiency of driving.However,because of long-term operation and aging of signal equipment,the on-board signal equipment inevitably fails.Based on this,the fault detection and handling of on-board signal equipment become highly important to subway operation.Nowadays,the maintenance of on-board signal equipment still limits in the state of off-line diagnosis and manual diagnosis.In order to improve the efficiency of fault handling of on-board signal equipment and reduce the possibility of metro operation delay,it is necessary to establish a set of real-time on-board signal equipment fault automatic diagnosis system.This paper presented a data collection method based on Windows message mechanism and MSSIM,an improved learning algorithm of Bayesian network structure and a four-layer fault diagnosis model.On this basis,the automatic fault diagnosis system of on-board signal equipment is realized.The main work of this paper is as follows:Automatic collection of fault data is the first step of automatic diagnosis,which also ensures the integrity and correctness of the data.First,the problem of exclusive use of serial port was solved and alternating collection of data between human and computer was successfully realized by expanding collection interface of LZB700 M train control system.The relevant messages were then sent by using Windows message mechanism to fault-receiving software provided by the manufacturer,which could realize automatic data collection.After this,MSSIM algorithm was employed to judge whether fault data collection was completed.Second,a four layer fault diagnosis model based on Bayesian network was proposed according to the causal relationship between fault and fault symptoms.It included equipment component layer,class B fault code layer,class A fault code layer and emergency braking layer.This model realized early warning by top-down reasoning and diagnosis by bottom-up reasoning.Moreover,according to the dynamic characteristics of fault diagnosis model,a high-performance structure learning algorithm called K2 Plus was proposed based on the K2 algorithm and characteristics of fault diagnosis model.The experimental results showed that K2 Plus algorithm could quickly establish the Bayesian network structure under the condition of high accuracy.Therefore,it successfully shortened the modeling time and laid the foundation of rapid modeling.At last,an automatic fault diagnosis system for on-board signal equipment was designed and implemented.The fault data was automatically collected by Windows message mechanism and uploaded to the server through FTP.The server used the four-layer fault diagnosis model established by MATLAB and FULLBNT tools to carry out fault diagnosis and early warning in time.Client interacted data with server through WCF.This system realized real-time automatic fault diagnosis of on-board signal equipment,improved the efficiency of signal equipment fault processing and brought great convenience for the maintenance of subway on-board signal equipment.