Research On Fault Diagnosis Method Of RDC In CTCS-1 Train Control System

The Chinese Train Control System level 1(CTCS-1)is designed for new and upgrade railways with speed below 200km/h,which adopts distance-to-go mode to supervise train operation.As core ground equipment in CTCS-1,the Regional Data Center(RDC)is aimed to generate and send train movement data in train control system.The failure of RDC will affect the efficiency and safety.The RDC fault diagnosis method proposed in this thesis can classify RDC faults quickly and accurately,thus it is able to provide maintenance suggestions and further improve maintenance efficiency.As a RDC fault may belong to different kinds of fault types and vice versa,it is difficult to obtain the corresponding fault information accurately and quickly.Therefore,the uncertain relationship becomes one of the major factors affecting the maintenance work.In order to realize the accurate classification of RDC faults,a multi-layered fault diagnosis method based on case based reasoning and Bayesian network was proposed in this thesis.In this method,faults are classified into accurate and deterministic fault type by analyzing historical fault data and equipment fault attributes.In addition,this fault diagnosis method can provide corresponding fault maintain strategy,reduce maintain time and increase equipment utilization and availability.The main works of this thesis include:(1)Based on the analysis of RDC composition and function,the failure modes and effect analysis(FMEA)of RDC are obtained.There are four types of RDC faults including power failure fault,external channel fault,hardware fault and software fault.At the same time,the RDC historical fault cases are clustered to construct RDC fault case library.(2)The RDC shallow fault diagnosis model based on case based reasoning and Bag of Words Model(BoW)was proposed.The word-bag text representation model fused vector space was established for fault cases.The similarity calculation method integrating cosine function algorithm and Euclidean distance was designed to calculate the optimal matching fault type of the new fault in the case library.The experimental results show that the matching results based on case based reasoning and BoW fault diagnosis model are better than those of the traditional case based reasoning,and the former can provide more accurate fault types.(3)The RDC deep fault diagnosis model based on Rough Set theory and Bayesian network was proposed.The detailed types,attributes and attribute features of RDC faults were analyzed.According to the analysis,the fault attribute features were used as the condition attributes,and the fault type was used as the decision attribute to establishe the minimum decision table via rough set theory.The Bayesian network model with fault types and fault attribute features as random variables was established to calculate the the posterior probability and further diagnose the fault.The RDC deep fault diagnosis model can not only provide a diagnosis result,but also offer the probability of different fault types to provide a maintenance reference.(4)The RDC fault diagnosis system of CTCS-1 based on Microsoft Visual Studio 2013 platform is designed and implemented.Additionally,C#,Python and Matlab are used to realize the accurate diagnosis of RDC and provide maintenance solutions.There are 53 graphs,21 tables and 64 references in this thesis.