Research On Intelligent Fault Diagnosis Method Of Urban Rail Signaling Equipment System Based On Fuzzy Logic

The signaling equipment system of urban rail transit is one of the pivotal infrastructures of urban rail transit to realize safe operation,and undertakes the vital tasks of supervising,controlling,and coordinating the safe and efficient train operation in real-time.This specific role in ensuring the safety and operational efficiency of urban rail transit also puts forward strict requirements in the reliability and maintenance of signaling equipment system itself.The signaling equipment system of urban rail transit consists of numerous control and basic equipment of the on-board and ground parts.It is inevitable that some equipment or the whole system will fail,due to equipment aging,service life expiration,external environment erosion and other factors.Fault diagnosis,as an important way of detecting,isolating,and identifying faults,has always been the core content of operation and maintenance of urban rail transit signaling equipment system.In order to meet the current development needs of intelligent operation and maintenance,the fault diagnosis of signaling equipment system of urban rail transit is also moving towards automation and intelligence.In recent years,the research of intelligent fault diagnosis based on machine learning and deep learning has made remarkable achievements in fault detection as well as its classification and recognition.However,most of the existing studies only realize the simple judgment of whether there is a fault in the equipment system and classify the specific fault types,and they are incapable of analyzing and explaining the fault causes and relation autonomously.Additionally,existing studies fail to fully consider the effects of various cognitive and information uncertainties on fault diagnosis.By utilizing the significant advantages of fuzzy logic theory in model interpretability and accurately describing the uncertain information,combining machine learning and data mining techniques,this dissertation studies the intelligent fault diagnosis method that can autonomously learn and explain the fault relationship and causes of urban rail signaling equipment system,so as to realize fault reasoning and decision-making that fully consider the uncertain characteristics of diagnosis information.Specifically,the research is carried out from the following four aspects: failure mode and effects analysis,fault recognition and classification,fault causality mining,and fault reasoning and tracing.The main research work of the dissertation is as follows:(1)Research on failure mode and effects analysis considering interrelationship among risk factors under fuzzy linguistic environment.Aiming at the problem that the uncertain semantic evaluation information cannot be expressed precisely in the process of failure mode and effects analysis(FMEA)of signaling equipment system,and the risk assessment decision of failure mode fails to consider risk factor correlations leading to the results with the lack of objectivity,a fuzzy linguistic multi-attribute decision making method is proposed to realize the failure mode and effects analysis of signaling equipment system with reliable evaluation decision.This method accomplishes the accurate expression to the uncertainty of failure mode linguistic evaluation information by developing the q-rung orthopair fuzzy uncertain linguistic sets(Q-ROFULS).Furthermore,the partitioned Maclaurin symmetric mean(PMSM)operators are proposed to realize the evaluation and decision-making of failure modes considering interrelationships among risk factors.Through the experimental verification and comparison of turnout switch machine in signaling equipment system,the results show that the method proposed in this dissertation can provide more accurate and objective risk impact assessment results of failure mode.(2)Research on fault classification of signaling equipment system jointly considering fault classification accuracy and result interpretability.Aiming to the problem that most of intelligent fault diagnosis based on machine learning and deep learning only shows good performance on the dataset and their learning process and diagnosis results are lack of intuitive explanation,a broad learning-based dynamic neuro fuzzy systems(BL-DNFS)is proposed to realize fault classification of signaling equipment system with interpretable diagnosis results.This method can present the fault classification results as intuitive and readable fuzzy rules through BL-DNFS.Furthermore,the dynamic incremental learning algorithm(DIL)oriented to BL-DNFS is developed,which enables BL-DNFS to achieve a good classification performance using the most compact model structure.The comparison and verification of method are conducted through the dataset experiment of CI route cancellation fault of urban rail transit signaling system,the results show that BL-DNFS has a higher average diagnosis accuracy than SVM,RFA,ELM,OS-FELM,and FELM,meanwhile,they enable an intuitive explanation for fault classification results in the form of fuzzy rules,so as to realize the integration of fault classification and result interpretation.(3)Research on fault cause relationship mining of signaling equipment system with logical interpretability.In view of the problem that urban rail signaling equipment system faults involve multi-source and complex causes,and existing research is difficult to mine and analyze the fault causes and relations autonomously,a mining method of fault cause relationship for urban rail signaling equipment system is proposed based on associative rule analysis.The feature reduction strategy based on information gains(IG)is firstly proposed to realize the extraction of fault features with high information content in operation and maintenance data.Then,a fault cause mining algorithm based on association rule classification(ARC)is proposed,which can excavate fault causes from the perspective of logical correlation,so as to extract association rules that clearly express the causal relationship of faults.Finally,the Apriori-based feature relationship mining algorithm is proposed to further accomplish the correlation and causality mining between data features.Based on the data experiment of automatic turn-back fault of train in signaling equipment system,it is verified that the method proposed in this dissertation can autonomously mine and integrate the potential correlation between equipment faults in the system,and form fault association rules with diagnosis accuracy of more than 97% and intuitive expression of fault causality.They provide decision support for the multi-stage intelligent implementation integrating fault detection,isolation,and location.(4)Research on fault reasoning of signaling equipment system considering diagnosis uncertainty.In order to accomplish the accurate inference of equipment fault severity and fault cause possibility in the signaling equipment system,this dissertation focuses on the problem that existing fault reasoning methods cannot accurately express the uncertainty of reasoning process and fail to realize backward tracing reasoning,and first propose a q-rung orthoapir fuzzy Petri nets(Q-ROFPN)model to realize fault reasoning of urban rail signaling equipment system considering uncertainty.An intuitive expression of fault correlation,propagation mode,and fault state uncertainty of each equipment in signaling equipment system is accomplished by applying Q-ROFPN and their corresponding q-rung orthoapir fuzzy reversed Petri nets(Q-ROFRPN).The forward fault propagation reasoning algorithm and reverse fault tracing reasoning algorithm are proposed respectively,which ensures reliable accomplishment of inferring equipment fault severity and fault cause possibility.Based on the case experiment of automatic train turn-back fault,it is verified that the method proposed in this dissertation can infer the fault degree and fault cause of signaling equipment system from two dimensions of membership and non-membership degrees and provide an accurate and reliable reasoning results,which accomplish the intelligent fault reasoning that simulates human perception and recognizes fuzzy and uncertain information.