Fault Diagnosis And Safety Assessment Of Urban Bus Operation

The development of public transport systems is closely linked to safety,and the development of public transport systems is also related to the overall improvement of urban transport networks.Therefore,the diagnosis of faults in public transport vehicles and the proposal of optimisation solutions have become a hot topic of research for scholars.However,most of the current fault diagnosis solutions for public transport vehicles are based on highly subjective methods such as expert systems,or only on one component.At the same time,data-driven diagnosis is rapidly emerging in the field of fault diagnosis and has been applied in many areas with more satisfactory results.Therefore,this paper applies the data-driven fault diagnosis method to bus systems to complete the fault diagnosis of buses.In terms of optimisation,a more scientific and comprehensive evaluation method is used to calculate the safety scores of various types of faults and to propose corresponding optimisation solutions and countermeasures.The key elements of this paper can be summarised as follows.(1)To address the problem of local online modelling for bus fault diagnosis,a fault diagnosis method(RVM-LL)combining an improved Lazy Learning(LL)method and a Relevance Vector Machine(RVM)is proposed.RVM-LL uses a modified distance function to select local nearest neighbours,then uses RVM to build a local model and finally uses a cross-validation method to select the optimal local model to diagnose the fault.This method does not require an exact mathematical model,and it also does not require a lot of time for training and is suitable for small samples.The data collected from the physical bus system is used as input,and the simulation software is used to conduct simulation experiments and set up comparison experiments to verify its validity.(2)To address the global offline modelling problem of bus fault diagnosis,this paper combines the Improved Genetic Algorithm(IGA)with Backpropagation Neural Network(BPNN)to build the IGA-BPNN model.The BP neural network parameters were initially optimised using the IGA to determine the initial optimal parameters of the network,and then the BP neural network was optimised again using the advantages of the gradient learning algorithm,and the optimised network was used to complete the fault diagnosis of the bus vehicle.Finally,simulation experiments were completed in the simulation software using the actual collected data,and several sets of comparison experiments were conducted to test the fault diagnosis effect of the IGA-BPNN.(3)A comprehensive evaluation method is proposed for the bus safety scoring problem.Firstly,a comprehensive assignment method(PC-EVM)combining the Preferential Chart(PC)and the Entropy Value Method(EVM)is proposed to obtain the comprehensive weights of the bus safety indicators,and the safety evaluation scores are obtained using the Fuzzy Comprehensive Evaluation(FCE)and ImportancePerformance Analysis(IPA)methods,and corresponding optimization countermeasures are proposed.