Research On Training Method Of High-speed Railway ATO On-board Subsystem Based On Human Reliability Analysis

The automatic train operation(ATO)system of high-speed railway in China is developing rapidly.With the increasing level of intelligence in equipment,human error has become the main cause of human-machine system failure.At the initial stage of the system put into use,it is of great significance to study effective training methods to solve the problem of human error,which meets the requirements of the whole cycle of safety critical system.In this thesis,the human reliability analysis(HRA)method of on-board ATO subsystem of high-speed railway is used as the theoretical basis to study the effectiveness and sequence optimization of the cases involved in the training content,and the training sequence is generated based on the human error probability.The main tasks are as follows:(1)A training case generation method based on colored Petri net(CPN)and systemtheoretic accident model and process(STAMP)is proposed.Based on the analysis of the characteristics of high-speed railway ATO scene and the relationship between them,and according to STAMP,the model of ATO scene is established by taking human behavior and system state as CPN transition and place respectively.Then the model is transformed into state space diagram,and the training case is obtained by the main path search algorithm and path splitting algorithm.(2)A human reliability analysis method based on mixed causal model is designed.Firstly,a three-layer mixed causal model is established to identify the basic unit of human behavior,the cognitive failure mode(CFM),and behavior formation factor.Then,the quantitative reasoning of human error probability is carried out: the problem of expert opinion fusion is solved by D-S evidence theory,the problem of data combination explosion is solved by fuzzy reasoning.The human error probability of typical CFM is obtained by success likelihood index method.Finally,the human error probability of training case is obtained by HRA event tree.(3)The optimization method of training sequence based on improved genetic algorithm is studied.The objective function of joint optimization is defined by three constraints: coverage,redundancy and sequence balance.The optimization method uses genetic algorithm to increase individual update steps to ensure sequence effectiveness and designs cross operation to avoid too many unsuitable genotypes at decision-making nodes.The optimal individual with termination of evolution is the required training sequence set after decoding.(4)The training cases and sequences of on-board ATO subsystem of high-speed railway are generated,and the architecture design and partial implementation of the training platform are completed.For ATO system,a total of 36 training cases are generated,and the human error probability of each training case is calculated.On this basis,10 training sequences are generated by sequence optimization method,and the sequence comparison before and after optimization is carried out.Finally,the architecture design of training platform and the visual operation of some training sequences are completed.There are 77 pictures,24 tables and 84 references.