| As urban rail transit system develops,more and more cities use FAO(Fully Automatic Operation)systems.The FAO system focus on train operation and integrates several systems,such as signal,vehicle system,comprehensive monitoring system and communication system.It improves the automation level and system performance,reduces the labor intensity of operators,but at the same time it also increases the difficulty of operation management especially under the condition of fault degradation.As there is no driver on train,it is difficult to intuitively perceive the operating environment and equipment status by remote dispatchers.And,FAO as a whole is too complex to instantly locate faults during operation.The uncertainty of operating risk is dramatically increased.Therefore,it is significant to dynamically assess risk situation of train operation and provide decision support for dispatchers to truly realize unmanned driving,based on the existing equipment status of the FAO system.Combining the equipment status,train operation risk situation in specific scenarios has been assessed in this dissertation.To achieve this assessment,the knowledge-based method is used to describe the construction of train accident model in the specific scenario,which can be transformed into a Bayesian network for probabilistic reasoning.The main contributions of this dissertation are as follows:(1)Early warning signs for train operation in specific scenarios are identified based on the equipment status,and the basic control structure is extended based on the early warning signals to generate the possible hazard causation scenarios during the operation.EWa SAP(Early Warning Sign Analysis based on the STPA)is applied to identify early warning signals,determine the relationship between early warning signals and accident precursors,and then establish fault propagation relationships to generate hazard-causing scenarios.(2)The formal description of fault propagation relations is completed by the constructed ontology model and formal language.Ontology is used to define classes,the properties of classes,and individuals in specific scenario.The formal language OWL describes the interaction behavior between subsystems and the parameters of each scenario.(3)Probabilistic expansions are given for the representation of uncertainties in hazard causation scenarios.Then,the ontology model is transformed to a Bayesian network and the probability of occurrence of the hazard is inferred.According to the above method,an ontology instance model is constructed and converted into a Bayesian network,and the posterior probability inference of the transformed Bayesian network is performed using Netica,a visualization tool about Bayesian network.(4)A situation assessment index for operational risk is proposed and a calculation method is given.The vulnerability level of safety barriers is proposed for train operation,which is used as a risk factor that is not caused by equipment failure in the risk situation assessment.(5)The dissertation uses the proposed method to evaluate the situation of real accident cases.Through the establishment of risk-causing scenarios and risk situation assessment models,the operational risk situation of trains in specific scenarios is obtained.If dispatchers can perceive the risk situation before an accident occurs,the risk will be reduced.The results of the case study show that the method proposed in this paper provides a reference for solving the problem of unknown risk posture during train operation,and can be used as a theoretical basis for dispatchers’ decision making.Through the analysis of real accident case study,it can be seen that the results of its situation assessment provides a reference for solving the problem of unknown risk situation during train operation,and can also be used as a theoretical basis for dispatchers’ decision making.This dissertation contains 49 figures,17 tables and 78 references. |
