Research On Fault Prediction Of On-board ATP Based On Bayesian Network Model

With the rapid development of urban railay transportation,people's safety requirements for railway transportation are getting higher and higher.ATP(Automatic Train Protection)system is a subsystem of train operation control system.It mainly plays the more and more important role of train overspeed protection.Fault prediction as an advanced and efficient technology,the construction of a reasonable and applicable PHM algorithm model can play an important role in the prediction of the fault of the on-board ATP system,so this paper mainly made the following the study:Firstly,Fault prediction technology is divided into four kinds of technology models: failure physical model,data-driven model,traditional reliability model and statistical probability model from three perspectives: accuracy of prediction,extensiveness of application and amount of investment.Secondly,this paper introduces and analyses the characteristics and transformation principles and methods of on-board ATP,and completes the structure learning of Bayesian network model.Last,according to the constructed fault tree model and transformation principle and method,seven fault types,18 fault modes and on-board ATP system equipment failure rate are constructed as nodes,intermediate nodes and leaf nodes to construct a Bayesian network model,Using the Noisy-OR model to calculate the conditional probability of each intermediate node and leaf node to complete the parameter learning of the Bayesian network based on the fault data provided by a subway station is used as the prior probability of each root node.Testing and verifing the rationality of the model calculation by Netica software.Taking advantage of the Bayesian network predictive reasoning,calculating the posterior probability of each node pair under the premise of on-board ATP system equipment failure,finding the on-board ATP system which is easier to occur fault through the percentage change of prior probability,it is easy to get in the on-board ATP system Among them,the percentage change of "radar fault","sensor fault","BSA fault" and "MT communication fault" exceeds 4%,and put forward suggestion to corresponding modules according to the type of fault belongs to.