Information Fusion Method For Fault Diagnosis Based On Reliability Analysis Under Epistemic Uncertainty

The performance of train-ground wireless communication system has been greatly improved with the widely application of fault-tolerant technology.On the other hand,it significantly increases the complexity of the system structure and causes the lack of sufficient fault data and fault correlation,which significantly raised challenges in system diagnosis.Aiming at the unique fault characteristics of train-ground wireless communication system,it is of great significance to establish a fault diagnosis model and design a dynamic diagnosis algorithm based on the multi-source heterogeneous information to repair the faults quickly with the objective of fast and low-cost diagnosis.Firstly,aiming at the problem of fault correlation caused by the redundancy technologies used in train-ground wireless communication system,a dynamic fault tree is used to establish the fault model.In view of the epistemic uncertainty,the interval numbers are used to describe the fault distribution parameters of the basic events.Aiming at the shortage of traditional methods,a dynamic fault tree is mapped into a dynamic evidential network calculate some reliability parameters of train-ground wireless communication system.Secondly,a new fault diagnosis method is proposed to improve the diagnosis efficiency based on the importance of components and cut sets under epistemic uncertainty.In addition,a dynamic diagnostic decision-making algorithm is designed base on the importance,risk achievement worth,test cost and previous diagnosis result.Specifically,it includes the evaluation method of test cost,the construction of heterogeneous multi-attribute decision table and normalization method to avoid ranking confusion,definition of the generalized distance aggregation function,attribute weight determination method based on information entropy and distance-based VIKOR algorithm.Case study demonstrates the feasibility and efficiency of this method.Finally,on the basis of the above research,sensor information is further fused to optimize the system diagnosis process.A diagnosis sensor model is constructed basedon a dynamic evidence network.Furthermore,an optimal sensor placement method based on the component's importance is proposed.In addition,reliability parameters are updated according to the sensor information and previous diagnosis results.A novel dynamic diagnosis algorithm is used to obtain the optimal diagnostic sequence,which can improve the diagnosis efficiency of train-ground wireless communication system.