Service Status Identification And System Reliability Evaluation Method Based On Tensor Domain Theory For Rail Train With Monitoring Data

Safety is the core competitiveness and permanent goal of railway.The lack of technical safety assurance would be devastating blow to the railway industry,and lead to serious social issues.With the rapid development of railway technology,we have to face the complex technical composition of equipment,high-intensity operating conditions,nonlinear accelerated degradation of working state,and multiple uncertainties.Some scientific problems should be solved with multi-source information to satisfy the urgent industry demand:how to automaticly identify the service status of equipment,how to conduct the operating reliability assessment in both component level and system level,in addition,different from systematic maintenance strategy for series or parallel system,a network maintenance strategy model should be made so as to satisfy Complex networked system.1.Aiming at dealing with the multi-source heterogeneous data,a state evaluation theory based on tensor domain was proposed.The basic concept and content of the tensor domain were elaborated,and two key technical problems were put forward.An Integrating synthetic minority oversampling and gradient boosting decision tree for bogie fault diagnosis of rail vehicle was proposed for technical realization.2.A new way is proposed to deal with multi-sensor monitoring data,automatic feature learning and classification without human intervention by taking advantages of tensor representation and convolutional neural network(CNN).The study is in the full sense of learning from raw data,without any manual features and with respect to multi-dimensional data.It shows a good adaptability and high efficiency under various working condition by taking full use of the multi-sensor data,and has powerful ability in accuracy and convergence speed.Moreover,it is not as sensitive to data quantity as other deep learning algorithms do.Such superior characteristic made the model more suitable for practical application,because of the insufficient failure data.At last,it is an intelligent End-to-End model,performing automatic fault diagnosis without manual intervention.3.A reliability analysis model in component level based on condition monitoring information is conducted based on fuzzy security domain and time-varying semi-markov chain.Safety region theory is extended from two-state into multi-state with automatic division.It can not only effectively solves the irregular selection problem of the number of state in markov process,but also provides a scientific basis for the division of component performance in different degradation stage.The markov process based on time-varying transition probability matrix can precisely describe the variation trend of the component characteristics with the increase of time,and provides a necessary support for a more accurate component state reliability evaluation and remaining useful life prediction.4.A multistate network flow model has been proposed with consideration of components degradation level and functional interaction between them.Flow rate of each arc depends on both component’ health status and the task it undertakes.The relative probability importance of each basic component and system reliability with and without forehead information are given.The results show that the network flow model provides a more useful,practical and effcient way to deal with the reliability evaluation for complex system,and works well on CRH3 bogie,and can support as guidance of bogie system design,daily system operation and predictive maintenance.5.With respect to the complex networked system,an improved spectral clustering algorithm is proposed to deal with the networked group maintenance strategy making problem based on state reliability,as most of the traditional systematic maintanence strategy making models can only worked on series-parallel system.It verified on CRH2 bogie system.Result shows that it can divids components into proper groups more scientifically and reasonably according to their parameters,and solve the problem of making maintenance strategy for complex networked systems.