Autonomous Perception Based Train Integrity Detection Method

As the ’large artery’ of national economic development, railway transportation is currently playing an important role in China’s transport, with all-weather, large capacity,safe and on time, low unit capacity price and energy consumption and a lot of advantage.In order to ensure the safe operation of trains traveling, all the train vechiles must be guaranteed to be connected by couplers in physics. Train integrity detection is to check the integrity state of the train. Train Integrity Monitoring System as an important part of the Train Control System, the way of detection is changing from ground equipment based methods to on-board equipment based methods. How to detect train integrity by on-board equipment is also a problem to be solved at present, which also meets the next generation of train control system and the train control system for low density railwayline in central and western China.To sovle the train integrity detection problem, we first model the train integrity detection requirements in different application background of the train integrity monitoring system. A digital track map aided train state estimate method is developed for the train state estimation problem of head-of-train and end-of-train by autonomous perception. Based on head-of-train and end-of-train relative autonomous perception information, hybrid bayesian learning method is proposed for train integrity detection model parameters learning and train integrity state determination. Then the safety of train integrity detection using autonomous perception information is quantitatively analyzed base on the uncertainty probability model checking method. Based on the research work,the following innovation results have been achieved:Firstly, the autonomous perception based train integrity detection requirements model is presented according to different application background. Timming analysis is applied in timeliness calculation model base on train safe approach time. Train integrity detection risk distribution model is used depend on failure risk of train integrity monitoring system and its safety risk reduction needs. The results established the demand of detection timeliness and safety risk for this dissertation.Secondly, the digital track map aided train state estimation algorithm is proposed.We established digital track map information aided train state probability estimation model at first. According to the operational characteristics of the train track constraint state estimation model decomposition, then optimal train autonomous positioning state estimation method is raised under the framework of expectation propogation using the idea of state filtering. The result shows it can further improve the train state estimation accuracy and provide a more accurate decision-making for the train integrity train autonomous perception information independently.Thirdly, the train integrity decision determination method is proposed based on hybrid bayesian learning method. Relative head-of-train and end-of-train motion state hybrid dynamic linear system model is bulted up. We use hybrid bayesian learning ideas respectively for system model parameters learning, and calculation of train integrity the state by the a posteriori probability maximum likelihood solution to realize train integrity status of accurate decisions. The false detection rate in the train integrity detection method is reduced by the uncertainty noise in the actual operating environment of the train.Lastly, model cheching method for train integrity detection with uncertainty probability model is raised. In view of the uncertainty factors in autonomous perception information, we presented model checking method for uncertainty probability according to the property of train integrity monitoring system components and the function of the attribute security features analysis and verification. It solved the problem of model checking with random failure under the condition of system operation in system development stage.This paper used the train actual operation data and experiments data analysis based verification procedure simulation data to verify the validity of the proposed model algorithm. The results can provide a theoretical reference for the research and design of on-board equipment based train integrity monitoring system for the next generation of train control system and the train control sytem for low density railway line in central and western China.