Application of improved HMM model in train location unit

Train Location Unit plays a crucial role in train control system, whose reliability is directly related to the safe operation of trains. There are many train positioning methods in public mainly relates to GPS/ Odometer(ODO), which is the direction of the actual demand. In this paper, a fault diagnosis method based on improved HMM(Hidden Markov Model) is studied which the genetic algorithm(GA) is used to obtain the parameters of HMM to make the training speed reach the steady state quickly and get more higher training accuracy, rather than B-W algorithm. The ODO will be taken as an example to verify the validity of the proposed method. There are three statuses of wheels in the train running process. To implement the detecting procedure of fault state for train's wheels, firstly extracting the feature from Odometer data during the running course of train. The feature will be preprocessed with algorithm of amplitude normalized and scalar quantization, then establishing the hidden fault states of ODO model by training data. Finally the ODO fault condition classifier is set up. By inputting the observation sequence which matches with improved HMM models classifier, the status of ODO is obtained effectively. Experimental results show that the diagnosis model built with proposed method can provide high diagnosis performance for different fault types. The diagnostic accuracy is up to 100% between normal and fault conditions and the overall diagnostic accuracy is up to 96.67%.