The Design And Application Of The State Identification System Of High Speed Train

In recent years, with the rapidly development of the high-speed train of China, the increasing of operating line and speed, the running process of the trains is more sensitive, and the components of the train are more easily to wear. When the trains run at high speed, any small changes will bring a great risk. In order to ensure the safety of the high-speed train, it’s essential to use effective and timely methods to recognize their running states. The running gear is a key component of the train, it has a directly impact on the trains’ safety and traction, so it’s selected as the research object. The vibration signal is used to analyze the states of the high-speed train.This thesis focuses on the study of the steps and methods of high-speed train state recognition. The main methods include data file management, extraction of abnormal data, data pre-processing, data feature analysis, signal feature extraction, states’ recognition and decision-making, and so on. The methods to detect the abnormal data section and outlier are proposed.According to the main methods and steps of state recognition, the high-speed trains state identification system is designed and implemented. The system includes mainstream algorithms for each step of state recognition, namely data file management, judgment and extraction of abnormal data, data pro-treatment, data feature analysis, signal feature extraction, and state recognition. It can be used to analyze a variety of signals. The system also provides automatic recognition function, and reserve interfaces to add new algorithms. This measure improves the scalability of the system.In the aspect of the treatment for the big data file, the functions of abstracting abnormal section and detecting outliers are set in the abnormal detection module. The abnormal data section can be abstracted by using the method of distance. Bidirectional sliding window with multi-quartile-range (MQR) is used to detect the outliers which caused by the accidental error, and the method of adjacent point interpolation is used to correct the outliers.The application of the high-speed train’s state recognition system is researched. Signals in the first 32 channels are taken as the treatment objects. The time-domain features (like period, stationary, linearity) of the signal in each channel are satisfied. SVM is used to classify and recognize the signal, based on 4 abstracting features of time domain statistics, power spectrum, energy entropy and the singular value. The difference of the recognition rate due to the changes in speed, state, eigenvalue, and channel is compared. The sensibility of the channel for the faults and the match between the channel s and the methods are summarized.The comparative results of the real test data and simulation data show that the two kinds of data have some similarities, the experimental results of simulation data can be applied to the real test data. But there are still some differences between them, so they can’t be treated as the same.