Online Fault Detection And Diagnosis System For Urban Rail Train Gearbox

With the rapid development of economy and technology, the deepening of urbanization and the fast growing of urban population, the urban infrastructure construction, especially the construction of transport facilities and contradictory urban development is increasingly prominent. At present, China is rapidly building the urban rail transit, and is at the peak of the opening and operation. As the carrier of carrying passenger, the safety of urban rail vehicles directly relates to the life and property safety of passengers. Urban rail vehicle gearbox, as one of its key components, expanded effective online real-time monitoring and diagnosis, can not only avoid accidents, but also change the existing repair mechanisms: the condition based maintenance instead of fault maintenance and times maintenance. Thereby it reduces operating costs and improves the level of operation and maintenance.This paper studies urban rail vehicles gearbox, diagnosis and monitoring of the malfunction of urban rail vehicle gearbox, from the beginning of studying the vibration mechanism, in order to establish key steps of researching ideas such as signal processing and the amount of energy feature extraction.Firstly, aim at complex and hard working conditions of urban rail vehicles and low signal to noise ratio, we use wavelet packet analysis method to denoise, and verify the effect of eliminating noise through contrasting effect Chebyshev filter wavelet packet.Secondly, we use a multi-sensor fusion method to reduce partial disturbance of working transit vehicles, in order to increase accuracy of fault diagnosis. Through simulated signals and experimental data, this method can effectively reduce the signal mean square error, improve the sampling accuracy, and exclude the impact of individual faulty sensor or local interference caused by the fault diagnosis.Thirdly, aim at non-linear and non-stationary of vibration signals, we adopt empirical adaptive modal analysis method to extract power failure signal layers of intrinsic mode, using wavelet packet analysis breaks down signals to four layers of energy as the fault characteristic features for fault diagnosis, supplementing by the amount of time domain features, which achieves a good results.Finally, as the premise of accuracy and rapidity of fault diagnosis, we make use of the experimental data with the SA optimization training SVM, and verified by a test set of data, and achieve 98.75% diagnostic accuracy rate, higher than the traditional SVM. It verifies the feasibility that SA-SVM is applied in urban rail vehicle gearbox fault detection, and a detection and fault diagnosis system for urban rail vehicle gearbox based on SA-SVM method and data acquisition systems is established.