Application And Research Of Grey Neural Network In Fault Diagnosis System Of S700K Switch Machine

With the rapid development of railway in speed and load, S700 K electric switch machine puts into use in great quantities, which requires a more complete mechanism to monitor running state and intelligent fault diagnosis methods to improve equipment safety and reliability. For a long time, the fault diagnosis and positioning analysis of switch machine have been relied on professional knowledge and experiences of maintenance staffs. The information of the microcomputer monitoring system only serves as a basis for the artificial fault judgment, which takes long time, and sometimes makes misjudgment and mistakes.Based on the analysis and research of domestic and foreign electric switch machine fault diagnosis methods, the dissertation aims at problems of relatively backward fault identification means to S700 K electric switch machine and limitations existing in the pre-existing fault diagnosis methods. Based on S700 K electric switch machine operating power curves acquired by the microcomputer monitoring system, combined the grey correlation analysis method with neural network modeling, the grey neural network is designed to realize fault diagnosis of S700 K electric switch machine.Firstly, based on the basic structure and working process of S700 K electric switch machine, the dissertation analyses the failure mechanism and main performance parameters and researches the relationship between the main performance parameters. It is concluded that the power value of S700 K electric switch machine can be used as feature parameters of fault diagnosis. According to the acquisition principle of S700 K electric switch machine power curves, the process when S700 K electric switch machine completes a conversion is divided into five stages. The dissertation respectively analyses the parameter information and fault information of S700 K electric switch machine under the eight kinds of common faults.Secondly, the framework of switch machine fault diagnosis is given. Based on the grey neural network diagnosis flow, the grey neural network is applied to fault diagnosis of S700 K electric switch machine. Based on grey system and neural network analysis method, the grey correlation analysis is combined with neural network modeling method to establish grey neural network. According to the time series data processing of grey correlation analysis, the fault feature set is established to calculate grey correlation degrees between new power curve and fault feature set by neural network. Based on adoption principles of resolution coefficient value and fault recognition rate under different resolution coefficient values, the optimal resolution coefficient value is selected to use in diagnosis system. The S700 K electric switch machine fault power curve of a signalling depot is applied to verify functionality and performance of fault diagnosis system. The result is consistent with inspection result. It indicates that the method above-mentioned is available.Finally, the design of the diagnosis system is finished. The dissertation builds fault diagnosis system of S700 K electric switch machine based on the grey neural network, using Access 2003 to establish background database, using Microsoft Visual C++6.0 to realize the fault diagnosis of the whole system, including the management of knowledge base, network designing, fault diagnostic reasoning and human-computer interface, etc. Using multiple sets of test sample to finish verification, diagnosis results show that the system can effectively and quickly implement fault diagnosis of S700 K electric switch machine.