Study On The Wheelset Damage Identification Based On Neural Network

The research purpose of damage identification of structure is to ensure the reliability and security of engineering structure and supply references to life estimation and maintenance of engineering structures. This is a proceeding subject always concerned by engineering area. A lot of recent papers this area and different methods of damage identification have been summed up in this paper. But up to today, there is no satisfying one.Firstly, use the theory of vibration and modal analysis, and make use of the general finite element sofware of ANSYS to simulate the crack of the wheelsets.Secondly, based on the results and analyse the data. Finally gain the sample data needed in the paper.The principle structural damage detection using artificial neural networks (ANN) is expounded in the paper. Then, the development of ANN technique in application of structural detection is reviewed , some main achievements summarized, and the main obstacles encountered are raised. Secondly, the paper is verified theoretically that the natural frequency changes of wheelsets contain the information such as location and degree of the damage. Finally, based on this theory, use proper computer program to simulation different damage of wheelsets, extract the improved BP algorithm network, detect the structural damage by the trained networks.The result shows the effectiveness of this method. In the end, through a simple structural damage identification experiment, the above conclusion is farther verified.