A novel strategy for fault propagation study on complicated gear system based on network theory

This study aims at complicated gear systems whose fault propagation cannot be solved based on the conventional numerical calculation and simulation. The gear system is abstracted with network structure and the relationship between nodes is established. Firstly the vibration signal of gear with local fault is obtained base on FEA simulation, and the relationship between local fault and vibration signal feature is established after feature extraction by Wavelet Packet Analysis (WPA) and dimension compression by Principal Component Analysis (PCA). At the same time, a network structure for gear system is developed and divided into several elementary cell in line with the number of meshing gears. Transfer functions between nodes in each elementary cell are given by applying Radical Basis Function (RBF) Neutral Network, indicating the interrelationship between vibration signal features of meshing gears and the physical space-time fault propagation through system. At last a model evaluating fault condition of the whole gear system is constructed by Weighted Network modeling method. The strategy discussed above is applied examplarily to a practical case, identifying the effect of local fault on whole gear system.