Fault Diagnosis Of Gear-box Based On GA-BP Algorithm Model

With the development of society, the drive undertake an important mission as part of mechanical equipment, Gear failure will not only damage the gear itself, but also endanger the worker ,that may result in huge economic losses and social impacts. Therefore, the gear drive system condition monitoring and fault diagnosis technology has great significance, that can reduce maintenance time and costs, while improve the efficience.Genetic algorithms is a mechanism developed stochastic global search optimization method, which mimic natural biological evolution imitated Darwin's theory of evolution and genetics of Montesquieu said. It is an efficient, parallel and global search methods, it can automatically access and accumulate knowledge about the search space, and adaptively control the search process and get the optimal solution, not only has better global Convergence ability and convergence speed, high efficiency, but also can overcome the inherent problem of local minimum. we can make full use of the nonlinear approximation ability of neural networks and Genetic algorithms , in order to complement each other, while avoiding their shortcomings .In this paper, through studing the structure and the work process of the gear-box,we constructed a 13-4-5 type of genetic algorithm neural network fault diagnosis system gear-box, we Installed the selected sensor on the selected measuring point, in order to measure the vibration signals of gear-box, denoised in order to reflect the status of operation of gear-box, Through a series of training and testing and compared with traditional genetic algorithm and the BP neural network, Improved genetic algorithm need 18 and 44, while traditional genetic algorithm need 71, the training epochs are increased by 74.6% and 38.0%,Genetic algorithms can effectively and reliably be used in the fault diagnosis by the experiments and simulations.