Research of fault-characteristic extractive technology based on particle swarm optimization

In the work process of gearbox, because the responding signal is very complex, it is difficult to extract its sensitive fault attributive information. The sensitivity of the fault degree, fault position and fault type is very different, so the characteristic parameter set constructed by the traditional characteristic extraction and analysis method is voluminous. Therefore, how to define the reliable and effective fault characteristic parameter set and how to optimize the parameter set by the sensitive degree are the await solved problems to realize real time and online fault diagnosis. In this paper, the characteristic extractive method base on particle swarm optimization (PSO) is presented for the problem of gearbox failure characteristic selection. Then the technology is applied to analyze and process the vibration responding signal of gearbox, extract and optimize the fault characteristic parameter set. Finally the parameter set nearly related to the gearbox's fault is constructed and it is used to the fault diagnosis. It proves validity of the diagnosis result that PSO algorithm has good effectiveness, higher diagnosis precision and fast optimal speed than the traditional genetic algorithm, The experimental result indicates that the wavelet neural network training method based on the PSO algorithm is an effective training algorithm, and meanwhile it is also an available approach to solve fault diagnosis problems.