Research of Gearbox Failure Diagnosis Based On Nerve Network

Traditional diagnosis only focuses on signal collection and processing, and lacks effective noise processing. In order to solve this problem, the thesis studies the diagnosis parameters in frequency plane and time plane after processing on the basis of wavelet noise reduction processing. At the same time, according to the typical fault characteristics of gearbox, the BP neural network suitable for fault diagnosis is determined, and the model of BP neural network for gearbox fault diagnosis is established. The BP neural network analysis program is developed through the GUI module of MATLAB, and the energy of each frequency band obtained by wavelet packet processing is taken as the input of the neural network. The vibration data of ship windbreaker is used in the test to complete the training and establishment of BP neural network, or the relevant experimental demonstration is carried out. The experimental results indicate that this method is effective and has greater advantages than traditional methods.