Study of Planetary Gear Fault Diagnosis Based on Energy of LMD and BP Neural Network

摘要

Planetary gear box has the characteristics of small volume and large transmission ratio and is widely used in construction machinery. After a long period of operation, the gear fault occurs frequently, which has a great influence on the equipment. However, due to its complex structure, the fault signal is often submerged in the inherent signal of the gearbox. In order to extract the fault feature from the signal, a method based on energy of Local mean decomposition (LMD) and Back Propagation (BP) neural network is proposed to solve this problem in this paper. Original signal is decomposed by LMD into 6 product functions (PF). The energy of each PF component are calculated and defined as the input of the BP neural network. Optimal model of neural network can be obtained based on sample training. The result of experimental shows that the proposed method can achieve an overall recognition rate of 95.5%, which proves that it is an effective method for planetary gear fault diagnosis.