Study on fault diagnosis of gear fracture based on beamformer

The sound signal of gear contains a lot of information on gear running state. However, due to the influence of noise, the sound collected by microphones contains a lot of noise pieces, which greatly affects the application of sound signals in the field of fault diagnosis. To extract the information on gear running state effectively, this paper proposes a method combining GSC with FastICA to ameliorate that the original GSC can not effectively improve the expected signal when the direction of arrival(DOA) is inaccurate. Then the noise pieces in the signal are further suppressed by using VDM to decompose the signal extracted by GSC-FastICA. Finally, the SVM parameters optimized by PSO are fine-tuned, and a model with relatively good generalization performance is obtained. Taking fault-frequency ratio as the evaluation index, the experimental results show that this method makes an effective improvement compared with the original GSC algorithm; and compared with the decomposition effect of EEMD, VMD extracts gear fault information more effectively. (c) 2022 Published by Elsevier Ltd.