Underdetermined blind source separation based on density peak clustering for gear fault identification

In order to improve the noise robustness of blind source separation algorithm in estimating the number of vibration sources，an underdetermined blind source separation method was proposed based on density peak clustering.Single source points were extracted from signals preprocessed，and then the mixed matrix was estimated by clustering single source points with density peak clustering.The separated signals were obtained by reconstructing the source signals based on the compressed sensing model.In order to verify the separation accuracy and noise robustness of the proposed algorithm，the simulation signals under different signal-to-noise ratios were separated by the proposed algorithm.The results showed that when the signal-to-noise ratios was not less than 4 dB，the proposed algorithm can accurately separate the source signals，and the accuracy and robustness of the algorithm were verified.A rotating component fault diagnosis test-bench was designed to verify the effectiveness of the proposed algorithm in practical application，and the measured composite fault vibration signal was processed.The test results showed that the algorithm successfully separated the single fault characteristics of bevel gear and planetary gear from the observed signal，contributing to the fault diagnosis of rotating components in engineering.