Fault diagnosis of industrial robot gearboxes based on a deep Boltzmann machine

DOI:	10.11809/bqzbgcxb2023.04.023
Keywords:	Deep Boltzmann Machine; wavelet packet transform; feature extraction; industrial robot gearbox; fault diagnosis
Abstract:	Under the complex working conditions of multi speed and multi load, this paper proposes a fault diagnosis method based on a Deep Boltzmann Machine (DBM) to solve the problem that the fault signals of industrial robot gearboxes are difficult to accurately identify. Wavelet packet transform (WPT) is used to extract the statistical characteristics of the original vibration signals under each fault state as the input of the DBM model. Then, the DBM is pre trained in an unsupervised manner, which deeply explores the extracted statistical features to obtain more abstract reconstructed fault feature vectors. Finally, the fault diagnosis results are output by Softmax classifier. The proposed method is applied to the fault diagnosis of six degree of freedom industrial robot gearboxes, and compared with the current mainstream decision making classification methods. The results show that the average recognition rates of 94% and 92.176% are achieved respectively by using the DBM for the fault diagnosis of industrial robot gearboxes under single and complex conditions, which proves higher accuracy and robustness of the DBM.
Issue:	Vol. 44 No. 4 (2023)
Published:	2023-04-28
PDF HTML