Fault diagnosis for gear wear of planetary gearbox

To solve problems of planetary gearbox fault diagnosis method having higher professional requirements, complex calculation process and longer model training time, a new fault diagnosis method for planetary gearbox based on PCA-EDT-DBN was proposed. PCA was used to analyze vibration signals acquired with several sensors, select the first p principal components of each column of signals according to requirements, and arrange these p principal components into a one-dimensional (1-D) sequence. Euclidean distances between the first p principal components of each 2 columns data were computed to obtain a distance matrix. This matrix was sequentially expanded into a 1-D sequence. Two 1-D sequences obtained according to the mode mentioned above were synthesized into a 1-D one taken as a sample to be input into DBN for model training. Then, new samples were input into the trained model to output a classification result intelligently, and realize planetary gearbox's fault diagnosis. Additionally, in order to improve the accuracy of model diagnosis, the orthogonal test method was used to optimize parameters of DBN. The preset fault test data for planetary gearbox teeth wear were used to verify the effectiveness of the proposed method. The results showed that the proposed method has advantages of higher diagnosis accuracy, shorter training time and simpler calculation process. © 2019, Editorial Office of Journal of Vibration and Shock. All right reserved.