Fault diagnosis of gearbox based on Wilcoxon rank- sum tests and maximum amplitude selection

This paper proposes a novel method for gearbox fault diagnosis, which is capable of identifying both single faults (either in gears or bearings) and various types of compound faults. Vibration signals collected from a test platform were employed to validate the proposed method, where five operating states were configured, including: (1) healthy state; (2) single-tooth breakage of the fixed-axis gear; (3) single-tooth breakage of the planetary gear combined with bearing rolling element damage; (4) planetary gear wear coupled with rolling bearing outer ring damage; and (5) fixed-axis gear root crack, planetary gear wear, and bearing outer ring damage. The proposed method Wilcoxon rank- sum tests and maximum amplitude selection (WTMAS) was used as feature extraction method for vibration signals of different states and to establish the training samples and test samples. The K-Nearest Neighbor (KNN) algorithm was utilized as the classifier for fault type classification and identification. Experimental results demonstrate that the average recognition rate of the proposed method for the five states reaches 95.753 %, indicating that the method exhibits high recognition accuracy for different types of faults and is thus an effective approach for gearbox fault diagnosis.