Early Weak Fault Diagnosis of Gearbox Based on ELMD and Singular Value Decomposition

Abstract

As important transmission device in the entire mechanical system, gearboxes shoulder the responsible for transmitting motion and torque. Gearboxes early fault diagnosis technology can realize early warning of fault, improve the reliability of equipment operation, and avoid major accidents. However, the background noise of the gearbox early fault vibration signal is strong, and the weak fault features are often submerged and the feature information is difficult to extract. Aiming at these problems, a new early fault diagnosis method by combining ensemble local mean decomposition (ELMD) and singular value decomposition (SVD) is proposed. Firstly, the raw gearbox fault vibration signal is broken down into a large number of narrow-band product functions (PF) with the ELMD method. Then, the PF component containing the most abundant fault characteristics is selected as the sensitive feature to be analyzed. The SVD is applied to the sensitive PF component, and the singular value difference spectrum is obtained. The reconstructed signal order is determined in the singular value difference spectrum. Finally, the Hilbert envelope demodulation analysis is performed on the signal which is reconstructed, and the fault feature information in the envelope spectrum is extracted. By comparing with the theoretical value, the fault location of gearbox is determined. The effectiveness and superiority of the proposed method are verified by the actual fault data of the gearbox.