An Integrative Computational Method for Gearbox Diagnosis

Increasing demand on energy has accelerated research on improving the reliability of wind turbines. As a critical component in wind turbine drivetrains, the majority of gearbox failures have shown to initiate from bearing failures. The low signal-to-noise ratio and transient nature of bearing signals pose significant difficulty for bearing defect diagnosis at the incipient stage. For improved bearing diagnosis, this paper presents a new method that integrates ensemble empirical mode decomposition (EEMD) with independent component analysis (ICA) to effectively separate bearing and gear meshing signals, without requiring a priori information on rotating speeds or bandwidth. The method first decomposes sensor measurement into a series of intrinsic mode functions (IMFs) as pseudo multi-channel signals, by means of EEMD, to satisfy the requirement by ICA for redundant information. ICA is performed on the IMFs to separate defective bearing components from gear meshing signal. Enveloping spectrum analysis is then performed to identify bearing structural defects. Both numerical and experimental studies have demonstrated the merit of the developed new method in improving gearbox diagnosis.