Reassigned Wavelet Scalogram for Gear Fault Detection under Nonstationary Operational Conditions

Gear transmission systems have extensive use in wind turbines, helicopters and heavy trucks, and their fault detection is of vital importance for ensuring industrial safety. However, they often run under nonstationary conditions, which accelerate the mechanical deterioration and meanwhile result in time-varying characteristics which can barely be identified through time or frequency domain analysis. Time-frequency analysis is often utilized to provide intuitive presentation of time and frequency information concurrently, yet limited time-frequency resolution and cross-term/inner interferences hinder the accurate fault identification by conventional linear or bilinear methods. In this paper the reassigned wavelet scalogram, which has merits of fine time-frequency resolution and cross-term free nature, is applied to reveal potential gear fault under nonstationary conditions. Its advantage over traditional time-frequency representations is demonstrated using a numerical simulated gearbox vibration signal, and results of lab experimental evaluation indicate that real-world gear fault is successfully diagnosed.