A Multi-Fault Feature Detection and Separation Method for Multistage Gearbox

Abstract

Due to its unique advantages, Instantaneous Angular Speed (IAS) has been prevalently applied to a wide variety of rotary machines. Nevertheless, accurate IAS measurement poses several intrinsic challenges. Affected by the measuring accuracy of IAS estimation, IAS signal extracted from the output shaft of the multistage gearbox is usually treated as a combination of periodic signal and structure borne noise. Therefore, it can be very difficult to extract fault signatures without a dedicated signal processing, especially for the multifault pattern. Because of the complexity properties and interferences among different faults, little research has been devoted to the multi-fault detection for gearbox based on IAS signal. Focusing on this issue, a newly hybrid procedure involving Minimum Entropy Deconvolution (MED), Empirical Mode Decomposition (EMD) and Autocorrelation Local Cepstrum (ALC) is proposed to address the multi-fault detection issue using IAS signal. The effectiveness of the hybrid method is validated by a multistage gearbox in practice. The experimental validation presented in Section 4 further confirms the obtained results, in which fault characteristics corresponding to different gears are successfully detected and separated.