Kernel regression residual signal-based improved intrinsic time-scale decomposition for mechanical fault detection

Correct identification of the faults in mechanical components is the key to monitoring health conditions of machines. However, the interference of low-frequency signals and environmental noises greatly restricts the performance of traditional fault detection methods. Considering that the signals of interest (SOI) are usually carried by high-frequency signals and the interference of low-frequency signals, a novel preprocessing technique called one-level kernel regression residual decomposition (KRRD) is presented. Combined with the improved intrinsic time-scale decomposition (IITD) and Hilbert envelope analysis technique, a novel mechanical fault detection method is proposed. The method combines KRRD to extract the high-frequency signals from the raw vibration signals to track the faulty information, IITD to further extract the SOI from the high-frequency signals by removing the noises, and Hilbert envelope analysis to demodulate the denoised signals to detect faults in the rolling element bearings and gears. In order to verify the performance of the proposed method, the simulated data and real experimental data collected from faulty bearings and gears are analyzed. Three commonly used methods, namely empirical mode decomposition (EMD), local mean decomposition (LMD) and variational mode decomposition (VMD) are introduced and comparisons with the three combinations (KRRD and EMD, KRRD and LMD, KRRD and VMD) are given. The analysis results indicate that the proposed method is superior to other methods for detecting faults in mechanical components. (paper)