Weak Faults Diagnosis for Rolling Bearings Based on Variational Hilbert Huang Transform

Vibration signal is the most used monitoring signal in fault diagnosis. However, the location of the sensor will be restricted in some special cases. Therefore, non-contact microphone sensors are often used in these cases, which may cause some problems. For example, acoustic signals are always the low signal-to-noise ratio signals, which are difficult to accurately detect faults. Hilbert-Huang Transform (HHT) is a powerful method for bearing fault diagnosis, which includes Empirical Mode Decomposition (EMD) and Hilbert Transform (HT). However, mode aliasing and redundant modes are common in empirical mode decomposition, which are bad for detecting bearing faults. Aiming at the disadvantages of HHT, the Variational-Hilbert-Huang Transform (VHHT) is proposed for weak bearing fault diagnosis. Firstly, the signal is decomposed into several intrinsic mode functions (IMFs) through EMD. Then sensitive IMFs, which are chosen through the sensitivity evaluation algorithm, are decomposed by variational mode decomposition (VMD) to obtain several limited bandwidth intrinsic mode function (LBIMFs). Next, LBIMFs are reorganized according to the signal characteristics to obtain the fault component. Finally, the bearing fault is detected by HT. The proposed method is verified by simulation and experimental signals, and compared with HHT and VMD. The results show that VHHT is more prominent in detecting faults in low SNR rolling bearing signals.