Dual-Mode Noise-Reconstructed EMD for Weak Feature Extraction and Fault Diagnosis of Rotating Machinery

Abstract

The meaningful data-based fault diagnosis is beforehand revealing the potential faults to reduce the costly breakdowns, one challenging of which is extracting the weak features from the complicated signals. Ensemble noise-reconstructed EMD (ENEMD) is an intelligent method by the nice integration of adaptively decomposing and naturally denoising. However, ENEMD still suffers from such issues as the false possible noise-only IMFs and the universal minimax threshold, reducing the precision of the critical noise estimation for the weak feature extraction. Thus, the dual-mode noise-reconstructed EMD method is proposed for weak feature extraction and fault diagnosis of rotating machinery. First, the possible noise-only IMF selection rule is redesigned according to the noise characteristic and the correlation evaluation, to eliminate the redundant slowly oscillating IMFs mistakenly chosen for noise estimation. Second, the adaptive local minimax threshold is proposed in the noise estimation technique for the low SNR signal, to overcome the drawback of additionally keeping some critical but weak fault features into the estimation noise. Hereinto, the local threshold is respectively performed in each sliding window defined by the demodulated rotating-related feature frequency. Third, the proposed method is addressed with the flowchart. Finally, two engineering case studies are implemented to demonstrate the feasibility and effectiveness of the method. The analytic results show that the method could effectively extract the periodic impulses generating by the early local damage in the gearbox of a hot strip finishing mill. Meanwhile, the method could successfully reveal the weak rubbing-impact faults along with alleviating the mode mixing phenomenon in the refined results for fault diagnosis of a heavy oil catalytic cracking unit. Hence, the method could provide a promising tool for weak feature extraction and fault diagnosis of rotating machinery.