Fault Diagnosis of Rolling Bearings Based on Improved Empirical Mode Decomposition and Fuzzy C-Means Algorithm

The fault signals of rolling bearings have a very low signal-to-noise ratio (SNR), making it difficult to fully extract and reconstruct the fault signals. To solve the problem, this paper proposes a way to recognize rolling bearing faults based on improved variational modal decomposition (VMD) and fuzzy c-means (FCM) algorithm. Firstly, the measured vibration signals of rolling bearings were subject to VMD on different scales. Next, the FCM clustering was performed to classy and recognize the eigenvectors of sample signals. Then, the authors calculated the normalized energy ratio of the autocorrelation function for each model obtained by decomposition, and applied it to optimize and reconstruct modal signals. Finally, the eigenvectors of sample signals were classified and recognized through FCM clustering. Several experiments were carried out to compare the improved VMD with empirical mode decomposition (EMD) and local mean decomposition (LMD) in the fault recognition and classification of rolling bearings in different backgrounds of strong composite noises. The comparison shows that the improved VMD has a strong denoising ability; the stronger the additive noise, the more superior the improved VMD is to LMD and EMD, and the better its feature clustering effect. The experimental results fully manifest the effectiveness of the proposed method.