Bearing Fault Diagnosis Based on IPOA-VMD and SSA-HKELM
Baoxian Chang, Xing Zhao, Dawei Guo, Siyu Zhao, Jiyou Fei, Hua Li, Xiaodong Liu · IEEE Access · 2024
This study presents a novel fault diagnosis approach for rolling bearings that integrates the Improved Pelican Optimization Algorithm (IPOA) for optimizing Variational Mode Decomposition (VMD) and the Sparrow Search Algorithm (SSA) for optimizing the Hybrid Kernel Extreme Learning Machine (HKELM). The method aims to overcome challenges such as weak early fault signals and the complexities in extracting fault characteristics that often result in subpar fault classification outcomes. A novel comprehensive indicator is introduced as the fitness function during the parameter selection phase of IPOA. By utilizing IPOA, the optimal combination of VMD’s parameters, including the mode component K and penalty factor$\alpha $, is determined. Signal decomposition via VMD yields a set of Intrinsic Mode Functions (IMF). The Kolmogorov-Smirnov Distance (KSD) is employed as a measure to assess the correlation between each IMF component and the original signal. Subsequently, the KSD values of the IMFs are calculated to identify the optimal IMF components, with their Multi-scale Range Entropy (MRE) computed as a distinguishing feature. Lastly, the HKELM, enhanced through SSA optimization, is employed for the training and classification of rolling bearing faults, with the reliability and efficacy of the proposed methodology validated through simulation and empirical data.