Hybrid Kernel-Driven ELM for Intelligent Fault Detection in Motor Rolling Bearings

Abstract

Motor rolling bearings are critical components in industrial machinery, and their failure can lead to significant downtime and financial losses. Therefore, developing robust and intelligent fault detection systems is essential for ensuring operational reliability. This study presents a novel approach utilizing a Hybrid Kernel-Driven Extreme Learning Machine (HK-ELM) for intelligent fault detection in motor rolling bearings. The hybridization of multiple kernel functions enhances the learning capability and generalization of ELM by capturing complex nonlinear features from vibration signals. The proposed model integrates Gaussian and polynomial kernels to construct a composite kernel function, enabling adaptive feature extraction. The experimental evaluation was conducted on publicly available bearing datasets, where the hybrid kernel model was benchmarked against traditional ELM and other standard machine learning models. Results demonstrate that the HK-ELM model outperforms its counterparts in classification accuracy, training time, and fault recognition robustness, especially in the presence of noise and overlapping classes. This paper also discusses the impact of kernel parameter tuning and the hybridization strategy on diagnostic performance, concluding that the hybrid approach offers a highly scalable and efficient solution for real-time industrial fault detection applications.