Enhanced Fault Diagnosis of Motor Rolling Bearings Using an Improved Multi-Kernel Extreme Learning Machine Approach

Abstract

Fault diagnosis of motor rolling bearings is critical for maintaining the operational efficiency and safety of rotating machinery systems. With the increasing demand for precise and rapid fault detection, traditional machine learning methods often fall short in terms of adaptability and accuracy under noisy or non-linear conditions. This study proposes an Enhanced Multi-Kernel Extreme Learning Machine (EMK-ELM) approach for robust and accurate fault diagnosis of rolling bearings in electric motors. The methodology integrates multiple kernel functions in the ELM framework to better capture complex data characteristics while mitigating the impact of outliers and noise. The proposed model is trained and validated using the Case Western Reserve University bearing dataset, and its performance is benchmarked against conventional ELM and other advanced classifiers including Support Vector Machines and Convolutional Neural Networks. Experimental results demonstrate that EMK-ELM significantly improves classification accuracy, generalization, and training speed, making it an ideal candidate for real-time fault diagnosis applications. This research lays a solid foundation for scalable, efficient, and intelligent bearing condition monitoring systems.