Intelligent Fault Diagnosis Based on the EAO-VMD in Dual-Rotor Cylindrical Roller Bearings
Sharadchandra Patil, Vishal G. Salunkhe, Prashant S. Jadhav, Shravani R. Desavale, V. V. Shinde, R. G. Desavale
Abstract
Abstract The accurate diagnosis of localized defects in cylindrical roller bearings is crucial for ensuring the reliability and longevity of rotating machinery. However, fault detection is often hindered by harmonic interference, noise contamination, and complex signal components, making feature extraction challenging and reducing classification accuracy. To address these issues, this study proposes an enhanced Aquila optimizer (EAO)-based variational modal decomposition (VMD) framework for intelligent bearing fault diagnosis. The EAO algorithm, incorporating chaotic inverse learning, a sinusoidal search strategy, and an adaptive variation mechanism, enhances the optimization of VMD parameters, thereby improving the decomposition of vibration signals and preserving critical fault-related features. Experimental validation is conducted using a dual rotor-bearing test rig, where vibration signals from healthy and defective bearings with varying fault sizes are analyzed. The extracted fault features are classified using support vector machines, extreme learning machines, and deep extreme learning machines. The results demonstrate that the improved Aquila optimizer-variational modal decomposition framework achieves a diagnostic accuracy of 99.57%, significantly outperforming conventional methods. This research underscores the effectiveness of the proposed method for real-time condition monitoring and predictive maintenance, offering a reliable and robust approach for early fault detection in industrial rotating machinery.