Fault Detection using Opposition-based Learning Orca Predator Algorithm with Random Forest

Machinery equipment stability and safety operations heavily depend on the implementation of intelligent fault detection systems. The system enhances production efficiency while decreasing maintenance costs together with its most significant benefit of preventing safety accidents. Intelligent compound fault detection methods require additional research regarding their interpretability features. The detection results prove difficult to understand which impedes their general acceptance for industrial use in the real world. Therefore, this research proposes an Opposition-based Learning Orca Predator Algorithm (OLOPA) with Random Forest (RF) for fault detection. The OLOPA exploration-exploitation control improves worldwide search performance while reducing convergence time thus enhancing both accuracy and stability in feature extraction. High-dimensional data management and strong generalization abilities characterize the RF classifier during its performance. The combined hybrid method provides dependable and interpretable fault detection capabilities which enables its use in complex industrial environments. Furthermore, this research extracts meaningful features from signals through time-domain and frequency domain to enhance the feature detection and signal quality maintenance. The OLOPA-RF achieves 99.91% accuracy for CWRU dataset which is better than state-of-the-art methods.