A hybrid framework based on Teager Energy Spectrum Correlation and Whale-optimized decision tree classifier for fault diagnosis in wind turbine drivetrain bearing

Wind turbine drivetrain bearings operate under complex conditions, where ambient noise hinders accurate fault diagnosis. The Teager-Kaiser Energy Operator (TKEO) has proven effective in enhancing the energy content of non-stationary signals (such as those generated by drivetrain components), by revealing subtle transient features typically masked by noise. While TKEO has been successfully integrated with downstream classifiers for wind turbine fault diagnosis, the lack of hyperparameter tuning in previous studies has limited the overall effectiveness of such diagnostic frameworks. In response this study presents a hybrid framework based on Teager Energy Spectrum Correlation and Whale-Optimized Decision Tree for improved fault diagnosis in wind turbine drivetrain bearings. Results demonstrate that incorporating advanced optimization techniques significantly enhances fault diagnostic performance in wind turbine drivetrain bearings. The baseline decision tree model, trained without optimization, achieves the lowest classification performance. The application of hyperparameter optimization methods - Grid Search, Random Search, and Whale Optimization - improved model performance, with Whale Optimization, achieving the highest accuracy - 84.8%, with precision, recall, and F1-score improving to 84.2%, 85.8%, and 84.0%, respectively. Feature importance analysis highlights Spectral Flatness and Entropy as the most influential features in fault diagnosis, indicating that frequency-domain and complexity-based features contribute significantly to classification performance.