This research presents an AI-based predictive maintenance framework designed to monitor and detect
mechanical faults in wind turbines by analyzing vibration and acoustic signals. The system integrates
high-frequency accelerometers and microphones to capture raw data from turbine components, which
is then processed using signal enhancement techniques, including Butterworth filtering and feature
extraction methods such as Fast Fourier Transform (FFT) and Mel-Frequency Cepstral Coefficients
(MFCCs). A 1D Convolutional Neural Network (CNN) model was developed and trained to classify
multiple fault types—Normal, Gear Fault, Bearing Fault, and Imbalance—achieving a high classification
accuracy of 96.8%. Evaluation metrics such as precision (96.3%), recall (96.5%), F1-score (96.4%),
and RMSE (0.157) demonstrate the robustness of the model in identifying faults early and accurately.
Comparative analysis with other machine learning models like SVM and Random Forest further
validates the superiority of the CNN-based approach. This work highlights the practical applicability
of AI-driven diagnostics in enhancing the reliability and operational efficiency of wind energy systems.
