Skip to content
Back to Projects

ReneWind Predictive Maintenance

Neural network classifier predicting wind turbine generator failures with 99% overall accuracy and 85% recall on the rare failure class.

Shipping
ML Neural Networks · Python TensorFlow/Keras Scikit-learn Pandas NumPy
99% Test Accuracy
0.95 F1 Score (macro)
85% Recall (Failure Class)
7 Models Tested

Summary

Built a classification model to predict wind turbine generator failures using historical sensor data from 40 instruments per turbine. Tested 7 model configurations, with the best achieving 99% test accuracy and 85% recall on the rare failure class in unseen data.

Approach

Trained neural networks with systematic architecture search (varying depth, optimizer, dropout, and class weighting) on imbalanced sensor data. Used walk-forward validation and focused on minimizing false negatives (missed failures) given the cost asymmetry: replacement >> repair >> inspection.

Key Result

Model 4 (Adam optimizer, 14→7 architecture) is recommended for deployment — on held-out test data it catches 85% of true generator failures at 97% precision on flagged failures, the strongest recall/precision balance among the 7 configurations tested.