← All projects

Gearbox Fault Diagnosis

A PCA–CNN classifier that locates bearing faults in a wind turbine drivetrain from vibration data.

Problem
Find which bearing is damaged in a 5-MW reference drivetrain (main, low-speed planet or high-speed shaft bearing) under different environmental conditions.
Data
Vibration from a high-fidelity 5-MW NREL drivetrain model, published on Zenodo and Hugging Face.
Approach
Signals are split into frames, PCA reduces the dimension and extracts features, and a CNN classifies the fault location.
Tools
Python, PyTorch Lightning, Hugging Face Hub
Publication
Fault detection of offshore wind turbine drivetrains in different environmental conditions through optimal selection of vibration measurements. Renewable Energy, 2023.
Procedure: data acquisition from a wind turbine drivetrain, PCA-based feature extraction, and CNN classification
Overall procedure: data acquisition, feature extraction with PCA, and classification with a CNN.