Bearing Fault Detection Autoencoder
This is an unsupervised autoencoder model trained on normal bearing vibration data from the CWRU dataset.
- Task: Anomaly/fault detection (high reconstruction error = fault)
- Framework: PyTorch
- Dataset: Case Western Reserve University Bearing Data
- Training: Only on normal samples
- Usage: Load the .pth file and compute MSE on new segments
Trained in Google Colab as part of a Generative AI course project.
Files:
- bearing_fault_autoencoder.pth → model weights
