A New Predictive Maintenance Approach: Novel Integration of GMM-LSTM for Prediction of Latent State and Failure Location of Rotating Machinery.
Summary:
We propose a hybrid predictive-maintenance framework that couples Gaussian Mixture Models (for latent bearing-state classification) with LSTMs (for temporal prediction and failure-location inference). Using FFT and WPD features, the model achieves robust early-fault detection and precise localization on multiple bearing datasets.
Highlights:
- GMM → latent state classification; LSTM → time-aware failure-location prediction.
- FFT + WPD feature stack for vibration-signal representation.
- Demonstrated generalization with strong precision/recall while mitigating overfitting.
- Real-time-oriented pipeline for rotating-machinery health monitoring.
Duration: January 2023 – June 2024.
Associated Members:
