This framework combines unsupervised and semi-supervised machine learning for effective fault detection in rotor systems, highlighting accuracy improvements.
Key Points
The semi-supervised method achieved over 95% accuracy in distinguishing multiple fault types.
Features were extracted using a multi-layer autoencoder, employing unsupervised learning techniques.
K-means clustering was utilized for unsupervised fault detection in rotor systems based on vibration signals.
The integration of these methods demonstrates significant potential for addressing limited labeled data in fault diagnosis.