P28 - Enhancing Predictive Maintenance with Temporal Convolutional Networks
- Event
- SMSI 2025
2025-05-06 - 2025-05-08
Nürnberg - Band
- Poster
- Chapter
- Poster Session
- Author(s)
- F. Rieger, R. Schilling, F. Heinrich, F. Wenninger - Fraunhofer EMFT, Munich (Germany)
- Pages
- 286 - 287
- DOI
- 10.5162/SMSI2025/P28
- ISBN
- 978-3-910600-06-5
- Price
- free
Abstract
Predictive maintenance is a crucial technique for reducing machine downtime. One challenge is the absence of labeled run-to-failure data. We propose a semi-supervised anomaly detection approach using Temporal Convolutional Networks, a regression model that has multivariate data as input and estimates vibration data. Our study reveals that our sensor signal estimation is quite accurate for normal data. The estimation error serves as a score that is useful for identifying anomalies.