PT2.60 - Damage Localization in Mechanical Structures Based on a Virtual Sensor Approach
- Event
- EUROSENSORS XXXVI
2024-09-01 - 2024-09-04
Debrecen (Hungary) - Band
- Poster
- Chapter
- PT2 - Smart Systems and Artificial Intelligence in Sensing
- Author(s)
- L. Merzak, E. Saoutieff, C. Ott, S. Boisseau - Université Grenoble Alpes, Grenoble (France)
- Pages
- 257 - 258
- DOI
- 10.5162/EUROSENSORSXXXVI/PT2.60
- ISBN
- 978-3-910600-03-4
- Price
- free
Abstract
This paper introduces a new methodology for localizing damages in mechanical structures thanks to a Virtual Sensor approach. We combine physics-based models, finite element methods, real-time data and data-driven models using machine-learning algorithms to determine in real-time the “damage localization” on real-world mechanical structures with a reduced number of strain sensors. By leveraging finite elements, we have built synthetic database gathering the normal physical behavior of the mechanical structures under various force applications and abnormal behaviors induced by notches in the material. Machine-learning algorithms have been trained on synthetic data coming from simulations. The notches have been effectively localized in real-time on the real-world mechanical structure with a reduced number of strain sensors, validating the approach.