Online segmentation of acoustic emission data streams for detection of damages in composites structures in unconstrained environments
Résumé
An approach for unsupervised damage detection in ring-shaped Organic Matrix Composites (OMC) under loading based on acoustic emissions (AE) is proposed. It relies on a specific clustering algorithm called Gustafson-Kessel (GK) that manages fuzzy memberships to clusters and complex cluster's shape. A methodology is proposed to 1) make the algorithm robust to initialisation in order to obtain reproducible results and reliable statistical models representing OMC damages, 2) detect and assess AE activity (AEA) over time for AE data mining to emphasize the more relevant AE data in a huge amount of AE hits, 3) adapt the statistical models based on statistical process control using imprecise updating rate automatically tuned.
Domaines
Mécanique [physics.med-ph]
Origine : Fichiers produits par l'(les) auteur(s)
Loading...