Simulation and AI, two complementary tools for NDE4.0
Résumé
The rise of artificial intelligence techniques and machine learning promises tremendous advances in the field of NDE and SHM for both the assistance to the analysis or for automated diagnostics. The potential benefits concern detection, characterization and sizing of defects. However, the development of efficient algorithms often face challenges due the complexity of the NDE problem, the multiplicity and the variability of parameters impacting acquired signals or images, combined to a lack of fully representative data bases. It has already been pointed out that simulation can be a powerful mean to tackle these issues. Simulation which relies on physic-based models can be used to provide numerical data for training algorithms and can also be used to select the relevant features, to design and to evaluate ML based algorithms. With this belief, CEA List is working on a CIVA module allowing the NDE engineer to benefit of the potential of simulation to handle and implement data science tools. In this communication, we develop the complementarity between simulation and AI, with examples of recent achievements, applications and results.
Origine : Fichiers produits par l'(les) auteur(s)