Non destructive Eddy Currents inversion using Artificial Neural Networks and data augmentation

R. Cormerais (1, 2, 3) , Roberto Longo (1, 2) , A. Duclos (2) , G. Wasselynck (4) , G. Berthiau (4)

1 GSII - ESEO-GSII
2 LAUM - Laboratoire d'Acoustique de l'Université du Mans
3 IREENA - Institut de Recherche en Energie Electrique de Nantes Atlantique UR 4642
4 IREENAT - Institut de Recherches sur l'Evolution de l'Environnement Normatif des Activités Transnationales

Eddy Currents (ECs) for Non Destructive Testing (NDT) is a method to determine the presence of flaws in metal materials. The estimation of flaw parameters like position and size through physical models is usually difficult. This article offers an alternative technique based on machine learning algorithms such as Artificial Neural Networks (ANNs). This approach often requires simulated signals to build an exhaustive training data-set, leading to a considerable amount of calculation time and resources. To deal with this problem, this article proposes a new method based on data augmentation via Principal Component Analysis (PCA). The presented method is evaluated using different kinds of simulated and experimental signals.

Keywords: Eddy currents testing; Artificial neural network; Data augmentation; Principal component analysis

Domains: Engineering Sciences [physics]

R. Cormerais, Roberto Longo, A. Duclos, G. Wasselynck, G. Berthiau. Non destructive Eddy Currents inversion using Artificial Neural Networks and data augmentation. NDT & E International, 2022, 129, pp.102635. ⟨10.1016/j.ndteint.2022.102635⟩. ⟨hal-03666804⟩