Jensen-Shannon Divergence for Non-Destructive Incipient Crack Detection and Estimation

Nowadays industrial process needs more and more accurate nondestructive procedures for material crack detection and diagnosis. Early detection in that cases are very challenging issues: more the cracks are incipient and higher is the difficulty for its detection and estimation. Indeed, these incipient cracks which cause non-obvious changes in sensor measurements needs to be properly detected and estimated. For conductive materials measurement based on impedance maps obtained from Eddy Current Testing (ECT) are used but the presence of environmental noise can mask the crack information and induce missed detection and false estimation. In this paper, we highlight the limitation of classical techniques and address this problem using a methodology based on wavelet transform and Jensen-Shannon divergence in the framework of Noisy Independent Component Analysis (NICA). For our work, the impedance maps are considered as a mixture information. Then, source signals containing the fault features are obtained by the application of the Independent Component Analysis regarding the noise. A wavelet decomposition is then used and operates as a noise reduction operation. Jensen-Shannon (JSD) divergence is then proposed for the crack detection. Thanks to a theoretical derivation, the fault severity estimation is obtained. The performances are evaluated and the superiority validated regarding other techniques already used in the literature. The performances limits are evaluated for noise varying environments and the optimal diagnosis is obtained for several incipient cracks.