An adaptive Learning-by-Examples strategy for efficient Eddy Current Testing of conductive structures

An innovative inversion strategy is presented to address the non-invasive inspection of large conductive structures by exploiting eddy current testing (ECT) measurements. The arising inverse problem is formulated within the Learning-by-Examples (LBE) framework and is solved by means of an efficient strategy that combines Partial Least Squares (PLS) feature extraction with an adaptive sampling strategy in order to generate optimal training databases during the off-line phase, while exploits Support Vector Regression (SVR) during the on-line phase for achieving robust and accurate estimations with almost real-time prediction performances.