Real-time crack characterization in conductive tubes through an adaptive partial least squares approach

This work deals with the real-time non-destructive testing and evaluation (NDT/NDE) of conductive tubes. An innovative learning-by-examples (LBE) strategy is proposed to address the inversion of eddy current testing (ECT) data. The partial least squares (PLS) features extraction technique is combined with an output space filling (OSF) adaptive sampling strategy in order to collect as much as possible information about the input/output (I/O) relationship to model, mitigating the negative effects of the curse of dimensionality. Robust and accurate predictions are then performed by means of support vector regression (SVR). A preliminary numerical validation is shown to prove the effectiveness of the approach.