TY  - RPRT
KW  - Eddy current testing
KW  -  inverse scattering
KW  -  nondestructive testing and evaluation
KW  -  statistical learning
KW  -  learning-by-examples
KW  -  support vector regression
KW  -  output space filling
KW  -  partial least squares
KW  -  adaptive sampling
PB  - University of Trento
AV  - public
ID  - elediasc12716
A1  - Salucci, M.
A1  - Anselmi, N.
A1  - Oliveri, G.
A1  - Massa, A.
N2  - This document deals with the real-time retrieval of the position of a crack embedded within a conductive planar structure starting from eddy current testing (ECT) measurements in a non-destructive testing and evaluation (NDT-NDE) scenario. Towards this end, an innovative adaptive learning-by-examples (LBE) technique has been developed. It is based on the innovative combination of a Partial Least Squares (PLS) features extraction technique and an adaptive sampling strategy to generate optimal training sets. Such information is used to train a Support Vector Regressor (SVR) in order to build a fast but accurate predictor of the crack descriptors staring from previously-unseen ECT measurements during the on-line testing phase.
The proposed LBE inversion strategy, previously validated on numerical simulations, is here tested against real laboratory-controlled experimental data coming from the World Federation of NDE Centers (FNDEC) ?2008 Eddy Current Benchmarks?.
M1  - technical_report
UR  - http://www.eledia.org/students-reports/716/
Y1  - 2016///
TI  - An Innovative Adaptive LBE Technique for Real-Time Crack Characterization: An Experimental Study
ER  -