This paper describes real-time cracks characterization and localization inside a Structure Under Test (SUT) by exploiting Learning by Examples (LBE) strategy in the context of Eddy Current Testing (ECT). Within the framework of LBE, an optimal training set has been generated in offline phase by adopting Partial Least Squares (PLS) feature extraction combined with a customized version of output space filling (OSF). Support Vector Regression (SVR) algorithm is utilized for developing an accurate model based on the training set and subsequently real-time inversion (online phase) has been performed on unknown test set. The robustness of the proposed PLS-OSF/SVR approach is numerically assessed in presence of synthetic noisy test set.
Fast characterization of multiple cracks in conductive media ased on adaptive feature extraction and SVR / Ahmed, Shamim; Miorelli, Roberto; Reboud, Christophe; Calmon, Pierre; Anselmi, Nicola; Salucci, Marco. - STAMPA. - 43:(2018), pp. 191-198. (Intervento presentato al convegno 22nd International Workshop on Electromagnetic Nondestructive Evaluation, ENDE 2017 tenutosi a Saclai, France nel 6th-8th September 2017) [10.3233/978-1-61499-836-5-191].
Fast characterization of multiple cracks in conductive media ased on adaptive feature extraction and SVR
Anselmi, Nicola;Salucci, Marco
2018-01-01
Abstract
This paper describes real-time cracks characterization and localization inside a Structure Under Test (SUT) by exploiting Learning by Examples (LBE) strategy in the context of Eddy Current Testing (ECT). Within the framework of LBE, an optimal training set has been generated in offline phase by adopting Partial Least Squares (PLS) feature extraction combined with a customized version of output space filling (OSF). Support Vector Regression (SVR) algorithm is utilized for developing an accurate model based on the training set and subsequently real-time inversion (online phase) has been performed on unknown test set. The robustness of the proposed PLS-OSF/SVR approach is numerically assessed in presence of synthetic noisy test set.File | Dimensione | Formato | |
---|---|---|---|
Fast Characterization of Multiple Cracks in Conductive Media B ased on Adaptive Feature Extraction and SVR.pdf
accesso aperto
Tipologia:
Versione editoriale (Publisher’s layout)
Licenza:
Creative commons
Dimensione
431.97 kB
Formato
Adobe PDF
|
431.97 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione