Classification of weld defects detected by ultrasonics using signal processing techniques

Type : Article de conférence
Auteur(s) :  DRAI Redouane, Khelil Mohamed, Benchaala Amar
Année :  2001
Domaine : Electronique
Conférence: Ultrasonic International 2001, 2-5 juillet 2001 (Pays-Bas).
Résumé en PDF :  (résumé en pdf)
Fulltext en PDF :  (.pdf)
Mots clés :  NDT, Ultrasonics, defects classification, digital wavelets transform, neural networks

Résumé : 

Discriminatory features from temporal and spctral signals of detected echo are extracted. the compact feature vector obtained will be then classified by different methods: K Nearest Neighbour algorithm, statistical bayesian algorithm and artificial neural network. other disciminatory features using a multiresolution analysis technique, callled discret wavelet transform are also extracted. then, the obtained feature vector will be also classified by the same algorithms. experimental results obtained from a data bank constitued by echoes detected in welds will be compared and discussed.