Title :
The classification of defects from ultrasonic measurements
Author_Institution :
Div. of Mater. Phys. & Metall., Harwell Lab., Didcot, UK
Abstract :
Both Rumelhart and Hopfield algorithms have been used in the classification of the nature of the defects found in steel pressure vessels. Ultrasonic reflections from such defects show only minor feature differences so that classification is difficult. Both neural-network algorithms performed as well as weighted minimum distance algorithms
Keywords :
computerised signal processing; neural nets; pattern recognition; steel; ultrasonic materials testing; Hopfield algorithms; NDT; Rumelhart; US reflections; classification of defects; computerised signal processing; neural-network algorithms; pattern recognition; steel pressure vessels; ultrasonic measurements;
Conference_Titel :
Expert Systems for NDT, IEE Colloquium on
Conference_Location :
London
