Using artificial neural networks to quantify discontinuities in Eddy current testing

Abstract

This paper presents an artificial neural network for quantitative eddy current testing of materials. Time domain parameters that are functions of digitized in phase and quadrature components of eddy current probe imvedance are given as input to the neural network; the output of the network, in user defined units, is tested and displayed continuously. The performance of the neural network has been tested on austenitic stainless steel plates for detection and depth quantification of surface breaking machined notches in the presence of disturbing variables such as liftoff, material property variations and surface roughness. The applicability of the neural network method has been demonstrated to detect and size discontinuities in thin walled stainless steel tubes with periodic wall thickness variations. Because the output is available in user defined units, the neural network method can be used for online shop floor applications that employ simple threshold based accept/reject criteria.