Artificial neural network modelling for evaluating austenitic stainless steel and Zircaloy-2 welds

Abstract

Ferrite content in austenitic stainless steel welds is a measure of resistance to solidification cracking. Accurate estimation of ferrite content in austenitic stainless steel welds is important to ensure crack free welds. An artificial neural network (ANN) model has been developed to predict ferrite number with an improved accuracy. Eddy current (EC) testing is attractive due to high sensitivity and versatility for the detection of harmful surface defects. Artificial neural network modelling has been used to process the eddy current data for evaluating the defect depth so that on-line eddy current testing is possible in austenitic stainless steel welds. There is a necessity to develop on-line monitoring methods for evaluation the quality of spacer pad welds in cladding tubes made of Zircaloy-2 used in pressurized heavy water reactors (PHWR). Shear strength values of the individual coins is the measure of the quality of the welds. Prediction of shear strength values of the individual coins ensures their integrity. Artificial neural network model has been developed for prediction of shear strength of spacer pad welds of Zircaloy-2.