Modeling the correlation between the alloy composition and the ferrite number in duplex stainless steel welds using artificial neural networks

Abstract

Weld metal composition is thought to be an important factor in influencing the austenite/ferrite ratio of duplex stainless steel microstructures. To produce the required balance in the austenite/ ferrite ratio in the weldmicrostructure, the chemical composition of the welding consumables should be adjusted. In the present work, Bayesian neural network analysis has been employed to predict the ferrite number in duplex stainless steel welds as a function of composition. The technique accounts for modelling uncertainty, and automatically quantifies the significance of each input variable. In this paper, the influence of variations in the weld composition on the ferrite number have been quantified for two duplex stainless steels. Predictions are accurate compared to published methods. The role of Si and Ti in influencing the ferrite number in these alloys has been brought out clearly in this study while these elements are not given due considerations in the WRC-1992 diagram.