Bayesian neural network analysis of ferrite number in stainless steel welds

Abstract

Bayesian neural network (BNN) analysis has been used in the present work to develop an accurate model for predicting the ferrite content in stainless steel welds. The analysis reveals the influence of compositional variations on ferrite content for the stainless steel weld metals, and examines the significance of individual elements, in terms of their influence on ferrite content in stainless steel welds, based on the optimised neural network model. This neural network model for ferrite prediction in stainless steel welds has been developed using the database used to generate the WRC-1992 diagram and the first author's laboratory data. The optimised committee model predicts the ferrite number (FN) in stainless steel welds with greater accuracy than the constitution diagrams and the other FN prediction methods. Using this BNN model, the influence of variations of the individual elements on the FN in austenitic stainless steel welds is also determined, and it is found that the change in FN is a non-linear function of the variation in the concentration of the elements. Elements such as Cr, Ni, N, Mo, Si, Ti, and V are found to influence the FN more significantly than the other elements present in stainless steel welds. Manganese is found to have a weaker influence on the FN. A noteworthy observation is that Ti influences the FN more significantly than does Nb, whereas the WRC-1992 diagram considers only the Nb content in calculating the Cr equivalent.