Yield Strength Modelling for Ferritic Steel Welds: Predicting yield strengths of ferritic steel weld deposits using an artificial neural network

Abstract

Creating ferritic steel welding alloys that can keep up with the demands of modern steelproduction is no easy undertaking. Until an optimal composition and welding process are found, this hasoften been accomplished through a process of experimental trial and error. A shorter trial period couldmean less money and time spent on the process overall. In this study, we describe how an artificialneural network may be used to predict the yield strengths of ferritic steel weld deposits based on thematerials used, the welding parameters employed, and the post-welding heat treatments applied. Itdetails the creation of the General regression neural network (GRNN) models and the verification of theirmetallurgical underpinnings and correctness.