Rapid and accurate prediction of temperature evolution in wire plus arc additive manufacturing using feedforward neural network

Abstract

This article proposes an approach based on a feedforward neural network (FFNN-SM) and computational simulations to rapidly predict thermal cycles in multi-layer single-bead walls fabricated during wire arc additive manufacturing (WAAM). First, a finite element (FE)-based model for thermal simulation in the WAAMed part was developed. Second, a FFNN-SM was trained and validated using the data generated from thermal simulations with different heat input levels (Q). The results reveal that the developed FFNN-SM enables a accurate prediction of the temperature evolution with a global R2 value higher than 98% and within only 40 s.