The prospects of implementation of artificial intelligence for modelling of microstructural parameters in metal forming processes

The primary trend in modern metal forming can be characterised by the increase in the complexity of the technological processes and higher demand for the quality of the products. This naturally raises the requirements for the quality of modelling prediction of various aspects of metal forming process, such as tool wear, metal flow, fracture and defects formation, microstructure evolution and mechanical properties. However, various independent benchmarking studies [1] have shown that modelling predictions can be wrong even for well-calibrated models, and all the efforts with more detailed and metrologically better experiments didn't lead to any significant leap in the prediction quality. As an attempt to implement some alternative approach, this paper investigates the applicability of an Artificial Intelligence (AI) approach, in particular Deep Learning models. The example of a recurrent neural network model predicting recrystallisation during hot forging of Inconel 718 is presented. The model considers the entire thermo-mechanical history at every point and is trained and blind-tested using actual experimental data.