Artificial neural network modeling of composition–process–property correlations in austenitic stainless steels

This paper discusses the application of artificial neural network modeling in austenitic stainless steels research, including: (i) correlation between chemical composition, process variables and flow stress of austenitic stainless steels under hot compression; (ii) constitutive flow behavior of type AISI 304L stainless steel during hot torsion; (iii) microstructural evolution during dynamic recrystallization of alloy D9; (iv) correlation between chemical composition and tensile properties of alloy D9. Multilayer perceptron based feed-forward networks have been trained by comprehensive in-house datasets. Very good performances of the neural networks are achieved. Various effects are modeled, among which are: (i) influence of alloy composition and processing parameters on flow behavior of austenitic stainless steels; (ii) effect of strain rate on torsional flow behavior of 304L stainless steel; (iii) combined influence of temperature and strain on dynamic recrystallization behavior of alloy D9. The simulated results are found to be consistent with the metallurgical trends. Finally, the issue of neural network's “black box” approach to modeling is addressed.