A data-driven approach to model the martensitic transformation temperature in strain-induced metastable austenitic steels
Abhishek Kumar Thakur,Bhaskarjyoti Das,Sandip Ghosh Chowdhury
DOI: https://doi.org/10.1016/j.mtcomm.2024.109016
IF: 3.8
2024-05-10
Materials Today Communications
Abstract:Strain-induced metastable austenitic stainless steels form an important class of materials in metallurgical industries for their wide range of applications. These steels undergo austenite-martensite phase transformation at temperatures above martensite start temperature, Ms , at which martensite is formed on mechanical deformation, also known as the Md temperature. Md temperature depends on several factors related to the steel and is important from alloy design perspective. In the literature, there are quite a few equations based on composition of the steels for the prediction of Md temperature. However, it is well known that the transformation from austenite to martensite is dependent on the austenite grain size as well as deformation conditions i.e. strain, strain rate and temperature of deformation. In the present work, the role of those parameters has also been considered. The model is implemented using fourteen input parameters viz., composition, grain size, amount of strain, temperature of deformation, and strain rate. The architecture of the neural network model is optimized rigorously to predict the Md temperature on a par with actual value. It has been shown that grain size and strain rate have very negligible influence whereas strain and temperature of deformation have quite strong role. Md temperature is increased with increasing strain whereas the temperature of deformation shows opposite dependence on it. An empirical equation thus, has been established to calculate the Md temperature of a steel as a function of its composition, grain size, temperature of deformation and strain. The final optimized model is then deployed to predict the Md temperature of different steels and predictions are found to be in close agreement to the experimentally measured Md temperatures. The developed model is general and can be extended to include other parameters as well as various other steel alloys.
materials science, multidisciplinary