An Image Texture Descriptor based Machine Learning Framework for Prediction of Thermo-Mechanic Heat Treatment Process in Plain Carbon Steel

To bring about variation in quality/grade of a metal, it is made to undergo specific thermo-mechanical treatment procedures, that can be customized to ensure that the metal microstructure evolves in a desirable manner, so as to obtain specific desired properties of the material. This is because the evolution of microstructure happens differently in different heat treatment procedures. Often, in the field of computational material science, we need to predict the correct thermo-mechanical treatment procedure, to subject a metal to, in order to get desired properties of it. It is extremely crucial and challenging to determine what values of different parameters to be set, before starting the thermo-mechanical treatment. For example, what amount of pressure or strain to be imposed on the metal during a heat treatment like large strain warm deformation, or what value to set the temperature to, during warm deformation and subsequent annealing of the metal. In this work, we propose a machine learning classification framework, for prediction of desired thermo-mechanical treatment process to be followed for plain carbon steel, given an initial microstructure, to arrive at a desired target microstructure state. Our experimental results prove that the proposed model yields considerable high prediction performance in terms of model accuracy and F1 score.