Abstract

Alloy design and design of its manufacturing process needs exploration over a vast design space due to the presence of many compositional and processing variables, much of which is not explored well in practice. The choice of multiple processing routes feasible for a given set of targeted alloy properties further complicates the design process. In this talk, we present a novel generalized generative design framework, which can produce multiple design options of composition, and their manufacturing process routes for a given requirement. The framework employs distinct Deep Reinforcement Learning (DRL) models, trained using computational simulation models, to carry out composition selection and various manufacturing process steps. A novel reward function is developed that integrates sustainability, cost, and manufacturability considerations into the decision making. A generative design step is proposed to utilize the trained DRL models to produce multiple designs for a given requirement. The framework is applied to the case of a hot rolled steel sheet design with two feasible process routes: conventional casting and thin slab casting, producing promising results. A benchmark study is conducted to evaluate the framework's performance against materials engineers’ designs for three use cases, revealing the framework's superior performance.

Bio

Akash Bhattacharjee is a scientist in PREMAP & ICME Research and Innovation Program, TCS Research, Tata Consultancy Services based in Pune, India. He earned a B. Tech. in Metallurgy and Materials Engineering from VNIT, Nagpur, India in 2013 and a M.E. in Materials Engineering from Indian Institute of Science, Bangalore, India in 2015. He is primarily a computational material scientist with expertise in using various simulation tools across the length scale and applying AI/ML for advanced materials design. His research interest includes Alloy Design, Computational Thermodynamic, AI/ML, Additive Manufacturing, Phase Transformation, Microstructure Modelling, Integrated Computational Materials Engineering, and Precipitation Modelling. He has worked with diverse alloy systems, including steels, Ni-based superalloys, and Al alloys.