(onsite and online)

Join Zoom Meeting

https://hku.zoom.us/j/91638414587?pwd=NHYQ3CEC6DCWU5zbpqvwXqvpt8eNU2.1

Meeting ID: 916 3841 4587

Password: 599331

Abstract:

Medium manganese (Mn) steels are a class of advanced high-strength alloys that exhibit an excellent combination of strength and ductility, primarily due to mechanisms such as austenite retention and transformation-induced plasticity. Given their strong sensitivity to phase evolution during heat treatment, optimizing thermal processing routes is essential for the continued development and performance enhancement of these materials. Accelerating the design of such complex alloys is crucial for meeting the demands of next-generation engineering applications. Traditionally, materials design has relied heavily on trial-and-error experimentation, which is both time-consuming and resource-intensive. In this context, machine learning has emerged as a powerful tool to accelerate materials discovery by extracting patterns from data and guiding informed decision-making. However, most existing machine learning models focus solely on predicting properties from composition, while neglecting the critical role of processing history. Others rely heavily on simulation data, which are limited to equilibrium thermodynamic assumptions and often fail to represent real-world material behavior. In this research, we propose a novel artificial intelligence (AI)-driven framework that integrates large language models (LLMs) and reinforcement learning (RL) to bridge the gap between chemical compositions, processing parameters, and resulting material properties. It offers a scalable, interpretable, and experimentally validated pathway to not only predict but also design optimized processing strategies alongside compositions—advancing beyond traditional, simulation-reliant approaches toward truly intelligent material discovery.

ALL INTERESTED ARE WELCOME