Data‐Driven Approach to Accelerate ICME for Ceramics | CeramShield AI

ABSTRACT

Integrated computational materials engineering (ICME) is an emergent approach to accelerate materials design by connecting process, structure, properties, and performance through experimental and computational methods. ICME studies are highly specialized, with each analysis focusing specifically in connecting one phase of the framework to the next. High fidelity simulations and experimentation make this possible, driving the costs of material production up. For more novel applications of materials such as using silicon carbide for heat shielding, ICME falls short due to gaps within available documentation. Recent advances in large language model (LLM) accuracy and speed have introduced the possibility of utilizing data‐driven approaches to improve upon ICME methodology. LLMs have developed to the point where local deployment is possible for use a process known as retrieval‐augmented generation (RAG). RAG avoids the main pitfall of LLM operation by providing a robust curated database for reference, minimizing the occurrence of data hallucinations that plague data‐driven systems. Applying the data‐driven model, titled “ CeramShield AI ,” to design and manufacturing of ceramics has shown promising results from the standpoint of both time and accuracy. Comparisons between experimental ICME studies and CeramShield AI show parity in results, and exhibit no issues with hallucination as a result of RAG. In addition, the data‐driven methodology can be leveraged to apply ICME in reverse, a novel process that accelerates material design and allows for more varied applications of the framework.