AI-Centric Workflow for Multiphysics Simulation

This technical report presents an AI-centric workflow designed to enhance simulation processes in engineering teams. The workflow incorporates large-language-model agents that are specifically trained on computational physics reasoning. By utilizing these agents, the time required for configuring, solving, and interpreting multiphysics studies is significantly reduced from hours or days to mere minutes, all while maintaining model fidelity. The process begins with a plain-language prompt, which the agent uses to create a detailed execution plan, generate necessary code, run simulations, and extract results in an understandable format. The report evaluates the workflow across various multiphysics scenarios, demonstrating a median setup time reduction of 75-90% and maintaining accuracy within 5% of expert-built baselines. Additionally, it outlines the benefits of rapid deployment, built-in interpretation of results, seamless integration with existing data flows, and accessibility for non-specialists. The report concludes with a demonstration of the workflow and best practices for effective prompt design and API orchestration.