CAREER: Autonomous AI Agents for Translational Multidisciplinary Design Optimization

This Faculty Early Career Development Program (CAREER) project supports integrated research and education focused on creating autonomous artificial intelligence (AI) agents that enable computer-based designs to be reliably translated into real-world engineered systems. The ability to rapidly design and deploy high-performance engineered systems is essential for U.S. technological leadership and global competitiveness in national defense, advanced manufacturing, and energy. However, translating designs into practice remains slow and uncertain because computer-based designs often struggle to meet real-world performance requirements when their underlying physics models are imperfect. This research will address this critical challenge by creating autonomous AI agents that detect modeling errors, correct physics models, verify the generalizability of the corrections, and quantify the resulting design impact across various disciplines, including fluid mechanics, structural dynamics, heat transfer, acoustics, and multiphase transport. The research will expand our knowledge of (1) how to enable reliable design decisions in the presence of modeling errors, (2) how modeling errors propagate through multidisciplinary systems, and (3) how AI-driven models impact the design space, optimization speed, robustness, and scalability. The knowledge gained will deepen our understanding of design tradeoffs across multiple disciplines, enable efficient exploration of high-dimensional, real-world design spaces, and unlock multidisciplinary design optimization (MDO) at scales and complexities beyond current capabilities. In addition, this project will create sustainable education modules that strengthen critical thinking, support workforce development, and broaden STEM engagement in the era of AI, serving students from elementary school through graduate education. The research goal of this project is to enable AI-powered, translational MDO by bridging numerical design optimization with reliable physical validation. The autonomous AI agent framework will identify and correct defects in partial differential equations (PDEs) in physics models, enabling corrected models to extend beyond the training parameter space, transfer across multiple physical disciplines, and produce compact, physically interpretable correction terms. To evaluate the generalizability of the framework, this project will consider three challenging MDO problems involving six distinct PDE models from bio-inspired drones, battery packs, and magnetic metal manufacturing. This project will advance AI and translational science by creating trustworthy AI-enabled engineering design methods and translating them into practical workflows through collaborations with government, national laboratory, and industry partners. The AI framework, corrected physics models, and associated datasets will be released publicly to promote collaboration, reproducibility, and further development. These resources will also be integrated into the education modules to support hands-on learning, develop the ability to critically evaluate AI-generated results, and prepare a future workforce for AI-enabled engineering design. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2541867 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Ping He | Institution: Iowa State University, AMES, IA | Award Amount: $549,499 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2541867 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2541867.html