Department of Mechanical Engineering Graduate Seminar

PRESENTATION: Mechanical compatibility is a critical requirement for therapeutic patches applied to soft tissues and organs, where large deformation, nonlinear mechanical responses, and auxetic surface behavior may occur. However, conventional patch design often relies on expert-defined architectures, manual parameter tuning, and finite design libraries, making it difficult to explore the large design space of auxetic metastructures. In this talk, I will present an AI-assisted framework for the inverse design of organ-inspired auxetic metastructure patches. First, a multi-objective data-driven generative design framework combines finite element modeling, a convolutional neural network surrogate, and a denoising diffusion probabilistic model with guided diffusion to generate patch geometries that match prescribed stress–strain and strain-dependent Poisson’s-ratio responses while accounting for trade-offs between the two mechanical objectives. Building on this framework, I will further demonstrate how OpenAI Codex can serve as an AI agent to autonomously design organ-inspired patches. A workflow-agent system autonomously executes an established inverse-design procedure, while a goal-driven agent system is given only the design objective and scientific constraints and formulates, executes, and revises its own computational strategies.

PRESENTER: Yingbin Chen is a Ph.D. student in the Department of Mechanical Engineering at the University of Iowa, where he works as a Graduate Research Assistant in the Multiscale Computational Science and Engineering Laboratory under the supervision of Prof. Shaoping Xiao. His research interests include AI for Science, computational mechanics, metamaterials, and multiscale modeling. His current research focuses on developing AI-assisted computational frameworks for the inverse design of bio-inspired auxetic metastructures, including machine-learning surrogate models, generative diffusion models, and AI agents for inverse design and computational workflows. Before beginning his doctoral research, he worked as a CAE engineer in the automotive industry, where he specialized in finite element analysis and structural simulation.