Generative Design and Manufacturing

Abstract

This course teaches designing complex multiphysics systems using generative AI tools with hands-on, project-based learning. It covers single physics design, multiphysics integration, and testing. Assessments include reports, team projects, and a final showcase. Students experience advanced software, navigating design trade-offs, and optimizing performance under constraints.

Objective

1. Understand and Apply Single Physics Design Principles: - Students will be able to formulate and solve single physics design problems using computational tools and optimization techniques. - Students will incorporate manufacturing and assembly constraints into their designs. 2. Master Multiphysics Design and Optimization: - Students will select and integrate multiple physics phenomena into their design process. - Students will evaluate and optimize designs considering the interactions between different physical phenomena. 3. Develop Skills in Generative AI and Optimization Techniques: - Students will use generative design tools, such as GANs and Diffusion Models, to create innovative design solutions. - Students will apply reinforcement learning and transfer learning techniques to improve design performance under multiple constraints. 4. Collaborate and Integrate Multidisciplinary Solutions: - Students will work in teams to integrate different design solutions, balancing performance, cost, and manufacturability. - Students will navigate trade-offs and combine human and AI-generated solutions. 5. Evaluate and Validate Designs: - Students will manufacture and test their designs, adjusting predictive models to account for real-world discrepancies. - Students will present their final designs and performance results, demonstrating their ability to meet the challenge problem requirements.

Content

This is a project-based class where teams of students will design and manufacture a complex part or assembled system to meet the performance requirements of a challenge problem provided by an industrial partner. The course contains a mixture of the following learning activities: (1) tutorial sessions introducing students to State-of-the-Art techniques for design and optimization using computation or artificial intelligence, (2) interactions with manufacturers or manufacturing experts to realize their designs and understand design/manufacturing tradeoffs, (3) independent resources for student learning on subsets of multi-physics phenomena and how to simulate or analyze them, and (4) autonomous work sessions where team members will collaborate to integrate different design solutions and navigate performance/cost tradeoffs. The course concludes with an end-of-semester showcase where each team's manufactured design will compete against one another to see which one produces the best performance on the challenge problem. Part 1: Computer-Aided Single Physics Design and Manufacturing - Students study the challenge problem and design it under only a single physics objective, including how to analyze and evaluate the challenge problem and provided constraints. - Create candidate designs and discuss and resolve challenges to manufacturing or assembling the design given different processes. - Examples of skills covered: Formulating the challenge problem into amenable computational representations. Review of classical shape and topology optimization methods. Introduction of AI-based Generative Design and Inverse Design methods. Incorporating manufacturing or assembly constraints into solutions. Part 2: Multiphysics Design and Systems Considerations - Students select a secondary physics or constraint and specialize their learning by selecting one sub-topic to focus on with the independent learning resources and techniques provided by the instructional team or CAD/CAE partners. - Investigate coupling effects that occur when multiple physics or components interact with one another and alter the design performance or manufacturing/assembly considerations. - Re-design or optimize their Part 1 designs under expanded multi-physics performance criteria or constraints. - Examples of skills covered: combining results from different solvers or optimizers; navigating among different Pareto optimal designs to discern tradeoffs; building surrogate models of complex physics to make decisions; exposure to new computational multi-physics solution methods. Part 3: Integration, Tradeoffs, and Testing - Students collaborate in teams from across the Part 2 specialization areas to integrate different design solutions into a candidate design that satisfies all performance requirements. - Refine manufacturing and assembly considerations on the final design and develop appropriate cost and performance models for the chosen design and manufacturing method. - Manufacture and evaluate candidate designs on the complete challenge problem. - Examples of skills covered: adjusting predictive models for differences between simulation and reality (i.e., the sim2real gap); integrating multiple human- or computer-generated solutions into a common design; deciding how to combine or trade-off between different performance and constraints; combining computationally generated solutions with human feedback and engineering knowledge.

Resources