2026 ASEE Annual Conference & Exposition

Teaching mechanical design problems that involve competing stiffness and fatigue requirements challenges students to integrate multiple mechanics concepts such as stress concentration, deflection compatibility, endurance limits, and material sensitivity into a coherent design rationale. These challenges are fundamentally computational in nature, requiring students to reason across coupled relationships rather than apply isolated formulas. Traditional instruction often fragments these ideas, leading to procedural knowledge that does not readily transfer when constraints interact. This work introduces a computation-centered, AI-augmented learning module designed to support integrated reasoning and verification in multi-constraint mechanical design, demonstrated through stepped shaft design as a canonical example.

The module was implemented in two junior-level courses, MEEN 305 Solid Mechanics and MEEN 368 Solid Mechanics in Mechanical Design. Browser-based computational notebooks form the backbone of the learning environment, allowing students to manipulate geometry and loading parameters, visualize system response, and compute stiffness and fatigue metrics. Generative AI, exemplified by ChatGPT, is used as a complementary computational support that prompts explanation, interpretation of trade-offs, and articulation of assumptions rather than providing authoritative solutions. AI accuracy is addressed through explicit verification practices, including parameter sweeps, cross-checking against computational outputs, and consistency with expected mechanics trends.

Student learning is examined through design artifacts, reflective responses, and guided recitation discussions, with emphasis on integrated reasoning, evidence-based justification, and verification behavior rather than numerical outcomes alone. Preliminary evidence indicates that students more effectively explain competing constraints, interpret design sensitivity, and justify design decisions using both intuition and computational evidence. While demonstrated through stepped shafts, the computation-centered instructional framework is applicable to a broad class of mechanics and mechanical design problems involving coupled constraints.