MIT Researchers Build AI Design Copilot That Lifts Buckling Load 39% in 15 Extra Seconds
Updated
Updated · MIT News · Nov 13
MIT Researchers Build AI Design Copilot That Lifts Buckling Load 39% in 15 Extra Seconds
3 articles · Updated · MIT News · Nov 13
Summary
Communications Engineering results showed MIT’s human-in-the-loop topology optimization system produced more manufacturable structures and raised linear buckling load by 39%, while adding only about 15 seconds versus fully algorithmic design.
The copilot works by flagging regions of a structure for human feedback as the design evolves, letting engineers thicken or thin selected areas instead of waiting for a finished design and restarting from scratch.
Machine learning keeps the process fast by predicting where humans are most likely to intervene, using a synthetic training dataset built from thousands of designs and likely critiques on aesthetics, performance, and manufacturability.
MIT says the approach could widen topology optimization beyond aerospace and automotive into buildings and bridges, where engineers have been slower to adopt black-box designs they cannot easily control or trust.