MIT AI copilot brings engineer feedback into structural design
MIT researchers developed a machine-learning system that lets engineers adjust structures during algorithmic design. Demonstrations produced more manufacturable designs with higher buckling loads while adding only 15 seconds to design time.
AI Hubs United Editorial ·
MIT researchers have developed an AI copilot that brings engineers into the structural design process while an optimization algorithm is still running. According to MIT AI News, the system identifies areas where a designer is likely to want changes and presents them for feedback. The work, published in Communications Engineering, aims to address a practical drawback of automated design: engineers may otherwise have to wait for a finished structure before spotting problems and restarting the process.
The approach uses topology optimization, which progressively fills a design space to meet specified goals. Its machine-learning component was trained on synthetic feedback covering thousands of designs, including considerations such as appearance, performance and manufacturability. In demonstrations, the researchers reported easier-to-manufacture structures and a 39 percent improvement in linear buckling load. Total design time increased by only 15 seconds compared with fully algorithmic approaches.
For engineers, the practical benefit is an opportunity to apply domain knowledge before a design is complete, rather than simply accept or reject the final output. The current system limits changes to making structures thicker or thinner within identified regions. The researchers hope to expand those editing options and make optimization more accessible to engineering users.
AI Hubs United original short summary of the source linked above. It is not the original article; read the full story at the source.
