MIT researchers have developed an AI copilot system that, while topology optimization algorithms generate a structure, hands selected regions to a designer for feedback so that the structure is easier to manufacture, buckling load increases by 39%, and total design time is only 15 seconds longer than fully automated methods. The system uses machine learning to quickly pinpoint the regions humans are most likely to want to modify, and its model is trained on a synthetic dataset built from human feedback on thousands of designs. At present, users can only thicken or thin the structure within the identified regions; the research team hopes to add more modification options in the future and believes this collaborative approach can coexist with purely AI-driven design methods.

Source:MIT 研究新闻 · Read Original · 前沿热点
Leave a Reply