RESEARCH · RESEARCH · #3528
MIT researchers build a human-in-the-loop "copilot" for topology optimization
MIT researchers published a Communications Engineering paper describing a human-in-the-loop AI 'copilot' for topology optimization that uses a trained ML model to flag regions where designers would likely intervene. In demos the approach produced more manufacturable structures with a 39% increase in linear buckling load while adding about 15 seconds to total design time versus fully automated optimization.
KEY POINTS
- MIT researchers published a Communications Engineering paper describing a human-in-the-loop AI 'copilot' for topology optimization that uses a trained ML model to flag regions where designers would likely intervene.
- In demos the approach produced more manufacturable structures with a 39% increase in linear buckling load while adding about 15 seconds to total design time versus fully automated optimization.
- The method combines human domain knowledge with fast ML guidance to make topology-optimized designs more trustworthy and manufacturable, lowering a key barrier to wider industry adoption.
WHY IT MATTERS
The method combines human domain knowledge with fast ML guidance to make topology-optimized designs more trustworthy and manufacturable, lowering a key barrier to wider industry adoption.