MIT researchers have created an AI system that acts as a 'copilot' for structural engineers, enabling human feedback to be integrated into algorithmic design as it evolves rather than only after a final design is produced. The system, described in Communications Engineering, could make topology optimization more accessible and manufacturable for a broader class of design engineers.
Topology optimization is one of the most advanced algorithmic design techniques, but its black-box nature has made it less widely adopted in industry. It has been used less to design buildings and bridges, partly because engineers need to trust the output. Designers see only the final result and often must start over if changes are needed, and the process can lead to impractical structures.
The MIT team's approach uses a machine-learning model trained on a synthetic dataset of likely human feedback to identify regions of a design an engineer would want to alter. During the algorithm's run, the system highlights those regions and lets the user adjust thickness before the design is finalised.
In demonstrations, the human-in-the-loop approach improved manufacturability and linear buckling load by 39 percent while adding only about 15 seconds to the total design time compared with fully automated methods, the researchers reported.
“Right now, the human doesn’t have any agency or control over the design process as it’s playing out,” said lead author Dat Ha, a recent MIT civil and environmental engineering doctoral student who is now a structural designer at LeMessurier. “For these really complex designs, we wanted to also tap into our brain, which is the most complex machine there is, to solve these design problems in a way that’s more practical to industry.”
“On paper, topology optimization is really powerful and can achieve great, high-performing designs, but its black box nature has made it less widely adopted in industry,” Ha said.
Senior author Josephine Carstensen, an MIT associate professor of civil and environmental engineering, said the idea is to bring together the best of algorithmic and human approaches. “The idea is to tap into the brains of the designers and get them to infuse their domain knowledge as the design evolves,” she said.
The researchers hope to add other modification options in future versions; the system currently allows users only to make structures thicker or thinner in identified regions. The work was supported, in part, by the MathWorks Engineering Fellowship Fund, TATA Steel and the MIT Generative AI Impact Consortium.
Analysis
Why This Matters
- By incorporating human judgment into the design process, the system could make advanced algorithmic design accessible to engineers in sectors such as construction, where trust in automated output is low.
- The approach reduces the likelihood of producing structures that are high-performing on paper but impractical or difficult to build, potentially cutting redesign costs.
- If adopted, it could speed up the development of lighter, stronger components for aircraft, vehicles and infrastructure.
Background
Topology optimization uses formal algorithms to iteratively fill a design space in a way that best meets performance goals, such as minimising weight or maximising stiffness. While the technique has been applied to parts for F1 racing cars, ultralight aircraft and antennas, its widespread adoption, particularly for buildings and bridges, has been hindered by its opaque nature: engineers cannot see or influence the design as it develops, and final outputs sometimes contain features that are impossible to manufacture. Earlier attempts to allow human interaction were too slow to be practical. The MIT group's new system uses a machine-learning model trained on a database of likely human critiques to quickly identify where an engineer would intervene, adding only seconds to the computation.
What to Watch
- Whether engineering firms in construction and other sectors will trial the system in real projects, and how it performs on larger-scale designs.
- Future versions of the tool that may allow a broader range of modifications could further expand its applicability.
- Adoption by researchers and students in academia could accelerate the teaching of topology optimization in engineering curricula.
Checked
12 claims in this story checked against 1 source: 11 supported.
Show each claim
- SupportedMIT researchers developed an AI copilot to bring human feedback into structural design as it evolves.MIT News: “MIT researchers are making algorithmic designs more collaborative with a system that acts like an AI copilot. As the algorithm is generating the structure, it presents regions to the human designer for feedback.”
- SupportedThe system uses machine learning to predict regions where engineers would want to make changes.MIT News: “The researchers’ new system uses machine learning to quickly identify the most likely regions of an algorithmic design that a human would want to change.”
- SupportedThe system improved manufacturability and linear buckling load by 39 percent.MIT News: “In demonstrations, the researchers showed their human-in-the-loop approach improved manufacturability and linear buckling load by 39 percent while only modestly increasing the total design time.”
- SupportedThe system added about 15 seconds to total design time compared with fully automated methods.MIT News: “In an open-access Communications Engineering paper describing the system, the researchers showed it led to more easily manufacturable structures with higher buckling loads while adding only 15 seconds to the total design time compared to all-algorithmic approaches.”
- Not in the sourcesThe system highlights the regions identified for possible modification.MIT News: “As the algorithm is generating the structure, it presents regions to the human designer for feedback.”
- SupportedThe system lets users adjust thickness in identified regions before the design is finalised.MIT News: “Right now, the approach only allows users to make structures within identified regions thicker or thinner, but they hope to add other options for modification in future versions.”
- SupportedThe system could make topology optimization more accessible and manufacturable for a broader class of design engineers.MIT News: “The researchers believe the approach, which uses a design technique known as topology optimization, could make algorithmic design more accessible and manufacturable for a broader class of design engineers.”
- SupportedTopology optimization’s black-box nature has made it less widely adopted in industry.MIT News: ““On paper, topology optimization is really powerful and can achieve great, high-performing designs, but its black box nature has made it less widely adopted in industry.””
- SupportedTopology optimization has been used less to design buildings and bridges, partly because engineers need to trust its output.MIT News: “Topology optimization has gotten traction in the automotive and aerospace industries, but it’s been used less to design buildings or bridges, partly because design engineers need to know they can trust the output.”
- SupportedDat Ha said, “Right now, the human doesn’t have any agency or control over the design process as it’s playing out.”MIT News: ““Right now, the human doesn’t have any agency or control over the design process as it’s playing out,” explains lead author Dat Ha PhD ’26, a recent MIT civil and environmental engineering doctoral student who is now a structural designer at LeMessurier.”
- SupportedDat Ha said, “For these really complex designs, we wanted to also tap into our brain, which is the most complex machine there is, to solve these design problems in a way that’s more practical to industry.”MIT News: ““For these really complex designs, we wanted to also tap into our brain, which is the most complex machine there is, to solve these design problems in a way that’s more practical to industry.””
- SupportedJosephine Carstensen said the idea is to tap into designers’ brains and have them infuse domain knowledge as the design evolves.MIT News: ““The idea is to tap into the brains of the designers and get them to infuse their domain knowledge as the design evolves,” Carstensen says.”
