Machine learning for fast design integration
Investigating surrogate and pattern-learning approaches that can return useful feedback early enough to shape a design—not merely verify it at the end.
Research track 03
ExperimentalStructures.
This track explores methods that shorten the distance between structural intuition and design feedback—from machine-assisted evaluation to geometric representations of force.
Experimental studies
New methods, anchored in structural behaviour.
Graphical statics for concrete foundations
Studying force geometry as a way to reason about load transfer and form in bridge foundation concepts, with emphasis on interpretability.
Hybrid engineering workflows
Connecting analytical models, learned approximations, and visual reasoning so each method contributes where it is strongest.
Design loop
Feedback must arrive while the design can still change.
01
Frame
Define the structural question and useful limits.
→02
Sample
Generate or collect meaningful design states.
→03
Learn
Build an interpretable approximation or relation.
→04
Decide
Use fast feedback to guide the next design move.
Research position
Speed is useful only when the result remains legible.
Can learned models support early structural choices?
The opportunity is not to replace analysis, but to expose tendencies quickly enough for designers to compare directions before detailed models become expensive to change.
Can force geometry make behaviour more visible?
Graphical statics offers a complementary path: a direct visual language for equilibrium that keeps the relationship between form and force in view.