Machine Learning Truss

Optimized a 32-inch span balsa truss for maximum load-carrying capacity across a 10-dimensional design space with coupled constraints.

• Maximize predicted failure load under a single point load 2 inches above the support line
• All 11 members within 8.5–14 inch length bounds
• Total member length under 120-inch material budget

Balanced force distribution, buckling capacity, and joint buildability throughout.

Machine LearningMATLABRoot Cause AnalysisStructures
Sole developer
Dec 2025

Project Complete

Genetic algorithm explored 39,280 candidate designs over 217 generations. Final geometry predicted a failure load of 56.20 ± 4.45 oz (3.51 lb), exceeding the 32 oz minimum by 76% with 97.7% material utilization. The optimizer discovered an asymmetric load point at 75% span and a shallow top-chord arch peaking at 11.22 inches, geometric features that would have been impractical to find through manual iteration.