AI-guided materials framework links optimisation to manufacturability

Researchers at Fudan University and ShanghaiTech reported a framework combining topology-preserving construction, finite-element analysis, a physics-inspired graph neural network and reinforcement learning for fibrous materials. Authors report roughly 50% higher strength and 20% lower mass than initial designs, with stereolithography and fused-deposition demonstrations.

Why it matters: The work connects generative optimisation to fabrication constraints, a key gap in industrial AI. Results remain author-reported and limited to the demonstrated materials and fabrication settings.

Nature Communications