Inventory-constrained optimization of grid shells driven by reuse

This work presents a novel design approach to enhance grid shell structures using steel beams from dismantled buildings. The objective is to improve grid shell performance with minimal aesthetic changes by minimizing the cut-off waste from reused elements. Given an input grid shell, the optimization combines Integer Linear Programming (ILP), differentiable optimization and graph neural networks to address two key objectives. First, it promotes the use of recycled components from a donor inventory, maximizing the reuse of beams over new elements and minimizing cutting processes. Practically, the target is trying to reuse the beams as-is. Second, it updates the nodal coordinates of the grid shell to reduce a loss function based on strain-energy, improving structural efficiency. Results show topology-preserving shape modifications and global form changes that enhance structural response. Life Cycle Assessment (LCA) confirms the method’s viability as a sustainable development strategy.