An integrated FE–ILP optimization framework for steel hall structures

Structural steel design optimization traditionally minimises steel mass subject to the ultimate and serviceability limit states. Cutting bar stock, produced in standard lengths, into assembled members generates trim waste, but this step is usually treated as secondary. When stock is finite and heterogeneous, the statically lightest section is rarely available in a way that yields minimal trim loss. To bridge this gap, we couple automated finite-element sizing with an exact integer linear programming (ILP) formulation of the one-dimensional cutting stock problem over a finite, heterogeneous inventory. A commercial finite-element solver is driven through its Python-accessible Component Object Model interface to build, analyse and post-process designs satisfying the ultimate-limit-state member checks of EN 1993-1-1 and EN 1993-1-5. Each candidate assignment is screened against the inventory and solved with the open-source CBC engine, so that a weighted objective combining trim loss and installed mass governs the choice of cross-sections rather than being evaluated afterwards. On a hall with an irregular grid, four member groups and mixed demand lengths, trim loss falls from 19.0% to 4.2% while the purchased stock mass decreases, and the selected assignment is not the statically lightest one. A constructed benchmark isolates the same mechanism at a single catalogue step, cutting trim loss from 37.52% to 1.64%; this figure is illustrative and specific to that inventory. For both cases we report the complete candidate ranking, the non-dominated set and the weighting sensitivity. On a standard laptop the pipeline runs in about eight to ten minutes for the larger instance.