Answers are grounded in this paper and cite their sources.
Early-stage aircraft structural design choices strongly influence performance, yet preliminary sizing often relies on empirical methods which may be inaccurate for composite structures and unconventional layouts. This paper presents an automated surrogate-assisted framework for coupled layout and size optimization of composite wingbox structures, demonstrated in a low-dimensional case study. A nested approach optimizes the number of ribs, number of stringers, and spar-cap width in an outer loop, while component sizing is handled by a commercial gradient-based solver in an inner loop. A staged multi-run sizing strategy reduces structural mass by up to 450 kg relative to the first feasible sizing result while reducing run-to-run variability. A Gaussian-process surrogate, trained on a limited number of sizing results, guides the outer-loop search with a noise-aware improvement-based acquisition function. Three Bayesian optimization repetitions were performed for each of three initial datasets, with 30 layout evaluations per study. All nine studies reduced the initial best mass, returning structural masses of 1735.62–1778.82 kg. The lowest mean mass was obtained with the largest allocation of the budget to adaptive sampling. Selected layouts typically featured 50 mm wide spar caps and 13–14 stringers, while rib counts varied. Results were benchmarked against a global gradient-free optimizer.
No takes yet. Share an insight, caveat, or question.
Ziakos et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: