Abstract
<title>Abstract</title> <p>A surrogate-assisted Bayesian optimization framework is developed for composite laminate stacking-sequence design under manufacturing constraints. The design problem is formulated as a finite discrete sequence-search problem considering symmetry, balance, minimum ply percentage, contiguity, and disorientation constraints. A classical laminate theory (CLT)-based analytical evaluator is used to compute a uniaxial compressive buckling index and a bending–twisting coupling index, which are integrated into a scalar objective function through a coupling penalty term. A position-weighted stacking-sequence distance is then constructed to represent the ordered sequence structure and the larger contribution of outer plies to bending-dominated responses. This distance is embedded into a sequence-kernel Gaussian process surrogate model. Sequential candidate selection is driven by an upper confidence bound acquisition function, and a minimum within-batch sequence-distance constraint is introduced to reduce candidate redundancy during batch sampling. Numerical studies involving different ply numbers, ply-angle sets, surrogate representations, coupling penalty coefficients, and evaluation budgets show that the proposed method improves search efficiency and maintains consistent candidate ranking in the discrete stacking-sequence space. Abaqus linear buckling analyses of representative laminates further support the consistency between the CLT-based screening results and finite element buckling responses.</p>