Trust-region genetic optimization and extreme sea-state sensitivity analysis of offshore jacket structures under code constraints

Deepwater jacket sizing must satisfy code constraints across multiple sea states despite costly finite-element evaluations. This study develops a feasibility-first trust-region genetic algorithm (TR-GA) for group-level sizing of an East China Sea jacket. TR-GA combines group screening, state-dependent discrete neighborhoods, feasibility repair and SACS evaluation across 16 load cases. Against a conventional full-space GA with matched variables, starting design, evaluator, random seeds and evaluation cap, both methods yielded independently validated feasible designs in 20 paired runs. TR-GA produced the lighter design in every pair, with a mean paired weight reduction of 6.404%. For the governing EL06 load case, Morris rankings remained consistent before and after optimization, while nested Sobol analysis identified wave height as the dominant influence on maximum and weighted unity checks. These findings support state-dependent search and prioritize wave height in extreme sea-state assessment for this jacket.