Hull-form stochastic optimization via computational-cost reduction methods

The paper shows how cost-reduction methods can be synergistically combined to enable high-fidelity hull-form optimization under stochastic conditions. Specifically, a multi-objective hull-form optimization is presented, where (a) physics-informed design-space dimensionality reduction, (b) adaptive metamodeling, (c) uncertainty quantification (UQ) methods, and (d) global multi-objective algorithm are efficiently and effectively combined to achieve high-fidelity simulation-based design optimization (SBDO) solutions. The application pertains to the multi-objective optimization for resistance and seakeeping (operational efficiency and effectiveness) of a destroyer-type vessel. Two hierarchical multi-objective SBDO problems are presented, with a level of complexity decreasing from the most general (stochastic sea state, heading, and speed) to the least general (deterministic regular wave, at fixed sea state, heading, and speed). Design-space dimensionality reduction is based on a generalized Karhunen-Loève expansion of the shape modification vector combined with low-fidelity-based physical variables. A multi-objective deterministic particle swarm optimization algorithm is applied to a stochastic radial-basis-function metamodel that provides objective predictions. UQ methods include Gaussian quadrature and metamodel-based importance sampling. Numerical simulations are based on unsteady Reynolds-averaged Navier-Stokes and potential flow solvers. The paper shows and discusses the joint effort of computational-cost reduction methods in enabling high-fidelity SBDO, providing guidelines for future research directions in this area.