High-fidelity solvers in the design optimization of ships

The use of high fidelity, CPU time expensive solvers is now quite common in the trial-and-error design process. Their use in a simulation-based design process is however still reduced by the large amount of computational effort needed in a design problem. To reduce the computational cost of running complex engineering simulations, different techniques may be adopted to obtain global approximation of the high-fidelity solvers, as the widely used metamodelling techniques. An alternative approach is the Variable Fidelity Modeling (VFM), which combine less accurate but also incomparably less expensive solvers with high-fidelity models. The use of trust region strategies and of consistency conditions finally ensure the accuracy of the optimal solution. Applications of metamodels and of VFM with trust region strategies are presented for both single and multi-objective complex design problems. Shape optimization of existing surface ships is performed through the solution of CFD solvers, included a high-fidelity finite-volume RANSE code for the prediction of the free surface flow past the ship free to dynamically sink and trim. In the adopted formulation, the accuracy of the metamodels evolve during the optimization cycles. Results show up to a factor of 5 in terms of CPU time.