MULTI-OBJECTIVE HULL-FORM OPTIMIZATION OF A SWATH CONFIGURATION VIA DESIGN-SPACE DIMENSIONALITY REDUCTION, MULTI-FIDELITY METAMODELS, AND SWARM INTELLIGENCE

A multi-objective simulation-based design optimization (SBDO) is presented for the resistance reduction and displacement increase of a small water-plane area twin hull (SWATH). The geometry is realized as a parametric model with the CAESES software, using 27 design parameters. Sobol sampling is used to realize design variations of the original geometry and provide data to the design-space dimensionality reduction method by Karhunen-Loeve expansion. The hydrodynamic performance is evaluated with the potential flow code WARP, which is used to train a multi-fidelity metamodel through an adaptive sampling procedure based on prediction uncertainty. Two fidelity levels are used varying the computational grid. Finally, the SWATH is optimized by a multi-objective deterministic version of the particle swarm optimization algorithm. The current SBDO procedure allows for the reduction of the design parameters from 27 to 4, resolving more than the 95% of the original geometric variability. The metamodel is trained by 117 coarse-grid and 27 fine-grid simulations. Finally, significant improvements are identified by the multi-objective algorithm, for both the total resistance and the displacement.