Propulsion-Driven Surrogate-Assisted CFD Optimization for Preliminary Stage Ship Hull Form Design

This thesis presents a novel methodology for optimizing hull forms based on their propulsion performance. An automated optimization framework is provided in such a way that applies to any potential ship type as follows. Hull geometries are to be parameterized to define the design exploration space. A set of constraints shall be included to establish the hull feasibility requirements. Designs are evaluated employing self-propulsion unsteady Reynolds-averaged Navier-Stokes (URANS) simulations that adapt to the input hulls based on their characteristic dimensions. Thus, the single-objective value to optimize for is computed as an operational profile-based weighted function of delivered power requirements estimated per accounted loading condition. A Bayesian optimization algorithm iteratively models the black-box constraints and objective function data using Radial Basis Function surrogates. Consequently, an acquisition function that balances exploration and exploitation proposes the most promising hull solution. A proof-of-concept is carried out by applying the introduced methodology to optimize a Trailing Suction Hopper Dredger hull of 96.2 m length between perpendiculars and 26.3 m breadth with two azimuth ducted propellers. Given the characteristics of this type of vessel, this case study is expected to highlight the advantages of optimizing for propulsion performance compared to hydrodynamic drag. The design exploration space is defined by six decision parameters: five concerning the hull surface and a sixth that sets the longitudinal position of the propellers. Three sailing conditions are considered to derive the objective value: fully loaded, intermediate and ballast situations. To begin with, a preliminary study conducts self-propulsion simulations on the starting hull under design conditions. According to the established numerical setup, full and symmetric computational domain simulations are compared to determine the suitability of the latter for the subsequent optimization process. Verification is assessed by performing a grid convergence study on both configurations, which confirms a consistent behaviour in the estimated delivered power. Negligible deviations are observed between the configurations, justifying the selection of the symmetric approach. Following the presented optimization methodology, 40 valid hulls are discovered. The most optimized design solution is achieved after 33 iterations, achieving a 17\% decrease in objective value from the starting hull. Lastly, the research compares the performance of the presented propulsion-driven optimization towards a resistance-driven approach that evaluates hull drag. The results demonstrate that the propulsion-driven optimization yields similar optimal hull solutions to those obtained through a resistance-driven technique. Given the close outcomes and the inherently lower computational costs of resistance simulations compared to self-propulsion simulations, resistance-driven optimization remains a more efficient option. Nonetheless, this dissertation only addresses one case study and further research in propulsion-driven optimization methodologies is recommended.