Journal Article

Design Optimization for Self-Propulsion of a Bulk Carrier Hull Using a Discrete Adjoint Method


P. He, G. Filip, J. R. R. A. Martins, and K. J. Maki


Computers & Fluids, 192104259, 2019



Computational fluid dynamics (CFD) based optimization is becoming increasingly popular in hydrodynamic design of ship hulls because it provides a fully automatic framework with a shorter design cycle than a human-supervised design tool. Despite the above advantage, CFD-based optimization requires careful attention to relevant design considerations, such that the final design is useful in practice. These considerations include all relevant objectives (such as drag and wake distortion) and constraints (such as volume, thickness, and curvature). Although constraints have been included in previous hull shape optimization studies, these studies have typically considered only one objective. To address this shortcoming, we conduct design optimization for self-propulsion by simultaneously considering drag and propeller-wake distortion. We use a gradient-based optimization framework that includes a discrete adjoint method for efficient derivative computation, which allows us to use a large number of design variables to parameterize the complex hull shape and thus gain a large amount of freedom for geometric modification. We impose appropriate geometric constraints (volume, thickness, and curvature) on the hull surface to ensure a practical design. In addition, we use a weighted objective function that includes drag and wake distortion to construct a Pareto front with five optimizations. We also consider hull-propeller interaction by comparing optimization results with and without a propeller. We use the Japan bulk carrier (JBC) as the baseline model and focus on optimizing the stern region. We find that optimizing for only one objective results in a large penalty on the other objective, whereas a weighted objective balances the penalty and achieves simultaneous improvement in drag and wake distortion. Moreover, we observe that the suction effect of the propeller suppresses the flow separation near the bilge tube and smooths out the velocity distortion at the propeller plane; these are effects that would end up affecting the optimized shapes. Our results demonstrate that it is necessary to simultaneously consider drag and wake distortion in hull-shape-optimization studies, and that constrained shape optimization with a large number of design variables is possible with the discrete-adjoint method.