ABSTRACT

This paper introduces a method of using Bayesian Optimization to build an approximate model called Multilayer Perceptron(MLP) in the process of automatic ship hull form optimization. An automatic optimization procedure for ship hull integrates Computer Fluid Dynamics(CFD) solver, ship hull modification tool, optimization algorithm and approximation model. RBF (Radial Basis Function) interpolation is built as a modification tool to modify hull shape. Sensitivity analysis is conducted to the selection of control points. The Multi-island Genetic Algorithm (MIGA) is applied to the optimization. The MLP regression is built as an approximation model to improve the efficiency of the optimization for solving computationally expensive numerical simulations. MLP model has many parameters called hyperparameters, which largely determine the accuracy of the model. The paper presents a training method called Bayesian tuning to obtain these parameters. By training the data collected through Latin Hypercube sampling, the MLP model composed of optimized hyperparameters has higher regression accuracy. The KRISO Container Ship(KCS) has been used as a verification model. As a result, the total resistance of KCS at service speed is reduced by 1.8%. The result shows that this procedure has better effectiveness and less time consumption in the ship hull optimization and the optimized hull form applying this method have lower resistance than the original one.

INTRODUCTION

Ship line design is a complex and comprehensive technology, which involves many technical fields and has strong constraints. It is the basis premise and core link in the general design of ships. It has a decisive impact on the navigation performance. However, the traditional line design method strongly depends on the designer's experience, and consumes a lot of model test and calculation resources. Therefore, it is urgent to develop automatic optimization without manual participation.

SBD(Simulation Based Design) technology opens up a new situation for ship line design. This technology can help design engineers to explore the design space under constraints and automatically get the optimal line design plan. A large number of studies show that using SBD technology to reduce resistance is rapidly growing interest in lines optimization (Campana et al. 2006; Ginnis et al, 2015; Cheng et al, 2018). Fig.1 shows the automatic optimization loop which integrates a parametric modeler, a solver and optimization algorithm. In the process of optimization, the solver performs huge calculations, which are often time-consuming. Therefore, the surrogate model composed of a small amount of data can be used to replace the solver to improve the efficiency.

This content is only available via PDF.
You can access this article if you purchase or spend a download.