The paper explores Kriging-based surrogate model combined with Weighted Expected Improvement approach and for the ship hull form optimization. The training dataset of the Kriging-based surrogate model is obtained by sampling the design space (Design of Experiments, DOE) and performing expensive high-fidelity computations on the selected points. Expected Improvement (EI) is used as a criterion to select one additional sample point in each iteration. The Weighted Expected Improvement (WEI) is derived from EI by adding a tunable parameter which can adjust the weights on exploration and exploitation in the Efficient Global Optimization (EGO).
The proposed method selects more than one new sample point by changing the weight parameter for each optimization iteration, thus it can be performed by parallel computation or multi-computer runs which improves the computational efficiency distinctly. This makes it possible not only to improve the accuracy of the surrogate model, but also to explore the global optimum much more quickly. The present method is applied to mathematical test function and a ship hull form optimization design in order to find the optimal hull form with best resistance performance in calm water in different speeds.
The result shows that the criterion of WEI can be applied in EGO for optimization design and can be easily extended to other hull form optimization design problems based on computational fluid dynamics.