International Conference

Jin-Hyeok Kim, Myung-Il Roh, In-Chang Yeo, Ki-Su Kim, "Estimation of the Hydrodynamic Performance of the Parameterized Hull Forms Using Deep Learning", Proceedings of G-NAOE 2022, Changwon, Korea, 2022.11.06-10

by SyDLab posted Jul 27, 2022
?

단축키

Prev이전 문서

Next다음 문서

ESC닫기

크게 작게 위로 아래로 댓글로 가기 인쇄
Extra Form
Abstract The CFD (Computational Fluid Dynamics) analysis is normally used to evaluate the hydrodynamic performance of hull forms in the process of hull form design. The hull form candidates are improved through the iterative process of editing them if their performance is not good enough. However, since the CFD analysis is computationally expensive, the estimation of the hydrodynamic performance using it requires a long time. Due to the limited time for the design, the hull form iterations are not enough to find the optimal hull form. To solve this problem, we proposed a method to estimate the hydrodynamic performance of the hull forms using deep learning. The use of deep learning has the disadvantage that it takes a long time to accumulate data and learn but has the advantage that it takes only a very short time to run the model and obtain the result once training is complete. In this study, the hull form of a small ship was first parameterized with dozens of parameters to generate various hull forms. Then, a number of hull forms were created by randomly generating thousands of parameter sets that determine the hull form. Finally, the hydrodynamic performance for the ground truth was derived by performing CFD analysis on the hull forms. We considered multiple deep learning models to estimate the performance more accurately and selected the best model among them. The proposed method was applied to a small ship. As a result, with the proposed model, the hydrodynamic performance of the hull forms can be estimated shortly with a certain level of error.
Publication Date 2022-11-06

Jin-Hyeok Kim, Myung-Il Roh, In-Chang Yeo, Ki-Su Kim, "Estimation of the Hydrodynamic Performance of the Parameterized Hull Forms Using Deep Learning", Proceedings of G-NAOE 2022, Changwon, Korea, 2022.11.06-10