| Abstract | Surrogate modeling is a method that uses a simplified model of an actual model to drive results with reduced computational cost while maintaining accuracy as much as possible. The surrogate model can predict the results quickly without performing complex analysis. In this study, surrogate modeling method is applied to ship flooding analysis. If a ship is damaged on the sea, the ship may sink and/or cause oil spillage, resulting in human casualties and environmental impact. The casualties of flooding can be minimized if the ship’s stability and the time to reach equilibrium is predicted within a short time when the damage occurs. However, since the flooding analysis is a time domain simulation with complexity, it takes too long time to be used in such circumstance. To reduce the computational cost of estimating ship’s stability, surrogate modeling method is used. To construct the surrogate model, various types of a deep artificial neural network are tried. Each result of Kim’s flooding analysis is used to train the neural network. In result, accurate estimations are obtained by the surrogate model within a dramatically reduced time. |
|---|---|
| Publication Date | 2017-12-15 |
Jisang Ha, Myung-Il Roh, Sung-Jun Lee, Ki-Su Kim, Seung-Min Lee, "Toward Rapid Analysis Using Surrogate Model by Deep Learning and Application to Ship Flooding Analysis", Proceedings of ISCDE(International Symposium on Computational Design and Engineering) 2017, Ho Chi Minh, Vietnam, pp. 1-2, 2017.12.13-16