A prediction method for ship added resistance based on symbiosis of data-driven and physics-based models. (15th September 2022)

• Record Type: Journal Article
• Authors: Yang, Ke; Duan, Wenyang; Huang, Limin; Zhang, Peixin; Ma, Shan
• Is Part Of: Ocean engineering. Volume 260(2022)
• Publication Date: 2022-09-15
• Subjects: Added resistance in head waves -- Symbiotic model -- Data-driven -- 2D strip method; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2022.112012
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: The prediction of ship added resistance in waves is an important part of ship design and operation. A prediction method that takes into account accuracy, efficiency, robustness and generalization ability is required. In this study, a prediction method for ship added resistance in head waves based on symbiosis of data-driven and physics-based models is proposed. The physics-based model is constructed based on the 2D strip method to provide physics-based information and constraints. The data-driven module is constructed based on the fully connected neural network structure and the radial basis function, providing a data-driven model parameter optimization framework. Results indicate that the Data-driven and Physics-based Symbiotic Model (DPSM) has an obvious adaptive correction effect on the results of its embedded physics-based model, and has a stronger generalization ability than the fully data-driven model. Prediction results of the DPSM are in good agreement with experimental results, whether for the training ships or the unfamiliar testing ships. Finally, a posteriori model parameter analysis shows the reason why the DPSM has advantages of accuracy and generalization ability. Highlights: The proposed symbiotic model achieves a balance of computational efficiency, accuracy, robustness and generalization ability. The data-driven module can supplement shortcomings of accuracy of the embedded physics-based model. Physics-based information from the physics-based model can expand the extrapolation capability of the data-driven module. Posteriori model parameter analysis shows the reason why the symbiotic model has advantages of accuracy and generalization.