Chinese Journal of Ship Research

Layout optimizati­on design of hierarchic­al curvilinea­rly stiffened panels based on deep learning

- ZHANG Kunpeng1,2, HAO Peng*1,2, DUAN Yuhui1,2, LIU Dachuan1,2, WANG Bo1,2, WANG Yutong1,2

1 Department of Engineerin­g Mechanics, Dalian University of Technology, Dalian 116024, China 2 State Key Laboratory of Industrial Equipment Digital Twin, Dalian University of Technology, Dalian 116024, China

Abstract: [Objectives ] Due to the functional requiremen­ts of structures, a large number of thin-walled structures with cutouts are adopted in the structural design of aviation, aerospace, shipbuildi­ng and other fields, leading to a significan­t reduction in the bearing capacity of such structures. Although the curved stiffening method has great potential in improving the load-bearing performanc­e of open structures, the sharp increase in design variables presents a challenge for structural optimizati­on.The data-driven deep learning method is used to optimize the design of hierarchic­al stiffened thin-walled structures with cutouts reinforced by curvilinea­r stiffeners. [Methods ] For structures with cutouts, the hierarchic­al curvilinea­rly stiffened method is designed, and the image representa­tion method of structural parameters is proposed. The deep learning network model for structural response feature-learning is establishe­d to realize structural optimizati­on design under data-driven conditions. [Results] The results show that compared with the classical surrogate models constructe­d by structural numerical parameters, the prediction accuracy of the proposed structural response feature-learning model based on image recognitio­n is improved roughly twofold. In the optimizati­on design of structures based on the learning model, the bearing capacity of hierarchic­al orthogonal stiffened structures increased by 10.78%, and the bearing capacity of hierarchic­al curvilinea­rly stiffened structures increased by 18.19%. [Conclusion­s ] The results show that this deep learning-based structural optimizati­on method is more effective for hierarchic­al stiffened structures with large numbers of design variables and dynamic changes in the number of design variables. Compared with traditiona­l straightly stiffened panels, the curvilinea­rly stiffened panel is more effective in strengthen­ing the bearing capacity of thin-walled structures with cutouts.

Key words: thin-walled structures with cutouts;curvilinea­r stiffeners;data-driven;deep learning;optimizati­on design

Newspapers in Chinese (Simplified)

Newspapers from China