ABSTRACT
Aerodynamic performance analysis is crucial for the design and maneuvering of sailboats. The calculation of aerodynamic coefficients using Computational Fluid Dynamics (CFD) methods requires a significant amount of time and computational resources. Therefore, this article proposes a Multi-Scale Residual 3D UNet (MSR UNet) model to rapidly predict the overall aerodynamic coefficients (i.e. the longitudinal force coefficient and lateral force coefficient) of a sailboat. First, multi-scale residual learning is applied to extract multi-scale features of the pressure field, channel and spatial attention are introduced to enhance the perception of key features such as pressure field boundary and pressure distribution. Subsequently, Residual-Subsampled Blocks (RSB) as well as Residual-Inception Blocks (RIB) are designed to extract local and global information of the pressure field by merging feature maps of different scales. Finally, using the high-dimensional pressure field as a transition, a mapping from the operating parameters (i.e. the ship relative wind angle, main sail angle and foresail angle) to the aerodynamic coefficients is established based on the feedforward neural network (FNN). The model is trained and tested using pressure fields and aerodynamic data obtained from 3D steady RANS CFD analysis. Experimental results demonstrate that our model not only possesses stronger multi-scale feature extraction performance but also achieves high-precision and real-time prediction of sailboat aerodynamic coefficients.
- Heebum Lee, Mi Yeon Park, Sunho Park, and Shin Hyung Rhee. 2016. Prediction of velocity and attitude of a yacht sailing upwind by computational fluid dynamics. Int. J. of Nav. Arch. And Oc. Eng. 8, 1 (January 2016), 1-12. https://doi.org/10.1016/j.ijnaoe.2016.01.003Google ScholarCross Ref
- S. Arunvinthan, V.S. Raatan, S. Nadaraja Pillai, Amjad A. Pasha, M. M. Rahman, and Khalid A. Juhany. 2021. Aerodynamic characteristics of shark scale-based vortex generators upon symmetrical airfoil. Energies 14, 7 (March 2021), 1808. https://doi.org/10.3390/en14071808Google ScholarCross Ref
- M. S. Eldred and D. M. Dunlavy. 2006. Formulations for surrogate-based optimization with data fit, multifidelity, and reduced-order models. In Proceedings of the 11th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, Portsmouth, VA, USA, 6–8 (September 2006), 7117. https://doi.org/10.2514/6.2006-7117Google ScholarCross Ref
- Ricardo Vinuesa and Steven L. Brunton. 2022. Enhancing computational fluid dynamics with machine learning. Nat. Comput. Sci. 2, 6 (June 2022), 358–66. https://doi.org/10.1038/s43588-022-00264-7Google ScholarCross Ref
- Yanxuan Zhao, Chengwen Zhong, Fang Wang, and Yueqing Wang. 2022. Visual Explainable Convolutional Neural Network for Aerodynamic Coefficient Prediction. Int. J. of Aero. Eng. (December 2022). https://doi.org/10.1155/2022/9873112Google ScholarCross Ref
- Tun Zhao, Weiqi Qian, Jie Lin, Hai Chen, Houjun Ao, Gong Chen, and Lei He. 2023. Learning Mappings from Iced Airfoils to Aerodynamic Coefficients Using a Deep Operator Network. J. of Aero. Eng. 36, 5 (May 2023), 04023035. https://doi.org/10.1061/JAEEEZ.ASENG-4508Google ScholarCross Ref
- Rupert Pache and Thomas Rung. 2022. Data-driven surrogate modeling of aerodynamic forces on the superstructure of container vessels. Eng. App. of Comput. Fl. Mech. 16, 1 (March 2022), 746-763. https://doi.org/10.1080/19942060.2022.2044383Google ScholarCross Ref
- Kuijun Zuo, Zhengyin Ye, Weiwei Zhang, Xianxu Yuan, and Linyang Zhu. 2023. Fast aerodynamics prediction of laminar airfoils based on deep attention network. Physics of Fluids 35, 3 (March 2023). https://doi.org/10.1063/5.0140545Google ScholarCross Ref
- Renkun Han, Yixing Wang, Yang Zhang, and Gang Chen. 2019. A novel spatialtemporal prediction method for unsteady wake flows based on hybrid deep neural network. Physics of Fluids 31, 12 (November 2019). https://doi.org/10.1063/1.5127247Google ScholarCross Ref
- Olaf Ronneberger, Philipp Fischer, and Thomas Brox. 2015. U-net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, 234-241. https://doi.org/10.1007/978-3-319-24574-4_28Google ScholarCross Ref
- Mateus Dias Ribeiro, Abdul Rehman, Sheraz Ahmed, and Andreas Dengel. 2020. DeepCFD: Efficient steady-state laminar flow approximation with deep convolutional neural networks. Fluid Dynamics (May 2020). https://doi.org/10.48550/arXiv.2004.08826Google ScholarCross Ref
- Li-Wei Chen, Berkay A. Cakal, Xiangyu Hu, and Nils Thuerey. 2021. Numerical investigation of minimum drag profiles in laminar flow using deep learning surrogates. Journal of Fluid Mechanics (June 2021), A34. https://doi.org/10.1017/jfm.2021.398Google ScholarCross Ref
- Sam Jacob Jacob, Markus Mrosek, Carsten Othmer, and Harald Köstler. 2021. Deep learning for real-time aerodynamic evaluations of arbitrary vehicle shapes. arXiv preprint. https://doi.org/10.4271/15-15-02-0006Google ScholarCross Ref
- Yuxin Wu and Kaiming He. 2018, Group Normalization. In Proceedings of the European conference on computer vision (ECCV). Munich, Germany, 3-19. https://doi.org/10.48550/arXiv.1803.08494Google ScholarCross Ref
- Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. 2018. Cbam: Convolutional block attention module. In Proceedings of the European conference on computer vision. Munich, Germany, 3-19. https://doi.org/10.48550/arXiv.1807.06521Google ScholarDigital Library
- Juncheng Li, Faming Fang, Kangfu Mei, and Guixu Zhang. 2018. Multi-scale residual network for image super-resolution. In Proceedings of the European conference on computer vision. Munich, Germany, 517-532. http://dx.doi.org/10.1007/978-3-030-01237-3_32Google ScholarDigital Library
- Panqu Wang, Pengfei Chen, Ye Yuan, Ding Liu, Zehua Huang, Xiaodi Hou, and Garrison Cottrell. 2018. Understanding convolution for semantic segmentation. In 2018 IEEE winter conference on applications of computer vision. Lake Tahoe, NV, USA, 1451-1460. http://dx.doi.org/10.1109/WACV.2018.00163Google ScholarCross Ref
- Nabil Ibtehaz and M. Sohel Rahman. 2020. MultiResUNet: Rethinking the U-Net architecture for multimodal biomedical image segmentation. Neural networks 121 (January 2020), 74-87. https://doi.org/10.1016/j.neunet.2019.08.025Google ScholarDigital Library
- Özgün Çiçek, Ahmed Abdulkadir, Soeren S. Lienkamp, Thomas Brox, and Olaf Ronneberger. 2016. 3D U-Net: learning dense volumetric segmentation from sparse annotation. Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, 17-21. https://doi.org/10.48550/arXiv.1606.06650Google ScholarCross Ref
- Ozan Oktay, Jo Schlemper, Loic Le Folgoc, Matthew Lee, Mattias Heinrich, Kazunari Misawa, Kensaku Mori, Steven McDonagh, Nils Y Hammerla, Bernhard Kainz, Ben Glocker, and Daniel Rueckert. 2018. Attention u-net: Learning where to look for the pancreas. arXiv preprint. https://doi.org/10.48550/arXiv.1804.03999Google ScholarCross Ref
Index Terms
- Fast Prediction of Sailboat Aerodynamic Coefficients using Multi-Scale Residual 3D Unet
Recommendations
A multi-scale network method for two-phase flow in porous media
Pore-network models of porous media are useful in the study of pore-scale flow in porous media. In order to extract macroscopic properties from flow simulations in pore-networks, it is crucial the networks are large enough to be considered ...
Multidisciplinary and multi-scale computational field simulations-Algorithms and applications
This paper describes the development and the validation of two computational fluid dynamics (CFD) frameworks for multidisciplinary and multi-scale simulations. The first framework is an overset grid topology with a six-degree of freedom rigid body ...
Comments