ABSTRACT
In aerodynamics related design, analysis and optimization problems, flow fields are simulated using computational fluid dynamics (CFD) solvers. However, CFD simulation is usually a computationally expensive, memory demanding and time consuming iterative process. These drawbacks of CFD limit opportunities for design space exploration and forbid interactive design. We propose a general and flexible approximation model for real-time prediction of non-uniform steady laminar flow in a 2D or 3D domain based on convolutional neural networks (CNNs). We explored alternatives for the geometry representation and the network architecture of CNNs. We show that convolutional neural networks can estimate the velocity field two orders of magnitude faster than a GPU-accelerated CFD solver and four orders of magnitude faster than a CPU-based CFD solver at a cost of a low error rate. This approach can provide immediate feedback for real-time design iterations at the early stage of design. Compared with existing approximation models in the aerodynamics domain, CNNs enable an efficient estimation for the entire velocity field. Furthermore, designers and engineers can directly apply the CNN approximation model in their design space exploration algorithms without training extra lower-dimensional surrogate models.
- M. Ahmed and N. Qin. Surrogate-Based Aerodynamic Design Optimization: Use of Surrogates in Aerodynamic Design Optimization. Aerospace Sciences and Aviation Technology, ASAT-13, 2009.Google Scholar
- J. Anderson. Computational Fluid Dynamics. 1995.Google Scholar
- V. Balabanov, A. Giunta, O. Golovidov, B. Grossman, H. Mason, T. Watson, and T. Haftkalf. Reasonable design space approach to response surface approximation. Journal of Aircraft, 1999.Google Scholar
- M. Batill, A. Stelmack, and S. Sellar. Framework for multidisciplinary design based on response-surfaceapproximations. Journal of Aircraft, 1999.Google Scholar
- Y. Bengio. Learning deep architectures for AI, Foundations and trends in Machine Learning. 2009. Google ScholarDigital Library
- L. Chittka, P. Skorupski, and E. Raine. Speed-accuracy tradeo s in animal decision making. Trends in Ecology and Evolution, 2009.Google Scholar
- D. Daberkow and D. N. New Approaches to Conceptual and Preliminary Aircraft Design: A Comparative Assessment of a Neural Network Formulation and a Response Surface Methodology. World Aviation Conference, 1998.Google ScholarCross Ref
- A. Dosovitskiy, P. Fischer, E. Ilg, P. Hausser, C. Hazrba, V. Golkov, P. v.d. Smagt, D. Cremers, and T. Brox. Flownet: Learning optical ow with convolutional networks. In IEEE International Conference on Computer Vision (ICCV), 2015. Google ScholarDigital Library
- D. Eigen, C. Puhrsch, and R. Fergus. Depth map prediction from a single image using a multi-scale deep network. In Advances in neural information processing systems, 2014. Google ScholarDigital Library
- H. Fang, M. Rais-Rohani, Z. Liu, and M. Horstemeyer. A comparative study of metamodeling methods for multiobjective crashworthiness optimization. Computers & Structures, 2005. Google ScholarDigital Library
- G. Forsythe and W. Wasow. Finite-Difference Methods for Partial Di erential Equations. 1960.Google Scholar
- A. Giunta. Aircraft Multidisciplinary Design Optimization Using Design of Experiments Theory and Response Surface Modeling Methods. 1997.Google Scholar
- S. Gupta, R. Girshick, P. Arbelßez, and J. Malik. Learning rich features from RGB-D images for object detection and segmentation. In Computer Vision-ECCV. 2014.Google ScholarCross Ref
- D. Hartmann, M. Meinke, and W. Schörder. Differential equation based constrained reinitialization for level set methods. Journal of Computational Physics, 2008. Google ScholarDigital Library
- V. Heuveline and J. Latt. The OpenLB project: an open source and object oriented implementation of lattice Boltzmann methods. International Journal of Modern Physics, 2007.Google Scholar
- S. Jeong, M. Murayama, and K. Yamamoto. Efficient optimization design method using kriging model. Journal of Aircraft, 2005.Google Scholar
- Y. Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Girshick, and T. Darrell. Caffe: Convolutional architecture for fast feature embedding. In Proceedings of the ACM International Conference on Multimedia. 2014. Google ScholarDigital Library
- A. Keane and P. Nair. Computational approaches for aerospace design. 2005.Google ScholarCross Ref
- S. J. Leary, A. Bhaskar, and A. J. Keane. Global approximation and optimization using adjoint computational uid dynamics codes. AIAA journal, 2004.Google Scholar
- Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 1998.Google ScholarCross Ref
- M. Mawson, G. Leaver, and A. Revell. Real-time flow computations using an image based depth sensor and GPU acceleration. 2013.Google Scholar
- G. R. McNamara and G. Zanetti. Use of the boltzmann equation to simulate lattice-gas automata. Physical Review Letters, 1988.Google Scholar
- A. Mohamad. Lattice Boltzmann Method. 2011.Google ScholarCross Ref
- S. Patankar. Numerical heat transfer and fluid flow. CRC Press, 1980.Google Scholar
- G. Russo and P. Smereka. A remark on computing distance functions. Journal of Computational Physics, 2000. Google ScholarDigital Library
- J. Schmidhuber. Deep learning in neural networks: An overview. Neural Networks, 2015. Google ScholarDigital Library
- R. Socher, B. Huval, B. Bath, C. D. Manning, and A. Y. Ng. Convolutional-recursive deep learning for 3d object classification. In Advances in Neural Information Processing Systems, 2012.Google Scholar
- D. Tran, L. Bourdev, R. Fergus, L. Torresani, and M. Paluri. Deep end2end voxel2voxel prediction. arXiv preprint arXiv:1511.06681, 2015.Google Scholar
- J. Turner, W. Clough, C. Martin, and J. Topp. Stiffness and deflection analysis of complex structures. Journal of the Aeronautical Sciences, 1956.Google Scholar
- N. Umetani, Y. Koyama, R. Schmidt, and T. Igarashi. Pteromys: interactive design and optimization of free-formed free-flight model airplanes. ACM Trans. Graph., 2014. Google ScholarDigital Library
- K. Valderhaug. The Lattice Boltzmann Simulation on Multi-GPU Systems. 2011.Google Scholar
- S. Wilkinson, G. Bradbury, and H. S. Reduced-Order Urban Wind Interference, Simulation: Transactions of the Society for Modeling and Simulation International. 2015 Google ScholarDigital Library
Index Terms
- Convolutional Neural Networks for Steady Flow Approximation
Recommendations
A numerical method for three-dimensional gas-liquid flow computations
A numerical method for multiphase flow computations based on a finite-difference formulation with a fixed grid is described. The method combines ideas from front tracking and the Ghost Fluid Method, with a numerical technique for velocity extrapolation ...
Comments