2018/3/11 Flow Analysis around Wing/Body 圆上活大学 Numerical Optimization 国上清大学 Definition of optimization problem Subject to =argminf() and Aerodynamic Shape Optimization g(x)20 hs)=0 f) CFD Optimizers Classification of Optimization Trade off between Accuracy and Methods 圈上活大坐 Costs 国上大坐 Accurate CFD codes (RANS)are computational expensive to run RANS code necessary for flows over complex geometries,high AOA, flow separation,etc. rithm ·Solutions -Fast mathematical models using interpolation or approximation Simulated Annealing functions Hill-Climbing -High performance computing,using facilities such as compute clusters, based on mon,widely used Quasi-Newton Ootiaization ://www-fpmaoc/ide// oerandenn ong Some issues using CFD 国上清大学 Software and resources 园上海发大坐 。 Grid Convergence Check CFD software was built upon physics,modeling,numerical Comparison with experimental data when available methods Use of Multiple CFD code for cross-validation Two types of available software Automation for Design Optimization -Commercial (e.g.,FLUENT,CFX,Star-CD) Automatic,robust mesh generation Research (e.g.,CFDSHIP-IOWA,U2RANS) -Calculation of sensitivity information More information on CFD can be got on the following website: CFD Online:http://www.cfd-online.com/ CFD are putational Dynamics:http://www.cd.co.uk/ CFX/AEA:http://www.software_aeat.com/cfx/ -Grid generation software .Gridgen:http://www.pointwise.com GridPro:http://www.gndpro.com/ Visualization software 132018/3/11 13 © Shanghai Jiao Tong University – Dr. Wenbin Song School of Aeronautics and Astronautics Flow Analysis around Wing/Body © Shanghai Jiao Tong University – Dr. Wenbin Song School of Aeronautics and Astronautics Numerical Optimization • Definition of optimization problem • Subject to • and • Aerodynamic Shape Optimization CAD Parametric model CFD Optimizers Initial values x x opt  argmin f gx  0 hx  0 f x  © Shanghai Jiao Tong University – Dr. Wenbin Song School of Aeronautics and Astronautics Classification of Optimization Methods http://www-fp.mcs.anl.gov/OTC/Guide/OptWeb/ Commonly used methods: • Evolutionary algorithms • Genetic Algorithms • Simulated Annealing • Hill-Climbing • SQP • Newton • Quasi-Newton • etc. © Shanghai Jiao Tong University – Dr. Wenbin Song School of Aeronautics and Astronautics Trade off between Accuracy and Costs • Accurate CFD codes (RANS) are computational expensive to run – RANS code necessary for flows over complex geometries, high AOA, flow separation, etc. • Solutions – Fast mathematical models using interpolation or approximation functions – High performance computing, using facilities such as compute clusters, based on • MPI, common, widely used • OpenMP, certain type of applications • GPU, developing fast © Shanghai Jiao Tong University – Dr. Wenbin Song School of Aeronautics and Astronautics Some issues using CFD • Grid Convergence Check • Comparison with experimental data when available • Use of Multiple CFD code for cross-validation • Automation for Design Optimization – Automatic, robust mesh generation – Calculation of sensitivity information © Shanghai Jiao Tong University – Dr. Wenbin Song School of Aeronautics and Astronautics Software and resources • CFD software was built upon physics, modeling, numerical methods • Two types of available software – Commercial (e.g., FLUENT, CFX, Star-CD) – Research (e.g., CFDSHIP-IOWA, U2RANS) • More information on CFD can be got on the following website: – CFD Online: http://www.cfd-online.com/ – CFD software • FLUENT: http://www.fluent.com/ • CFDRC: http://www.cfdrc.com/ • Computational Dynamics: http://www.cd.co.uk/ • CFX/AEA: http://www.software.aeat.com/cfx/ – Grid generation software • Gridgen: http://www.pointwise.com • GridPro: http://www.gridpro.com/ – Visualization software • Tecplot: http://www.amtec.com/ • Fieldview: http://www.ilight.com/ 78