SHAPE OPTIMIZATION USING THE ONE-SHOT ADJOINT TECHNIQUE ON GRAPHICS PROCESSING UNITS

This paper presents the implementation of the one-shot technique for the solution of design/shape optimization problems on NVIDIA Graphics Processing Units (GPUs), for incompressible fluid flow problems. One-shot optimization techniques are based on the simultaneous solution of the flow, adjoint and shape correction equations. They are efficient alternatives to standard gradient-based algorithms in which, within each cycle, the flow and the adjoint equations are solved the one after the other, followed by the update of the shape using the computed gradient of the objective function. The aim of this paper is to superimpose the parallel speedup gained by GPU-enabling the corresponding software to the gain in efficiency offered by the one-shot algorithm. To this end, the experience of the authors’ group in porting Navier-Stokes solvers for compressible fluid flows to GPUs is exploited. Key features of the new method are the use of the flow equations for incompressible fluids and the solution of the coupled flow-adjoint equations on the GPU. The programmed software is used for the shape optimization of the tubes of a heat exchanger and an elbow duct.

[1]  Jonathan Cohen,et al.  Title: A Fast Double Precision CFD Code using CUDA , 2009 .

[2]  P. Roe.  Approximate Riemann Solvers, Parameter Vectors, and Difference Schemes , 1997 .

[3]  A. Jameson,et al.  Aerodynamic Shape Optimization of Airfoils in Ultra-Low Reynolds Number Flow using Simultaneous Pseudo-Time Stepping , 2007 .

[4]  Kyriakos C. Giannakoglou,et al.  Continuous adjoint approach to the Spalart–Allmaras turbulence model for incompressible flows , 2009 .

[5]  Yong Zhao,et al.  A general method for simulation of fluid flows with moving and compliant boundaries on unstructured grids , 2003 .

[6]  Antony Jameson,et al.  One-Shot Pseudo-Time Method for Aerodynamic Shape Optimization Using the Navier-Stokes Equations , 2007 .

[7]  Kyriakos C. Giannakoglou,et al.  CFD-based analysis and two-level aerodynamic optimization on Graphics Processing Units , 2010 .

[8]  Michael B. Giles,et al.  The harmonic adjoint approach to unsteady turbomachinery design , 2002 .

[9]  Rainald Löhner,et al.  Running unstructured grid‐based CFD solvers on modern graphics hardware , 2011 .

[10]  K. Giannakoglou,et al.  Aerodynamic Shape Optimization Using First and Second Order Adjoint and Direct Approaches , 2008 .

[11]  W. K. Anderson,et al.  Aerodynamic design optimization on unstructured grids with a continuous adjoint formulation , 1997 .

[12]  Kyriakos C. Giannakoglou,et al.  Unsteady CFD computations using vertex‐centered finite volumes for unstructured grids on Graphics Processing Units , 2011 .

[13]  Kyriakos C. Giannakoglou,et al.  A continuous adjoint method with objective function derivatives based on boundary integrals, for inviscid and viscous flows , 2007 .

[14]  Eric Darve,et al.  Large calculation of the flow over a hypersonic vehicle using a GPU , 2008, J. Comput. Phys..

[15]  Jostein R. Natvig,et al.  Solving the Euler Equations on Graphics Processing Units , 2006, International Conference on Computational Science.

[16]  P. Spalart.  A One-Equation Turbulence Model for Aerodynamic Flows , 1992 .

[17]  Kyriakos C. Giannakoglou,et al.  CUDA Implementation of Vertex-Centered, Finite Volume CFD Methods on Unstructured Grids with Flow Control Applications , 2012 .

[18]  Graham Pullan,et al.  Acceleration of a 3D Euler solver using commodity graphics hardware , 2008 .

[19]  Antony Jameson,et al.  Aerodynamic design via control theory , 1988, J. Sci. Comput..

[20]  Jaime Peraire,et al.  Aerodynamic design using unstructured meshes , 1996 .

[21]  K. Giannakoglou,et al.  The continuous direct-adjoint approach for second order sensitivities in viscous aerodynamic inverse design problems , 2009 .