Loading...

Development of Parallel Algorithm for Adjoint Optimization of Turbine Blade

Salehi, Hadi | 2012

588 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43545 (45)
  4. University: Sharif University of Technology
  5. Department: Aerospace Engineering
  6. Advisor(s): Mazaheri, Karim
  7. Abstract:
  8. This study develops an aerodynamic shape optimization code using parallel processing capability based on a gradient-based adjoint method. Calculation of the gradient of the objective function with respect to design variables is the most costly part of the gradient-based optimization algorithms. Applying adjoint methods, gradients can be calculated with solving some additional equations known as the "adjoint equations", instead of direct calculation. In this study, the blade shape optimization is performed by inverse design method and using steepest descent optimization algorithm. The objective function of inverse design problem is the desired blade surface pressure distribution. In each design cycle, both of the flow and adjoint equations are solved. In this regard, a flow solver and an adjoint solver are developed. The computational cost of solving both of flow and adjoint equations is almost the same. Flow equations and adjoint equations are discretized respectively by AUSM+ and Jameson methods and they are solved in multi-block H-O-H type grid. The geometry of blade airfoil is parameterized by Bezier curves and its control points are considered as design variables. In the first step, the validity of flow solver is tested using numerical and experimental data. Then, inviscid and viscous inverse design problems are solved. In the inviscid and viscous inverse design problems, maximum error of final blade in comparison with target blade is 0.00108% and 0.049%, respectively. By running the inverse design code on 12 processors, running speed was more than 9 times that on single processor. Also, applying simultaneous convergence and gradient extrapolation methods, inverse design convergence is accelerated
  9. Keywords:
  10. Inverse Design ; Parallel Processing ; Turbine Blades ; Adjoint Equations ; Flow Equations

 Digital Object List

 Bookmark

No TOC