Abstract
Designing efficient cooling systems for the highly thermally loaded blades and vanes of high-pressure-turbines requires minimizing cooling air consumption while maintaining cooling effectiveness. The complex interactions of the multiple design parameters makes the deterministic identification of the optimal cooling design extremely challenging. The application of mathematical optimization algorithms is key to automatically identifying the optimal design accompanied with enormous potential to significantly reducing development time.
In the present study, the cooling mass flow rate of a three-pass cooling system with rib-roughened walls is minimized while approximately preserving the initial temperature distribution in the blade component. Since the geometrical optimization is applied in the preliminary design phase, a 1D, correlation-based flow solver is coupled with a simplified 3D thermal model for the numerical evaluation of the cooling system performance. Prior to the optimization, a Monte-Carlo based sensitivity analysis of the system behaviour to geometric variability is performed. For the optimization the evolutionary, metamodel-assisted algorithm AutoOpti, developed at DLR, is used. Hereby, two sampling strategies, different base population sizes as well as the influence of a reduction of the parametric model are discussed. In all optimizations a significant reduction of the cooling mass flow rate of up to 12 % was achieved, whereas interestingly different design changes were observed between the optimizations.
