Modern gas turbine endwall is operating in harsher conditions for the application of low NOx combustor. Non-axisymmetric endwall has been extensively studied for aerodynamic performance improvement, because endwall contouring can decrease the pressure gradient between the pressure side (PS) and the suction side (SS) in the blade passage. In addition to the influence of pressure gradient on aerodynamic losses, the vortical structures induced by pressure gradient are also the sources of high heat transfer regions in the passage. Consequently, thermal loads might be reduced by decreasing the pressure gradient thus weakening the strength of the secondary flows. In terms of engineering applications, distribution of thermal load is very important for the design of endwall cooling scheme, and it is necessary to take both aerodynamic and heat transfer performances into consideration for the endwall profile design. In this work, aero-thermal coupled design optimization of a turbine blade endwall was carried out. The endwall contour was obtained by multiplying heights of two curves in the streamwise and pitchwise directions. The streamwise curve was controlled by non-uniform B-spline (NUBS) and the pitchwise one was obtained by employing the sinusoidal function. The optimization method adopted in this research was the multi-objective genetic algorithm (MOGA) coupled with Kriging (KRG) model, which has been validated by benchmark functions. Numerical validation shows that static pressure coefficients on the blade surfaces and the Nusselt number (Nu) on the endwall agree well with the experimental results. The design variables were the endwall profile parameters, and the objective functions were maximizing total pressure recovery coefficient (ξ) at the blade outlet and minimizing the Nu on the endwall. Two optimal cases were selected from the Pareto front and analyzed in detail. It is indicated that the turbine blade aerodynamic performance can be improved while the heat transfer is restrained simultaneously. For the optimal Case I, mass flow-averaged ξ increases by 0.88%, and for Case II, area-averaged Nu reduces by about 7.78%.

This content is only available via PDF.
You do not currently have access to this content.