Aerodynamic Design Optimization of an Axial Flow Compressor Stator Using Parameterized Free-Form Deformation

Abstract

AbstractThis paper describes an aerodynamic design optimization of a highly loaded compressor stator blade using parameterized free-form deformation (FFD). The optimization methodology presented utilizes a B-spline-based FFD control volume to map the blade from the object space to the parametric space via transformation operations in order to perturb the blade surface. Coupled with a multi-objective genetic algorithm (MOGA) and a Gaussian process-based response surface method (RSM), a fully automated iterative loop was used to run the optimization on a fitted correlation function. A weighted average reduction of 6.1% and 36.9% in total pressure loss and exit whirl angle was achieved, showing a better compromise of objective functions with smoother blade shape than other results obtained in the open literature. Data mining of the Pareto set of optimums revealed four groups of data interactions of which some design variables were found to have skewed scatter relationship with objective functions and can be redefined for further improvement of performance. Analysis of the flow field showed that the thinning of the blade at midspan and reduction in camber distribution were responsible for the elimination of the focal-type separation vortex by redirecting the secondary flow in an axially forward direction toward the midspan and near the hub endwall downstream. Furthermore, the reduction in exit whirl angle especially at the shroud was due to the mild bow shape which generated radial forces on the flow field thereby reducing the flow diffusion rate at the suction surface corner. This effect substantially delayed or eliminated the formation of corner separation at design and off-design operating conditions. Parameterized FFD was found to have superior benefits of smooth surface generation with low number of design variables while maintaining a good compromise between objective functions when coupled with a genetic algorithm.