Design optimization of the last stage of a 4.5-stage compressor using a POD-based hybrid model

Abstract

The paper presents a hybrid model method based on proper orthogonal decomposition (POD) and its application to aerodynamic design optimization of the last stage of a low-speed 4.5-stage compressor preliminary designed by an in-house design procedure. The stator blade of the last stage is redesigned to reduce the flow separation zones that appear in the suction-endwall corners. POD, recognized as a useful tool to construct the reduced order models, is employed to reconstruct the flow solutions with respect to a given aerodynamic shape in the system space. The flow solutions are described in a weighted summation form of the basis modes. In order to improve the description performance, the weights of basis modes are obtained by a nonlinear regression method in the present paper, rather than the widely used linear regression. The implementations of POD-based hybrid model are presented and the description performance is evaluated and validated firstly. Then the optimization is performed by using the hybrid model through modifying the spanwise stagger angle and the camber of stator blade, while maintaining the mass flow rate. The effects of stagger angle change and camber modification on the performance improvements are presented in detail, demonstrating that the POD-based hybrid model is an easy implemented but valuable and efficient method in aerodynamic design optimization. Finally, data mining technique is employed to discover the design knowledge of the optimization. POD, due to its ability on extracting the dominant characteristics, is used to illustrate the primary basis modes of design parameter changes, further demonstrating the effects of blade redesign on the performance variations.