Prescreening surrogate-model-assisted multi-objective aerodynamic optimization design of highly loaded axial compressor in heavy-duty gas turbine

Abstract

To address the challenges of numerous design variables and high evaluation costs in the aerodynamic optimization of advanced highly-loaded axial compressors, this paper presents an axial compressor aerodynamic optimization platform based on directly manipulated free-form deformation (DFFD) parameterization method, a data-driven multi-objective differential evolution optimizer, and computational fluid dynamics. The DFFD method allows for direct geometric manipulation of the target compressor, enabling efficient parameterization of the blades and flow path and reducing the number of design variables. A prescreening surrogate-model-assisted multi-objective differential evolution (Pre-MODE) optimizer is developed to obtain competitive solutions within a limited number of iterations. The surrogate model in Pre-MODE conducts real evaluations only on selected individuals in each generation, thereby reduce evaluation costs while ensuring prediction accuracy. Additionally, a parallel task management system is integrated to improve optimization efficiency. To verify the effectiveness of the optimization platform, a 1.5-stage highly-loaded compressor is aerodynamically optimized. After optimization, the adiabatic efficiency at the design point and the surge margin are increased by 2.42% and 9.04%, respectively. These results confirm the effectiveness and speed of the aerodynamic optimization platform, promoting the establishment of an integrated optimization design system for axial compressors.