Integrated HPC OGV and Pre-Diffuser Design Optimization Using Kriging Surrogate Model

Abstract

The integration of high compressor outlet guide vane (OGV) and combustor pre-diffuser requires some radial turning to be performed within the OGV passage. However, the enhanced loading of OGV leads to the increase in adverse pressure gradient within the OGV passage. Consequently, both the end-wall and blade boundary layers are thickened which could lead to boundary layers separation. In this work, an adaptive global optimization process is applied for the OGV/pre-diffuser system, which combines design of experiment (DOE), Kriging surrogate model and micro genetic algorithm. The meridional flow passage of OGV/pre-diffuser system is parameterized using Bezier curves with the combination of mean line and thickness distribution. In order to prevent the OGV corner separation, the bowed design is applied to the OGV to help delay flow separation. A composite curve combined with two straight lines and a conic Bezier curve is used to represent the OGV stacking line along circumference so that the bowed blades could be parameterized. Aerodynamic performance evaluations of the compressor are performed using a three dimensional Reynolds-averaged Navier-stokes computational fluid dynamics solver — NUMECA. In the optimization process, expected improvement sample criteria is adopted for balancing the exploration and exploitation with Kriging surrogate model. Reasonably high performance is confirmed by comparing the baseline and optimal designs. This study gives some insights into design optimization of an integrated OGV/Pre-diffuser for axial compressor.