Incorporation of High-Fidelity Flow Field Information into Preliminary Design of Multi-Stage Axial Compressors

Abstract

This thesis establishes an axisymmetric methodology that incorporates pre-performed high-fidelity CFD into the performance estimation of multi-stage axial compressors during preliminary design. Its key differentiator is that radial non-uniformity, inferred from three-dimensional CFD and represented using orthonormal basis functions, replaces empirical correlations of blockage, loss, and deviation as well as simplified models of flow features, such as boundary-layer growth, spanwise mixing, and endwall-corner separation. The methodology includes the effects of changes in radial non-uniformity and in blade geometry on the axisymmetric flow field. The approach can supersede current throughflow methods, increasing the fidelity of preliminary design.

The primary impact of the methodology is a new capability for power gas turbine compressors to rapidly assess off-design matching at different spanwise locations along the blade height, enabling early-design choices, such as the annulus-area scheduling, based on the fidelity of CFD. Over a range of off-design conditions from near stall to near choke, the massflow capacity of a four-stage compressor was estimated within 1.2% and its efficiency within 1.5 percentage points compared to CFD at equal loading. The estimation of quasi-one-dimensional performance and the characterization of the flow close to the endwalls are improved relative to estimations of a legacy streamline curvature method since radial non-uniformity is inferred from high-fidelity flow field information. The methodology is demonstrated to be suitable for incorporation into compressor design systems.