Manufacturing and Build Variations Modelling for Multi-Stage Axial Compressors: CFD Aerodynamic Predictions

Abstract The numerical prediction of multi-stage compressor flows has historically been a challenge for CFD models, due to their complex, turbulent and three-dimensional nature. High-fidelity models such as LES and DNS have shown promising results, but their application is limited by the high computational cost, while steady-state RANS simulations are still used for the majority of applications in the industry. Over the years, engine manufacturers have developed best-practices and guidelines for multi-stage axial compressor simulations, providing reliable CFD models capable of predictions comparable to test data, both with regards to performance and surge margin at different operating conditions. However, even with state-of-the-art models, several discrepancies can be observed and are generally addressed with calibration based on operational experience. These discrepancies are partially due to known limitations of steady-state CFD, as well as to modelling assumptions and the use of idealised geometries. This paper demonstrates how the accuracy of CFD predictions can be greatly improved by modelling the effect of manufacturing and build variations, such as tip clearances, surface roughness and in-tolerance geometry variations. The resulting engine-build specific models show greatly improved agreement with engine test data, with no increase in computational cost. These models can be used as part of the design process, to develop robust axial compressor designs desensitized to these variations. Moreover, the results from the analyses can be used to re-define tolerance ranges and drawing specifications, to reduce the potential impact on build-to-build engine performance scatter.