Advanced CFD modelling of multi-stage axial compressors

The axial compressors of power-generation gas turbines have a high stage count, blades with low aspect ratios and relatively large clearances in the rear section. These features tend to promote the development of intense secondary flows that, coupled with the severe diffusion characterizing these machines, make the CFD modelling of axial compressors a great challenge. The focus of this thesis is on the use and tuning of modern CFD methods able to assess a reliable numerical setup to be used by the industry as a base for the design of efficient and high-performing multi-stage axial compressors. The first part of the work includes the details about the development and implementation of a highly conservative non-reflecting mixing plane model in the in-house CFD code TRAF, able to handle both the perfect and the real gas case. Later, the numerical setup for steady-state multi-stage compressor simulations is presented, providing also details about some crucial aspects of the compressor modelling, namely shroud leakages and clearances. The setup has been validated against experimental data on a GT compressor designed by Ansaldo Energia. To determine the influence of mixing plane models on performance prediction, unsteady full-annulus simulations have been performed at two different operating conditions: design point and near-stall. Finally, the last part of the thesis is dedicated to well-known phenomenon of radial mixing in axial compressors. The physical causes of radial mixing are discussed in depth, leading to the conclusion that a state-of-the-art, unsteady calculation of the full compressor is able to provide very strong evidence of radial mixing. A special attention is devoted to the evaluation of what is lost in the compressor modelling due to the assumption of a steady-state picture of the flow. In order to do this, the high-pressure section of a heavy-duty axial compressor of the Ansaldo Energia fleet, characterized by really high clearances, is considered. The results of steady and unsteady RANS simulations are compared with experimental data, showing that only adopting an unsteady approach, the enhanced radial mixing of this peculiar application can be properly captured. On the contrary, the steady-state modelling leads to a strong underestimation of the radial transport phenomenon. A possible explanation for this is provided after examining what occurs across the inter-row interfaces for RANS and URANS solutions: the stream-wise vorticity associated with clearance flows is one of the main drivers of radial mixing and restraining it by pitch-averaging the flow at mixing planes is the reason why the RANS approach is not able to properly predict the radial transport of fluid properties in the rear part of the axial compressor.