Influence of Axial Compressor Model Simplification and Mesh Density on Surge Margin Evaluation

The menace of surge occurrence in the compressors is taken very seriously and its avoidance became a fundamental for the design of any modern jet engines. Nowadays, a problem with appropriate evaluation of the compressor surge margin while considering different simplifications of three-dimensional CFD model is still present. For that purpose, this article presents a comparison between the measurement data and several variants of 3D CFD models characterized by a specific mesh density. To calculate all the results on which the comparisons and conclusions are based, an 8-stage axial compressor is taken into account. Flow conditions of the machine are computed for three part load speeds: The low, the mid and the high one respecting the variable guide vanes schedule fitted to the specific load. For each of speed variants a four mesh configurations were generated: coarse, medium, fine and extra-fine. All speed configurations were treated with two different turbulence models – Wilcox k-ω and Menter’s SST k-ω, giving ultimately 15 CFD models, calculated with the TRACE solver using an initialization based on a circumferentially averaged flow solution delivered by the Streamline Curvature Method. During the study an additional assessment of reference grid independence was performed and the mesh convergence has been achieved. A comparison between turbulence models and the measurement proves that SST turbulence model is not well distributed through the speeds in compare to the measurement data and the Wilcox turbulence model. Inconsistency of sensitivity in the mesh coarsening for different rotational speeds was found. Increasing the mesh roughness level has to be executed for each speed separately. Overall compressor map shows that shift of the Pressure Ratio and the Mass Flow decreases with lower rotational speed. Neglecting the system add-ons like labyrinth sealing volumes, bleed-ports and other leakages has a visible influence on deviations from the measurements. Because of intended future use in design and optimization the “Medium” grid with Wilcox k-ω turbulence model was chosen, being a good representation of the Rig characteristics with reduction of the computing time.