Large Eddy Simulations in a Transonic Centrifugal Compressor

Abstract

In an attempt to validate a Large Eddy Simulation (LES) approach, computations of a transonic centrifugal compressor with a backswept, unshrouded impeller followed by radial and axial vaned diffusers are performed. A sector composed of one main blade and one splitter blade, two radial diffuser vanes and six axial diffuser vanes is simulated including all the technological effects of the experimental rig. The LES methodology to simulate the rotor/stator configuration is introduced. Emphasis is put on the best trade-off between accuracy of the simulation and affordable CPU cost. A law-of-the-wall boundary condition is used to reduce the mesh size, with a target of y+ around a hundred for all walls except in the tip leakage with y+ around five. Computation of one entire characteristic line is obtained continuously in time: the transient from the flow at rest to the converged points at blockage, peak efficiency, near surge and path to deep surge is computed increasing progressively the outlet pressure as in the experiments. First, LES results are compared to experiments and show excellent agreement both in terms of overall performance and time-averaged internal flow fields previously obtained by Laser Doppler Anemometry. Then, a focus is proposed on the complementary information LES provide in the rotor. The key findings are that contrary to previous URANS studies in this centrifugal compressor, LES capture influential details of the flow structures in the rotor: secondary structures, shock/boundary layer interaction and boundary layer separation at the tip of the impeller. Moreover, it is clearly shown that the tip leakage vortex increases in size and intensity from peak efficiency to surge and becomes much more erratic. Emphasis is put on the causes and consequences of the tip leakage spillage in the neighbouring rotor channels. Pressure fluctuations were also found to increase from peak efficiency to surge downstream the splitter blade leading edge. The whole results finally show that LES with a law-of-the-wall provides excellent results in such a complex case.

Design, Plant Test and CFD Calculation of a Turbocharger for a Low-Speed Engine

“…The SST turbulence model and the ANSYS CFX program were used. Calculated and test data comparison demonstrated significant overestimation of the calculated pressure ratio and a shift of the calculated characteristics to the right, which is typical for a number of other comparisons of CFD calculations with experiments [7][8][9][10][11][12][13][14][15][16][17][18][19][20]. We can also observe unusually large overestimation of calculated efficiency.…”

Topology Rule-Based Methodology for Flow Separation Analysis in Turbomachinery

“…Both rotor and stator domains are run simultaneously within the code and primitive variables are interpolated and exchanged using overset grids with the CWIPI library of Onera [21]. The setup is similar to the one used in [22] where the convective operator is discretized by the Lax-Wendroff scheme (2nd-order accurate) and an explicit time advancement [23]. The Sigma [24] Sub-Grid Scale model (SGS) is used and standard log-law is applied on all solid boundaries.…”

Towards the Large-Eddy Simulation of a full engine: Integration of a 360 azimuthal degrees fan, compressor and combustion chamber. Part I: Methodology and initialisation

“…about 100 for a uniform refinement of 1 mm and a fan tip-gap discretisation of 17 cells. The CoHP mesh has a uniform mesh refinement of 0.25 mm, a wall refinement of 0.125 mm and between 7 to 9 cells in the tip-gap of the impeller following the recommendation by Dombard et al (2018). This leads to average y þ values of about 150 on the hub and shroud and 75 on the walls of the blades and vanes.…”

“…Once the transient finished, data was acquired during 10 fan rotations for flow analysis. Note that as of today, one brute force possibility to initialise compressor simulations is to start with the flow at rest and then increase progressively the rotational speed and outlet pressure as explained by Dombard et al (2018). Using this approach, convergence would be achieved within 10 revolutions.…”