UNSTEADY FULL ANNULUS MULTI-STAGE COMPRESSOR CALCULATIONS - DETAILS ON CFD-EXPERIMENT COMPARISON

Abstract

The unsteady flow in a multi-stage compressor has been simulated using time-domain methods to gain a better insight into the complex flow physics and address the computational bottlenecks associated with large scale numerical simulations in the time domain. For these investigations the time-periodic flow in the DLR research compressor Rig250 has been simulated. Rig250 is a 4.5 stage axial flow compressor with a swan neck that includes struts for support. In the time-domain simulation the struts, IGVs and first two stages are computed time-accurately using a full annulus mesh to fully capture the blade row interactions. The remaining two stages are included in the simulation by solving, with single passages, for the time-mean flow. In total, the low Reynolds full annulus mesh comprises nearly 1 billion nodes distributed over 6000 blocks. For an operating point close to maximum efficiency, comparisons are made between the time-accurate full annulus calculation and experimental test data to see how Kiel probe stagnation pressure measurements can be resolved. It is shown, that with larger efforts in simulations also the comparison between numerical and experimental data has to be conducted with greater care. The numerical and experimental results are in good agreement.The unsteady flow in a multi-stage compressor has been simulated using time-domain methods to gain a better insight into the complex flow physics and address the computational bottlenecks associated with large scale numerical simulations in the time domain. For these investigations the time-periodic flow in the DLR research compressor Rig250 has been simulated. Rig250 is a 4.5 stage axial flow compressor with a swan neck that includes struts for support. In the time-domain simulation the struts, IGVs and first two stages are computed time-accurately using a full annulus mesh to fully capture the blade row interactions. The remaining two stages are included in the simulation by solving, with single passages, for the time-mean flow. In total, the low Reynolds full annulus mesh comprises nearly 1x109 nodes distributed over 6000 blocks. For an operating point close to maximum efficiency, comparisons are made between the time-accurate full annulus calculation and experimental test data to see how Kiel probe stagnation pressure measurements can be resolved. It is shown, that with larger efforts in simulations also the comparison between numerical and experimental data has to be conducted with greater care. The numerical and experimental results are in good agreement.