Experimental And Numerical Verification of an Optimization of a Fast Rotating High Performance Radial Compressor Impeller

Abstract

Outgoing from a well-proven radial compressor design which has been extensively being tested in the past known as SRV4 impeller (Krain impeller), an optimization has been performed using the AutoOpti tool developed at DLR’s Institute of Propulsion Technology. This tool has shown its capability in several tasks, mainly for axial compressor and fan design as well as for turbine design. The optimization package AutoOpti was applied to the redesign and optimization of a radial compressor stage with a vaneless diffusor. The numerical results of this optimization were presented by Voss et al. [1] and by Raitor et al. [2]. The optimization was performed for the SRV4 compressor geometry without fillets using a relatively coarse structured mesh in combination with wall functions. The impeller geometry deduced by the optimization had to be slightly modified due to manufacturing constraints. In order to filter out the improvements of the new so-called SRV5 radial compressor design, two work packages were conducted: The first one was the manufacturing of the new impeller and its installation on a test rig to investigate the complex flow inside the machine. The aim was, first of all, the evaluation of a classical performance map and the efficiency chart achieved by the new compressor design. The efficiencies realized in the performance chart were enhanced by nearly 1.5 %. A 5 % higher maximum mass flow rate was measured in agreement with the RANS simulations during the design process. The second work package comprises the CFD analysis. The numerical investigations were conducted with the exact geometries of both, the baseline SRV4 as well as the optimized SRV5 impeller including the exact fillet geometries. To enhance the prediction accuracy of pressure ratio and impeller efficiency the geometries were discretized by high resolution meshes of approximately 5 million cells. For the blade walls as well as for the hub region the mesh resolution allows a low-Reynolds approach in order to get high quality results. The comparison of the numerical predictions and the experimental results shows a very good agreement and confirms the improvement of the compressor performance using the optimization tool AutoOpti.