Study of the Variation in Component Level Characteristics of a Centrifugal Compressor Across the Compressor Map Using CFD and Experiments With Detailed Instrumentation on a Gas Stand

Abstract

Abstract A compressor map with variations in component level characteristics is introduced. This novel approach to plot component level characteristics identify and give context to the critical design parameters such as flow angles, pressure coefficients, and loss coefficients across the entire compressor map along with the conventional performance parameters. Compressor design involves a significant amount of CFD at a stage level, but the decisive importance of component performance, design and analysis are many times overshadowed by 3D CFD, both steady RANS and unsteady RANS/LES/DES. Although the authors support the development of high-fidelity computational tools, validating these computational results with experimental measurements is essential. A detailed analysis of the Scania’s in-house designed compressor stage tested in the Scania Gas stand shows that in order to create an optimal design, component level performance assessment is necessary for the impeller, diffuser and volute. The variation of performance parameters along a speed line and across varied tip speeds are shown along with indications for the onset of instabilities at a component level. This is a crucial input for the design of a compressor stage. In addition to detailed analysis of experimental results, this work includes predictions from two different CFD solvers and discussion of deviations between them. Both the CFD solvers can be used to predict performance parameters at a stage level for most parts of the map. For operating conditions close to surge and higher tip speeds (565 m/s), there are major deviations in the CFD solvers’ predictions. The impeller pressure rise and efficiency could be predicted well with the CFD solvers for most of the tip speeds. However, deviations between measurements and predictions increase with increasing tip speeds. The diffuser performance is predicted well by one of the solvers. The other solver underpredicts the diffuser and overpredicts the volute performance. These are also important inputs for 1D design and also for positioning of the best characteristics on the compressor map for the optimal operation of the machine. The goal of this work is to quantify performance predictions on a component level accurately, so that stage level optimization is worth executing. Further, such validated models could aid reliable design predictions and an optimization of the entire compressor stage.