Physics-based surge point identification for unsupervised CFD-computation of centrifugal compressor speed lines

Abstract

We present a speed line calculation model using physics-based surge point identification. When using surrogate model-based design of experiment (DOE) techniques, the large number of computations demands an automated, unsupervised process to calculate the speed line and to identify its significant points. A high-fidelity CAE workflow is used to calculate the machine design, create the CAD geometry, generate a mesh, and prepare a CFD setup. The choke point, the surge point and the peak efficiency point are sequentially identified using sophisticated identifiers and optimization algorithms. The model was used to calculate the speed line of 50 different centrifugal compressor designs with 50 geometry variations each. From these 2,500 designs, 2,449 speed lines were calculated successfully, without specific settings having been assigned to individual designs. This proves that the model enables the stable, reproducible, reliable and effective CFD calculation of the speed line of centrifugal compressors.