Centrifugal compressors represent a significant challenge for modern computational methods.  Traditional simulations are calibrated and validated to match with experimentally determined overall performance measurements such as total-pressure ratio, efficiency, and mass flow rate.  Modelling decisions which offer marginal predictions improvements with high computational cost are often discarded to facilitate the rapid turnaround required for use in the design process.

This paper approaches the modelling process from a different perspective stemming from two tenets: that factors exist in experimental hardware which cannot yet be perfectly recreated in a computational setting and the goal of CFD use in research is fundamentally different from applications in design.  Design applications must prioritize computational speed and flexibility whereas research applications can prioritize accuracy at the expense of speed.  Results will be presenting showing the modeling development process from this perspective with a focus on the accurate prediction of trends and flow profiles—rather than absolute values of performance parameters—and the more accurate inclusion of real-world flow physics—even at the expense of computational cost.  The results show that physics-based criteria for modeling decisions results in a model that is more-predictive of trends and flow behavior over the operating range of the compressor stage. 

This model is then used for detailed aerodynamic analysis of loss mechanisms within the diffuser passage of the CSTAR centrifugal compressor.  The compressor is uniquely instrumented to allow detailed characterization of internal flow phenomena.  Most notably, the compressor contains multiple total pressure and total temperature rakes at the inlet and outlet of the impeller and the diffuser along with extensive static pressure taps throughout the flow field.  These allow detailed comparisons between numerical and experimental results.

The influence of heat transfer, fillets, wall roughness, and turbulence model choice on flow development is analyzed as well.  The inclusion of heat transfer did not significantly impact the loss development through the passage however the stage total temperature ratio was shifted closer to the experimentally determined values.  Fillets caused flow redistribution around the passage and a decrease in the extent of flow separation along the diffuser vanes.  Compared to experimentally measured values, the fillets shifted the location of the peak total pressure (at the exit of the diffuser passage) away from the pressure surface of the vane and closer to the measured location.  Accurate inclusion of surface roughness caused a large decrease in the predicted total pressure ratio.  Additionally, roughness effects significantly shifted the location of the peak total pressure at the diffuser exit by 18% of the pitch.  All three of these effects caused detailed aspects of the flow development to more accurately match experimental data.

The Spalart-Allmaras model, compared to the other models tested without heat transfer, fillets, or roughness, was shown to most accurately match the experimental value of efficiency and other overall performance metrics.  However, more detailed flow measurements show that this was non-physical in nature as a significant over-prediction of flow separation, with the concomitant losses in total pressure, was compensating for the models’ typical over-prediction of stage total pressure rise.  Two equation models (SST, k-ε, and k-ω) showed an improvement with the k-ε, and k-ω models showing marginal improvements over the SST model in terms of diffuser exit profiles.  Explicit Algebraic Reynolds Stress Models (EARSM) based on both k-ε, and k-ω formulations were also tested.  These models showed improvements over their baseline formulations, however the k-ω-based EARSM model more accurately matched the shape of the experimentally measured performance speedline.  Finally, two Reynolds Stress Models (RSM) were implemented.  The results showed some improvements, however the numerical difficulty in achieving reliable convergence prevented a more thorough study and limited their usefulness from a research perspective.


The Impact of Various Modelling Decisions on Flow Field Predictions in a Centrifugal Compressor

Division / Sponsor:
International Gas Turbine Institute (IGTI)

Paper Proceedings: https://asmedigitalcollection.asme.org/GT/GT2020/volume/84058

Turbo Expo - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.