Computational Fluid Dynamics (CFD) is becoming fundamental to predict turbomachinery performance. However, only using advanced numerical models coupled with high fidelity grid generation is possible to reach a very good matching with test data. In this regard, secondary flow modeling plays a critical role in the accuracy of performance prediction for centrifugal compressor stages. This study analyses the effects of cavity models on centrifugal compressor stages performance across the full range of impeller flow coefficients used in common industrial applications. Both bi-dimensional low flow coefficients with splitter and non splitter blades and three-dimensional high flow coefficients stages have been used as test cases to compare the numerical prediction with test data. Furthermore the effects of secondary flows modeling have been assessed when comparing detailed flow features with advanced experimental data both in terms of 1D profiles and 2D maps. The effects of cavity flows modeling is growing, as expected, moving to very low flow coefficients, reaching several points of difference in efficiency calculation with respect to simpler models. Furthermore, the agreement with experimental data is very good both in terms of overall performance and detailed flow features. Finally, the high fidelity CFD is capable to give deep in-sides into the flow evolution inside the machine allowing aero designers to design centrifugal compressor stages with higher performance. It should be remarked here that a good matching of CFD prediction with test data is possible only by using high fidelity models.

The increasing trend of high stage pressure ratio with increased aerodynamic loading has led to reduction in stable operating range of centrifugal compressors with stall and surge initiating at relatively higher mass flow rates. The casing treatment technique of stall control is found to be effective in axial compressors, but very limited research work is published on the application of this technique in centrifugal compressors. Present research was aimed to investigate the effect of casing treatment on the performance and stall margin of a high speed, 4 : 1 pressure ratio centrifugal compressor through numerical simulations using ANSYS CFX software. Three casing treatment configurations were developed and incorporated in the shroud over the inducer of the impeller. The predicted performance of baseline compressor (without casing treatment) was in good agreement with published experimental data. The compressor with different inducer casing treatment geometries showed varying levels of stall margin improvement, up to a maximum of 18%. While the peak efficiency of the compressor with casing treatment dropped by 0.8%–1% compared to the baseline compressor, the choke mass flow rate was improved by 9.5%, thus enhancing the total stable operating range. The inlet configuration of the casing treatment was found to play an important role in stall margin improvement.

The sizing of surge protection devices for both compressor and surrounding system may require the knowledge of performance curves in 2nd quadrant with a certain level of accuracy. In particular two performance curves are usually important: the pressure ratio trend versus flow rate inside the compressor and the work coefficient or power absorption law. The first curve allows estimating mass flow in the compressor given a certain pressure level imposed by system, while the second is important to estimate the time required to system blow down during ESD (emergency shutdown). Experimental data are routinely not available in the early phase of anti-surge protection devices and prediction methods are needed to provide performance curves in 2nd quadrant starting from the geometry of both compressor and system. In this paper two different approaches are presented to estimate performance curves in 2nd quadrant: the first is a simple 1D approach based on velocity triangle and the second is a full unsteady CFD computation. The two different approaches are applied to the experimental data more deeply investigated in part I by Belardini E.[3]. The measurement of compressor behavior in 2nd quadrant was possible thanks to a dedicated test arrangement in which a booster compressor is used forcing stable reverse flow. The 1D method showed good agreement with experiments at design speed. In off-design condition a correlation for deviation angle was tuned on experimental data to maintain an acceptable level of accuracy. With very low reverse flow rates some discrepancies are still present but this region plays a secondary role during the dynamic simulations of ESD or surge events. The unsteady CFD computation allowed a deeper insight into the fluid structures, especially close to very low flow rates when the deviation of the 1D method and the experimental data is higher. An important power absorption mechanism was identified in the pre-rotation effect of impeller as also the higher impact of secondary flows. These two methods are complementary in terms of level of complexity and accuracy and to a certain extent both necessary. 1D methods are fast to be executed and more easily calibrated to match the available experiments, but they have limited capability to help understanding the underlying physics. CFD is a more powerful tool for understanding fluid structures and energy transfer mechanisms but requires computational times not always suitable for a production environment. 1D method can be used for standard compressor and applications for which consolidated experience have been already gathered while CFD is more suitable during the development of new products or up to front projects in general.