A Reynolds-averaged Navier-Stokes computational analysis of a novel multi-nozzle annular air-jet pump

Morrall, Andrew and Campobasso, Sergio and Campobasso, Sergio (2021) A Reynolds-averaged Navier-Stokes computational analysis of a novel multi-nozzle annular air-jet pump. PhD thesis, Lancaster University.

Text (2021MorrallPhD)
2021MorrallPhD.pdf - Published Version
Restricted to Repository staff only until 1 February 2026.
Download (190MB)

Abstract

This thesis presents a thorough investigation of the flow field of single-phase swirl inducing multi-nozzle annular jet pumps, in which a system of discrete jets flows into the pump bore through nozzles fed by compressed air accumulated in an annular plenum chamber. The investigation is based on Reynolds-averaged Navier-Stokes (RANS) Computational Fluid Dynamics (CFD) simulations validated by means of measured static pressure field measurements. These measurements are obtained using a bespoke test rig developed for this project. The swirling flow component makes the RANS analysis of this problem particularly challenging, and, for this reason, a thorough cross-comparison of the predictive capabilities of the popular two-equation k-omega shear stress transport (SST) turbulence model and the computationally more demanding but generally more reliable Reynolds Stress Transport (RSM) method is carried out for the swirling flow problem at hand. This comparative analysis shows that the use of the RSM approach results in overall better predictions of the experimental observations. It is also found that including in the physical domain the external region surrounding the inlet of the suction duct of the pump is necessary to further improve the prediction of the pump flow field. This is due to the high streamline curvature at the duct inlet that results in a small separation at the inner lip of the intake duct, and a relatively thick low-speed annular flow region between this toroidal recirculation zone and the pump jet system. The adoption of this so-called boxed intake in the simulations results in a more realistic flow field configuration at the duct inlet, because the external flow far field boundary conditions imposed on the box far field boundaries allow a more realistic adjustment of the flow field to the duct inlet geometry than possible by using an internal far field boundary condition at the intake of the suction duct, which forces a quite unlikely straight flow pattern. Validation of all turbulence modelling and geometric CFD set-ups is accomplished by cross-comparing measured and computed static pressure fields of two pump designs for three values of the total pressure of the compressed air feeding the discrete jet systems. The geometry of the two pumps differs only for the axial inclination of the nozzles and, thus, their discrete jet systems, with one pump (design 1) featuring a lower value of this parameter than the other (design 2). Both pumps are tested at three operating conditions characterised by different strength of the swirling jets, varied by changing the total pressure of the air in the plenum chamber feeding the jets. At the lowest swirl level of design 1, the SST and RSM simulations are in fairly good agreement, and they both agree well with the measured pressure fields. As the flow swirl increases, however, the two CFD results start diverging from each other, with the RSM prediction remaining in better agreement with the measured static pressure for all three swirl levels. At the highest swirl level, the SST analysis predicts weaker dissipation of the jet energy and stronger mixing of injected and pumped streams, resulting in higher performance predictions than obtained with the RSM analysis. Design 2 is characterised by a stronger weight of the swirling flow component than design 1 for a given total pressure of the compressed air, due to the lower axial velocity component resulting from the higher inclination of the jets on the bore axis. The higher swirl ratio of design 2 results in both SST and RSM predictions presenting more pronounced discrepancies with the measured data, due to expected difficulties of the RANS approach to resolving flow fields with high levels of swirl. The RSM prediction, however, provides a closer match of the measured data also for the three regimes of design 2. Both CFD and measurements indicate higher performance and efficiency of design 1, providing initial evidence that high swirl levels reduce the overall performance of the considered fluidic device. The aforementioned CFD simulations of designs 1 and 2 serve the purpose of not only validating and cross-comparing the predictive capabilities of the SST and RSM approaches for different swirling regimes, but also explaining the key flow physics of the considered pump concept and provide new knowledge for its design optimisation. More specifically, the simulations are used to a) explain the fluid mechanic features accounting for the suction and energy transfer capabilities of the pump, and b) assess and explain the sensitivity of the device performance and efficiency to design parameters such as nozzle diameter and circumferential inclination, and compressed air total pressure. Building upon the numerical and experimental findings on the fluid mechanics, performance and efficiency analyses of designs 1 and 2, a simulation-based parametric study assessing the sensitivity of the pump performance and efficiency to diameter, number, circumferential and axial orientation of the nozzles and total pressure of the compressed air is performed. This study, based on the use of a simplified sector model to reduce the computational burden of the parametric study, leads to the definition of a third device (design 3) which is built and tested. It is found that despite deliberately not featuring flow swirl and notwithstanding some discrepancies between the nominal and manufactured design 3 geometry (small differences in surface finish, diameter and length of the nozzles), design 3 has higher performance and efficiency than both designs 1 and 2, a conclusion arrived at with both a full-annulus CFD simulation accounting for all discrepancies between nominal and manufactured geometries, and pressure field measurements of the manufactured design 3. One of the unexpected findings of this research is that the swirl ratio required to possibly improve the performance and efficiency of the considered device is small and could not be quantified precisely in this research.