Performance of Air-Air Ejectors with Multi-ring Entraining Diffusers

Abstract

This research study considered subsonic short air-air ejectors with multi-ring entraining diffusers. Many references can be found for the design of air-air ejectors with solid diffusers. However, a limited amount of work has been published specially addressing the performance of short ejectors with entraining diffusers. This study was an experimental and computational investigation of how ejector performance is affected by ejector geometry (i.e. nozzle, mixing tube and diffuser), flow inlet swirl conditions and flow temperature. Ejector performance was quantified in terms of pumping, pressure recovery, wall temperature and velocity and temperature distribution at the diffuser exit. The experiments were conducted on one cold flow wind tunnel and one hot gas wind tunnel. In total, eight ejector systems were tested for this research. Five different swirl conditions and two primary air flow temperatures were studied. Ejector inlet conditions were measured using four fixed 7-hole pressure probes in the annulus. Ejector exit flow conditions were measured using a traversing 7-hole pressure probe with a thermocouple. A parallel computational study was conducted along with the experimental study. The commercial CFD packages, Gambit 2.3 and Fluent 6.2, were selected for meshing and flow solutions. The objective of the computational study was to determine the utility of RANS based CFD model for predicting device performance as design changes were implemented. The computational study was intended to provide practitioners with guidance as to when CFD will provide practical answers to specific questions relating to the ejector performance including ejector pumping, pressure recovery, wall temperatures and velocity and temperature distribution at the diffuser exit.

In total, twenty-six complete cold flow experiments and twenty complete hot flow experiments have been completed. A detailed CFD model study has been performed to select the suitable computational domain, mesh density, boundary conditions, turbulence model and near wall treatment. Twenty-four CFD cases were selected to compare with the corresponding experimental data. The experimental results showed that the inlet swirl conditions and the diffuser bent angle had significant effects on the ejector performance. In general, the maximum ejector performance was achieved with the 20° inlet swirl condition. This level of swirl enhanced pressure recovery in the ejector. As the diffuser bent angle increased, the total pumping decreased due to the flow impingement in the diffuser. The oblong ejector generally had better flow mixing performance than the round ejector. For the CFD simulations, the Realizable k-ε turbulence model was found to give reasonable predictions for most of the bulk flow properties such as the total pumping, velocity profiles, swirl levels and back pressure. These were achieved at a reasonable cost in terms of the human efforts and computational resources. The RSM was able to give slightly improved predictions but at a much higher cost in terms of the efforts and computing resources. All of the turbulence models had difficulty predicting the pressure recovery in the mixing tube and diffuser because of their inability to accurately predict flow separation in the core of the swirling primary flow. As a result of this, the turbulence models considered in this work overpredicted the pumping of the mixing tube and underpredicted the pumping of the entraining diffuser. This unresolved issue with the CFD models is an important consideration when designing such devices.