UFR 4-02 Best Practice Advice
Confined coaxial swirling jets
Underlying Flow Regime 4-02 © copyright ERCOFTAC 2004
Best Practice Advice
Best Practice Advice for the UFR
The first advice is that confined coaxial swirling jets flow type cannot be calculated by standard k-ε model because it is not adequate for such types of flow. Also other models based on k-ε does not give accurate results and the prediction of flow field parameters is insufficient. Numerical calculations with Reynolds-stress transport model or algebraic Reynolds stress model give better predictions than k-ε based models, however discrepancies between numerical and experimental results are still noticeable. The new ASM model applied by Lei et al.  gives good predictions of mean velocities, static pressure and axial turbulence intensity, however in the region close to the axis, especially at downstream locations it overpredicts axial velocity and underpredicts static pressure and fails in prediction of turbulence intensity.
It could be concluded that RSM and ASM can be used successfully to predict swirling confined jets in all applications where the mean velocity and pressure are of primary importance. However, these models lead to significant discrepancy with experimental data as far as turbulence structure is important. In all industrial applications where mixing intensity is of a major importance to predictions quality, like for example in combustors or chemical reactors, these models are not recommended. By contrast it seems that LES method offers very attractive features for this complex flow type, as shown by Schlüter et al. , predicting perfectly both mean velocity field and turbulence intensity. It must be kept in mind that CPU time required to predict swirling confined jets by LES method is huge, although LES should be recommended as the only tool to predict this flow type with sufficient accuracy in applications where turbulence structure and mixing is of primary importance.
It should be stressed also that 2D LES, as could be expected, leads to completely wrong results, since there are no arguments to impose axi-symmetric condition for resolved velocity field. This assumption is justified in RANS approach as mean velocity field is axi-symmetric, but fails in LES method. Actually some successful 2D DNS and LES was performed for combusting flows but this method cannot be applied to flows without combustion. Comparison of 2D axisymmetric and full 3D calculations presented by them disqualifies 2D predictions which completely differ from measurement, however, the price to be payed for this is large increase in computing time and computer resources. In the case of LES calculations important fact is that great care has to be taken with respect to the formulation of boundary conditions as well as to the simulation time.
© copyright ERCOFTAC 2004
Contributors: Stefan Hohmann - MTU Aero Engines