Large eddy simulations (LES) of inclined dense jets in coastal waters
Date of Issue2016
School of Civil and Environmental Engineering
Nanyang Environment and Water Research Institute
Desalination plants are increasingly common to augment the freshwater supply to ease the pressure on the demand of freshwater due to the rapid global urbanization. In view of potential environmental impacts, the brine outfall of desalination plants must be properly designed to ensure good mixing with the surrounding ambient seawater. Many previous studies had been performed experimentally on the mixing behavior of inclined dense jets, and analytical models had also been developed based on these experimental results to aid the outfall design. However, the accuracy of predictions by these models still needs further improvement. The Computational Fluid Dynamics (CFD) approach offers another alternative to predict the mixing behavior of inclined dense jets in coastal waters. So far, only a few CFD studies using the Reynolds Averaged Navier-Stokes (RANS) approach have been reported in the literature. In the present study, the mixing behavior of the inclined dense jet is investigated with the Large Eddy Simulation (LES) approach, which is theoretically more advanced in terms of resolving the large coherent eddies involved. To build capability and experience, LES was first carried out to investigate two related topics: (a) flow separation around a circular cylinder (for wall interactions), and (b) a circular turbulent wall jet without buoyancy (for concentration spreading along the wall boundary). In (a), the velocities within a square pipe with an embedded cylindrical tube were simulated by both RANS and LES. Experimental measurements were also carried out. In comparison, RANS with the k-ε closure and wall function yielded the least accurate results particularly in the wake region; while the k-ω closure was in satisfactory agreement with the experimental data near to the cylinder but grossly over-predicted in the far field. LES with near wall modelling provided the best predictions of the experimental data. Thus, the results illustrated that LES with near wall modelling can better predict the flow field around boundaries including possible flow separation. In (b), LES with near wall modelling was able to predict the strong anisotropy of the concentration spreading of walls jet near the boundary, while RANS with different turbulence closures and the wall function failed to do so. The LES results also agreed well with the available experimental data in the prediction of turbulence intensity near the wall. Given the experience, LES with near wall modelling was then employed to predict the mixing behavior of inclined dense jets in coastal waters. The results showed that LES predicted the geometrical characteristics of the inclined dense jet reasonably well in general, and the accuracy was much better than the existing integral models. The LES predictions on dilution were also much better than the existing integral models, but still with an under-prediction of ~ 20% compared with the experimental data. The turbulence intensity was predicted satisfactorily by LES in the outer half of the inclined dense jet but not in the inner half, suggesting that the current mesh scheme was not able to fully capture the convective turbulence by the unstable stratification in the inner half. With the bottom impact, LES with near wall modelling was able to reproduce the localized concentration build-up at the impact point reported by Abessi and Roberts (2015), while RANS with the k-ε closure and wall function was unable to do so. The profiles of normalized mean concentration and concentration fluctuation along the spreading layer predicted by LES were close to previous experimental data, showing self-similarity. However, the dilution and thickness were under-predicted. Overall, the results in the present study demonstrated that, with the current mesh scheme, LES with near wall modelling was a better alternative to predict the mixing behavior of inclined dense jets in coastal waters than the existing integral models and RANS with the k-ε closure and wall function compared herein.