k-ε Predictions of the initial mixing of desalination discharges

Abstract

Predictions from a k-ε model are compared with recently acquired experimental data from inclined negatively buoyant discharges. The k-ε model is part of a standard computational fluid dynamics package (CFX). Two approaches are taken when implementing the model. One involves using an essentially standard form of the model to predict flow behaviour. The other approach involves calibrating the model, through adjustment of the turbulent Schmidt number in the tracer transport equation, to achieve reasonable predictions for positively buoyant vertical discharges and then applying it to inclined negatively buoyant discharges. While the calibrated approach improves the predictions of some bulk parameters (notably the tracer spread and dilution) when compared to predictions from the standard model, the overall effect on the quality of the predictions is small. Comparisons with experimental data indicate that predictions from both the standard and calibrated simulations compare favourably with trajectory data, but integrated dilution predictions at the centreline maximum height are conservative (mean-integrated concentrations are over-predicted). The standard and calibrated k-ε predictions confirm the importance of buoyant instabilities on the lower (inner) side of the flow, the effects of which are clearly evident in the mean concentration profiles. However, these simulations have a tendency to overestimate the influence of stabilizing density gradients on the upper (outer) side of the flow and are unable to effectively predict the cross-sectional distribution of a tracer. In contrast to a previous study, the above comparisons indicate that predictions of bulk parameters from such models can be poor and indeed are no better than those obtained from relatively simple analytical solutions.