Numerical predictions of the bifurcation of confined swirling flows

Abstract

The bifurcation of confined swirling flows was numerically investigated by employing both the k--E and algebraic stress turbulence models. Depending upon the branch solution examined, dual flow patterns were predicted at certain swirl levels. In the lower-branch solution which is obtained by gradually increasing the swirl level from a low-swirl flow, the flow changes with increasing swirl number from the low-swirl flow pattern to a high-swirl flow pattern. In the upper-branch solution which is acquired by gradually decreasing the swirl level from a high-swirl flow, on the other hand, the flow can maintain itself in the high-swirl flow pattern at the swirl levels where it exhibits the low-swirl flow pattern in the lower branch. The bifurcation of confined swirling flows was predicted with either the k-e model or the algebraic stress model being employed. Both the k--E and algebraic stress models result in comparable and sufficiently good predictions for confined swirling flows if high-order numerical schemes are used. The reported poor performance of the k--E model was clarified to be mainly attributable to the occurrence of the bifurcation and the use of low-order numerical schemes.

Investigation of Radial Swirler Effect on Flow Pattern inside a Gas Turbine Combustor

“…The results showed that the standard and the Durbin k-ε models gave the best agreement with the experimental data. This supported the finding of Jiang and Shen (1994) and Durst and Wennerberg (1991) where good agreement between k-ε model predictions and experimental result were reported.…”

The three-dimensional numerical aerodynamics of a movable block burner

“…The reverse flow was much more pronounced in the RNG k-ε model, which penetrated almost the entire extension of the annular duct, reaching an axial velocity of -4 m/s at the annular duct exit and a peak of approximately 22 m/s, in comparison with 19 m/s for the k-ε model. The reverse flow in the annular region was practically imperceptible with the standard k-ε model, and, despite the observations of Jiang and Shen (1994) for rotating flow in a tube expansion, each turbulence model showed distinct results. However, the effects of using the results obtained for the upwind scheme as a necessary initial condition to obtain convergence in the higher-order upwind scheme requires a more extensive investigation.…”

“…Conversely, the lack of agreement between the experimental data and the predictions regarding swirling flows could be attributed, as mentioned by Jiang and Shen (1994), to the phenomenon of bifurcation, in which one given swirl number can be associated with two distinct conditions of steady state flow. This phenomenon of non-unique steadystate behavior occurs as a consequence of the existence of two paths through which the desired swirl number can be obtained, that is by the gradual increase in the intensity of the swirl, the lower branch, or, alternatively, by the gradual reduction in the intensity of the swirl from a higher condition, the upper branch, such that the predicted swirl number can differ significantly from the experimental swirl number depending on the path.…”

“…Jiang and Shen (1994) claimed that it was necessary that the numerical simulations followed a path identical to that in the experiment for a correct corroboration. In the simulations presented it is probable that the lower branch was used; there is no information about that in the experiments report.…”

“…The authors observed a complex and transient behaviour in spiral, despite the fact that the boundary configuration was only a duct where the velocity components were fixed. Jiang and Shen (1994) attributed the failure of the predictions regarding swirling flows much more to a phenomenon called bifurcation than to the k-ε turbulence model. In the bifurcation phenomenon one given swirl number can be associated with two distinct conditions of steady state flow.…”