Two phase flow 1D turbulence model for poly disperse upward flow in a vertical pipe

For an adequate simulation of poly-disperse bubbly flows the consideration of the bubble size distribution is required, since the interaction between the liquid and the gas bubbles sensitively depends on the bubble size. In the frame of multi-fluid models this can be done by dividing the gas phase into a number of bubble classes. To test sub-models suitable for implementation into CFD codes a simplified test solver suitable for vertical pipe flow was developed and successfully applied in the last years. It considers a number of bubble classes, but resolves volume fractions of the single bubbles classes, velocities and turbulence parameter only in radial direction. Progress was achieved especially for the models reflecting the momentum transfer between gas and liquid phases (models on bubble forces). The situation is still unsatisfying regarding the simulation of bubble coalescence and break-up. One of the reasons may come from shortcoming in the calculation of turbulence parameters. Local bubble coalescence and break-up rates strongly depend on local turbulence parameter, namely and i. Bubble induced turbulence is presently considered in most CFD codes as well as in the Test-Solver by an additional term for the turbulent viscosity to obtain more accurate velocity calculations. In order to improve the local values of turbulence parameter it is necessary to include source term into the equations of the turbulence model itself. For the case of the Turbulence model such source term can be found in Literature. A simplified turbulence model was implemented in the test solver. In this work this simple model was replaced by two-phase model along with a source term out of the literature. The model was numerically solved and successfully implemented into the Test-Solver. The new calculated results were validated against old simple model calculations as well as against experimental data from MTLoop (DN50) and TOPFLOW (DN200). New calculations with model and source term bring clear improvements upon old ones and resolves significant discrepancies appeared in old calculations in DN200 data. A comparison with CFX 5.7 calculations for some points in DN50 is presented as well. The comparison shows the need for implementation of the source term in CFD calculations.