TRANSIENT SIMULATION OF CAVITATION IN PLANAR VENTURI USING TWO-FLUID MODEL

A cavitating venturi operates with an oscillating two-phase cavity that evolves through a combination of different flow mechanisms and is also closely linked to changes in geometrical configuration. Numerical prediction of the transient flow features of the venturi could become a convenient tool for venturi sizing and design for specific applications. The use of Eulerian two-phase mixture models, along with suitable cavitation models and Reynolds averaged Navier-Stokes (RANS) turbulence model, is known to over-damp the transient oscillatory nature of the cavitation zone due to the overproduction of turbulent viscosity. It is found from the literature that a modified turbulent viscosity equation formulated as a function of the two-phase density, after appropriate tuning of the model constant, is able to predict the transient phenomenon of cavitation in internal flows. However, tuning the model is highly case-specific, and generality regarding the correct frequency predictions is not always guaranteed. The current work presents the steady-state and transient numerical simulations using a two-fluid Eulerian model for the two-phase field, the Schnerr-Sauer model for cavitation and the RANS model for turbulence (without modifying the turbulent viscosity). Commercial software Ansys Fluent is used for the simulations. The model's steady-state predictability of cavitation length is benchmarked using the axisymmetric venturi data from the literature. A parametric study was also conducted to choose appropriate interfacial closure models. Systematic transient simulations were then carried out for a range of pressure ratios (Pr, the ratio of the absolute pressure at the outlet to that at the inlet of the venturi) representing the three different regions of experimentally obtained frequencies reported in the previous work of the present authors. The dynamic behavior predicted by the two-fluid modeling indicates two distinct regions of cavity oscillations. Although the numerically predicted frequencies deviate from the experimental predictions, distinct frequencies are predicted, indicating distinction in the dynamics at different pressure ratios. The current results from two-fluid models definitely provide pointers towards realistic dynamic predictions. Appropriate model improvements could offset the need for computationally expensive large eddy simulation (LES) and direct numerical simulation (DNS) models.