An assessment of eddy viscosity models on predicting performance parameters of valves

The major objective of the present study is to evaluate the performance of a range of turbulent eddy viscosity models in the prediction of macro-parameters (flow coefficient (C-Q) and force coefficient (C-F)), for certain types of valve, including the conic valve, the disk valves, and the compensated valve. This has been achieved by comparison of numerical predictions with experimental measurements available in the literature. The examined turbulence models include most of the available turbulent eddy viscosity models in STAR-CCM + 12.04. They are the standard k-epsilon model, realizable k-epsilon model, k-omega-sst model, V2F model, EB k-epsilon model and the Lag EB k-epsilon models. The low-Re turbulence models (k-omega-sst, V2F, EB k-epsilon and Lag EB k-epsilon) perform worse than the high-Re models (standard k-epsilon and realizable k-epsilon). For the conic valve, the performance of different turbulent models varies little; the standard k-epsilon model shows a marginal advantage over the others. The performance of the turbulence models changed greatly, however, for prediction of C-Q and C-F of the disk and compensated valves. In general, the realizable k-epsilon model is demonstrated to be a robust choice for both valve types. Although the EB k-epsilon may marginally outperform it in the prediction of C-F at large disk valve opening. The effects of the unknown entry flow and initialization conditions are also studied. The predictions are more sensitive to the entry flow condition when the valve opening is large. Additionally, the uncertainties caused by unknown entry conditions are comparable to overall modelling errors in some cases. For flow systems with multiple stable flow-states coexisting in the flow domain, the output of the numerical models can also be affected by the initialization conditions. When the streamline curvature and secondary flow is modest like conical valve flow, the nonlinear modification of the standard k-epsilon model and k-omega-sst model, as well as the curvature correction in the realizable k-epsilon model, will not have visible effects on the numerical prediction. Once the strong streamline curvature and secondary flow exit in the domain, such as the disk valve flow, the non-linear modification of the standard k-epsilon model will greatly improve the numerical outputs, however, the non-linear modifications of k-omega-sst model only have minor effects. Moreover, the curvature correction in the realizable k-epsilon model will jeopardise the accuracy of outputs in the same case.