Quantification of numerical uncertainty in computational fluid dynamics modelling of hydrocyclones

Large Eddy Simulations of the flow through a hydrocyclone are used to demonstrate that the Grid Convergence Index (GCI) is a practical method of accounting for numerical uncertainty. The small values of GCI (<7.2%) associated with the tangential velocity predictions suggest that numerical uncertainty due to discretization error does not greatly contribute to the disagreement between simulation and experiment in the tangential direction. The large values of GCI (<303.2%) associated with the axial velocity predictions imply that uncertainty due to discretization error is significant and further mesh refinement can yield better agreement in the axial direction. This was demonstrated through additional grid refinement which produced a reduction in the GCI of as much as 256.6% and a drop in the overall average difference between simulation and experimental of more than 36%. Overall, these results suggest the GCI is a useful tool for quantifying numerical uncertainty in CFD simulations.

► We performed a Large Eddy Simulation of the flow in a hydrocyclone.
► We compared the numerical velocity predictions with experimental data.
► We quantified numerical uncertainty using the Grid Convergence Index (GCI).
► High GCI values in the axial direction suggested that mesh refinement was needed.
► GCI is a useful tool for quantifying numerical uncertainty in CFD simulations.