Abstract
The water mixing experiment in the Generic Mixing Experiment (GEMIX) facility performed at the Paul Scherrer Institute is used as a benchmark case to investigate the influence of the main uncertain parameters on the turbulent mixing under isokinetic flow conditions. The benchmark experiment features two horizontal water streams with the same inlet velocity that merge together to form a mixing flow inside the larger horizontal square channel. The turbulence intensity and the velocity profile at the inlet were used as the main uncertain input parameters. The selected set of computational fluid dynamics (CFD) simulations based on different combinations of values for uncertain parameters has been performed with the code NEPTUNE_CFD that solves the Reynolds Averaged Navier Stokes (RANS) equations with the k-e turbulence model. To investigate the influence of the uncertain parameters over a wide range of values, the surrogate model called optimal statistical estimator (OSE) was used to generate the response surface of the results. It has been demonstrated that the OSE method can be successfully applied to build the response surface from a limited set of simulation points. For the two-parameter problem of the current study, only a few CFD simulation points are found sufficient to construct the quality response surface.
Highlights
Computational fluid dynamics (CFD) simulations are affected by uncertainties, due in large part to the increasing complexity of physical models and the inherent introduction of random model data
The sensitivity of the simulation results to the range of different combinations of input parameters can be evaluated by hundreds of CFD simulations
The sensitivity to the inlet turbulence intensity β is presented for three different values of inlet velocity profiles (α = 0, 0.4 and 1)
Summary
Computational fluid dynamics (CFD) simulations are affected by uncertainties, due in large part to the increasing complexity of physical models and the inherent introduction of random model data. That the uncertainty in the choice of the turbulent model parameters has to be considered, and the physical properties of the fluid and the experimentally provided boundary conditions can be largely affected by uncertainties. One way to alleviate this burden is the construction of a surrogate model that mimics the behaviour of the CFD simulation as closely as possible while being computationally much cheaper. Such models are based on a complex mathematical response function that is constructed on a limited number of intelligently selected data (CFD) points. The optimal statistical estimator (OSE) method [6] is used for the response surface generation
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
