Narrow Results

Numerical flow simulations are an important part of the design and verification of measurement configurations. Three examples are presented: variable density flow, particle flow behind pipe junctions and turbulent buoyancy-driven flow. Potentially it can used for uncertainty estimations.

Uncertainty evaluation in metrological applications with computationally expensive model functions can be challenging if it is not clear if the model can be locally linearized and the law of the propagation of uncertainty of the Guide to the Expression of Uncertainty in Measurement can be applied. The use the Monte Carlo method as presented in GUM supplement 1 is not practical as it requires a vast number of model evaluations, which can be very time consuming in case of computationally expensive model functions. For this type of model functions smart sampling approaches can be used to assess the uncertainty of the measurand. In this paper a computational fluid dynamics model of sonic gas flow through a Venturi nozzle is studied. Various smart sampling methods for uncertainty quantification of the model's output parameter mass flow rate are assessed. Other sources of uncertainty of the model are briefly discussed, and a comparison with measurement data and with the results of a 1-dimensional simplified model are made.