Uncertainty Quantification for the Eulerian-Lagrangian simulation of evaporating sprays
for the Eulerian-Lagrangian simulation of evaporating sprays
Evaporating sprays can be almost routinely simulated using an Eulerian-Lagrangian approach which relies on a RANS modeling of the continuous phase and a Lagrangian description of the discrete phase, including a turbulent dispersion model to express the effect of turbulent fluctuations within the carrier phase on the spray particles and an evaporation model for the spray droplets. Both descriptions are coupled through a two-way approach which accounts for the effects of the continuous phase on the droplets and the retroaction of the droplets on the carrier phase. Several experiments available in the literature have been used to calibrate the physical models involved in the numerical prediction of evaporating sprays (see (Sommerfeld & Qiu, 1998), (Gounder et al., 2012)). The present study focuses on the numerical prediction of evaporative sprays taking into account existing experimental and modeling uncertainties. Due to the significant cost of the numerical predic-tion, a surrogate model is used to quantify the effect of these uncertainties on quantities of interest. The non-intrusive Polynomial Chaos Expansion (PCE) technique available in the UQLab environment is coupled with the deterministic CFD solver to yield statistical outputs (typically mean value and variance) of key quantities, such as the liquid mass flow rate for instance. The study will analyze the sensitivity of these outputs to various uncertainties: experimental uncertainties on the continuous phase inputs (gas inlet mass flow rate, velocity distribution…) and/or the discrete phase inputs (liquid mass flow rate…); modeling uncertainties, in particular regarding the evaporation model.