Uncertainty quantification analysis of RANS of spray swirling jets undergoing vortex breakdown

The CFD-based design of combustion chambers for novel aircraft engines is a challenging task, accompanied by several sources of computational uncertainty. This work highlights the benefits which follow from the use of variable-fidelity approaches and Bayesian uncertainty quantification (UQ) at the preliminary stage of the design process. A prototypical non-reactive acetone/air aerosol swirling jet is investigated through an Eulerian-Lagrangian approach. Two direct numerical simulations (DNS) provide reference data, coping with two swirling regimes. Consequently, exploiting a set of Reynolds-averaged Navier-Stokes (RANS) simulations, a polynomial chaos expansion (PCE) representation is adopted to propagate the uncertainty associated with the spray dispersion model and the turbulent Schmidt number, providing the sensitivity of the output variance to each uncertain input. This way, the most significant sources of model uncertainty may be identified and eventually reduced via a calibration on higher-fidelity data, thus improving confidence in predictions during design optimization.