Kinetic uncertainty quantification in turbulent flame simulations: A compact approach

Uncertainties in the predictions brought about by inherent uncertainties in the kinetic constants need to be assessed to examine the reliability of the simulations. As the volume of the parameter space grows exponentially with dimension, any method that attempts to fully characterise the parameter space suffers from the curse of dimensionality. In this work, we present a compact approach to quantifying the kinetic uncertainty in turbulent flame simulations and apply it to Sandia flame D. Firstly, we prescribe the quantities of interest (QoIs) a priori. Then, the most influential subsets of sensitivity matrices of the QoIs are isolated. The subset isolation is found to be effective as it has identified the reactions that have significant contributions to the uncertainties given by the polynomial chaos expansion (PCE) a posteriori. Lastly, we propagate the uncertainties of the kinetic constants falling in the subset into the turbulent flame simulation predictions, and by constructing ordinary least-squares regression-based sparse PCEs between the predictions and these kinetic constants, uncertainty quantification is performed. The sparse PCEs have demonstrated their ability to identify the most significant region of uncertainties in the computational domain with small sample sizes while preserving lower cross-validation errors in such a region.