COMPUTATIONAL CHALLENGES IN SAMPLING AND REPRESENTATION OF UNCERTAIN REACTION KINETICS IN LARGE DIMENSIONS

This work focuses on constructing functional representations of quantities interest (Qols) an uncertain ignition reaction in high dimension. Attention is focused the delay time iso-octane air mixture, using a detailed chemical mechanism with 3811 elementary reactions. Uncertainty all rates directly accounted for associated uncertainty factors, assuming independent log-uniform priors. A Latin hypercube sample (LHS) times was first generated, and resulting database then exploited to assess possibility polynomial chaos (PC) terms canonical random variables parametrizing rates. We explored two avenues, namely, sparse regression (SR) LASSO, coordinate transform (CT) approach. Preconditioned variants both approaches were also considered, logarithm as Qol. Both resulted similar representation errors. However, CT approach able reproduce better empirical distribution underlying LHS ensemble, preserved positivity time. When preconditioned however, performances obtained SR representations. The results revealed that yield consistent global sensitivity estimates. finally used test reduced dimension representation, outline potential extensions work.