Bayesian Error Propagation for a Kinetic Model of n-Propylbenzene Oxidation in a Shock Tube

摘要: We apply a Bayesian parameter estimation technique to chemical kinetic mechanism for n-propylbenzene oxidation in shock tube propagate errors experimental data Arrhenius parameters and predicted species concentrations. find that, the methodology successfully, conventional optimization is required as preliminary step. This carried out two stages: First, quasi-random global search using Sobol low-discrepancy sequence conducted, followed by local means of hybrid gradient-descent/Newton iteration method. The concentrations 37 at variety temperatures, pressures, equivalence ratios are optimized against total 2378 observations. then study influence uncertainties measurements on some model well Markov chain Monte Carlo algorithms employed sample from posterior probability densities, making use polynomial surrogates higher order fitted responses. conclude that provides useful tool analysis distributions responses, particular their correlations. Limitations method discussed. For example, we second-order response surfaces assuming normal propagated largely adequate, but not always.