Uncertainty analysis in automatic reaction mechanism generation : neopentyl + O₂

Abstract

The process of building accurate chemical mechanisms for hydrocarbon oxidation systems is difficult since these mechanisms can have hundreds of species and thousands of reactions. Computer programs have recently been developed to construct these models automatically, but until this work, these programs did not include tools for the propagation of uncertainty. Rate constants and thermodynamic properties are not known precisely, and this can lead to large errors in model predictions. This work presents tools for sensitivity analysis and uncertainty propagation within an automatic reaction mechanism generator. A function for calculating first-order sensitivity coefficients with respect to rate and thermodynamic parameters and initial conditions is implemented in the MIT Reaction Mechanism Generator (RMG). An algorithm for generating error bounds on model output using first-order sensitivity coefficients and uncertainties in model parameters is also implemented. These tools are applied to an automatically generated model for the oxidation of the neopentyl radical, and results are compared to experimental observations.

(cont.) Comparison of the model with experimental data allowed identification of two rate constants. At 673 K and 60 Torr, kC5H11+O2-->OH+C5HI0O = 1.9x 10-14 ± 6x 10-15 cm3/molecule-s, and kOH+C5H1I-C5HOI+H20 = 3.1 x 10-12±1 .5 x 10-2 cm3/molecule-s.The computer-generated model is consistent with two prior literature studies.