Uncertainty quantification and reduction for combustion kinetic Modeling: A case study of NH3/H2 models
Gongrui Huang, Hongxin Wang, Tian Liang, Oskar Haidn, Nadezda Slavinskaya
Abstract
Variations in the selection of parameters for reaction rate constant (RRC) contributes to uncertainties in combustion kinetic models. To quantify and reduce these uncertainties, this study proposes an efficient framework integrating sensitivity analysis and Monte Carlo simulation capable of simultaneously considering numerous experimental conditions, while incorporating the probabilistic distribution of simulation errors and RRCs. In this framework, a vast number of modified models, generated based on the initial uncertainty bounds of the RRCs for highly sensitive reactions identified through a comprehensive sensitivity analysis, are used to simulate experimental measurements and obtain the distribution of prediction errors. The posterior probability distributions for each RRC are further derived, ultimately leading to the determination of reduced uncertainty bounds. The proposed framework was successfully applied to reduce the uncertainties in ammonia (NH 3 )/hydrogen (H 2 ) models, utilizing over 2,500 experimental data points, including ignition delay times, premixed laminar flame speeds, and species concentrations. By conducting a comprehensive sensitivity analysis on 11 representative NH 3 /H 2 models, 52 highly sensitive reactions contributing significantly to the uncertainty were identified. The RRCs for these 52 reactions from measurements, theoretical calculations, and reviews were collected, with their initial uncertainty bounds determined statistically. A new comprehensive NH 3 /H 2 combustion kinetic model with 42 species and 346 reactions was developed and extensively validated over a wide range of conditions. Following this, the reduced uncertainty bounds of RRCs for 52 reactions were obtained through Monte Carlo simulation, resulting in the uncertainty reduction of NH 3 /H 2 models, which provides valuable insights for model optimization. This framework offers a robust tool for uncertainty quantification and reduction in combustion kinetic models and can be broadly applied to other fuel systems.