An improved particle swarm algorithm-based method for kinetic modeling study of ammonia/air laminar flame speed

Abstract

In recent years, ammonia has garnered increasing attention as a promising carbon-free fuel. Laminar flame speed is a critical property of ammonia fuel that has been extensively studied by researchers using chemical kinetics mechanisms. However, some deviations still remain in the numerical predictions. To further improve the prediction accuracy of the laminar flame speed for ammonia/air flame, an ammonia kinetic mechanism is developed in this work. Reactions from Gri-Mech 3.0, Li et al. mechanism, and Han et al. mechanism that have significant impact on laminar flame speed were first assembled to the Okafor et al. mechanism, and H/OH sub-mechanisms were merged to establish a kinetic mechanism containing 43 species and 142 reactions. A particle swarm algorithm with improved inertia weights is proposed and the fitness function for outputting laminar flame speed prediction error is customized based on Cantera codes, then forming the algorithm framework used to optimize the kinetic mechanism. The pre-exponential factor A, the temperature exponent n and the activation energy Ea of the four reactions in the merged mechanism that have high sensitivity to the laminar flame speed are selected as the independent variables for optimization, and the reactions rate constants corresponding to the high prediction accuracy of the laminar flame speed is finally obtained. The numerical prediction indicates a reduction in mean absolute percentage error of laminar flame speed from 22.563 % to 10.649 % using the optimization mechanism. The results of the sensitivity and reaction pathways analyses demonstrated that more H-related reactions were considered in the optimized mechanism, and the relative ROP of H-related reactions were adjusted at different equivalence ratios, resulting in higher or lower predictions of laminar flame speed compared with the unoptimized Okafor et al. mechanism. Ignition delay time as well as the species distribution of NH3, NO, and N2O was studied. The ignition delay time predictions using the optimized mechanism are in better agreement with the experiment data at P = 11 atm and 30 atm. The optimization mechanism demonstrated accurate predictions for NH3 and NO while exhibiting an overestimation for N2O, but the N2O prediction error is smaller compared to the Okafor et al. mechanism.