Construction of reduced chemical mechanisms orientated toward specific applications: a case study of primary reference fuel

Abstract

Due to the specific prediction requirements in combustion simulations, the reduced chemical mechanism developed focusing on particular applications is usually needed. A systematic method for chemical mechanism reduction orientated toward specific applications is proposed in this work. The reduction process is divided into two parts for large-molecule fuels, including the reduction of the fuel-specific sub-mechanism and the reduction of the C0–C4 sub-mechanisms. In the first part, path sensitivity analysis and global sensitivity analysis are used to identify the key reaction classes in the fuel-specific sub-mechanism, and then the rate of production (ROP) analysis and genetic algorithm are utilised to recognise the representative reactions with the maximum flux in the key reaction classes and optimise the reaction rate coefficients within their uncertainties. For the second part, the directed relation graph with error propagation and sensitivity analysis (DRGEPSA) is employed to reduce the C0–C4 sub-mechanisms. And then, the genetic algorithm with binary variables is applied to further reduce the pathways of the small-molecule reactions targeted at ignition delay times in shock tubes, major species (fuel, O2, CO, and CO2) profiles in jet-stirred reactors, and laminar flame speeds, respectively. Finally, reaction lumping is conducted by identifying the quasi-steady-state (QSS) species. Based on a detailed mechanism of primary reference fuel (PRF) consisting of 2,870 species and 9,233 reactions, three reduced models orientated toward specific applications, including ModelST with 42 species and 82 reactions targeted at the ignition timings in shock tubes (ST), ModelJSR with 43 species and 79 reactions targeted at the major species concentrations in jet-stirred reactors (JSR), and ModelLPF with 41 species and 69 reactions targeted at laminar flame speeds in laminar premixed flames (LPF) are constructed, respectively. The validation results indicate that all the three reduced mechanisms can achieve good predictions in their featured specific applications.