A global pathway selection algorithm for the reduction of detailed chemical kinetic mechanisms

A Global Pathway Selection (GPS) algorithm for the reduction of detailed kinetic mechanisms is proposed and validated. The approach consists of (1) the construction of element flux graphs for each considered element from the simulation data obtained using detailed chemical mechanisms, (2) the selection of important species that act as hubs that transfer significant element flux, and (3) the subsequent identification of the global pathway for each hub species by searching a certain number of shortest paths with the constructed element flux graphs from the initial reactants to the final products through the considered hub species. The skeletal mechanism is then obtained by removing the species and reactions that are not important to any identified global pathways. Validations are performed to generate skeletal mechanisms for the combustion of two fuels: n-dodecane, and the mixture of toluene, iso-octane, and n-heptane. For both cases, a series of mechanisms are generated using GPS, and the maximum error of ignition delays is similar or smaller comparing to the skeletal mechanisms generated using Path Flux Analysis method (Sun et al., 2010) with similar number of species. GPS is further compared with two methods using sensitivity analysis, which is usually effective but time consuming. By comparing simulation results of ignition delay, perfect stirred reactor (PSR) temperature, and laminar flame speed over a wide range of operating conditions, it is observed that a 35-species skeletal mechanism generated by GPS for n-dodecane shows similar overall accuracy compared with the 31-species skeletal mechanism obtained by Direct Relation Graph Aided Sensitivity Analysis (Vie et al., 2015). Further validation is conducted for the combustion of the mixture of toluene, n-heptane and iso-octane. A 276-species mechanism generated using DRG with error propagation followed by sensitivity analysis and reaction elimination (Niemeyer et al., 2014) shows larger errors for PSR temperature and ignition delay with fuel composition different from the raw database. However these errors can be reduced significantly if GPS-generated mechanism of similar size is tested.