Machine learned compact kinetic models for methane combustion

Chemical kinetic models are an essential component in the development and optimisation of combustion devices through their coupling to multi-dimensional simulations such as computational fluid dynamics (CFD). Due significant level detail contained within, detailed chemical computationally prohibitive for use CFD. Therefore, low-dimensional which retain good fidelity reality needed, production requires considerable human-time cost expert knowledge. Here, we present a novel automated compute intensification methodology produce overly-reduced optimised (“compact”) models. The Machine Learned Optimisation Kinetics (MLOCK) coded algorithm systematically perturbs each four model components discover what combinations terms results with high calculations. A virtual reaction network comprised n species is first obtained using conventional mechanism reduction procedures. Once lower than threshold value, performance typically low fidelity. To adjust this, weights (virtual rate constants) important connections reactions) between node (species) numerically across sequential phases replicate select version MLOCK (MLOCK1.0), simultaneously all three Arrhenius constant parameters assesses suitability new objective error functions, quantify calculations candidate set targets, In this study, demonstrated by automatic creation compact archetypal case methane/air combustion. It shown that NUGMECH1.0 2789 reliably compacted 15 (nodes), whilst retaining overall 79–90% at Industry-defined target calculations, outperforming prior state-of-art.