Reaction network reduction for distributed systems by model training in lumped reactors: Application to bifurcations in combustion.
A new methodology is presented to derive reduced reaction mechanisms for distributed reacting flows by model training in a lumped parameter system (a continuous-stirred tank reactor). The method identifies the relevant transport time scales in the reaction zone of a distributed system along with the local composition vector, over a range of operation conditions. A training box in the parameter space of pressure-transport time scale-composition is then identified. Sensitivity and principal component analyses are subsequently performed at bifurcation points in a lumped parameter system at representative conditions of the training box. The most inclusive chemistry derived in the lumped system captures the proper transport-chemistry coupling and is suitable for the distributed reactor. Application to ignition of hydrogen/air and methane/air mixtures is presented and validated for premixed and diffusion flames in a stagnation flow geometry. (c) 1999 American Institute of Physics.