Automatic Analysis and Reduction of Reaction Mechanisms for Complex Fuel Combustion

Abstract: This work concentrates on automatic procedures for simplifying chemical models for realistic fuels using skeletal mechanism construction and Quasi Steady-State Approximation (QSSA) applied to detailed reaction mechanisms. To automate the selection of species for removal or approximation, different indices for species ranking have thus been proposed. Reaction flow rates are combined with sensitivity information for targeting a certain quantity, and used to determine a level of redundance for automatic skeletal mechanism construction by exclusion of redundant species. For QSSA reduction, a measure of species lifetime can be used for species ranking as-is, weighted by concentrations or molecular transport timescales, and/or combined with species sensitivity. Maximum values of the indices are accumulated over ranges of parameters, (e.g. fuel-air ratio and octane number), and species with low accumulated index values are selected for removal or steady-state approximation. In the case of QSSA, a model with a certain degree of reduction is automatically implemented as FORTRAN code by setting a certain index limit. The code calculates source terms of explicitly handled species from reaction rates and the steady-state concentrations by internal iteration. Homogeneous-reactor and one-dimensional laminar-flame models were used as test cases. A staged combustor fuelled by ethylene with monomethylamine addition is modelled by two homogeneous reactors in sequence, i.e. a PSR (Perfectly Stirred Reactor) followed by a PFR (Plug Flow Reactor). A modified PFR model was applied for simulation of a Homogeneous Charge Compression Ignition (HCCI) engine fuelled with four-component natural gas, whereas a two-zone model was required for a knocking Spark Ignition (SI) engine powered by Primary Reference Fuel (PRF). Finally, a laminar one-dimensional model was used to simulate premixed flames burning methane and an aeroturbine kerosene surrogate consisting of n-decane and toluene. In general, detailed calculations of temperature, pressure, concentration and flame velocity show excellent agreement with measurements. Skeletal mechanisms for PRF were constructed for the SI engine case, reproducing autoignition well on removal of reactions pertaining to 15% of the species. QSSA reduction was tested on the staged combustor and the engines, using pure and weighted lifetime indices. Monitoring NO concentrations in the staged combustor and ignition timing in the engines, good reproduction is possible while approximating about 70% of the species. However, some species have to be manually retained for accuracy and numerical stability. For improved ranking, sensitivity was added to the index applied to the premixed flames, in addition to necessary molecular transport information. The maximum atomic mass fraction occupied by a certain molecular species was also constrained to limit the mass and energy deficiency caused by QSSA. For methane, the laminar flame velocities as well as concentration profiles are well predicted by the most strongly reduced mechanism with five global reaction steps. For the kerosene surrogate mechanism, QSSA involving 50% of the species was successfully attempted.