Reduction of Reaction Mechanisms

摘要: Increases in both chemical kinetics knowledge and the capacity of computers have led to availability very large detailed kinetic mechanisms for many problems. These may contain up several thousand species ten reaction steps. For computational reasons, however, still cannot be used spatially 2D or 3D fluid dynamics simulations, where applied mechanism typically requires less than 100 species. Also, within such mechanisms, key processes can masked by presence steps only marginal importance. A first step reducing size a is identify which do not need included order accurately predict target outputs model. Such methods lead so-called “skeletal” schemes. This chapter discusses different identification redundant mechanism, including those based on sensitivity Jacobian analyses, comparison rates, trial error calculated entropy production. Another family development skeletal schemes investigation graphs. We discuss here directed relation graph (DRG) method its derivatives, path flux analysis (PFA) method. Mechanism reduction also optimisation minimise an objective function related simulation between full reduced models, subject set constraints (e.g. numbers required). Integer programming genetic algorithm-based been are discussed here. From these schemes, subsequent reductions achieved via either lumping. Chemical mathematical approaches lumping with applications combustion, atmospheric biological systems. Reduction timescale separation then introduced starting classic quasi-steady-state approximation (QSSA). Computational singular perturbation (CSP) described as means informing derivation analytically models. Further efficiency gains obtained using numerical place more traditional descriptions source terms The generation models original differential equations thermodynamics problem deduced from results. Using any methods, has meet special requirements, evaluated quickly provide accurate approximation. series approaches, tabulation artificial neural networks (ANNs) various types polynomials, that all tested context modelling.