A New Global Mechanism for MILD Combustion Using Artificial-Neural-Network-Based Optimization

Abstract

A new global mechanism of combustion called the GM-ANN mechanism is proposed for MILD combustion, with its reaction parameters being optimized by artificial neural network (ANN). More specifically, the GM mechanism is first obtained by selecting well-performed global reactions from Jones and Lindstedt ( Combust. Flame 1988, 73, 233) (named “JL” mechanism) and Westbrook and Dryer ( Combust. Sci. Technol. 1981, 27, 31) (named “WD” mechanism). Then, its parameters are optimized using ANN to achieve the results best matching those from experiments and/or numerical simulations using the detailed mechanism GRI-Mech-3.0 (abbreviated as GRI-3.0). The GM-ANN mechanism is tested by comparing its performance with those of GRI-3.0 and JL and WD mechanisms in zero-dimensional perfectly stirred reactor (PSR), nonpremixed CH4/H2 jet-in-hot-coflow (JHC) flame, and premixed and nonpremixed combustion in furnace. Results obtained demonstrate that the GM-ANN mechanism performs better than the JL and WD mechanisms for various cases of MILD combustion. Therefore, the GM-ANN mechanism should be a better choice than the JL and WD mechanisms for high-cost computations of MILD combustion by large eddy simulation (LES) and direct numerical simulation (DNS) that need to use global mechanisms.