Abstract: In order to optimize the ignition delay time and CO mole fraction, a new optimization mechanism for methane oxygen-enriched combustion based on artificial-neural-network (ANN-OMOC) was proposed by simplification and optimization for the detailed methane oxygen-enriched combustion mechanism USC mech2.0, using the directed relational graph with error propagation, full species sensitivity analysis and artificial neural network (ANN). The results of simulation calculation and comparative analysis for methane oxygen-enriched combustion show that the prediction errors of ignition delay time and laminar flame velocity are reduced from 2.53%, 24.38% to 0.50%, 14.41% by use of ANN-OMOC, compared with the prediction error of the simplified mechanism FSSA for methane oxygen-enriched combustion. Meanwhile, compared with the simplified mechanisms DRGEP and FSSA for methane oxygen-enriched combustion, the optimized mechanism ANN-OMOC has the best prediction results for ignition delay time, peak mole fraction of OH and peak mole fraction of CO, with relative errors of less than 10%.