Comments on the GRI-Mech 3.0 Optimization Process

[A separate commentary describes the optimization procedure.]

GRI-Mech 3.0 began with 81 targets and 103 variables. The final optimization considered 77 targets and varied 32 parameters. Four targets were omitted when found inconsistent with other optimization targets, and 71 variables were frozen at starting values when judged to provide only insignificant improvements to overall target fits if varied.

The majority of high temperature methane shock tube and flame speed targets drive an increase in only 3 rate constants. These are the methane recombination and two CH3 + O2 rate parameters, which assume optimized 3 0 values close to those in mechanism 1.2. In the final optimization run we decided to limit the multiplier for CH3 + O2 = OH + CH2O to that for the O + CH3O channel. It was not necessary or particularly advantageous to allow the optimization to adjust other key sensitive rate constants: H + O2, OH + CO, O + CH3, HCO decomposition, CH2 + O2, and OH + CH3 remain at their initial values. If the H + O2 chain branching step and OH + CO are included, the optimization would decrease these rate constants, and produce improved flame speed predictions at 1 atm, but at the expense of greater disagreement at high pressure and for shock tube species profiles. We thus decided not to alter these rate constants. Including the OH + CH3 and HCO + M reactions among active optimization variables results in suggested multipliers of 1.0 - no change. For O + CH3 and CH2 + O2, the optimization is more inclined to alter the product branching ratios than overall rates, but still gives negligible improvement.

Many of the other reaction rate constant changes in the optimization are driven by major improvements in matching smaller subsets of targets. Propane results are improved by decreasing k(312), and ethane targets influence k(74), k(158), and k(159) for example, although the ethane kinetics also actuate the methane (especially CH3) targets. This isolation of select target-reaction subsets is still apparent in the limited list of nitrogen targets, as it was for version 2.11, although several new results have been added. Compare the individual target page sensitivities for details.

In order to accommodate the new CH2O targets in the 3.0 optimization, it was only necessary to optimize the HO2 + CH2O and HCO + O2 rate constants. Adjustments to the key H + CH2O and decomposition reactions (reparameterized for the 3.0 round starting mechanism) were not needed. Eiteneer et al (1998) needed to suggest some modifications to these rate constants when they first considered their data as an addendum to version 1.2 optimization results.

Another cluster of rate parameters (49, 125, 126, 289, 240, 246-8) influences the optimized results for new CH measurements and prompt NO targets. The starting kinetics were also revised based on new data, especially regarding the CH + O2 and H2 reaction systems.

We did permit some rather target-specific optimization in limited cases where considerable improvement could be made. The increase in k(119) improves the prediction of the IG.St1a ignition delay and the high pressure flame speeds, as does lowering OH + HO2. This speeds up high pressure, low temperature chain branching. Flame speed improvements also largely dictate the modifications to the HCO + O2 and H2O rate constants. Changing H + O2 + H2O and the H + HO2 branching ratio improve methane and hydrogen flame speed predictions. These four alterations all serve to reduce the chain branching and lower the atmospheric pressure flame speed overpredictions.

In some cases, usually involving these target-specific rates or some nitrogen targets, rates constants were adjusted to the full limits we judged prudent, although larger changes would improve agreement more. However, changes to the allowed limits did not occur in the optimization for the key core hydrocarbon kinetics (reactions 52, 119, 155, 156, 158, 159, and 168). In fact, including the CH3 + CH3 steps in the optimization resulted in smaller changes to the other important rate parameters.

We originally included some additional targets involving ignition delays in propane fuel mixtures, from Borisov et al., but large deviations were observed for all optimization runs for 2 of the four targets. Within the context of our very limited C-3 kinetics, these results are inconsistent with the other targets and hence were omitted.

The table below compares rate constant values at 1 atm 1500K for many of the key reactions, giving a ratio for GRI-Mech 3.0 vs. 2.11. Significant changes are highlighted, and may be due to new kinetic results, the new optimization, or a combination.