Optimization Results

The GRI-Mech 3.0 optimization produces generally better agreement with the targets than the previous 1.2 and 2.11 mechanisms it replaces. The average ignition delay target predictability is improved by 6%. Some of the shock tube species profile data are less well fit by about 10%, however. The agreement is as before, of course, much better than the base starting kinetics mechanism.

Where new data have been included in the 3.0 targets, much better agreement is observed - see the ethane and CO/H2 flame speeds, the formaldehyde data sets, and the new prompt NO targets. While the CH measurement in a lean premixed flame is less well fit, the CH kinetics for GRI-Mech 3.0 have been substantially revised as a result of new rate constant measurements on CH + O2 and H2. Among flame speed data, the high pressure results seem about 7% less well described. However, the calculations for version 1.2 appear to be in error on the high side by around this amount due to a wrong sign in the thermal diffusion term in the computation, so the 3.0 results should in fact be considered an improvement.

A few comments are noted below where either substantial disagreement or notable agreement of model with validation experiment was observed, or differences with the results of the previous mechanisms are significant.

Methane (and other) flame speeds still are slightly overpredicted, especially on the lean side for methane, and little kinetically can be found as a solution. Larger overpredictions for ethene, propane, and rich methanol are as might be expected, given that the optimization was not intended to predict their performance. CO/H2 flame speeds are better predicted.

HCO low pressure premixed flame concentrations are now underpredicted in the new mechanism, compared to previously. This is a consequence of the reparameterization of key rate constants for CH2O destruction, introducing the CH2O targets into the optimization, and increasing HCO + O2.

The new prompt NO target increases predictions of rich flame NO, and substantially reduces disagreement of the model with the observations of Riesel et al. for NO in rich atmospheric pressure ethane flames, and for various methane flames from the Purdue group.

One consequence of adding acetaldehyde-vinoxy product chemistry to the C-2 oxidation mechanism is the observed prediction of cool-flame-like oxidation for ethylene. This extensive behavior is not reported in Glarborg's reactor experiments. Overall, methane and ethane high temperature oxidation appears adequately modeled. Lower temperatures demand caution.

The hydrogen second limit is now about 40K high; it was 40K low for GRI-Mech 1.2. New rate expressions for the H + O2 reactions are responsible.

The mechanism somewhat overpredicts the amount of NO reburn in reactor experiments, but underpredicts the amount of HCN. Matching the temperature dependence is sometimes a problem. Yet in low pressure flames, NO reburn is underpredicted while CH loss is overpredicted. Thus it appears that some experimental discrepancies or mechanistic shortcomings remain.

Pure propane fuel ignition delays tend to be predicted a bit too fast, but this should not be surprising because the optimization was only designed for propane as a minor constituent and the chemistry was truncated without stable intermediates like propene.

Methane ignition delay data, particularly the sets from Seery and Bowman, Frenklach and Bornside, and Spadaccini and Colket (including the targets), are very well predicted. Compared to GRI-Mech 1.2, methane ignition is slightly faster and ethane slightly slower.

An absolute OH time profile at 80 atmospheres in a shock tube, as well as low pressure flame OH data, are extremely well modeled. The mechanism performs well over a wide pressure range.