OCR Text |
Show constant modification, say from a new kinetics measurements or someone's mechanistic deductions from a potential future target experiment, is consistent with the current target set and possible mechanisms. Optimization runs can be conducted with selected rate parameters frozen at any specified values within the originally determined bounds of the response surfaces. Several conclusions may be drawn from these runs. First, some optimization of the rate parameters is necessary to reduce the error in predictability from the basic mechanism. However, only a few of rate constant adjustments are needed in order to substantially improve matters to the point justified by the accuracy of the data. The system is still overdetermined, and in need of more and different targets. W e see a similar pattern to the results of previous optimization efforts, favoring an increase in the rate of H + C H 3 recombination, and a value for the C H 3 + 0 2 rate constants close to previous values, based on many targets. Some rate constants are optimized on the basis of the lesser number of targets they influence: lower rates for ethyl and propane decomposition and an increase in the C H + N 2 reaction are indicated. A n examination of the optimization is also fertile ground for suggesting which rate constants and product branching paths need improved measurement (if possible) in order to constrain the range of the key variables. Some, such as H + 0 2 are about as well determined as possible, while high temperature measurements and product determinations of reactions such as O, O H , and H 0 2 + C H 3 would be very helpful. Others, such as C H + 0 2 and H + 0 2 + M , were recently measured and now have new values and smaller ranges than in prior optimizations. Performing the optimization also reveals the needs for additional targets, and was a driving force behind our inclusion of new C H targets in this version. But several rate constants are still being optimized based on 1-3 targets, which leads to a lower reliability until some different targets are generated as well. Prompt N O remains such a case. In mechanism 2.11, w e had available a single highly uncertain low pressure flame target, and lowered the C H + N 2 rate constant on that basis. The newer, more precise target n o w reverses this tendency, but it is critical to have other reliably modeled experiments to use as targets. In general, the entire nitrogen target database remains too sparse. The optimization process also can reveal inconsistent results, since it samples what we judge to be the domain of allowable mechanisms. W e originally included some pure propane ignition delay targets, for example, but difficulties optimizing several of these points led to their exclusion, and to the conclusion that either our limited propane mechanism is inadequate or there are errors in the data. As another example, the current runs as well as the 1.2 optimization all overpredict a 3 bar flame speed but underpredict the 5 bar target. This is an expected inconsistency since the latter measurement is uncorrected for stretch. GRI-MECH CALCULATOR A n on-line calculator is available at our W e b site and via a link from the GRI/Net home page. It permits users to simulate simple kinetics problems by specifying desired input conditions. N o knowledge of code and machine operation, input/output formats, or the details of kinetics modeling is required. The Calculator 2.2 version offers the options of a well-stirred reactor or a plug flow reactor (a straightforward time-dependent one-dimensional kinetics problem), models the problem using the appropriate Chemkin-II code with the current GRI-Mech mechanism, and provides the output in an interactive graphical (or numerical) fashion to the user. In addition to specifying the input concentrations and conditions, the user may choose various problem options such as constant pressure or volume, and adiabatic or an imposed linear 7 |