The project objectives are to understand what controls the process of post-flame oxidation of unburned hydrocarbons in spark ignition engines. The questions to be addressed by the project are: (a) under what conditions (engine operation and relevant temperatures) the production of products of incomplete combustion (particularly important ozone precursors and toxic products) are most likely, (b) the relative roles of mixing and reaction on oxidation during the different phases of post flame oxidation and (c) how well predictions of unburned hydrocarbon product hold against experimental data.
Hydrocarbon emissions from spark-ignition engines are responsible for a large fraction of ozone precursors in the atmosphere as well as some important regulated toxic hydrocarbons. Over the last several years, a very large effort has been invested by automotive and oil companies in the U.S. and abroad to identify and quantify the effect of fuel composition, engine design and control strategies on hydrocarbon emissions, and on a wider scale, to provide solutions for attaining the restrictive ULEV emissions. This massive effort has led to considerable improvement in the characterization of the effect of fuel formulation on current fleet technology. Nevertheless, the state of the art in understanding the physical and chemical phenomena leading to the production of products of incomplete combustion in spark ignition remains largely correlative and semi-empirical.
The process of oxidation of unburned hydrocarbons in spark ignition engines involves the emergence of hydrocarbons from cold engine walls (crevices, oil layers and deposits) and subsequent diffusion (turbulent and molecular) through the thermal boundary layer into the hot burned gases. Experimental evidence and calculations indicate that oxidation is fast during the gas expansion phase, when burned gas temperatures are very high, proceeding at a progressively slower rate during the exhaust displacement phase, eventually freezing when bulk gas temperatures reach about 1200 K. In order to characterize the hydrocarbon oxidation process, and identify what phases control the formation of potent ozone precursors and toxic hydrocarbons, one must combine chemical kinetic models to fluid mechanical mixing models. Given the complexity of the task, it is necessary to simplify the fluid mechanics during each phase of the process (expansion and displacement), and to identify simplifications to the chemical kinetic models that might be useful in more complex fluid mechanical simulations of the gas reaction.
The processes of hydrocarbon emergence and oxidation taking place inside the engine cylinder and exhaust system involve molecular and turbulent diffusion away from the walls into the hot core gases, total or partial reaction and subsequent convective transport out of the cylinder. The simulations are primarily concerned with the oxidation during the expansion process, during which hydrocarbons emerge from the various cold wall and oil sources, and are confined to a fairly thin (sub-millimeter) layer adjacent to the wall. This allows the simulations to be kept one-dimensional, so that detailed chemistry computations are feasible.
The model is a one-dimensional mixing layer model that includes convective and diffusive transport of species and energy, as well as chemical production and destruction of species and chemical energy release. The initial and boundary conditions are set to match conditions expected around the engine walls. The evolution of the species released at different times during the expansion process can be investigated, and the contributions of the different factors (reaction, diffusion and convection) to the overall process can be investigated.
We have improved the numerical efficiency and portability of the code by developing an in-house adaptive gridding strategy which has reduced computational times by a factor of five, while removing the constraint of using proprietary libraries (formerly from Numerical Algorithms Group). This has allowed a larger number of computations to be made, as well as the use of different machines. Further, we have generalized the code to different one dimensional geometries (cylindrical and spherical), thus allowing computations of emissions due to incomplete vaporization of droplets (as an example). Finally, the code has been adapted to the problem of determining emissions resulting from vaporization of liquid fuel still present in the cylinder during post flame expansion.
Key results of the simulations, and comparison with the experiments are listed below:
The original computations assumed that the temperature profile encountered by the unburned hydrocarbons near the wall was a step function during all times in the expansion of the cases. This allowed the analysis of the phenomena of diffusion and reaction starting from a well-characterized initial condition. However, the predictions made are clearly not realistic for comparison with experiments, as the thermal boundary layer grows during expansion to a size of the order of a couple of millimeters. This colder layer of gases prevents oxidation, leading to higher surviving fractions of hydrocarbons at the end of the expansion cycle. The calculations for all different cases where experiments were available have been redone with the developing boundary layers. The results show much closer agreement with the experimental values.
The model has been able to capture the effect of operating conditions: the changes in the survival rate (emission rate) of hydrocarbons as a function of operating conditions (equivalence ratio, wall temperature and fuel type) agree well with experimentally measured values. The effect of equivalence ratio is related primarily to the change in temperature. When changed from the baseline conditions (f=0.9) to f=1.0, the temperature of the burned gases increases, so that oxidation is faster, even though oxygen concentrations decrease. In the case of wall temperatures, the effect is twofold: the lower density of the gases for increasing temperatures lead to lower overall storage of mass. Further, the higher temperatures close to the wall lead to increasing diffusivity of the gases, which contributes to enhanced oxidation. The effects are well captured by the model. Finally, in the case of fuel type, the emitted amounts with isooctane increase markedly relative to propane. Analysis of the simulation results indicates that the primary reason for this effect is the relative long kinetic pathway of isooctane to final products, leading to overall slower reaction. Although agreement is not perfect in the case of fuel changes, it is to be expected from two factors: imperfections in the chemical kinetic mechanism, which has only been validated for a few conditions, and the fact that no sources other than crevice gases have been considered Ð neglecting other potential sources such as liquid fuel presence.
A comparison of the measured products remaining after oxidation shows that the case of propane, the split between fuel and non-fuel is well captured, including the minor species present. In the case of isooctane, however, the model is unable to capture the product distribution. This could result from several reasons: the model assumes only one source (crevice) for isooctane, while it is known that liquid fuel presence also contributes to HC emissions; further, the chemical mechanism for isooctane has not been validated as thoroughly as the mechanism for propane. Finally, in both cases there are questions about how accurately the current simplified diffusion model can capture the processes inside the cylinder. Given these uncertainties, we find the agreement to be quite reasonable.
The original plan has been altered to: (a) eliminate 2D and 3D simulations in favor of more detailed 1D simulations, (b) additional time was spent on optimizing the speed of the numerical computations and (c) computations of emissions due to liquid fuel vaporization and oxidation have been added. These tasks, except for the liquid vaporization and oxidation model, have been completed.
The contract has ended in September 1998 and has not been renewed.