Modeling of Pollutant Formation in Lifted Diesel Jets

May Yen, Purdue University

Abstract

The accurate modeling of soot and NO in reacting diesel sprays can aid diesel engine designers as they strive to reduce these pollutants. A literature review shows, however, that the accurate modeling of soot is elusive. Empirical and semi-empirical models with tunable constants can be calibrated to match some features of measured soot distribution in diesel sprays and atmospheric turbulent jets, but they are, at best, suited for limited range of engine operating conditions. More detailed kinetic models are less amenable to calibration, but they are not able to reproduce the measured soot distribution in diesel sprays and atmospheric turbulent jets. In general, the models predict the axial location of the peak soot volume fraction to be upstream of the measured axial location and the radial spread to be narrower than in measurements. Quantitative disagreement of predicted and measured soot volume fraction is often greater than an order of magnitude. ^ In this work, a detailed kinetic soot model is evaluated in diesel sprays. An Unsteady Flamelet Progress Variable (UFPV) model is employed to account for turbulence/chemistry interactions. When modeling soot, the residence time is estimated using tracer particles that are injected with the fuel. An analysis of the results shows that, in general, the total predicted soot in the spray correlates with the flame lift-off height, as expected from measured results. The measured soot distribution in the spray is, however, not reproduced well by the model. The axial location of the predicted peak soot volume fraction is upstream of the measured axial location. To determine if these differences arise from inaccuracies in employing the kinetic model at higher pressures than for which the kinetic model was originally validated, atmospheric turbulent jets are simulated. The peak axial location is, however, again predicted to be upstream of the measured location. This is consistent with prior predicted results in the literature. Hence, it is concluded that the differences observed in the diesel sprays and in the atmospheric turbulent jets may be on account of inaccuracies in the turbulence-chemistry interaction model; but, simulations carried out in laminar jets as part of this work suggest that inaccuracies in the kinetic model may also be a contributing factor. ^ Since modifying the kinetic model is challenging without detailed experimental data about species concentrations and understanding about chemistry, it was decided to employ semi-empirical modeling of soot which has resulted in better predictions in laminar flames than when employing the kinetic model. A review of two-equation models, however, shows that there is wide variation in the sub-models and their constants employed for soot nucleation, surface growth, number density, and radiation. A baseline model was developed and was shown to predict results that agree quite well with the measured results of temperature and soot volume fraction distribution. Sensitivity of the predictions to model constants is studied, showing that changing one sub-model or model constant affects all processes. This soot model coupled with a simple one-step combustion model is then used to compute soot in turbulent flame at atmospheric conditions. The agreement between the measurements and computations is better than prior results reported in the literature for the same measured flame although soot is still found upstream of that of the measurements. The semi-empirical model coupled with a simple one-step combustion model is then used to predict soot in diesel jets. When compared with the earlier predictions that employed a detailed kinetic soot model coupled with an unsteady flamelet progress variable (UFPV) model, results predicted with the simpler models showed better agreement with the measured results. ^ This work shows that further work is needed in developing more accurate kinetic models for soot formation and oxidation. Furthermore, work is also needed to develop turbulence-chemistry-radiation interaction models that can be employed for soot predictions.^

Degree

Ph.D.

Advisors

John Abraham, Purdue University.

Subject Area

Engineering|Mechanical engineering

COinS