Numerical Study of Flame Lift-off and Soot Formation in Diesel Fuel Jets

Numerical Study of Flame Lift-off and Soot Formation in Diesel Fuel Jets Song-Charng Kong*, Yong Sun and Rolf D. Reitz Engine Research Center, Department of Mechanical Engineering University of Wisconsin Madison Madison, WI 537, USA Abstract A detailed chemistry-based CFD model was developed to simulate the diesel spray combustion and emission process. A reaction mechanism of n-heptane is coupled with a reduced NOx mechanism to simulate diesel fuel oxidation and NOx formation. The soot emission process is simulated by a phenomenological soot model that uses a competing formation and oxidation rate formulation. The model is applied to predict the diesel spray lift-off length and its sooting tendency under high temperature and pressure conditions with good agreement with of Sandia. Various nozzle diameters and chamber conditions were investigated. The model successfully predicts that the sooting tendency is reduced as the nozzle diameter is reduced and/or the initial chamber gas temperature is decreased, as observed by the. The model is also applied to simulate diesel engine combustion under PCCI-like conditions. Trends of heat release rate, NOx and soot emissions with respect to EGR levels and start-of-injection timings are also well predicted. Both and models reveal that soot emissions peak when the start of injection occurs close to TDC. The model indicates that low soot emission at early SOI is due to better oxidation while low soot emission at late SOI is due to less formation. Since NOx emissions decrease monotonically with injection retardation, a late injection scheme can be utilized for simultaneous soot and NOx reduction for the engine conditions investigated in this study. Introduction Diesel engine manufacturers are facing stringent emission regulations and a better understanding of the diesel spray combustion process is crucial to help design low emission diesel engines. Experimental data have been used to construct a conceptual diesel spray combustion image that depicts the flame structure and soot and NOx distributions []. It has been shown that the details of the flame structure are crucial to the soot formation process during the mixing-controlled combustion phase [,3]. The lifted flame consists of a diffusion flame at the periphery of the fuel jet (where NOx is formed) and a rich reaction zone located downstream of the lift-off length in the central region of the fuel jet (where soot is formed). The lift-off length determines the time for fuel-air mixing prior to ignition and entering the reacting zone, and thus will affect the sooting tendency of diesel fuel jets. As a complement to optical soot and NO diagnostics, predictive numerical models can also help understand the diesel spray combustion process and provide insights to the details of flame structure. Development and applications of engine CFD models have become increasingly important and effective in analyzing the complex diesel combustion process [ 7]. The use of detailed chemistry is also essential to better predict fuel oxidation and emission formation, especially for the low-temperature HCCI combustion process which is of much interest [,7]. This study develops a numerical model that uses detailed chemical kinetics to simulate the diesel lift-off flame, and its combustion and emission formation. The model is validated using experimental combustion and emission data from a combustion vessel and from a heavy-duty diesel engine under various operating conditions. Model Formulation The CFD code is a version of KIVA-3V [8] with improvements in various physical and chemistry models developed at the Engine Research Center, University of Wisconsin Madison. The major model improvements include the spray atomization, drop-wall impingement, wall heat transfer, piston-ring crevice flow, and soot formation and oxidation models [9,]. The RNG k-ε turbulence model was used for incylinder flow s. Since detailed reaction mechanisms for n-heptane were used to simulate diesel fuel chemistry, the CHEMKIN chemistry solver [] was integrated into KIVA-3V for solving the chemistry during multidimensional engine s. The chemistry and flow solutions were then coupled. * Corresponding Author

A skeletal reaction mechanism for n-heptane [] was used to simulate diesel fuel chemistry due to their similar ignition characteristics and cetane number. A new NO Mechanism was obtained by reducing the Gas Research Institute (GRI) NO mechanism [3]. The resulting NO mechanism contains only four additional species (N, NO, NO, N O) and nine reactions that describe the formation of nitric oxides. Note that the sum of NO and NO is used to compared with the engine-out NOx emissions measurements in this study. Soot emissions are predicted using a phenomenological soot model [9] that was incorporated into the KIVA/CHEMKIN code. Two competing processes are considered in this model, namely soot formation and oxidation. Experiments Sandia Combustion Chamber Experiments conducted in an optically accessible, constant-volume combustion chamber under simulated quiescent diesel engine conditions were used for model validations [,3]. The high-temperature and highpressure environments are created by burning a specified premixed mixture before the start of fuel injection. Caterpillar Diesel Engine Engine performed on a Caterpillar heavy-duty diesel engine were also used for model validations []. The engine is a single-cylinder engine whose specifications are listed in Table. The parameters that were varied included start-of-injection timings and EGR. The fuel injector was a production style Caterpillar electronic unit injector (EUI). Table Caterpillar 3E engine specifications [] Bore Stroke 37. mm 5. mm Compression Ratio.: Displacement. L Connecting Rod Length. mm Piston Geometry Mexican Hat Max Injection Pressure 9 MPa Number of Nozzle Holes Nozzle Hole Diameter. mm Included Spray Angle 5 Experimental conditions for model validation EGR (%) 8, 7, Start-of-injection (ATDC) -, -5, -, -5,, +5 Results Sandia Combustion Chamber Experimental results of the Sandia combustion chamber [,3] were used to validate the present models. The baseline experimental conditions for model validations are listed in Table. A typical image of predicted fuel spray and gas temperature distributions of a free diesel lift-off flame is shown in Fig.. The injector is located at the top of the image. It can be seen that the liquid fuel undergoes atomization, vaporization, and mixing with entrained air before the lift-off location and then enters the reaction zones. To put in the context of a transient injection process, chemical reactions take place once the fuel is injected and mixes with air, and lead to autoignition at a certain location as seen in Fig. (a). Note that in Fig. (a), the light colors seen between the spray tip and the ignition location indicates a continuous temperature rise as a result of pre-ignition chemical reactions. The ignition location is approximately where the steady-state flame is stabilized in most cases, i.e., the lift-off location. Table Experimental conditions for model validations Fuel # Diesel Injection system Common-rail Injection profile Top-hat Injector orifice diameter 5µm, µm, 8µm Orifice pressure drop 38 MPa Discharge Coefficient.8,.8,.77 Fuel temperature 3K Ambient temperature 85~3K Ambient density 7.3,.8, 3. kg/m 3 O concentration % (a). ms ASI Preignition heat release (b).7 ms ASI Figure Sample images of the predicted fuel spray and gas temperature distributions for d nozz = µm, T amb =9 K, P amb =38 MPa, ρ amb =.8 kg/m 3. Color scale: 9 to K. Planar laser-induced incandescence (PLII) images of soot along a thin plane of the fuel jet were compared with model predictions, as shown in Fig.. The injector orifice is located at the far left center of each image,

with fuel being injected to the right. The flame lift-off length was determined from the OH chemiluminescence images in the [,3]. It is defined as the axial distance between the orifice and the location where the OH chemiluminescence intensity is approximately 5% of that just downstream of the initial rapid rise in the OH chemiluminescence. The cross-sectional average equivalence ratio at the flame lift-off length was estimated and is given on the left of the PLII images. The present simulated soot mass fraction distributions are given in Fig. (b) to compare with the PLII images. The predicted equivalence ratio at the lift-off length is also given on the left of the images. The color scale of the predicted soot mass fraction is also shown in Fig.. Both and s show that as the ambient gas temperature decreases, lift-off length increases, soot concentration decreases, and equivalence ratio at the lift-off length also decreases. The conditions with ambient temperature and 9 K are found to be sooting cases while no soot production is observed for the 85 K case. In the s, the two sooting cases are found to have soot mass fraction of the order of.e-5 while the predicted maximum soot mass fraction is only about.e-8 for the 85 K case. Other comparisons between model results and experimental images suggest that a soot mass fraction of.e-5 can be used as the criterion to specify sooting and nonsooting conditions in the s. This criterion is used later in this paper to determine the sooting limit of injectors with different orifice diameters. ASI PLII Simulation (a) Experimental lift-off length and PLII images K Φ=3. 9K Φ=. 85K Φ=. Lift-off Figure 3 Time sequence (ASI in ms) of PLII images and predicted soot mass fraction contours. The lift-off length and x=5 mm positions are shown on the images with vertical dashed and solid lines, respectively. d nozz = µm, P inj =38 MPa, T amb = K, ρ amb =.8 kg/m 3. (b) Predicted lift-off length and soot mass fraction Figure Comparisons between PLII images and predicted soot mass fractions at the central plane of the fuel jet at 3. ms ASI. The equivalence ratios were estimated at the lift-off length. Relative PLII camera gain is given in brackets. d nozz = µm, P inj =38 MPa, ρ amb =.8 kg/m 3. The temporal evolution of a typical injection and combustion event is illustrated in Fig. 3. Time after start of injection (ASI) for each image is given on the left. The distance from the injector is shown at the bottom. The dashed vertical line shows the lift-off length (8.3 mm) and the solid line shows the x=5 mm position, which was found in the to have the peak soot emissions at 3. ms ASI. The axial distributions of soot along the centerline of the fuel jet were also compared. Figure shows 3

comparisons of measured time-averaged KL factors and predicted soot mass fraction at 3. ms ASI. The KL factor is an indication of optical thickness derived from laser-extinction soot measurements [3]. The KL factor is proportional to the mass of soot along the line of sight of the extinction measurement, so it can be compared with the predicted soot mass that is integrated along the same line...8.... 8 K Figure Comparisons of measured time-averaged KL factors and predicted soot mass along the central axis of the fuel jet for the same conditions as in Fig. 3. Both measured and predicted data were normalized to allow qualitative comparison. Results were acquired at 3. ms ASI for d nozz = µm, T amb = K, P inj =38 MPa, ρ amb =.8 kg/m 3. Optical thickness data were acquired at multiple axial locations along the centerline of the fuel jet at a certain time after start of injection. Due to the different nature of the KL factor from measurements and the integrated soot mass from the s, only qualitative comparisons can be made to assess the model performance. Thus, both measured KL factors and predicted soot mass are normalized to allow qualitative comparisons, as shown in Fig.. Comparisons between the measured KL factors and predicted soot mass along the central axis at 3. ms ASI were also presented in Fig. 5 for other ambient gas temperature conditions of 95,, and 3 K. The measured radial soot distribution 5 mm downstream of the injector (location indicated by the vertical solid lines in Fig. 3) was also compared with s in Figure. Optical thickness data were acquired at multiple radial locations 5 mm from the injector at 3. ms ASI for the same conditions as in Fig. 3. Note that a 3-D cubic mesh with mm grid size was used for the calculation such that it would be easier to integrate the soot mass in the radial direction. As before, both measured KL factors and predicted soot mass were normalized to allow qualitative comparisons. Figure also indicates that the results match the very well even for such a small length scale (note that the length scale is mm in the radial direction while it is mm for the axial direction)....8... 95K..8... K.8... 8. 8...8.... K 8..8.... 8 3K Figure 5 Comparisons of measured time-averaged KL factors and predicted soot mass for various ambient temperatures, 95K, K, K and 3K. Both measured and predicted data were normalized to allow qualitative comparison. Results were acquired at 3. ms ASI for d nozz = µm, P inj =38 MPa, ρ amb =.8 kg/m 3.. - -8 - - - 8 Radial distance from center (mm) Figure Comparisons of measured time-averaged KL factors and predicted soot mass as a function of radial distance from the jet centerline at an axial distance of 5 mm from the orifice (vertical solid line in Fig. 3). Both measured and predicted data were normalized to allow qualitative comparison. Results were acquired at 3. ms ASI for the same conditions as in Fig. 3. The ultimate goal of the numerical model is to predict the sooting tendency of a diesel injector under different operating conditions. Figure 7 shows comparisons of the measured and predicted sooting tendency of diesel injectors with different orifice diameters in an ambient density temperature domain. To the left of

each curve are the non-sooting regimes, and to the right are sooting regimes. In the, the sooting limit is determined by the visibility of soot in the PLII images. In the s, the maximum soot mass fraction of.e-5 is used as the criterion, as discussed earlier. Ambient Density [kg/m 3 ] 3 8µm µm Non-Sooting 5µm Sooting Experiments 8µm µm 5µm Simulation 8µm µm 5µm 7 75 8 85 9 95 5 5 5 Ambient Temperature [K] Figure 7 Measured (solid lines) and predicted (dashed) sooting and non-sooting regimes as function of ambient gas temperature and density for P inj =38 MPa. For the conditions of each curve, non-sooting combustion occurs to the left and sooting combustion to the right of each curve. To determine the sooting limit in the, cases of different temperatures with a 5 K interval were simulated at a fixed ambient density. For example, non-sooting and sooting cases are marked with open and closed symbols, respectively, as shown in Figure 7. The average temperature between adjacent non-sooting and sooting cases is regarded as the sooting limit for a specific injector at the corresponding ambient density condition. As can be seen in Fig. 7, although there are discrepancies between the exact locations of the measured and predicted sooting curves, especially for the small orifice (5 µm) injector, the trends are well predicted. As ambient density and temperature increase, or as orifice diameter increases, the sooting tendency increases. The discrepancy between measurements and predictions for the small orifice is probably due to the significant difference in the spray atomization and mixing processes between orifices with a conventional size and a small size which may not be well captured by the present spray model. Pressure (MPa) 8 - - 8% EGR - - - - Crank Angle (ATDC) +5 35 3 5 5 Figure 8 Comparisons of measured (solid line) and predicted (dotted) cylinder pressure and heat release rate data for 8% EGR cases (SOI=, and +5 ATDC). Caterpillar Diesel Engine The present models were further applied to simulate combustion and emission processes in a heavy-duty diesel engine. Figures 8 and 9 show the measured and computed cylinder pressure and heat release rate data for selected cases. The model is seen to perform well over a wide range of engine conditions. The heat release rate data does not exhibit the distinct premixed and diffusion burn characteristics of conventional diesel engines. The highly premixed burned features of the present PCCI is captured well by the model. Pressure (MPa) 8 - - % EGR - +5 5 35 3 5 5 5 HRR (J/deg) HRR (J/deg) - - - Crank Angle (ATDC) Figure 9 Comparisons of measured (solid line) and predicted (dotted) cylinder pressure and heat release rate 5

data for % EGR cases (SOI=, and +5 ATDC). NOx (g/kg-f) 8 7% % NOx Emission vs. SOI 8% EGR Expt NOx Model NOx -5 - -5 - -5 5 SOI (ATDC) Figure Measured and predicted engine-out NOx emissions for cases listed in Table. The predicted soot and NOx (i.e., sum of NO and NO ) emissions were also compared with the measurements, as shown in Figs. and. It can be seen that the overall trends of soot and NOx are captured very well with respect to the start-of-injection timing. It is of interest to note that engine-out soot emissions reach a peak value when fuel is injected near top-dead-center. The present model also predicts correctly the soot reduction seen at further retarded injection timing (e.g., SOI=+5 ATDC) for all different EGR levels. Soot (g/kg-f)...8... Expt Soot Model Soot Soot emission vs. SOI % EGR 7% 8% -5 - -5 - -5 5 SOI (ATDC) Figure Measured and predicted engine-out soot emissions for cases listed in Table....8... In-Cylinder Soot (g/kg-f) 5 3 % EGR SOI= SOI= - SOI=+5-5 5 5 Crank Angle (ATDC) Figure In-cylinder soot mass histories for % EGR cases at three different injection timings. Values at exhaust valve opening (3 ATDC) are shown in Fig.. The numerical model can explain the soot emission reduction seen as the injection is further retarded past TDC. This reduction is not seen in the conventional diesel combustion soot-nox trade-off with respect to injection timing. Figure shows the total in-cylinder soot mass evolutions for three different injection timings, i.e., SOI=,, +5 ATDC. The model results indicate that the lower exhaust soot emissions for SOI= ATDC is due to a better oxidation process as compared to that of SOI= ATDC. On the other hand, a lower soot emission for SOI=5 ATDC is because less soot is formed in the cylinder as a result of the low temperature combustion (as also can be seen from the low cylinder pressure in Figs. 8 and 9 for SOI=5 ATDC). The above low temperature combustion characteristics are consistent with results in HSDI diesel engines [5] as well, and can be further facilitated to achieve low-emission diesel PCCI operation. Conclusions A numerical model has been developed to simulate diesel fuel jet combustion in a combustion vessel and in a heavy-duty diesel engine. The model uses a skeletal reaction mechanism to describe the fuel oxidation and NOx formation processes and a phenomenological model to simulate soot formation and oxidation. The model successfully predicts the lift-off length of a free diesel diffusion flame under various ambient conditions. The model results indicate that chemical reactions prior to the lift-off location are important for the stabilization of the lift-off flame. The s also agree with the measurements in predicting the sooting tendency of diesel fuel jets that the sooting ten-

dency increases as the ambient gas density, temperature, or orifice diameter increase. Experiments conducted in a heavy-duty diesel engine under PCCI-like conditions were also modeled. The predicted heat release rate data, NOx and soot emissions agreed well with the measurements. The model results indicate that low soot emissions can be obtained at late injection timings (i.e., SOI past TDC) by suppressing the total soot formation as a result of low-temperature combustion. Since NOx emissions decrease monotonically as injection is retarded, such a late injection scheme can be utilized for simultaneous soot and NOx reduction for future low-emission diesel PCCI engines. Acknowledgments The authors acknowledge the financial support by the DOE/Sandia National Labs and Caterpillar, Inc. Experimental data provided by Drs. L. Pickett and D. Siebers (Sandia) for the model validation are greatly appreciated. References Package for the Analyses of Gas Phase Chemical Kinetics, Sandia Report, SAND 89-89.. Patel, A., Kong, S.-C., Reitz, R. D.,, Development and Validation of a Reduced Reaction Mechanism for HCCI Engine Simulations, SAE --558. 3. Smith, G.P., Golden, D.M., Frenklach, M., Moriarty, N.W., Eiteneer, B., Goldenberg, M., Bowman, C.T., Hanson, R.K., Song, S., Gardiner, W.C., Lissianski, V.V. and Qin, Z.,, http://www.me.berkeley.edu/gri_mech/.. Klingbeil, A. E.,, Particulate and NO X Reduction in a Heavy-Duty Diesel Engine Using High Levels of Exhaust Gas Recirculation and Very Early and Very Late Injection, MS Thesis, University of Wisconsin-Madison. 5. Miles, P.C., Choi, D., Pickett, L.M., Singh, I.P., Henein, N., RempelEwert, B.A., Yun, H. and Reitz, R.D.,, Rate-Limiting Processes in Late- Injection, Low-Temperature Diesel Combustion Regimes, Proc. THIESEL Conference, pp.9 7.. Dec, J.E., 997, A Conceptual Model of DI Diesel Combustion Based on Laser Sheet Imaging, SAE 97873.. Pickett, L.M. and Siebers, D.L.,, Non-Sooting, Low Flame Temperature Mixing-Controlled DI Diesel Combustion, SAE --399. 3. Pickett, L.M. and Siebers, D.L.,, Combust. Flame, 38, pp. 35.. Hergart, C., Barths, H. and Peters, N., 999, Modeling the Combustion in a Small-Bore Diesel Engine Using a Method Based on Representative Interactive Flamelets, SAE 999--355. 5. Tao, F., Golovitchev, V.I., Chomiak, J.,, Combust. Flame, 3, pp.7 8.. Kong, S.C. and Reitz, R.D.,, Proc. Combust. Inst. Vol. 9, pp.3-9. 7. Kong, S.C., Patel, A., Yin, Q. and Reitz, R.D., 3, Numerical Modeling of Diesel Engine Combustion and Emissions Under HCCI-Like Conditions with High EGR Levels, SAE Paper 3--87. 8. Amsden A.A., 997, KIVA-3V: A Block-Structured KIVA Program for Engines with Vertical or Canted Valves, LA-333-MS. 9. Han, Z., Uludogan, A., Hampson, G.J. and Reitz, R.D., 99, Mechanism of Soot and NOx Emission Reduction Using Multiple-Injection in a Diesel Engine, SAE 933.. Patterson, M.A. and Reitz, R.D., 998, Modeling the Effects of Fuel Spray Characteristics on Diesel Engine Combustion and Emissions, SAE 983.. Kee, R. J., Rupley, F. M. and Miller, J. A., 989, CHEMKIN-II: A FORTRAN Chemical Kinetics 7