A Comparison of Numerical Results for an Optically Accessible HSDI Diesel Engine with Experimental Data Way Lee Cheng, Robert Wang, Jared Zhao and Chia-fon F. Lee Department of Mechanical and Industrial Engineering University of Illinois at Urbana-Champaign ABSTRACT Comparisons between numerical results and experimental data for a homogeneous charge compression ignition (HCCI) engine with optical accessibility are being presented in this paper. The KIVA 3V program, a general purpose combustion routine, is being enhanced by implementing the optimized Shell model for the ignition process, applying the extended Zeldovich formulation for nitrogen oxide formation consideration. Soot formation and oxidation models are also included to better describe the soot emission characteristics. The numerical model was then used to simulate the DIATA engine operating at different conditions. Spray cones, pressure and heat release calculations and combustion luminescence images, the KIVA outputs were then compared to experimental data. By examining the results generated by KIVA, it was shown that the numerical scheme accurately predict the spray characteristic, combustion luminescence and also the pressure and heat release rate. Soot was found to be concentrated along the outer edge of the piston bowl. Inefficient oxidation of soot was due to the incapability of the fluid to move the soot to regions with high oxygen concentration, where the species could readily be oxidized. Nitrogen oxide was seen to be produced in regions with high local temperature. INTRODUCTION Homogeneous charge compression ignition (HCCI) engines are engines operating under conditions such that a mixture of fuel vapor and air is auto-ignited through compression of the piston, usually near the top dead center. Other characteristics of HCCI operations include globally lean fuel vapor and air mixture and a combustion process which is usually characterized by a single peak heat release rate. The main benefits of an HCCI engine is that nitrogen oxides and soot, the two major emission constituents being targeted by EPA, can be reduced simultaneously. High temperature flame does not exist in HCCI combustion since combustion takes place across the whole combustion chamber with homogenous charge [1]. Formations of the pollutants are also inhibited by the overall lean conditions. As combustion takes place at lower temperatures, the amount of nitrogen oxide produced is significantly decreased. Moreover, as fuel vapor and air are well-mixed prior to autoignition, locally rich regions, which lead to soot production, are not observed in HCCI combustion, unlike typical diffusion-type combustion. Therefore, amount of soot formed is significantly reduced. While there are many advantages to HCCI engines, there remain unresolved problems. These include ignition control, transient operation, high load operation, hydrocarbon and carbon monoxide emissions and cold start ability. Although HCCI operation of diesel engines was seriously examined both numerically and experimentally, the low temperature combustion process for a small bore HSDI diesel engine has not yet been thoroughly investigated. A numerical model is being applied for an optically accessible HSDI engine. Different strategies in operating the engine, including EGR fraction and various injection schemes are being examined. The results are then compared to experimental data. The simulation code applied towards the study is an augmented version of KIVA-3V Release 2 code developed by the Los Alamos National Laboratory [2]. MODEL FORMULATION Several submodels were enhanced or added to the KIVA 3V Release 2 code by members of the Spray and Engine Research Group at the University of Illinois at Urbana-Champaign (UIUC) in order to improve simulations for an optically accessible HSDI diesel engine. Since an optical engine is operated with no engine oil, a blowby model was appended to the KIVA code. Other submodels that were improved include the low temperature combustion, nitrogen oxide formation, soot production and oxidation. Blowby Model A circumferential flow model is appended to the crevice flow model and dynamic ring pack model, both developed by Namazian [3, 4]. The circumferential flow is usually neglected due to the existence of the oil film on the cylinder wall. This layer of oil acts as a seal on the side and peripheral ring sealing surface. As a consequence, fuel within the engine cylinder is prevented from leaking. However, due to the obstruction of the optical accessibility by an oil film, optical engines are operated without engine oil. Therefore, the circumferential flow must be accounted for accurately model the operation of an optical engine. The mass flow rate equation in the crevice flow model, with the area of the circumferential gap calculated from the geometrical dimensions of the cylinder liner, and, the pressure above and below the rings can then be used to determine the circumferential mass flow rate.
Ignition Model Diffusion-type combustion ceases to exist in HCCI operation as the fuel vapor-air mixture is relatively well mixed with no liquid fuel droplets, autoignition is controlled mainly by chemical kinetics. Only accounting for chemical kinetics in modeling ignition delay is accurate only in homogeneous charge, however, this does not preclude other physical parameters, including spray breakup, atomization, heating and vaporization and also mixing [5] from affecting autoignition. The Shell ignition model, a general multi-step kinetics model, developed by Halstead et al. [6] was implemented into KIVA to predict low temperature chemistry. The set of kinetic parameters applied in the model are chosen in order to minimize the ignition delay error over a wide range of operation conditions. Combustion Model A premixed/laminar timescale model is built into the KIVA code. This model is combined with the low temperature autoignition model for simulations of HCCI combustion processes. The switching temperature (or cutoff temperature) between the two models is set at 1150 K. The laminar timescale model is used when the local cell temperature is above 1150 K. Shazi [7] showed this cutoff temperature provides accurate ignition timing for the engine geometry mesh used in this study. Diesel fuel combustion is replaced by single step global reaction of tetradecane, C 14 H 30, shown in equation (1) as tetrad cane has comparable thermo physical properties with typical diesel fuels. 2 C 14 H 30 + 43 O 2 28 CO 2 + 30 H 2 O (1) As aforementioned, the fuel-air mixture is well mixed with no liquid fuel droplets exist upon ignition, thus, a diffusion type combustion model can be excluded from the numerical scheme. Emission Model: Nitrogen Oxide The extended Zeldovich mechanism was implemented to model the formation of nitrogen oxides. The main reactions in the mechanism involve: O + N 2 NO + N (2) N + O 2 NO + O (3) N + OH NO + H (4) By the steady state assumption, the temporal variation of nitrogen oxide concentration can be evaluated by combining the forward and backward reactions, shown in equations (2) to (4). Heywood [8] pointed out that nitrogen monoxide is the significant species, among the two oxides of nitrogen being produced in the process. The Zeldovich formulation is therefore not adjusted to account for the existence of nitrogen dioxide. There is no preset activation temperature to activate the model. Note that the relatively low combustion temperature observed in HCCI combustion prevents nitrogen oxides from forming in large quantities. Emission Model: Soot Soot content in the exhaust gas is the difference between formation and oxidation of the species within the engine cylinder. The Hiroyasu model [9] is employed to model the formation process while the Nagle and Strickland Constable model [10] is used to describe the oxidation process in the computation schematic in this study. Soot formation rate is proportional to the mass of fuel vapor multiplied by an Arrhenius rate constant in the Hiroyasu s model. The Nagle and Strickland-Constable model explains the mechanism of soot oxidation. It is assumed that there are two types of reaction sites on the particulate surface, one is highly reactive, while the other is less reactive. A highly reactive site can become a less reactive site through a thermal shifting. Reactions can take place at sites of either type and the reaction rates can be combined to determine the oxidation rate. COMPUTATIONAL MODELING OF THE HIGH SPEED DIRECT INJECTION DIESEL ENGINE The KIVA-3V Release 2 program with the aforementioned enhancement is utilized to simulate the combustion process for the Ford Motor Company s DIATA engine, the test unit for experimentation at UIUC. The specifications of the engine are shown in Table 1. A 60º axisymmetric mesh, created by Y. Zeng and later modified by J. Zhao to include the crevice region, is used to represent of the engine. The employed mesh was capable to capture the three dimensional fluid dynamic while optimizing the computational efficiency. Bore 7.0 cm Stroke 7.8 cm Connecting Rod Length 13.26 cm Displacement Volume 300 cc Compression Ratio 19.5:1 Number of Valves 4 Number of Nozzle Holes 6 Engine Speed 1500 rpm Injection Pressure 600, 1000 bar Equivalence Ratio 0.25-0.45 Swirl Ratio 2.5 Table 1: Specifications of the DIATA research engine
RESULTS Results from several engine operations are presented and compared to the experimental data. Table 2 summarizes the different conditions being examined. Testing Condition A 1 B C D IMEP Output [bar] 3 3 3 7 Engine Speed [RPM] 1500 1500 1500 1500 Injection Pressure [bar] 600 600 600 1000 EGR Fraction [%] 25 50 25 25 Injection Time [CA] 2 360 360 362 330, 372 Fuel Quantity [mm 3 ] 7.7 7.7 7.7 0.8, 15.1 Notes: 1. This is taken as the base case for comparison. 2. Top dead cetre corresponds to 360º crank angle. presented together with experimental data. Since the engine is being operated with no engine oil, piston blowby will have a substantial effect. This can be observed from Figure 2 by noticing the difference in predicting the pressure variation. Therefore, the crevice flow model must be included for accurate results. Good agreement between the experimental data and computational outputs are observed. The peak combustion pressure and ignition delay are both accurately described by numerical results, despite the fact that the pressure trace during compression is slightly delayed. There are some discrepancies between the peak heat release rate and also towards the end of the combustion process, however, the two data sets correspond well initially, including the initial drop at about 368º crank drop due to vaporization of fuel. Table 2: Experiment testing conditions Pressure / Heat Release Calculations Fang et al. [11] provide extra details regarding the experimental results for 3 bar IMEP output. Accurate modeling of the spray dynamics is essential since it has an important effect on autoignition and mixing process. The simulated spray cone for the testing condition of 3 bar IMEP, 25% EGR, injection at 360 crank angle is compared with the image captured from the bottom of the optical engine. Due to light refraction through the quartz piston in the optical engine, the location of the simulated liquid parcels are adjusted according to Snell s Law, assuming the refractive index for air is 1.0 and that for quartz is 1.54. The results are shown in Figure 1. The left halves of the images are obtained from experiments, and the right halves are KIVA generated outputs. It can be seen from the figure that the numerical model accurately describes the spray penetration. Figure 1. Comparison of experimental and computationally simulated and processed spray images The temporal variation of pressure and the apparent heat release, deduced from pressure, at 3 bars IMEP, 25% EGR, injection at 360 is shown in Figure 2. Results from KIVA with and without blowby consideration are being Figure 2. Comparison of pressure and heat release rate for 3 bar IMEP load, 25% EGR with injection at top dead center (360 ). The effect of increasing EGR to 50% and delaying injection until 362º crank angle can be seen in Figures 3 and 4, respectively. The pressure history is accurately described with EGR increased to 50%. Consistent match between the experimental and computational data for the initial drop after top dead center and the subsequent rise due to ignition can be seen in Figure 3. For delayed fuel injection at 362º, fuel is injected when the cylinder undergoes expansion. In spite of the fact that a larger pressure drop is predicted, as observed in Figure 4, the peak combustion pressure is still accurately predicted by KIVA. Also noted from Figure 4 is the optimized Shell Model [12] accurately predicts the ignition delay. Although the accuracy of the model has been improved with the introduction of the blowby subroutine, the pressure history tends to lag the data acquired in experiments for all the three testing conditions. However, the difference is made up around top dead center. Notice that the single combustion reaction of tetradecane tends to overpredict the peak heat release. The positive heat release rates after fuel vaporization are accurately predicted, however.
Figure 3. Comparison of pressure and heat release rate for 3 bar IMEP load, 50% EGR with injection at top dead center (360 ). Figure 4. Comparison of pressure and heat release rate for 3 bar IMEP load, 25% EGR with injection at top dead center (362 ). For the higher load operation of 7 bar IMEP, an initial injection of 0.8 mm 3 of fuel was introduced at 330º followed by injecting 15.1 mm 3 at 372º crank angle. Low temperature reaction from the initial injection can be observed from Figure 5. A small amount of heat is being released and this endures for a relatively long time. These low temperature reactions include low temperature combustion and breakdown of the fuel into smaller radicals. Due to the long duration of the main injection, a slight amount of liquid fuel remains at ignition. However, as seen in Figure 5, the single peaked heat release rate curve indicates the combustion process can still be regarded as HCCI combustion. As in all the previous cases, KIVA tends to overpredict the peak heat release. Delay in the evolution of in-cylinder pressure is, again, observed when compared to the experimental data. However, the numerical result from KIVA matches accurately the top dead center pressure, and the pressure rise during ignition and combustion, although the peak combustion pressure is slightly overpredicted. Figure 5. Comparison of pressure and heat release rate for 7 bar IMEP load, 25% EGR with initial injection at 330 and main injection at 372. Combustion luminescence images produced from KIVA outputs are being compared against with optical images obtained experimentally from the DIATA engine for the 7 bar IMEP load operation. Higher concentrations of soot was created in the combustion zone and because of stronger combustion observed in the engine. Radiative emission is being used to visualize combustion luminescence. The experimental images are obtained from the underside of the piston are being compared in with the contours for soot concentration ranging from 1 10-8 to 5 10-7 g cm -1 in Figure 6. It is seen that combustion starts from the outer edge of the piston bowl with multiple ignition points in both sets of image. The high injection pressure of 1000 bar allows the fuel to mix in the center region of the piston bowl, where combustion luminescence can be observed throughout the region and extends to parts of the squish area. The comparison shows that soot concentration at higher load operation might be correlated to the light emitted during combustion, even though further study on this is needed in order to determine the extent that soot imaging can assist in reproducing combustion luminescence. 380 deg 382 deg 384 deg 386 deg 388 deg Figure 6. Comparison of experimentally captured and computationally produced combustion imaging using soot contours for the 7 bar IMEP operation. Soot and Nitrogen Oxides Emission The temporal variation of soot for 3 bar IMEP operations is shown in Figure 7. The initial increase in soot concentration is due to soot formation and the eventual decrement in soot concentration can be attributed to soot oxidation. The total amount of soot generally reaches its
maximum value in approximately 15º crank angle for all the three operations being studied. However, with EGR increased from 25% to 50%, the soot amount increased slightly after the rapid initial formation due to inefficient combustion with heavy amount of EGR. The weaker oxidation observed in all three cases can be attributed to the fluid flow field within the engine cylinder during the expansion stroke. The soot concentration at 380º crank angle with the engine operating at 3 bar IMEP output, 25% EGR and injection at 360 is shown in Figure 8. Notice that most of the soot is deposited against the outer edge of the piston bowl wall. The flow within the cylinder during expansion is incapable of moving the soot from its locationl to the squish region where it can be effectively oxidized. Indeed, it was found that flows within the bottom half of the piston bowl will push the soot deeper into the bowl. This indicates the significance of the fluid flow field within the cylinder on the effectiveness of soot oxidation. injection reduces the combustion temperature and thus a longer ignition delay reduces the soot produced. The concentration of nitrogen oxide is shown in Figure 9. The species formed rapidly after autoignition and continues to form even after the main combustion event when compared to the single peaked heat release rate curve shown in Figures 3 to 5. As expected from the Zeldovich model, the concentration of nitrogen oxide freezes at the maximum value as the hot regions cool down during the expansion stroke. As seen in the figure, heavy usage of EGR reduced the amount of nitrogen oxide by over 50%. The main contribution to the reduction is the lowered combustion temperature, in addition to the decreased amount of oxygen within the engine cylinder. Late injection, at 362º crank angle, also reduces the production of nitrogen oxide by approximately 20%. Autoignition and combustion both occurs at a later instant for both cases, allowing the fuel vapor-air mixture to expand and cool and this reduces the peak combustion temperature. Inferred from the three-dimensional contours shown in Figure 10, one can observe that regions of high concentration of nitrogen oxides coincide with regions of high local temperature, typically above 2500 K. This verifies the aforementioned conclusion that the amount of nitrogen oxide decreases due to lowered combustion temperature by both increasing EGR and delayed fuel injection. Figure 7. Comparison of in cylinder soot concentration at 3 bar IMEP load operation, with different injection schemes Figure 9. Comparison of in cylinder nitrogen oxide concentration at 3 bar IMEP load operation, with different injection schemes Figure 8. Soot concentration at 380 during the peak of formation for 3 bar IMEP load, 25% EGR and injection at 360 crank angel. Comparing the three cases in Figure 7, one can realized that the oxidation process is much less efficient with high quantity of EGR. In spite of increasing EGR postponed ignition, thus, enhanced the mixing process, the decreased oxygen concentration results in richer local conditions during combustion. The oxidation process is thus less effective with 50% EGR. Late injection of fuel at 362º crank angle reduced the soot concentration by almost 50%. This shows that this injection scheme is being used to its advantage. Deferred CONCLUSIONS The KIVA-3V Release 2 program was modified for simulating a HSDI diesel engine with optical accessibility operating under HCCI conditions. The Namazian formulation for the ring-pack model and crevice flow model was enhanced to account for circumferential flow between the piston ring and the cylinder wall. This is essential for accurately modeling of the optical engine because it is run without engine oil. The optimize Shell ignition model was integrated to KIVA. The Zeldovich formulation was being implemented for describing the nitrogen oxide formations. The Hiroyasu Model, together with the Nagle and Strickland-Constable models were
implemented to characterize the soot formation and oxidation processes. Figure 10. Three dimensional contours for temperature, fuel, nitrogen oxide and soot concentrations of in cylinder imaging at 378º for 3 bar IMEP, 25% EGR and top dead center injection engine operation The spray cone images were processed, by relocating the liquid parcels according to Snell s Law to account for light refraction through the quartz piston and compared to experimentally generated images. Good agreement was observed between the two results. The pressure and heat release rate histories were compared against data obtained experimentally. The dominant effect of engine blowby in an optical engine is verified by comparing the results from the KIVA code with and without inclusion of the blowby model. The optimized Shell model is also shown to accurately predict the ignition timing. However, overprediction of the peak heat release is observed with the single reaction of tetradecane that represents the diesel fuel combustion. Combustion images obtained from the 7 bar IMEP operation of the HSDI engine were produced, by plotting the contour of soot concentration and were compared to the experimental images. Consistency was observed between the numerical and computational results. Nevertheless, nitrogen oxide and soot emissions were also studied. It was observed that the efficiency of soot oxidation significantly depends on whether the in-cylinder fluid can successfully relocates the soot to regions with high oxygen concentration. Increasing the EGR could significantly reduce the amount of nitrogen oxide produced. The reduced combustion temperature, together with decreased amount of oxygen, leads to the significant reduction of nitrogen oxide. Flows with Sprays, Los Alamos National Laboratory Report LA-11560-MS, 1989. 3. Namazian, M., Studies of Combustion and Crevice Gas Motion in a Flow Visualization Spark Ignition Engine, Ph.D. Thesis, Massachusetts Institute of Technology, 1981. 4. Namazian, M. and Heywood, J. B., Flow in the Piston Cylinder Ring Crevices of a Spark Ignition Engine: Effect on Hydrocarbon Emissions, Efficiency and Power, SAE Paper 820088. 5. Kong, S.C., and Reitz, R.D., Multidimensional Moeling of Diesel Ignition and Combustion Using a Multistep Kineticks Model, Journal of Engineering for Gas Turbines and Power, Vol. 115, 1993, pp. 781-790. 6. Sazhina, E.M., Sazhin, S.S., Heikal, M.R., Babushok, V.I. and Johns, R.J.R., A Detailed Modeling of the Spray Ignition Process in Diesel Engines, Combust. Sci. and Tech., Vol. 160, 2000, pp. 317-344. 7. Shazi, R, Modeling the Autoignition, Combustion, and Pollutant Formation in a High-Speed Direct Injection Diesel Engine, MS Thesis, Department of Mechanical and Industrial Engineering, University of Illinois at Urbana-Champaign, 2002. 8. Heywood, J.B., Internal Combustion Engine Fundamentals, McGraw-Hill, New York, pp. 572-577, 1988. 9. Hiroyasu, H. and Nishida, K., Simplified Three- Dimensional Modeling of Mixture Formation and Combustion in a D.I. Diesel Engine, SAE Paper 890269, 1989. 10. Nagle, J. and Stickland-Constable, R.F., Oxidation of Carbon between 1000-2000 C, Proceedings of the Fifth Carbon Conference, Vol. 1, Pergamon Press, London, 1962. 11. Fang, T., R.E. Coverdill, C.F. Lee, and R.A. White, Low temperature combustion within a small-bore high-speed direct-injection (HSDI) diesel engine, SAE Paper 2005-01-0919. 12. Hamosfakidis, V., and Reitz, R.D., Optimization of a Hydrocarbon Fuel Ignition Model for Two Single Component Surrogates of Diesel Fuel, Combustion and Flame, Vol. 132, 2003, pp. 433-450. REFERENCES 1. Zhao, F., Asmus, T.W., Assanis, D.N., Dec, J.E., Eng, J.A. and Najt, P.M., Homogeneous Charge Compression Ignition (HCCI) Engines: Key Research and Development Issues, Society of Automotive Engineers, Warrendale, 2003. 2. Amsden, A.A., O Rourke, P.J. and Butler, T.D., KIVA- II: A Computer Program for Chemically Reactive