Application of Detailed Soot-Particle Model to Simulations of Fundamental Spray Experiments and GDI Engine

Transcription

1 International Multidimensional Engine Modeling User's Group Meeting April 7, 2014, Detroit, Michigan Application of Detailed Soot-Particle Model to Simulations of Fundamental Spray Experiments and GDI Engine Long Liang, Chitralkumar V. Naik, Karthik V. Puduppakkam, Abhijit U. Modak, Ellen Meeks Reaction Design 5930 Cornerstone Court West, Suite 230, San Diego, CA The method of moments has been implemented in FORTÉ CFD to track the evolution of dispersed phase particles. In this paper, we demonstrate its usage along with a detailed soot-particle model. We have applied this soot-modeling methodology to simulations of fundamental spray combustion cases and of a real-world spray-guided gasoline direct injection (GDI) engine. The detailed soot kinetics model used in this paper is formulated based on targeted laboratory-scale soot experiments. The resulting gas-particle chemistry mechanism captures particle inception, growth, and oxidation. Computed soot particulate characteristics were compared to available experimental data. The soot model results were then further analyzed to demonstrate the in-depth physical/chemical information provided by the detailed soot-particle model. 1. Introduction Since the introduction of the EURO 5 emission standard, particulate matter (PM) emissions are no longer only a concern for diesel engines. The 4.5 mg/km limit of PM is being enforced for both diesel and gasoline direct injection (GDI) engines. In addition to the PM mass requirement, the new European legislation (EURO 6) will also implement a particulate number (PN) requirement for all spark ignition (SI) engines. To meet the progressively more stringent standard, it is crucial for engine powertrain developers to gain a fundamental understanding of the chemical and physical processes of soot formation in combustion. Numerous studies have shown that fuel composition has significant impact on PM emissions. For example, one research published in 2006 [1] studied the effect of various parameters on PM emissions in a spray-guided gasoline direction injection (SGDI) engine. Parameters studied include fuel type, air-fuel ratio, injection timing and spark timing. It was concluded that fuel type had the biggest impact on PM emissions changing the fuel from iso-octane to toluene caused the PM emissions to increase by one order of magnitude. Another study published in 2010 [2] investigated the correlation between gasoline properties and particulate number (PN) for a large number of gasoline samples collected in various countries. Their measurements over the New European Driving Cycle (NEDC) indicated that aromatics with a high boiling point and a high double bond equivalent value tended to produce higher PN. The PN value can be different by a factor of 10 among the fuels tested. These studies indicate that any numerical simulation of soot formation must be able to capture the fuel effects accurately in order to be used as a predictive design tool. Recent advances on several fronts have made it feasible for simulations to predict combustion and emissions 1 characteristics for various fuels under a wide range of operating conditions. First, accurate and validated detailed reaction mechanisms for many gaseous and liquid fuel components are available [3]. Using a multi-component fuelsurrogate approach [4], the chemistry of complex fuels can be well represented. Compared to using a single component surrogate (such as n-heptane for diesel), which may be adequate for capturing ignition characteristics, multicomponent surrogates can capture the fuel effects on both combustion and emissions. Second, the advances in Computational Fluid Dynamics (CFD) capabilities allow reliable modeling of sprays and hence high-fidelity simulations of spray dynamics and vaporization are possible. Third, advances in core solver technology has resulted in orders of magnitude reduction in simulation time even with direct usage of larger (more accurate) chemistry reaction mechanisms in CFD simulations [5]. In FORTÉ CFD, we have implemented the method of moments [6] for solving particle size and number density information [7]. With a detailed soot kinetics model, FORTÉ CFD can simulate soot formation and destruction processes, including gas-phase precursor formation, particle nucleation, surface growth and oxidation, as well as particulate coagulation and aggregation. Compared to phenomenological soot models, which often make restrictive assumptions such as a constant and uniform composition for soot, the present method simulates soot particle evolution in a more fundamental way in a CFD simulation. The goal of the present work is to demonstrate this soot modeling approach using simulations under engine conditions. Two sets of experiments were used, one is the set of constant volume spray experiments published by the Engine Combustion Network at Sandia National Laboratories [8-10], and the other case is a spray-guided GDI engine by IFP Energies nouvelles (IFPEN) [11].

2 2. Modeling Methodology 2.1 Detailed Soot-Particle Model The detailed soot-particle kinetics model used in the present work was developed and validated by Puduppakkam et al. [7], using a broad range of fundamental experimental data. This soot model includes nucleation of particles (PM) from pyrene in the gas phase and includes heterogeneous gassurface reactions for describing soot growth, oxidation, and soot-aging processes. These processes are illustrated in Fig. 1. Detailed fuel combustion (gas phase) mechanisms are used to describe the formation of gas-phase soot precursors. Nucleation describes how particle (or nucleus) is created from the gas-phase soot precursors. After the nuclei are formed, they start to interact with each other as well as with the gas mixture around them. While particle-particle interactions such as coagulation/aggregation are non-chemical processes, interactions between particles and surrounding gas mixture are chemical processes taking place on the particle surface. These surface processes are modeled using a surface chemistry mechanism. They may lead to particle mass growth or reduction and/or reconditioning of particle surface. 2.2 Particle Tracking Using the Method of Moments The application of the method of moments to soot-particle formation and tracking was first reported by Frenklach and coworkers [6]. The method of moments tracks the evolution of an aerosol system by the moments of its particle-size distribution function. The moments are used to represent the average properties of a particle population, such as number density, total particle volume fraction, total particle surface area density, and the average particle size. In discrete form, the particle-size moments are defined as M = j N (1) where M is the rth moment, j is the particle size, and N is the number density of the particles of size j. Using this definition, it can be shown that the zero-th moment is the total particle number density of the particle population, i.e, N = M (2) The total particle mass density can be expressed as m = m M (3) where m is the mass of the bulk species molecule comprising the particle core. The average particle diameter can be derived as d = d M / /M (4) where M / is the one-third moment and d is the diameter of the bulk species molecule. In FORTÉ CFD, the rth particle moment, M, is tracked using the following transport equation: + (u M ) = j D, N + (R + G + S ) (5) Thus, the rate of change of M in a computational cell comes from three contributing factors: source term, diffusion transport and convective transport. The R, G, S terms on the right-hand side are the source terms and correspond to nucleation, coagulation, and surface reactions, respectively. The second term on the right-hand side is the diffusion term, in which D, is size-dependent particle diffusivity. The second term on the left hand side is the convection term due to bulk flow velocity. Using the operator-splitting method of FORTÉ, the source terms are solved together with the gasphase chemistry. 3. Results and Discussion 3.1 Spray Combustion in Constant Volume Chamber Two spray combustion experiments from Sandia National Laboratories (referred to as Spray H and spray A, respectively) were modeled to test the detailed soot-particle model. The Spray H experiment published by Idicheria and Pickett [9] studied ignition and combustion of n-heptane jets at 1000 K and two ambient density levels (14.8 kg/m 3 and 30 kg/m 3, respectively). For each ambient density level, the volume fraction of O 2 is varied from 8% to up to 21%. The Spray A experiment by Kook and Pickett [10] studied three fuels, a jet fuel surrogate (Jet-SR) with a binary blend of n- dodecane/m-xylene: 77/23% by liquid volume; diesel No. 2; and world-average jet fuel (Jet-A). Among these, this paper presents results only for Jet-SR, whose composition is precisely defined. This choice is made to eliminate uncertainties associated with selecting a surrogate blend for complex fuels. In Spray A, the ambient density was fixed at Fig. 1 Detailed Soot-Particle Model 2

3 22.8 kg/m 3 and oxygen level was fixed at 15%. The initial ambient temperature was varied between 900 K and 1000 K. In both experiments, the key diagnostic parameters measured in these experiments include pressure profiles, OH chemiluminescence, Schlieren imaging, and soot amounts. The measurements were made using laser extinction and planar laser-induced incandescence. Conditions inside the chamber were nearly uniform and contained a mixture of O 2 /N 2 /CO 2 /H 2 O. For both spray experiments, the fuel injection profiles were square and the duration of injections was approximately 7 ms to achieve a quasi-steady state for the measurements. The nozzle exit diameters used in Spray H and Spray A are 0.1 mm and 0.09 mm, respectively. Transient calculations were performed in FORTÉ CFD using Reynolds Averaged Navier-Stokes (RANS) approach. The KH-RT breakup model and a multi-component fuel vaporization model were used for spray. The same default breakup model constants were used in all calculations. The nozzle discharge coefficients in Spray H and Spray A are 0.8 and 0.86, respectively, which match the measured values. A master mechanism that includes 4263 species and reactions was reduced for each fuel using the Reaction WorkBench [12]. The resulting reduced mechanisms for n- heptane and the Jet-SR contain 326 and 389 species, respectively. The constant volume chamber (edge length = 10.8 cm) is meshed using Cartesian mesh. Figure 2 shows the mesh cutting through the center of the spray structure. The cell size in the finest region is 1 mm. Results were found to be insensitive to further mesh refinement. Fig. 3 Predicted and measured lift-off lengths for spray H. The soot volume fractions (SVF) for Spray H at an ambient density of 14.8 kg/m 3 and varying oxygen concentrations are shown in Fig. 4. Corresponding plots for the ambient density = 30 kg/m 3 are shown in Fig. 5. Both cases show that with a decrease in the ambient oxygen concentrations, the SVF decreases and the high soot zone is shifted downstream of the flame. At higher ambient density, the soot volume fraction is nearly six times higher than that at lower density. In the case of high ambient density and 10% O 2, the downstream shift of the peak soot region is underpredicted. This discrepancy is likely due to imperfections in the chemistry in capturing the impact of oxygen on soot oxidation. Overall, the sooting characteristics were reproduced well by simulation for all Spray H conditions. Fig. 4 Measured and Computed soot volume fraction in Spray H (ambient density = 14.8 kg/m 3 ). Fig. 2 Mesh used in Spray H and Spray A simulation. Spray H Results Both spray and combustion characteristics were well captured in the simulation for the spray H conditions, including liquid/vapor penetration lengths, ignition delay, and lift-off length. As an example, the calculated and measured lift-off lengths are compared in Fig. 3. In this comparison, liftoff length is determined by the distance between the injector and the axial location of the sharp increase in hydroxyl radical (OH) concentrations. Fig. 5 Measured and Computed soot volume fraction in Spray H (ambient density = 30 kg/m 3 ). 3

4 Spray A (Jet-SR) Result The SVF for Spray A at both 900 K and 1000 K are shown in Fig. 6. The simulation accurately predicted the influence of ambient temperature on soot. As the ambient temperature increases from 900 K to 1000 K, the peak SVF is increased by a factor of five, and the high soot region moves slightly upstream of the flame. is initiated at 339 CA ATDC, i.e., 1 CA degrees after the end of injection. Simulation using FORTÉ CFD was carried out from intake valve opening (IVO) to exhaust valve opening (EVO), covering the intake charge, compression, spray, combustion and expansion processes. Automatic mesh generation with local mesh refinement was used to handle moving boundaries. In Fig. 8 mesh snapshot is shown on a cut plane. The number of cells in the entire mesh varied from a maximum of 580,000 cells to a minimum of 74,000 cells. The Linearized Instability Sheet Breakup (LISA) model implemented in FORTÉ was used to model the spray. The G-equation model was used to calculate the turbulent (premixed or partially premixed) flame propagation in the cylinder. Fig. 6 Measured and Computed soot volume fraction in Spray A for the binary jet fuel surrogate (Jet-SR). Distribution of predicted particle size (diameter) in the plane of laser sheet is shown in Fig. 7. The location of larger particles coincides with the location of high soot regions. This is expected since particles upstream of the peak soot zone are still growing and those downstream are being oxidized quickly. Predicted peak particle size is slightly larger at 1000 K than that at 900 K. The particle diameter averaged over the whole chamber is on the order of 10 nm in this jet flame. Fig. 8 Geometry and configuration of the IFPEN GDI engine. A 7-component surrogate fuel was created using a Surrogate Blend Optimizer in Reaction Workbench software [12]. As listed in Table 1, the multicomponent surrogate matches the chemical and physical properties of the gasoline very well. The composition of the 7-component surrogate blend is shown in Table 2. Table 1 Properties of the European gasoline tested and the surrogate fuel used for the simulation. Fig. 7 Computed soot particle diameters (microns) in Spray A. 3.2 Spray-Guided GDI Engine The GDI engine simulated is a 4-stroke 4-valve sprayguided IFPEN engine with optical access [11]. The bore diameter, stroke, and connecting rod length are 82.7 mm, 93 mm, and 144 mm, respectively. The fuel injector was placed at the center of the cylinder head with an angle of 15 relative to the vertical axis. The distance between the nozzle and the piston surface at TDC is about 10 mm and the distance between the spark plug and piston surface at TDC is 8 mm. The engine geometry is shown in Fig. 8. The IFPEN GDI engine was operated using a standard European gasoline with average Octane rating of 91 at low-load conditions (1200 RPM, 4.3 bar IMEP). The overall equivalence ratio is 0.4. Fuel injection uses a hollow-cone injector. Injected fuel mass is 15.3 mg per engine cycle. Start of injection is at 334 CA ATDC and injection duration is 4 CA degrees. Spark Properties Gasoline Surrogate RON MON H/C molar ratio LHV, MJ/kg T10, K T50, K T90, K Table 2 Composition of the multicomponent surrogate fuel. Fuel Components Liq. vol % iso-octane 25.0 n-butane 3.3 n-pentane 23.3 Toluene ,2,4-Trimethyl Benzene 21.5 Methyl Cyclohexane Hexene 5.8 4

5 A 230-species, 1740-reaction mechanism reduced from a master mechanism was used to calculate gas phase chemical kinetics. First, the pressure and heat release rate traces are compared to measured data in Fig. 9. Measured pressure data from 500 cycles are also shown in the Figure to highlight the impact of cycle-to-cycle variation. As seen, the computed curves match the measured curves reasonably well. zone through wall film vaporization or liquid droplet vaporization. Computed in-cylinder averaged soot parameters indicate that the soot particle number density (the zero-th moment) reaches its peak value at 357 CA ATDC, but the soot volume fraction (which scales with the first moment) reaches its peak at 375 CA ATDC. This indicates the evolution of soot particles from inception to growth along with coagulation. After 375 CA ATDC, the soot mass and volume fraction start to decrease and this can be attributed to oxidation. 4. Summary Fig. 9 Computed and measured in-cylinder pressure traces of the IFPEN GDI engine. In the experiment, soot volume fractions are measured using a combination of planar laser-induced incandescence (PLII) and laser extinction method (LEM). The laser sheet is located at a vertical plane in-between the injector and the spark plug. Time evolution of the soot cloud is shown in Fig. 10 at eight selected crank angle locations. The soot cloud is represented by an iso-surface on which soot volume fraction equals 0.01 ppm. At each crank angle location, the solutions are illustrated using a top view and a side view, respectively. The geometries shown include the piston surface and a vertical cut plane, which are shaded using temperature. At 338 ATDC (which is at the end of injection and 1 CA degree before the advent of the spark), the edge of the spray structure impinges on the spark plug, causing wall wetting on the spark plug. The spark ignition kernel flame forms and starts to propagate at around 347 CA ATDC. At this crank angle location, the simulation predicts that a small soot cloud is formed outside the spark gap. Then the soot cloud keeps growing until 375 CA ATDC. During the expansion stroke, the soot cloud gradually shrinks and eventually gets separated from the cylinder head and moves downward with the piston. It can be clearly seen that the flame propagation speed is much faster than the growing speed of the soot cloud. This indicates that sooting is mainly a post-flame phenomenon in this GDI engine case. The flame propagation model does not directly interact with the soot model. Soot is produced when liquid fuel enters the high-temperature/low-oxygen-content post-flame Using FORTÉ CFD, simulations were performed for both fundamental spray combustion experiments and a sprayguided GDI engine. The sub-models incorporated in FORTÉ include method of moments to track evolution of soot particles. A detailed soot-particle kinetics model was used in these simulations. Computed soot volume fractions were in good agreement with the measured data in all spray chamber cases. Soot formation and oxidation processes in the IFPEN GDI engine were also analyzed. The capability of using fundamental models to predict soot particulate characteristics should be able to help engine combustion developers to gain useful insights into sooting problems as they take on the challenge of meeting more stringent emission regulations. References 1. P. Price, R. Stone, T. Collier, M. Davies, SAE (2006). 2. K. Aikawa, T. Sakurai, J. Jetter, SAE (2010). 3. K. V. Puduppakkam, C. V. Naik, E. Meeks, SAE (2010). 4. C. V. Naik, K. V. Puduppakkam, C. Wang, J. Kottalam, L. Liang, D. Hodgson, E. Meeks, SAE Int. J. Engines, 3(1), pp (2010). 5. FORTÉ CFD, Reaction Design, San Diego, CA. 6. M. Frenklach, H. Wang, in Soot Formation in Combustion: Mechanisms and Models, H.Bockhorn (Ed.), Springer-Verlag, pp (1994). 7. K. V. Puduppakkam, A. U. Modak, C. V. Naik, E. Meeks, Proc. ASME TurboExpo, GT (2014) Combustion Network, C. A. Idicheria, L. M. Pickett, SAE (2005). 10. S. Kook, L. M. Pickett, SAE Int. J. Fuels Lubr., 5(2), pp (2012). 11. C. Lacour, et al. Analyse de la fraction volumique de suies par couplage PLII/LEM dans un moteur IDE. in Congrès Francophone de Techniques Laser, CFTL Vandoeuvre-lès-Nancy, France. 12. Reaction Workbench, Reaction Design, San Diego, CA. 5

6 Fig. 10 Time evolution of soot cloud (iso-surface with SVF=0.01 ppm). Cut planes are shaded using temperature. 6