Comparison of diesel spray combustion in different high-temperature, high-pressure facilities Lyle M. Pickett, Caroline L.

Genzale Sandia National Laboratories Gilles Bruneaux and Louis-Marie Malbec IFP Caspar Christiansen Technical University of Denmark Abstract Diesel spray experiments at controlled high-temperature and high-pressure conditions offer the potential for an improved understanding of diesel combustion, and for the development of more accurate CFD models that will ultimately be used to improve engine design. Several spray chamber facilities capable of high-temperature, high-pressure conditions typical of engine combustion have been developed, but uncertainties about their operation exist because of the uniqueness of each facility. For the IMEM meeting, we describe results from comparative studies using constant-volume vessels at Sandia National Laboratories and IFP. Targeting the same ambient gas conditions (900 K, 60 bar, 22.8 kg/m3, 15% oxygen) and sharing the same injector (common rail, 1500 bar, KS1.5/86 nozzle, 0.090 mm orifice diameter, n-dodecane, 363 K), we describe detailed measurements of the temperature and pressure boundary conditions at each facility, followed by observations of spray penetration, ignition, and combustion using high-speed imaging. Performing experiments at the same high-temperature, highpressure operating conditions is an objective of the Engine Combustion Network (http://www.ca.sandia.gov/ECN/), which seeks to leverage the research capabilities and advanced diagnostics of all participants in the ECN. We expect that this effort will generate a high-quality dataset to be used for advanced computational model development at engine conditions. Introduction The Engine Combustion Network (ECN) is an experimental and modeling collaboration dedicated to the improvement of engine CFD modeling. The ECN contains an internet data archive library (http://www.ca.sandia.gov/ECN/) with well-documented diesel spray experiments at engine conditions. Quantitative comparisons of spray evaporation and mixing at high-temperature, high-pressure conditions are now possible, providing new opportunities for CFD model improvement at realistic engine conditions. The dataset also currently includes reacting and non-reacting data such as the liquid and vapor penetration versus time, liquid length, ignition delay and pressure-rise rate, lift-off length, quantitative soot volume fraction, and various high-speed movies of combustion. Various groups have used the dataset for model validation (e.g. [1-8]). Until recently, the ECN data archive was derived from a single spray chamber facility at Sandia National Laboratories. However, many ECN users responded favorably to this initiative and are now working collaboratively to create future datasets of interest, utilizing multiple facilities throughout the world. The ECN working group has identified a few experimental conditions that will be the focus of modelers and experimentalists who wish to voluntarily participate in the ECN. The experimental collaboration is made possible by the donation of "identical" injection systems by Robert Bosch LLC. The ECN group has agreed to experiment at the same injector and ambient conditions, thereby leveraging the expertise of each participant, and enabling direct comparison between different facilities. Significant effort is being made to verify the boundary conditions at each facility, as well as to measure the rate of fuel delivery and internal movement of the injector. In the end, all groups would benefit by sharing information at the same operating conditions. This is the same vision guiding the Turbulent Nonpremixed Flame (TNF) workshop [9], where years of quantitative experimental effort on a particular flame (e.g. Flame D) have created a standard library for computational comparison. The ECN seeks to generate a similar database at engine conditions. We expect that the dataset developed will become a serious focal point for model validation and further advanced diagnostics.

"Spray A" Conditions Following the tradition of the TNF workshop, we have chosen to name the first target condition "Spray A". Specifications for the injector and ambient operating conditions for this condition are given in Table 1. The first target experimental condition is a low-temperature combustion condition relevant to engines that use moderate EGR and have minimal NOx emissions. The injector specifications are relevant to modern advanced injection systems with high pressure capability. Bosch donated 10 different injectors for the initiative, including 5 single-hole injectors with an orifice on the axis of the nozzle, as well as 5 three-hole injectors with an 145 included angle. A comparison between single and multi-, side-hole injectors will therefore be a part of the Spray A condition. The common rail and injector tube length are specified to ensure consistent hydraulic effects on injection. A single-component fuel, n-dodecane, is chosen to enable a complete specification of the physical and chemical properties of the fuel. Detailed chemical kinetic mechanisms exist for n-dodecane, allowing full treatment of the combustion chemistry. Finally, although the boiling point temperatures of n-dodecane are lower than that of diesel fuel, ndodecane has boiling point temperatures that are sufficiently high to emphasize spray mixing processes. (It also is compatible with production fuel pumping systems.) Table 1. Spray A Experimental Conditions.
Ambient gas temperature Ambient gas pressure Ambient gas density Ambient gas composition Common rail fuel injector Fuel injector nominal nozzle outlet diameter Nozzle K factor Nozzle hydro-erosion Spray full included angle Fuel injection pressure Fuel Fuel temperature at nozzle Common rail volume/length Distance from injector inlet to common rail Fuel pressure measurement Injection duration Approximate injector driver current 900 K 6 MPa 22.8 kg/m3 15% O2, 75.1% N2, 6.2% CO2, 3.6% H2O Bosch solenoid-activated, generation 2.2 0.090 mm 1.5 { K = (dinlet doutlet)/10 [use m] } Discharge coefficient = 0.86 with 100 bar P. 0 (single axial hole) or 145 (three-hole nozzle) 150 MPa n-dodecane 363 K (90C) 22 cm3/ 28 cm (Use GM rail model 97303659) 24 cm 7 cm from injector inlet / 24 cm from nozzle 1.5 ms 18 A for 0.45 ms ramp, 12 A for 0.345 ms hold

Objective Spray data of the caliber needed for detailed model development and validation requires extensive characterization of all of the boundary conditions needed as model inputs. However, the type of spray chambers that operate at high-temperature, high-pressure conditions typical of engine combustion are uncommon and uncertainties about their operation exist because of the uniqueness of these facilities. Nevertheless, ECN experimental participants are expected to measure the boundary conditions (e.g., the parameters specified in Table 1) for the ambient and injector, and state uncertainties associated with the boundary conditions and other data. Our approach to better understand each facility, and to minimize uncertainties, is to directly compare results at standardized experimental conditions, i.e., Spray A. Once these boundary conditions are better understood, variation in parameters beyond those of Spray A can proceed with higher confidence. The objective of this report is to present concurrent measurements of Spray A boundary conditions and spray penetration and combustion data at two different facilities, a combustion vessel at IFP and one at Sandia National Laboratories. We compare results from each facility and share the first detailed measurements of Spray A. Because the experimental campaigns occurred at the same time, we used two different injectors (axial, single-hole). Therefore, the results also address questions about repeatability from injector to injector. A final objective of this report is to demonstrate steps that are needed for other interested groups to participate in ECN spray research.

Ambient Gas Characterization IFP and Sandia use a preburn-type combustion vessel to generate high-temperature, high-pressure gases. The ambient pressure, temperature and species at the time of injection are varied by igniting a combustible-gas mixture that burns to completion, and waiting for the product gases to cool to the desired state at the time of spray injection. A detailed description of this type of facility is found in [10]. Figure 1 shows high-speed schlieren imaging of the preburn and cool down process relative to the time of spark. The images show a premixed flame propagating from a pair of spark plugs at the top of the vessel. The flame moves through the charge to the bottom of the vessel, completing the premixed burn by approximately 400 ms after the time of spark. At 200 ms, the flame envelope is midway through the combustion chamber, and clear zones are found on the top and the bottom of the schlieren images. The bottom zone is fresh reactants that have yet to burn. The top zone contains combustion products that are no longer burning across the entire span of the chamber. Lacking strong gradients in temperature that are characteristic of a premixed flame, the top zone also appear clear, at least at first. With increasing time, however, schlieren disturbances appear at the top and move throughout the chamber. These schlieren effects are caused by heat transfer to the cooler vessel walls, which forms temperature gradients in boundary layers near the walls [11]. Gas temperature measurements were performed during the preburn and cool down to understand the temperature distribution of gases that will mix with the spray upon injection. An adjustable probe with fine wire (50-m diameter) platinum/platinum-rhodium (type-R) thermocouples was used to characterize this thermal environment. Figure 2 shows measurements of gas temperature at a position 55 mm in front of the injector, a position that is also indicated by a red marker in each image of Fig. 1. The measured temperature shown in Fig. 2 is not corrected for radiation and heat transfer transients (e.g. [12]), but these corrections were applied to the data near the time of injection to improve accuracy at the time of interest. The bulk temperature is also shown, as determined from the real-gas equation of state using the measured pressure and initial mass-averaged bulk density (bulk density is constant for an experiment). The gas temperature measurements show an initial rise caused by compression heating of unburned mixtures ahead of the flame. The flame arrives at the thermocouple at approximately 250 ms, whereupon there is a rapid increase in temperature similar to the expected flame temperature. After the
2000 1800 1600

Vessel position for TC Axial x = 55 mm Horiz. y = 0 mm Vertical z = 0 mm

Temperature [K]

1400 1200 1000 800 600 400 500

Tbulk
Flame arrival

Tbulk =

P MW Ru Z
3000
7

Compression Heating
1000 1500 2000 2500 Time after spark [ms]

Figure 1. Schlieren imaging of preburn and cool down until the time of injection for Spray A conditions. Flame begins at two spark plugs at the top. Injector is at left, distances are relative to injector in mm. Sandia.

Figure 2. Gas temperature measurements by thin-wire (0.05 mm) thermocouple for Spray A conditions (green). Vessel bulk temperature, derived from the measured pressure and bulk density prior to the preburn (blue). Thermocouple probe is at the same vertical position as the injector, as indicated in the legend, and drawn as a red marker in Fig. 1. Data shown is not corrected for radiation and transient heat transfer, but these corrections are later applied to the data range at the time of injection (3000 ms). Sandia.

time of peak pressure (or bulk temperature), the gases slowly cool by heat transfer with the surrounding walls. A mixing fan runs during the entire process (shown at the upper right in the schlieren images) to produce a more uniform mixture temperature [13]. Significant fluctuations occur for gas temperatures higher than 1300 K, but the gas temperature settles thereafter as it cools to the Spray A temperature (900 K) at 3 s. Note that the core temperature, the zone right in front of the injector, is higher than the bulk temperature because cooler, higher density gases exist in boundary layers or in the bottom of the vessel. This point is often overlooked in engine or combustion vessel experiments. These measurements provide a relationship between the core and bulk temperatures so that the core temperature surrounding the fuel spray can be predicted from the pressure measured during subsequent experiments. Figure 3 shows the average gas temperature at various positions in this core region of the vessel, relative to predictions based on the measured pressure. Along the axis of the spray, the gas temperature is uniform within 1%, except for a measurement point in the boundary layer only 1 mm in front of the injector tip. Temperature is also uniform along the same vertical plane (z = 0) outside of the spray. The most significant variation is in the vertical direction where the temperature varies by 4% for a vertical change of 15 mm. This vertical stratification is obviously caused by buoyancy effects. Despite the vertical variations in temperature, systematic combustion asymmetries in the vertical direction (i.e., ignition site, or lift-off length) have not been found. Similar gas temperature measurements were performed at IFP to establish the core temperature in front of the injector apart from the bulk temperature. Injector Characterization Figure 4 shows the mass flow rate of injection for the injector used at Sandia, along with fuel pressure and injector driver current at both Sandia and IFP. By specifying the same common rail, and rail-toinjector tube length (see Table 1), the hydraulic fluctuations in fuel pressure are shown to be similar at each institution, even when using different injectors. The fluctuations in fuel pressure appear to coincide with measured fluctuations in the rate of injection, with a phase shift consistent for the fuel speed of sound (approximately 1600 m/s) and the length between the nozzle and the pressure measurement location. Despite some differences in the absolute fuel pressure (caused by limitations with the fuel pressure control systems), the similarity in measured fuel pressure fluctuations means that subtle changes in fueling rate during the course of injection will likely be represented at each institution. The injector driver current is also reasonably similar, even though different custom drivers are used at each institution. Spray Results Useful diagnostics for initial comparison of spray behavior include high-speed schlieren, Miescatter, and chemiluminescence imaging. Figure 5 shows a time sequence of schlieren images, with the liquid boundary of the spray from simultaneous Mie-scatter imaging overlaid in blue [11]. Image pairs are shown with raw schlieren and also with background correction from successive images in order to more easily visualize the actual jet structure, as opposed to background schlieren
1.06 1.04 1.02 1 0.98 0.96 0.94 0 10 20 30 40 50 60 70 Axial distance from injector nozzle [mm] 80 TC1, y = -15, z = 0 TC2, y = 0, z = 0 TC3, y = 0, z = -15 TC4, y = 0, z = +15 TC5, y = +15, z = 0

Figure 3. Average gas temperature in the core region of the Sandia combustion vessel at Spray A conditions.
5

Tcore,measured/Tcore,predicted

IFP

160 150

4 Sandia 3 2 1 0 0 500 1000 Time ASI [s] 1500

Pressure
IFP

Mass flow

Sandia

140 130 120 110 2000

Sandia

Current/10

Figure 4. Spray A mass flow rate of injection and comparison of measured fuel pressure and injector driver current at IFP and Sandia. Mass flow rate is derived using force (momentum) measurement, along with the total mass of fuel injected (3.5 mg).

Fuel Pressure 7 cm from injector [MPa] 26 cm from nozzle

Mass flow rate [mg/ms]

effects discussed in conjunction with Fig. 1. Characteristics displayed in the schlieren imaging include: separation between vapor and liquid phases by 100 s after start of injection (ASI); cool flame appearance which appears to erode the head of the jet by 300 s; high-temperature ignition by 500 s; bright soot luminosity at the jet head by 1200 s; end of injection at 1500 s; flame flashback to the injector by 2000 s; and soot burnout by 2500 s. See Ref. [11] for further discussion about interpretation of these phenomena in schlieren imaging. Figure 6 shows high-speed chemiluminescence imaging performed at Sandia (top) and OH chemiluminescence performed at IFP (bottom). After ignition, the flame lift-off length is slightly less than 20 mm in both facilities. The similarity in measured lift-off length is significant because of the strong dependency of flame lift-off on ambient gas temperature [14], which suggests that a similar ambient temperature exists at each facility. Finally, we compare the liquid and vapor penetration for reacting (0% O2) and non-reacting (0% O2) conditions, performed by high-speed Mie scatter and schlieren imaging at both Sandia and IFP. Results are given in Fig. 7, which shows good agreement between the liquid, vapor, and reacting flame penetration at each facility. In fact, differences between the IFP and Sandia vapor penetration datasets, including some variance in the IFP data, may be caused only by differences in the image processing methods. As shown in Fig. 5, removal of background schlieren effects is necessary to identify the vapor border of the jet and the image processing routines are still being improved. Based on the liquid penetration, the injection duration also appears slightly longer at IFP. This difference is not surprising, given that different injectors were in use. However, the injection duration can be controlled easily by trimming the injector driver current to minimize this difference. Another interesting difference is that of the liquid penetration during the transient start of injection. The IFP data shows an overshoot in liquid
Broadband Chemiluminescence

-15 -10 -5 0 5 10 15

OH Chemiluminescence (1100-1800 s ASI)

10

20

30

40

50

60

70

80

90

Figure 5. Schlieren imaging with liquid border in blue. Image pairs are shown without (In) and with (In-In-1) background correction [11]. Sandia.

Figure 6. Broadband chemiluminescence high-speed imaging, along with liquid Mie-scatter, Sandia. OH chemiluminescence imaging time-averaged from 1100-1800 s ASI. IFP.

100 90

Maximum Penetration [mm]

80 70 60 50 40 30 20 10 0 0

Sandia IFP

Reacting Penetration Vapor Non-reacting

Liquid
500

1000 1500 2000 2500 3000 3500 Time ASI [s] Figure 6. Comparison of liquid, vapor, and reacting flame penetration at IFP and Sandia. Liquid length threshold is 3% of maximum. Vapor penetration as in Naber and Siebers, 1996.

penetration prior to relaxation to a steady value near 10 mm while this is not shown in the Sandia data. Differences in nozzle tip temperature caused by preburn gas heating may be one explanation that will be explored in future collaboration. Conclusions The Engine Combustion Network has established a standardized set of test conditions (Spray A) for future experimental research as a pathway for CFD model improvement and validation. Initial results obtained at two different facilities, IFP and Sandia National Laboratories, show reasonable similarity despite the challenges of providing matched boundary conditions at these unique facilities. These encouraging findings suggest that parallel research, and leveraging experimental work at different facilities, can accelerate the development of a high-quality dataset and improved understanding of spray combustion.
1.

References

2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

Vishwanathan, G, and Reitz, R. Numerical Predictions of Diesel Flame Lift-off Length and Soot Distributions Under Low Temperature Combustion Conditions, SAE 2008-01-1331. Campbell J, Hardy G, and Gosman AD. Analysis of Premix Flame and Lift-Off in Diesel Spray Combustion using MultiDimensional CFD, SAE 2008-01-0968. Karrholm FP, Tao F, Nordin N. Three-Dimensional Simulation of Diesel Spray Ignition and Flame Lift-Off Using OpenFOAM and KIVA-3V CFD Codes, SAE 2008-01-0961. DErrico G, Ettorre D, Lucchini T, Simplified and Detailed Chemistry Modeling of Constant-Volume Diesel Combustion Experiments, SAE 2008-01-0954. D'Errico, G., Lucchini, T., Montenegro, G., Onarati, A., and Etorre, D., "Fundamental and applied studies of detailed chemistry based models for diesel combustion," THIESEL 2008, Valencia, Spain. Lucchini, T., d'Errico, G., Etorre, D., and Ferrari, G., "Numerical Investigation of Non-Reacting and Reacting Diesel Sprays in Constant-Volume Vessels," SAE Paper 2009-01-1971. Vishwanathan, G. and Reitz, R., "Modeling soot formation using reduced polycyclic aromatic hydrocarbon chemistry in nheptane lifted flames with application to low temperature combustion," Journal of Engineering for Gas Turbines and Power, pp. 032801-2009. Abraham, J. and Pickett, L.M., "Computed and Measured Fuel Vapor Distribution in a Diesel Spray," Atomization and Sprays, in press, 2010 Turbulent Nonpremixed Flame Workshop. http://www.ca.sandia.gov/TNF/abstract.html . Engine Combustion Network data archive. http://www.ca.sandia.gov/ECN/ Pickett, L.M., Kook, S., and Williams, T.C., "Visualization of diesel spray penetration, cool-flame, ignition, hightemperature combustion, and soot formation using high-speed imaging," SAE Paper 2009-01-0658. Miles, P.C. and Gouldin, F.C., "Determination of the time constant of fine-wire thermocouples for compensated temperature measurements in premixed turbulent flames," Combust.Sci.Tech. 83, pp. 1-19, 1992. Siebers, D.L., "Liquid-phase fuel penetration in diesel sprays," SAE Transactions 107, pp. 1205-1227, 1998. Siebers, D.L. and Higgins, B.S., "Flame lift-off on direct-injection diesel sprays under quiescent conditions," SAE Transactions 110, pp. 400-421, 2001.