ILASS Americas 27th Annual Conference on Liquid Atomization and Spray Systems, Raleigh, NC, May 2015 Characteristics of Spray from a GDI Fuel Injector Using TRF Gasoline Fuel Surrogate Libing Wang 1, William Roberts 2, and Tiegang Fang 1, * 1 Department of Mechanical and Aerospace Engineering North Carolina State University Raleigh, NC, 27695-7910 USA 2 Clean Combustion Research Center King Abdullah University of Science and Technology, Thuwal, Saudi Arabia Abstract Characterization of the spray angle, penetration, and droplet size distribution is important to analyze the spray and atomization quality of fuel injectors. In this paper, the spray structure development and atomization characterization of a gasoline direct injection (GDI) fuel injector was investigated. The experimental setup included a fuel injection system, a high-speed imaging system, and a droplet size measurement system. Spray images were taken by using a high speed camera for spray angle and penetration analysis. Sauter mean diameter, Dv(10), Dv(50), Dv(90), and particle size distribution were measured using a laser diffraction technique. The fuel injection event was synchronized with the instruments for high-speed imaging and droplet size measurements by using a pulse generator. The fuel used is a gasoline surrogate fuel, called toluene reference fuel (TRF). Fuel under 100 bar (10 MPa) pressure was injected into ambient atmospheric pressure (1 atm) and temperature (298 K). The injection process is very consistent and the spray angles are consistently near 96º during the relatively stable stage. The front edge of the spray remains relatively flat during the spray process at the early stage. The spray front penetration speed has a peak of 66 m/s, then decreases until the end of the injection and stays relatively stable. The Sauter mean diameter (SMD) shows a decrease with the increase of the horizontal distance from measuring center at the axis of the nozzle to the spray edge. The particle size with time and the time-averaged particle size are also analyzed and discussed. * Corresponding author: tfang2@ncsu.edu
Introduction Liquid fuels, with the advantages of high energy density and easy transportability, have been playing an irreplaceable role in industrial and commercial use. However, petroleum, the most important source for liquid fuels, is estimated to be depleted in the near future. Total world proved oil reserves reached 1687.9 billion barrels at the end of 2013, sufficient to meet 53.3 years of global production [1]. On the other hand, excessive consumption of fossil liquid fuels generates pollutant emissions and greenhouse gases. Due to the fact that the majority of greenhouse gas emission is related with combustion of fossil fuels [2], studying and optimizing the combustion process in internal combustion engines is becoming more and more important because optimizing the process can help reduce fuel consumption and pollutant emission. Liquid fuel atomization is critical to combustion and emissions in gasoline engines. Characterization of the spray angle, penetration, and droplet size distribution is important to analyze the spray and atomization quality of fuel injectors. Studies of the fuel spray behavior can be on either macroscopic or microscopic scale [3]. Macroscopic parameters, such as spray penetration and cone angle, can be measured by direct visualization methods. Microscopic parameters such as droplet velocity, droplet size and size distribution can be measured through laser diffraction particle analyzer systems [4]. In order to facilitate the study of properties of the fuel, reference fuel was introduced. Edgar [5] first proposed using n-heptane and iso-octane as reference fuels to measure the knocking characteristics of gasoline, resulting in the octane number. Since then, scientists have tried to use different combinations of fuels as reference fuel to represent complex liquid fuels. Sturgis [6] used di-isobutylene and benzene to represent the olefins and aromatics, respectively, for antiknock studies of gasoline-like hydrocarbon mixtures. Because of the development of engine study, the demand for better surrogate models to better represent fuels is significant. Conventionally, gasoline is represented using a binary mixture of the primary reference fuels (PRFs): n- heptane and iso-octane, which represent linear and branched alkanes, respectively [7]. Recently, adding an additional components such as toluene, a representative of aromatics, have been considered to make the mixture better match the reactivity of gasoline [8 9]. This resulting mixture is called three-component toluene reference fuels (TRFs). In addition, gasoline direct injection as an emerging technology is increasingly found in recent sparkignition engines. It has the advantages of lower fuel consumption and higher power output [10]. This work investigates the spray characterization of a gasoline direct injection (GDI) fuel injector by means of spray visualization and droplet size measurements. The fuel used is a gasoline surrogate fuel, which is a toluene reference fuel (TRF), composed of isooctane, n- heptane, and toluene. The fuel injection activation pulse duration is 3 ms and the fuel injection pressure is 100 bar (10 MPa). Spray structure and cone angle were analyzed and the droplet size distribution and characterization parameters of atomization including Sauter mean diameter (SMD), Dv10, Dv50 and Dv90 were obtained in this work. Experimental Setup The experimental setup mainly consists of the fuel injection system, high-speed imaging system, and droplet size measurement system, which will be introduced separately as follows. The fuel injection system includes a low pressure fuel supply system, a high pressure pump, a fuel rail and a fuel injector. The high pressure pump is driven by a 2 horse power (HP) electric motor. The rail pressure is controlled by a Labview program using a PID control scheme. This system can provide a pressure of up to 200 bar (20 MPa) in the common rail. A piezoelectric fuel injector is controlled by a Direct Injection injector driver from Drivven, a subsidiary of the National Instruments, Inc. Fuel is injected into atmospheric pressure (1 atm) and temperature (298 K). A Stanford pulse/delay generator (Model No.DG535) produces a pulse to control the fuel injection duration and trigger the driver of injection, the high-speed camera, or the Malvern SprayTec particle size analyzer simultaneously. Thus the fuel injection event can be synchronized with high-speed imaging and droplet size measuring. The rising edge of the triggering pulse is used for all the devices. The high speed imaging system mainly consists of a Phantom V4.3 camera from Vision Research Inc, a light source and an image acquisition computer. Fig.1 shows a schematic of the fuel injection system and imaging system. For the current investigation, the trigger pulse from the pulse generator was used for external trigger. A light source (PAR64 1000 W lamp) was used to volumetrically illuminate the spray. A 50 mm fixed focal length lens was used. The F# was set at 2.8 and the exposure time was 10 μs. The camera was set at 8000 fps with a resolution of 256 256. The corresponding physical image resolution is approximately 0.464 mm/pixel. The Mie-scattering signal from the droplets were recorded by the camera. Ten sets of high-speed videos were taken for each condition and saved for later process. For a typical run, 200 frames were stored to analyze the whole injection process. The droplet size measurement system measures the particle size distribution of the spray by using a com-
mercial SprayTec system with a laser diffraction technique. Laser-diffraction technique is a commonly used technique for droplet size and its probability distribution measurements, especially for the transient spray process. In this work, this technique is used for measuring the droplet size and its distribution at different locations. The SprayTec particle size analyzer was operated at a sample rate of 10 khz for transient spray analysis. The equipment has a lens of 300 mm, capable of measuring particle size diameter from 0.1 to 900 micrometer. The measurement system was triggered by the same pulse as for the fuel injector. Image and Droplet Size Data Processing Spray image processing program was developed by using Matlab software to read the original cine file frames to computer and calculate the spray angle and spray front penetration. The images were further enhanced by adjusting the intensity and contrast. A threshold was then used for spray edge detection. This method proved to be effective for detecting the visible edge. Next operation was to use an algorithm to find the left and right edges of the spray for each image. Then a linear fitting was applied to the spray edges and the slopes of the two fitted straight lines were used to calculate the spray angle at each time step. A similar edgefinding method was used to obtain the penetration of the spray front. The spray penetration length was defined as the averaged value of the spray front. The spray angle was calculated only during the fuel injection process with liquid fuel signal appearing at the injector orifice. The spray penetration was calculated for up to 4.8 ms, well beyond the end of fuel injection. The standard operation procedure setup for the SprayTec system is described as follows: 1) sampling rate is set at 10 KHz for rapid measurement; 2) the analyzer is triggered from an external source (the pulse generator); 3) the measuring duration is 20 ms; and 4) 10 runs were operated for each location. The measured locations were in the plane 15 mm below the injector orifice. Five locations were measured covering from the axis to 20 mm away from the axis with a step of 5 mm. By using a linear approximation, the refractive index for the TRF mixture is calculated at 1.4188 and the density is 736.25 gm/liter. During the laser diffraction droplet size measurements, the entire ranges of detectors (total 36) were used. Certain detectors were disabled when processing the results. This is because the TRF mixture is a volatile fuel, and fuel evaporation during the injection process can produce a beam-steering effect and as a consequence fake large particle sizes can be produced in the results. In order to avoid the beam-steering effects and remove the fake particles, certain inner detectors were disabled during the processing of the results. On the other hand, abnormally small diameter size between 0.1 and 1 micrometer was also observed in the measurements and these particles were removed by disabling the outer detectors. The Sauter mean diameter (SMD or D32), Dv10, Dv50 and Dv90 at each time step were processed and averaged over ten repeatable runs. Laser diffraction technique can provide the volume weighted particle size probability distribution and particle diameters. Dv10 means the particle diameter below which 10% of the sample volume exists, while Dv50 and Dv90 indicate the particle diameters below which 50% or 90% of the volume exists, respectively. The Spray- Tec software provides the concentration weighted average particle size based on the transient scattering data. One should notice that the laser diffraction technique is a line-of-sight technique and the results represent the droplet size feature across the entire laser beam path length. Results and Discussion In this section, the spray macroscopic features will be discussed first and followed by the atomization results. The ten sets of videos of the high-speed Miescattering technique were processed and used for the calculation of spray angle, penetration. To demonstrate the development of the spray, a typical set of images are shown in Fig. 3. In the images, besides the spray signal other added information includes a time stamp showing the time of the image after the activation of the injection pulse, a white thick bar showing the scale of the image, a thin white line showing the injector orifice, a red curve for the left edge of the spray, a blue curve for the right edge of the spray, and a green curve showing the spray penetration front. From the spray images, it is observed that the spray develops very smoothly at the beginning and then spread out later. The cone angle stays rather stable after the first image. There is little fluctuation at the spray front during the early fuel injection process. For this case, the spray angle was calculated up to the frame of 3.077 ms, because at 3.131 ms the fuel has stopped flowing from the orifice. The spray images show a few interesting features: 1). There is an increase of fuel liquid signal at frame of 3.256 ms, which occurs after the injection duration. This bit of fuel is believed to be from the fuel channel between the pintle and the nozzle body. During the shut-off process, the fuel in the channel is squeezed out when the pintle goes back to the closing position. 2). After 1.0 ms, the spray front starts moving downward unevenly at different locations and the non-uniform structure in the circumferential direction starts to appear in the spray images. 3). There exist some vortices on the edge of the spray after 1.5 ms, which might also cause the linear fit for spray angle to change slightly. Fig. 4 shows the spray angle development throughout the process. The first frame shows a smaller spray
angle, which is mostly due to the poor definition of the spray edges at a very short spray penetration. The last two frames also show large variation and poor definition of spray angles. During the most time of the fuel injection process from 0.327 to 2.952 ms the spray shows very consistent spray angles with little variation among different fuel injection events, while there is a slight rise at about 1.702 ms which should be caused by the vortices on the edge of the spray. The time averaged spray angle is about 96.16 degree during this period. The spray penetration length is shown in Fig. 5. At the beginning, the spray penetration increases very fast. After around 1.0 ms, the penetration slows down. The spray penetration velocity can be estimated based on the penetration length and the time step as shown in Fig. 6. It is found that spray penetrates at a relatively small velocity at the start (~45 m/sec) then the velocity increases rapidly to a peak of about 66 m/sec. Then the velocity begins to decrease until the end of the injection (3 ms) and stays relatively stable at a speed lower than 5m/sec after the end of fuel injection. Results of the droplet size parameters, including SMD (D32), Dv10, Dv50, and Dv90, are shown in Fig.7 for time dependent data. In each plot, data from different locations are compared. One thing to note is that for the location of 10 mm from axis, there are several points (0.9 ms, 1.2 ms to 1.6 ms) without data which is due to the dense spray blocking the laser beam. All the droplet sizes show similar trend with time. At the beginning, the droplet sizes vary by a certain amount. After 1.2 ms, the droplet sizes become relatively stable until the end of the injection (3 ms), except the droplet sizes at 20 mm which rise a little bit. At around 3.3 ms there is a big sudden drop for all the droplet sizes at different locations. During the late stage, D32, Dv50 and Dv90 increase a little bit and then decrease, while Dv10 remains increasing. On the other hand, the effects of locations can be observed in each set of droplet size. The droplet sizes at the axis, 5 mm and 10 mm from the axis have a close value, while droplet sizes become smaller when the location shifts from 10 mm to 20 mm away from axis during most of the injection time. In order to compare the overall atomization quality at different locations, the time dependent data were further averaged over time. The averaged result is obtained from the SprayTec software, called concentration weighted average diameter, which is calculated based on the laser diffraction results and considering all the particles during the transient process. The concentration weighted average D32, Dv10, Dv50, and Dv90 are shown in Fig.8. It is found from the plots that weighted average D32, Dv10, Dv50, and Dv90 have similar trends. In general, the averaged diameter first increases a little bit and then decreases with the increase of the distance from the axis, while the peak value appears at 5 mm for every averaged diameter. The extent of decrease is bigger for Dv50 and Dv90. Conclusion A series of experiments was conducted to investigate the characterization of sprays from a GDI fuel injector. The spray development was visualized by a high-speed Mie-scattering imaging technique and the transient droplet size was measured by a laser diffraction technique. The average spray angle is about 96.16 degree and remains very stable during the fuel injection process. Spray penetration variation is rather small at the beginning and then increases by a certain amount during the fuel injection process. Spray penetration speed has a peak of 66 m/s, then decreases until the end of the injection (3 ms) and stays relatively stable afterwards. Because the ambient temperature is at room temperature, the fuel vaporization is not as strong as those under elevated ambient temperature and the spray keeps penetrating till the momentum decays away. The droplet sizes become relatively stable from 1.2 ms after trigger to the end of the injection (3 ms), except the droplet sizes at 20 mm which rise slightly. At around 3.3 ms there is a big sudden drop for all the drop let sizes. During the late stage, D32, Dv50 and Dv90 increase slightly and then decrease, while Dv10 remains increasing by a small amount. The effects of locations can be observed in each set of droplet size. The droplet sizes at the axis, 5 mm and 10 mm from the axis have similar values, while droplet sizes become smaller when the location shifts from 10 mm to 20 mm (away from axis) during most of the injection duration. Acknowledgements This material is based upon work supported in part by the Saudi Aramco R&D Center through the Clean Combustion Research Center of the King Abdullah University of Science and Technology. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the funding agencies. References 1. BP Statistical Review of World Energy, June 2014. 2. Inventory of US greenhouse gas emissions and sinks: 1990 2009. US Environmental Protection Agency; 2011. 3. Chen, P.C., Wang, W.C., Roberts, W.L., Fang T.G., Fuel 103: 850 861 (2013). 4. He, C., Ge, Y., Tan, J., Han, X., International Journal of Energy Research 32:1329 1338(2008). 5. Edgar, G., Industrial & Engineering Chemistry Research. 19: 145 146 (1927).
6. Pahnke, A.J., Cohen, P.M., Sturgis, B.M., Industrial & Engineering Chemistry Research. 46: 1024 1029 (1954). 7. Niemeyer, K.E., Sung, C.J., Combustion and Flame 161: 2752 2764 (2014) 8. Knop, V., Pera, C., Duffour, F., Combustion and Flame 160: 2067 2082(2013). 9. Rapp, V.H., Cannella, W.J., Chen, J.Y., Dibble, R.W., Combustion Science and Technology. 185: 735 748(2013). 10. Überall, A., Otte, R., Peter, E., Krahl, J., Fuel, 147, : 203-207(2015)
Pulse generator Common rail Fuel tank High Speed Camera Injector Light Imagine Acquisition Figure 1.Schematic of the fuel injection system and imaging system. Pulse generator Common rail Fuel tank Injector Laser beam Data Acquisition Figure 2.Schematic of droplet size measurement system.
Figure 3. Development of fuel spray at different time steps after the trigger signal.
Figure 4. Spray angle development during the fuel injection process. Figure 5. Spray penetration length with time. Figure 6. Spray penetration velocity at different time steps.
(a) (b) (c) (d) Figure 7.Droplet size parameters (D32, Dv10, Dv50, and Dv90) as a function of time for different locations.
(a) (b) (c) Figure 8.Averaged droplet size parameters (D32, Dv10, Dv50, and Dv90) over time for different locations. (d)