Application of an Imaging-based Diagnostic Technique to Quantify the Fuel Spray Variations in a Direct-Injection Spark-Ignition Engine

Transcription

1 SAE TECHNICAL PAPER SERIES Application of an Imaging-based Diagnostic Technique to Quantify the Fuel Spray Variations in a Direct-Injection Spark-Ignition Engine David L. S. Hung, David M. Chmiel and Lee E. Markle Technical Center Rochester, Delphi Energy & Chassis Systems, Delphi Corporation Reprinted From: Direct Injection SI Engine Technology 2003 (SP-1746) 2003 SAE World Congress Detroit, Michigan March 3-6, Commonwealth Drive, Warrendale, PA U.S.A. Tel: (724) Fax: (724) Web:

2 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE. For permission and licensing requests contact: SAE Permissions 400 Commonwealth Drive Warrendale, PA USA permissions@sae.org Fax: Tel: For multiple print copies contact: SAE Customer Service Tel: (inside USA and Canada) Tel: (outside USA) Fax: CustomerService@sae.org ISSN Copyright 2003 SAE International Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in SAE Transactions. Persons wishing to submit papers to be considered for presentation or publication by SAE should send the manuscript or a 300 word abstract of a proposed manuscript to: Secretary, Engineering Meetings Board, SAE. Printed in USA

3 Application of an Imaging-based Diagnostic Technique to Quantify the Fuel Spray Variations in a Direct-Injection Spark-Ignition Engine David L.S. Hung, David M. Chmiel and Lee E. Markle Technical Center Rochester, Delphi Energy & Chassis Systems, Delphi Corporation Copyright 2003 SAE International ABSTRACT This paper presents an imaging-based diagnostic technique to quantify the pulse-to-pulse variability of macroscopic fuel spray characteristics for Direct-Injection Spark-Ignition (DISI) engine applications. The analysis approach is based on the construction of a spray ensemble image, reflecting the regions of probability of liquid presence for a set of images. Overlaying of an individual spray boundary on the probability-based ensemble image can further enhance the twodimensional visualization of the pulse-to-pulse spray variations. Spray structures at three experimental conditions were examined: room ambient, and early and late injection conditions inside an optical engine. While the spray structure was observed to be considerably different for the three conditions, the magnitudes of variation of global spray shape and spray tip penetration distance were found to be of similar order. This methodology is capable of capturing the magnitude of pulse-to-pulse variability in penetration and spray geometry and displaying the variations in a twodimensional representation. It is a useful diagnostic tool for fuel injector spray evaluation and can be used in support of the fuel injector development and application process for DISI engines. INTRODUCTION In the pursuit of creating environment-friendly automobile engine systems with reduced emissions and improved fuel economy, fuel injector manufacturers are continuously utilizing advanced spray atomization technologies and testing diagnostics to develop better and more robust injectors for the enhancement of fuel mixture preparation and optimization of combustion performance. Fuel injectors are key elements in an engine system. In DISI engines, gasoline liquid fuel is injected directly into the engine cylinder; hence the quality of fuel spray injection is critical. One significant challenge associated with the fuel mixture preparation is for the injectors to produce repeatable spray structure in a highly cyclic and hostile combustion environment. Zhao et al. [1] reviewed the general requirements for gasoline DI fuel injectors, which include accurate fuel metering, minimal spray skew, minimum pulse-to-pulse variation, among others. Misfires, poor combustion, and excessive emission products may be linked to the pulse-to-pulse variation of the liquid gasoline entering the cylinder, wetting the spark plug and cylinder surfaces, and building up liquid films on the piston top. Spray variations could adversely influence the fuel mixture preparation and result in combustion instability. A wide variety of analyses for the different aspects of fuel mixture fluctuations leading to combustion variations have been well documented [2-4]. In particular, several researchers have also applied laser diagnostics to understand the cyclic variability of pulsed injection sprays. Wright and Drallmeier [5] used the twowavelength infrared extinction technique to evaluate the cycle-to-cycle variation of the pulsed spray vapor fields. In addition, Nagayama et al. [6] applied the Laser Doppler Velocimetry technique to investigate the variation of the injection velocity from a pulsed gasoline injector. Over the past few years, extensive improvements have been made in spray diagnostics, allowing more detailed spray measurements to be obtained for testing validation or used as input conditions for CFD modeling and simulation. With the recent advances in image diagnostics, advanced imaging algorithms have also been developed to precisely determine the key parameters such as the liquid phase fuel penetration in diesel sprays [7], and the flame front development in an optically accessible two-stroke research engine [8]. Characterization of macroscopic fuel spray variations can also be achieved through optical imaging techniques and image post-processing. Combining various optical techniques and image processing, Hentschel et al. [9] examined the DI fuel sprays inside a pressure chamber and in an optical engine. In general, they showed that

4 cycle-to-cycle variations are more pronounced in DI gasoline engine when compared to pressure chamber experiments. Even though spray visualization techniques for DISI fuel injection development have been well studied and documented, specific understanding of global spray development pulse-to-pulse variation is limited. Instead, researchers have worked to mitigate the effects of spray variations by acquiring a set of spray images to constitute a statistically valid sample, and creating ensemble average images. Some, though, have created spray deviation images for the determination of mean, maximum and minimum limits, and standard deviations of the spray characteristics such as the penetration and spray angle. The understanding of pulse-to-pulse variation requires more information than the ensemble average or deviation. Moreover, the literature reveals no work linking the repeatability of spray structure at the atmospheric conditions (bench testing) to the repeatability of spray structures in DISI engines. Therefore, the objective of this paper is to demonstrate a fully automatic imaging-based diagnostic that has been specifically designed for quantifying the pulse-to-pulse variability of the macroscopic fuel spray characteristics for DISI engine applications. These characteristics include the spray structure, penetration, and spray angle. Spray visualization is performed using a customdesigned digital imaging system. The pulse-to-pulse variations of atmospheric condition spray structures are evaluated to demonstrate the principles of this technique. Then, variations of in-cylinder spray structure are evaluated to demonstrate a practical application of this technique. Ideally, a direct link would be established to correlate the variations in direct injection fuel sprays to the variations in cylinder combustion. However, to link these two phenomena is beyond the scope of this study. Possible future work could tie this type of fuel variation analysis in a given in-cylinder location at a specific engine crank angle to the coefficient of variation (COV) of indicated mean effective pressure (IMEP), or other combustion metric. ANALYSIS PROCEDURE FOR PULSE-TO- PULSE VARIATIONS OF SPRAY STRCUTURES Spray imaging via Mie scattering techniques can be conducted to visualize the macroscopic spray characteristics. The standard spray image analysis consists of a series of fully automated steps to extract the macroscopic spray characteristics such as the penetration tip distance and the spray angle. Details of such steps have been described by Hung and Markle [10] and are not repeated here. An additional imaging methodology for evaluating the macroscopic pulse-to-pulse spray variations is based on the technique of constructing a Presence Probability Image (PPI) for a set of images recorded at a fixed time after the start of injection (SOI), as described in details by Grimaldi et al. [11]. Grimaldi et al. had applied this technique successfully to determine the average location of the liquid-phase of the spray and to examine the spray penetration and global shape of the jet. Smallwood and Deschamps [12] had also developed a similar probability approach to evaluate the ensemble presence of flame front in a SI Engine. Hung [13] has extended the basic principle of this technique into an image diagnostic tool that can be used to evaluate the magnitude of macroscopic spray variation. The key advantage of this PPI technique is that the pulse-to-pulse variation of the spray structure can be clearly visualized in a twodimensional perspective. In the following section, the methodology of applying this image processing technique to quantify the spray variation is discussed. IMAGE ANALYSIS ALGORITHM FOR THE DETERMINATION OF LIQUID PRESENCE First, the region of the liquid presence in each image is identified using a user-defined thresholding technique. The image is then binarized based on the threshold selection, as illustrated in Figure 1. The liquid spray region is displayed as white pixels with a grayscale value of 1, and the remaining non-liquid area as black pixels with a grayscale value of 0. Second, the grayscale value of this binarized image at every pixel location is first multiplied by the maximum luminance intensity of the image format (for example, the maximum luminance value for an 8-bit image is 255). These two steps (binarization and grayscale multiplication) are then repeated for each image. Finally, to form the PPI, the grayscale value for every pixel location is then added up pixel-by-pixel for all images, and then divided by the total number of images in the whole set. At the end, for an 8- bit grayscale image, the pixels with a grayscale value of 255 on the PPI correspond to locations of 100% probability of liquid presence, and the pixels with a value of 0 represent no liquid presence and a zero probability. The intensity of gray pixels corresponds to the liquid presence probability at their locations. Figure 2 demonstrates an illustrative example of how a PPI can be formed based on two individual binarized spray images. The overlapped region corresponds to the

5 enclosed area of the liquid that is present in both images. Therefore, it has a probability of 100% liquid presence (two occurrences out of two images). The rest of the spray region that is not overlapped now has a probability of 50% (one out of two). Anywhere outside the liquid region has a 0% probability. images. The enhancement of the visualization of liquid presence is achieved by adding a pseudo-color or grayscale scheme that represents a specific range of probability. The extent of the color change across the spray boundaries indicates the magnitude of the spray variation. Spray Region Spray region : Presence Probability = 100 % Everywhere outside: Probability = 0 % Mie-Scattered Spray Image #1 As expected, the pulse-to-pulse variation can be found mostly around the boundary and particularly near the tip of the spray, due to the influence of the ambient air interaction. It is evident from Figure 3 that there is almost no color variation along the initial periphery of the spray where the spray angle may be determined. Hence, the spray angle variation is minimal. Further away from the injector tip, the PPI shows a gradual increase of color variation near the tip and across the edges. Most of the variation is found near the center of the spray tip. More pronounced variation can also be seen near the spray recirculation zones. The 100% probability boundary represents the core region where the liquid is always present. Similarly, the outermost periphery indicates where the liquid may be least likely to exist beyond this border. The probability presence images constructed and presented in this paper are based on a total of thirty images for each measurement condition. In general, any practical number of images may be used for the similar analysis, as long as there is a statistically reasonable level of confidence. 100% 50% 0% Probability Scale Binarized Spray Image #1 Figure 1. Determination of the liquid region Total of 30 images Figure 3. A presence probability image based on thirty binarized spray images Figure 2. An illustrative example of constructing a presence probability image Figure 3 depicts a typical presence probability image, constructed with a total of thirty individual binarized spray

6 100% 50% 0% Probability Scale Image #1 Image #2 Image #3 Figure 4. An illustration of three different spray boundaries overlaid on the presence probability image In essence, the PPI is an ensemble image of liquid presence, consisting of a set of binarized spray images. It provides a new way to examine the spray variation in terms of a probability defined for the presence of the liquid region. However, since the image is binarized, there is no account for the different amount of liquid present within a spray. It is simply a measure of whether liquid is present or not at that location. In addition, the PPI combines the variations of the spray shape from all the images considered, and therefore it does not necessarily resemble the shape or the appearance of the individual sprays. Once the PPI is formed, it is not possible to follow what the shape of a particular spray might be in relation to the overall probability among the total number of images. To resolve this issue, an individual spray outline can be superimposed onto the PPI. This reveals how a particular spray boundary and maximum spray tip location may be identified on the overall ensemble probability map. Figure 4 shows three images with different spray boundaries overlaid on the ensemble probability image. It is easily seen that the maximum variation occurs along the leading edge of the spray. This method provides a two-dimensional qualitative visualization of the spray variations and it allows the intermittent changes of both spray tip penetration and spray angle to be visualized simultaneously. EXPERIMENTS, RESULTS AND DISCUSSIONS SPRAY VISUALIZATION AT ATMOSPHERIC CONDITION Fuel Injector and Fuel System Gasoline sprays into atmospheric condition (room pressure and temperature) were generated from a prototype inwardly-opening Direct Injection-Gasoline (DI- G) injector with an experimental director plate. Isooctane was used for fuel at a pressure of 10 MPa and temperature of 23 C. A compressed air driven pump was used to generate the fuel pressure, as shown in Figure 5. A hydraulic noise suppressor and accumulator were used to dampen out the pressure pulsation from the DI-G injector so that fuel pressure and flow could be monitored. The DI-G injector was operated at a pulse width of 1.5 ms, which corresponds to a fuel delivery of 15 mg/pulse. Figure 5. Fuel system set-up and components

7 Image Acquisition Equipment A Mie scattering technique was used to visualize the spray structure. Figure 6 depicts the experimental setup. The fuel injector was mounted in a simple atmospheric test rig. The Nd:YAG laser light is diffused through a 60 full width half maximum (FWHM) light shaping diffuser. The laser emits light at 30 mj per pulse at 532 nm and has a 10 ns pulse duration. The light was used to side-illuminate the entire spray. A black velvet cloth was placed behind the injector to reduce the light reflection from the background surface. The images, covering a view of about 40 mm from the injector tip downstream, were captured with a 90 mm lens on a 12- bit grayscale CCD digital camera with a spatial resolution of 1280 by 1024 pixels. The camera and laser were sequentially synchronized relative to the start of injection (SOI) logic signal. To trigger the measurement, the injector was pulsed with a period of 160 ms, which corresponds to the engine running speed of 750 RPM. Each spray image was converted to an 8-bit format and was post-processed with Optimas image analysis software. In addition, a background image was recorded and it was subtracted from each spray image. This background correction step removed any extraneous objects and the uneven light spots present in the background. For the results presented here, the threshold value used to define the liquid presence region was selected to be 5 on a 256-level grayscale (or 2% of the available dynamic level). This threshold value was selected to reduce the background noise and to enhance the visualization of the spray structure. To ensure that the atmospheric condition testing was performed in a quiescent environment, a pitot tube was used to estimate the velocity of the ambient room air. It was held in various orientations at the location of the fuel injector set-up and no detectable stagnation pressure or velocity (less than 0.01 inches of water or 2 m/s) was observed. Atmospheric Condition Results and Discussion Figure 7 shows an instantaneous Mie-scattered spray image at 0.9 ms after SOI. At this early stage of injection, the spray is gradually forming into a conical structure. While the spray boundary near the injector tip is somewhat developed, the boundary near the spray tip is relatively blurred. The spray also exhibits a slight skew on the left side. Figure 8 depicts the variations of both the axial spray tip penetration distance and initial spray angle for all thirty images. For example, the range of the tip penetration distance was between 18.6 mm and 24.4 mm. The mean distance was measured to be 20.9 mm with a standard deviation of 1.3 mm. Similarly, the spray angle ranges between 32 to 47 with a mean value of about 39. The measurement uncertainties for the penetration and spray angle, based upon the resolution of the images, were estimated to be about 0.5 mm and 1.0 degree, respectively. Figure 7. An instantaneous spray image at atmospheric condition (0.9 ms after SOI) Figure 6. Imaging test set-up at atmospheric condition For atmospheric spray imaging, thirty images at 0.9 ms after SOI were recorded for the PPI analysis. This time was chosen because of the configuration of the optical access to the combustion chamber and bore. Though images taken at 0.9 ms after the SOI represent an interim point in the spray development, these images can provide an indication of the spray trajectory that may lead to wall impingement or piston wetting. Spray Tip Penetration (mm) Penetration 10 Penetration Ave 10 Spray Angle Spray Angle Ave Image Number Figure 8. Variations of spray tip penetration distance and spray angle at atmospheric condition Spray Angle (degrees)

8 Figure 9 displays the PPI image composed of thirty binarized images corresponding to this condition. It is evident that near the tip of the spray, there is more variation on the left side than on the right side, which corresponds to where the skew appears. The variation along the spray edge near the injector tip is relatively less than the spray tip. Axial Spray Tip Penetration (mm) Atmospheric condition Probability (%) Figure 10. Ensemble spray penetration variation as a function of the liquid presence probability SPRAY VISUALIZATION IN AN OPTICAL ENGINE Figure 9. Ensemble spray PPI at atmospheric condition (0.9ms after SOI) In addition to the visualization, it is also possible to quantify the pulse-to-pulse variability of other important spray parameters such as penetration and spray angle can be evaluated. For example, the variation of the spray penetration can be extracted as a function of the liquid presence probability. The area with a 100% probability can be considered as the core region of the spray structure where liquid fuel is present in all images. The minimum penetration reported would occur at the 100% probability and the maximum value would be at the minimum probability. The difference of penetration between the 100% and the minimum probability liquid presence is about 7.5 mm. The plot of liquid spray characteristics (i.e., axial penetration) versus probability, as shown in Figure 10, illustrates the potential errors that may be reported if the probability of presence is not determined. Without taking many images, it is impossible to know the possible variation of the reported penetration, spray angle, or other image-based spray parameters. It is very important to be able to visualize the DI-G injector fuel sprays in the combustion chamber of a motoring engine. In-cylinder conditions like pressure, temperature, and airflow can be monitored to determine their effects on spray penetration (axially and radially), spray angle, and atomization. Visualization can also show how injector sprays interact with the piston, cylinder walls, intake valves and spark plug. Injection parameters such as fuel pressure, pulsewidth, and injection strategies can be varied to determine their effects. These interactions and effects in a motoring engine cannot be readily estimated by spraying into atmospheric conditions. Optical Engine Background The optical engine used for this study was a four-stroke single cylinder Ricardo Hydra base engine with a bore and stroke of 86 mm. The Delphi designed cylinder head has a pent roof, one exhaust and two intake valves and centrally mounts the DI-G fuel injector, as illustrated in Figure 11. The spark plug is mounted in close proximity to the injector tip. The engine s parameters and experimental conditions are summarized in Table 1. The engine was made to be as similar as possible to a firing single cylinder engine with a few exceptions to allow for optical access. Visualization of the tumble plane is gained through a 20 mm quartz cylindrical window, placed just below the cylinder head and through two pent windows that are installed in the front and back of the cylinder head, as shown in Figure 12. This provides an unobstructed view (except where blocked by a thin copper gasket) of the DI-G injection event from the tip of the injector to the interaction with the piston and cylinder walls (approximately 35 mm).

9 Figure 11. DI-G injector/spark plug/valve placement in cylinder head Engine Speed Bore & Stroke Piston Type 750 RPM 86 mm Flat Cylinder Head Pent roof Intake Valve Opens (IVO) 26 BTDC Intake Valve Closes (IVC) 240 ATDC Exhaust Valve Opens (EVO) 240 BTDC Exhaust Valve Closes (EVC) 26 ATDC Geometric Compression Ratio 10.1:1 Effective Compression Ratio 7.2:1 Oil Temperature 72 C Coolant Temperature 78 C Table 1. Engine parameters and experimental condition Figure 12. View of piston through pent/cylinder windows (top), optical engine (bottom), Swirl plane optical access is accomplished by using an extension piston with interchangeable, windowed, piston tops that sits on top of a lower, lubricated piston, as shown in Figure 13. The interchangeable piston tops use two non-lubricated Teflon impregnated rings and one rider-piston ring that are mounted low on the top piston to seal to the cylinder wall. By mounting the rings lower on the top piston, friction and scuffing of the engine s cylinder window can be avoided. A stationary mirror is installed between the uprights of the extension piston to provide visual access in the swirl plane. A flat top piston top with a 60 mm diameter quartz window was used for all experiments. Figure 13. Flat top piston on extension piston with a quartz insert Image Acquisition Set-up The DI-G injector that was used for the atmospheric condition experiments was installed into the optical engine cylinder head. The fuel system and imaging equipment were the same as used for the atmospheric condition testing. To image the fuel spray inside the optical engine, the setup is very similar to the imaging technique for the atmospheric condition images, as shown in Figure 14.

10 The camera is placed in front of the engine and the Nd:YAG laser on the side of the engine. The laser light is diffused through a 60 FWHM light shaping diffuser off the mirror and up through the flat piston optics. For swirl plane imaging, the camera and laser switch places so that the camera takes images up through the flat piston optics and the laser light is diffused through the front of the engine s cylinder and pent windows. The optical engine test set-up is capable of visualizing both the tumble plane and swirl plane images. However, for this demonstration of the analysis technique, only the tumble plane images are used. monitored continuously with a combustion analysis system (CAS) to determine how the engine operates. Note that 720 marks the top dead center (TDC) of the compression stroke. As in the atmospheric condition testing, the fuel injector was operated at a pulsewidth of 1.5 ms, and thirty images at 0.9 ms after SOI were recorded for the PPI analysis. Late vs. Early Injection Fuel Injection (CA) MAP (kpa) Cylinder PP (kpa) SOIcp (kpa) EOIcp (kpa) Table 2. Parameters for the engine experiments Optical Engine Results and Discussion Figure 14. Imaging test set-up for the optical engine To trigger the imaging system, the optical engine uses a crank angle encoder that provides 3600 pulses per revolution (PPR) and 1PPR of the crank shaft. This encoder and a 1PPR camshaft encoder allow an engine setpoint computer to send out triggers at any crank angle/duration to the engine s fuel injector and spark plug, and to the imaging system. Two different experiments were run in the optical engine. These were late and early injections. These conditions were chosen because they represent the extremes in the engine for when fuel injection can occur. Most GDI engines must be able to provide both late injection (stratified charge) and early injection (homogeneous) conditions for when the engine calls for part and full loads, respectively. The parameters that changed between the experiments were the fuel injection crank angle (CA) and manifold absolute pressure (MAP). Table 2 summarizes these two conditions and also gives average cylinder pressure information taken during the experiments. Important parameters such as the start of injection cylinder pressure (SOIcp), end of injection cylinder pressure (EOIcp), and cylinder peak pressure (PP). Cylinder pressure is measured using a pressure transducer that is mounted in the cylinder head near the back of the pent roof. The signal from the transducer is Figure 15 displays the instantaneous spray structure recorded at 0.9 ms SOI, for the three conditions: room ambient, and early and late injection conditions inside an optical engine. For early injection during the intake stroke (Figure 15b), the intake valve appears on the left side and the spark plug is on the right. The dark line in the middle indicates the position where the gasket interface between the pent roof and cylinder window is located. For late injection (Figure 15c), the spark plug is still visible, the intake valve has been closed and the top surface of the piston is located 6 mm from the bottom of the image. Additional images taken of the spray in the swirl plane show that both intake valves and the spark plug do not interfere with the spray structure. As illustrated in Figure 15, the spray structure and penetration largely depend on the time of the injection and the resulting pressure condition inside the cylinder. In the early injection mode, the cylinder pressure is below 100 kpa absolute and the air density is below atmospheric condition, resulting in a more diffused spray structure and a longer penetration. In the late injection mode, the increased cylinder pressure causes the air density to increase, thus forcing the spray to become more compact with a shorter penetration. During the intake stroke (early injection condition), the fuel temperature was initially set at 20 C, and the maximum cylinder temperature was about 70 C. It is also noted that the spray structure in this condition appears with soft edges. This is likely due to the less pronounced flux gradient around the spray periphery as a result of air mixing during the intake stroke. In addition, since the depth of field was optimized for the compact spray structure at higher cylinder pressure condition (late injection), this limited depth of field might have led to a slightly blurred spray edge resulting in a more diffused spray structure observed during the early injection condition.

11 40.0 Axial Spray Tip Penetration (mm) Early Injection Late Injection Image Number a. Room ambient condition Figure 16. Axial spray tip penetration distance comparison: early vs. late injection Figure 16 shows the penetration data for the early and late injection cases. The variation is similar for these two conditions. The average penetration difference between early and late injection is about 11 mm. The imagebased macroscopic spray characteristics illustrate that increasing ambient pressures decrease penetration. The images taken for the early injection case exhibit more penetration than the atmospheric condition case as shown in Figure 8. The atmospheric pressure spray images exhibit more penetration than the observed late injection images. b. Early injection in an optical engine c. Late injection in an optical engine Figure 15. Instantaneous spray images at 3 different experimental conditions, recorded at 0.9 ms after SOI Figures 17 and 18 depict the probability ensemble images for both the early and late injection conditions. Again, they exhibit a similar type of variation as the atmospheric condition displayed in Figure 9. Even though the spray structure was different for these two conditions, the magnitudes of variation of global spray shape were found to be similar. Again, the color variation along the spray edges illustrates how much spray variation is exhibited. More variation is found along the tip of the spray. Figure 19 shows the comparison of penetration variations for the early and late injection conditions, based on the probability of the liquid presence. Supposedly, for applications employing spray-guided combustion, it is necessary to specify the requirements of spray structure, penetration and spray angle at different time intervals after the start of injection. These specifications are to ensure that the liquid portion of the fuel injected does not impact any surfaces in the combustion chamber, piston, or cylinder walls. For wallguided combustion, adherence to these specifications work to ensure the adequate delivery of the fuel as dictated by the designed combustion process. Therefore, it is very important to include the extent of pulse-to-pulse variability of spray structures into such design specifications. For those circumstances, the outermost border in the PPI may be used as the

12 maximum permissible boundary (lowest percentage of probability) of the desired spray structure. The corresponding spray structure, penetration and spray angle can be used as input values for combustion chamber design or CFD analysis. Figure 17. Ensemble spray PPI for early injection condition in optical engine Figure 18. Ensemble spray PPI for late injection condition in optical engine Axial Spray Tip Penetration (mm) Probability (%) Early Injection Late Injection Figure 19. Ensemble spray penetration as a function of the liquid presence probability (early vs. late Injection) The cylinder head used for this work was designed for spray-guided mixture preparation studies, so it generates little tumble or swirl. Steady state flow bench testing was employed in the development of this "quiescent" head. In addition, in-cylinder Particle Image Velocimetry (PIV) measurements were performed to determine the magnitude of the air motion that may influence the spray shape. The air velocity measured in the plane containing the injector tip and cylinder diameter, orthogonal to the crankshaft axis. Existing PIV results for crank angles near the positions of early and late injection were evaluated. At the crank angle of 450 during the intake stroke, the maximum velocity observed at the measured plane was about 23 m/s, which was less than the measured spray penetration velocity of about 65 m/s. For the late injection case, earlier flow structures have become less pronounced and the peak velocity magnitude at the crank angle of 630 has decreased to 7.5 m/s, again, less than the spray velocity that was measured to be about 35 m/s at this time. Swirl plane PIV measurements for these conditions showed lower air velocities than the tumble measurements. The pressure differential between the liquid and the gas determines the initial bulk liquid velocity. The bulk liquid encounters the ambient velocity (zero for a quiescent environment) and ambient density. The bulk liquid and subsequent droplets are acted on by the aerodynamic forces that tend to overcome the liquid surface tension to atomize the fuel. For the conditions examined here (early and late cylinder injection), the fuel pressure differentials are small, and the differences in the bulk liquid velocities are relatively minor. However, the ambient density differences are large. So, even though the bulk fuel velocity is decreasing with increased ambient pressure, the higher ambient density increases the aerodynamic breakup. Discussion Sprays from automotive fuel injectors, including gasoline direct injection injectors, are difficult to describe. They are pulsed sprays, so a description of their geometry must start with an indication of the time during the fuel injection event. The apparent spray characteristics are largely dependent on the measurement technique used to evaluate them. Whether intrusive or in-situ, or based on diffraction, interferometry, or imaging, the reported spray is a function of how it is measured. While Mie-scattering and shadow imaging techniques provide useful qualitative and even some quantitative descriptions of spray and mixture formation processes, they do not provide complete information about the spray. Also, Mie-scatter images can be particularly deceiving as the small particles present in gasoline direct injection sprays have diameters approaching the wavelength of the light used to observe them and

13 therefore, scatter light disproportionately to their diameter or mass. Determination of the location and concentration of liquid and vapor mass requires the use of additional diagnostics such as the laser induced fluorescence techniques. Quantification of spray geometry from images requires the determination of the spray boundaries, which are often ambiguous. Different types of spray with different spatial distributions will not scatter light in the same manner. Distinguishing the spray region from the background by setting a threshold value is a very important step towards the quantification of spray variations. A fixed threshold value was used in this study to illustrate the principle of this analysis algorithm. In general, the criterion of selecting a threshold intensity value, either based on a fixed value or a user-defined algorithm, may be optimized for a specific type of spray structure. The investigation of how to threshold an image is beyond the scope of the present study. However, various algorithms based on the statistical approaches of image histogram and brightness values of pixel may be available to improve the determination of a threshold scheme for automotive spray structures [14]. Finally, the PPI approach seems to be universally applicable to other types of fuel sprays generated from the port fuel injectors or diesel injectors. Their pulse-topulse variability of spray structures may also be characterized in a similar methodology. CONCLUSION An image analysis technique has been demonstrated to evaluate the macroscopic spray variations of a DI-G injector, both at an atmospheric condition and inside an optical engine. In addition to the standard spray characteristics, such as the spray tip penetration and spray angle, the two-dimensional global spray structure variation was also revealed by using the probability approach. This methodology is useful for evaluating the magnitude of variation. Overlaying an individual spray boundary on the PPI enhances the visualization of the spray variation by providing a two-dimensional perspective of the spray structure. It allows the variation of multiple spray characteristics to be examined simultaneously. This technique creates a useful tool for injector design and application development. ACKNOWLEDGMENTS The authors gratefully acknowledge the significant contributions and support from those who assisted in this work. Their help was invaluable in discussion, hardware preparation and testing. Particular thanks are extended to: F. Brado, S. Maczynski, P. VanBrocklin, and D. Varble. REFERENCES 1. Zhao, F-Q, Lai, M-C., and Harrington, D.L., A review of mixture preparation and combustion control strategies for spark-ignited direct-injection gasoline engines, SAE Technical Paper, No , Davis, L.I., Feldkamp, L.A., Hoard, J.W., Yuan, F., Connolly, F.T., Daw, C.S., and Green, J.B., Controlling combustion cyclic variations in leanfueled spark ignition engines, SAE Technical Paper, No , Helie, J., Duclos, J-L., Baritaud, T., Boinsot, T., and Trovue, A., Influence of mixture fluctuations on combustion in direct injection spark ignition engines simulations, SAE Technical Paper, No , Zhao, F., Kadota, T., and Takemoto, T., Temporal and cyclic fluctuation of fuel vapor measurement in a spark ignition engine, SAE Technical Paper, No , Wright, J.H., and Drallmeier, J.A., Cyclic variability of pulsed spray vapor fields, Exp. In Fluids, Vol. 25, pp , Nagayama, H., Shimazu, M., Ikeda, Y., and Nakajima, T., Cyclic variation measurements of fuel injection by LDV, Proc. 11 th Internal Combustion Engine Symposium Japan, pp , Siebers, D.L., Liquid-phase fuel penetration in diesel sprays, SAE Technical Paper, No , Schiebl, R., Dreizler, A., and Mass, U., Comparison of different ways for image post-processing: detection of flame fronts, SAE Technical Paper, No , Hentschel, A., Homburg, A., Ohmstede, G., Muller, T., and Grunefeld, G., Investigation of spray formation of DI gasoline hollow-cone injectors inside a pressure chamber and a glass ring engine by multiple optical techniques, SAE Technical Paper, No , Hung, D.L.S., and Markle, L.E., Macroscopic spray characterizations of DI gasoline injectors using automatic image analysis, Proc. 12 th ILASS- Americas, Indianapolis, IN., Grimaldi, C.N., Postrioti, L., Stan, C., and Troger, R., Analysis method for the spray characteristics of a GDI system with high pressure modulation, SAE Technical Paper, No , Smallwood, G., and Deschamps, B., Flame surface density measurements with PLIF in an SI engine, SAE Technical Paper, No , Hung, D.L.S., An imaging-based analysis tool for evaluating the macroscopic spray variations of a direct injection-gasoline injector, Proc. 15 th ILASS- Americas, Madison, WI., Russ, J.C., Thresholding images, J. Computer- Assisted Microscopy, Vol. 7, No. 3, pp , 1995.

14 CONTACT David L.S. Hung, Ph.D. Technical Center Rochester Delphi Energy & Chassis Systems Delphi Corporation P.O. Box 20366, Rochester, NY Phone: (585) Fax: (585) MAP Nd:YAG PIV PP PPI PPR RPM SOI SOIcp TDC TTL WOT Manifold absolute pressure Neodymium yttrium aluminum garnet Particle image velocimetry Peak pressure Presence probability image Pulse per revolution Revolution per minute Start of injection Start of injection cylinder pressure Top dead center Transistor-transistor logic Wide open throttle DEFINITIONS, ACRONYMS, ABBREVIATIONS Arbitrary ATDC After top dead center BTDC Before top dead center CA Crank angle CAS Combustion analysis system CCD Charged coupled device CFD Computational fluid dynamics COV Coefficient of variation DI Direct injection DI-G Direct injection-gasoline DISI Direct-injection spark-ignition EOIcp End of injection cylinder pressure EVC Exhaust valve closes EVO Exhaust valve opens FWHM Full width half maximum IMEP Indicated mean effective pressure IVC Intake valve closes IVO Intake valve opens Prefixes n nano, 10-9 µ micro, 10-6 m milli, 10-3 k kilo, 10 3 M Mega, 10 6 Quantities and Units g Gram s Second m Meter N Newton, 1 kg. m/s 2 Pa Pascal, 1 N/m 2 J Joule, 1 N. m Degree