J Vis (2011) 14:149–160 DOI 10.1007/s12650-011-0083-0 R E G UL A R P A P E R Mayank Mittal • David L. S. Hung Guoming Zhu • Harold J. Schock • Fuel spray visualization and its impingement analysis on in-cylinder surfaces in a direct-injection spark-ignition engine Received: 17 June 2010 / Accepted: 19 February 2011 / Published online: 23 April 2011 Ó The Visualization Society of Japan 2011 Abstract Experiments are performed to investigate the effects of fuel spray on in-cylinder mixture preparation and its impingement on cylinder walls and piston top inside a direct-injection spark-ignition engine with optical access to the cylinder. Novel image processing algorithms are developed to analyze the fuel impingement quantitatively on in-cylinder surfaces. The technique is useful to optimize the fuel pressure, injection timing and the number of injections to minimize the fuel impingement on in-cylinder surfaces. E85, which represents a blend of 85% ethanol and 15% gasoline (by volume) is used in this study. Two types of fuel injectors are used; (i) low-pressure production-intent injector with fuel pressure of 3 MPa, and (ii) high-pressure production injector with fuel pressures of 5 and 10 MPa. In addition, the effects of split injection are also presented by maintaining the same amount of fuel used in single injection. It is found that the split injection is an effective way to reduce the overall fuel impingement on in-cylinder surfaces while maintaining a reasonably good air–fuel mixture in the cylinder. Keywords Spray visualization  Fuel impingement  Direct-injection spark-ignition engine 1 Introduction Improvement in fuel efficiency and reduction in exhaust emissions are the main goals behind the new developments in internal combustion engines. The concept of the direct-injection spark-ignition (DISI) engine has the potential to achieve such goals. In this technology, fuel is directly injected into the engine cylinder, which offers great flexibility to control the fuel injection timing, its duration and the number of injections. Note that the fuel–air mixture preparation in the combustion chamber is one of the key factors that influence the in-cylinder combustion characteristics and hence the engine performance (Hung et al. 2007). Therefore, optimizing the fuel–air mixture homogeneity is an important parameter for the engine designers. In general, a homogeneous fuel–air mixture is achieved by injecting the fuel during the intake stroke. However, due to in-cylinder injection and higher injection pressures, the fuel impingement levels on M. Mittal (&)  G. Zhu  H. J. Schock Department of Mechanical Engineering, Michigan State University, East Lansing, MI, USA E-mail: mittalma@msu.edu Present Address: M. Mittal Caterpillar Inc., Mossville, IL, USA D. L. S. Hung Department of Mechanical Engineering, University of Michigan-Shanghai Jiao Tong University Joint Institute, Shanghai, China 150 M. Mittal et al. in-cylinder surfaces in DISI engines are typically higher than those in port-fuel injection (PFI) engines (Pereira et al. 2007). This result in an increase in the levels of un-burned hydrocarbons and smoke emissions, which reduces the potential fuel economy benefits associated with the direct-injection engines. Therefore, it is important to control the fuel injection timing precisely in order to minimize the fuel impingement on in-cylinder surfaces. Several studies have been reported on fuel spray pattern visualization and its influence on mixture formation inside the cylinder of direct-injection systems. Grimaldi et al. (2000) studied the spray characteristics inside a gasoline direct-injection system with high-pressure modulation. The authors used a laser sheet technique for spray visualization, where the radiation of Nd-YAG pulsed laser was scattered by the spray droplets lying on the lighted plane and collected by a CCD camera. The spray images were analyzed in terms of spray penetration and global shape. Hung et al. (2003) used a Mie scattering technique for spray pattern visualization inside a DISI engine. The authors used the presence probability image (PPI) technique (Grimaldi et al. 2000) to quantify the pulse-to-pulse variability of the macroscopic fuel spray characteristics. Kawajiri et al. (2002) investigated the spray behavior in a constant-volume cylindrical vessel, in which a swirling gas motion similar to that in a DISI engine cylinder was generated. The experimental results were also compared with the numerical simulation results by the authors. They found that the main part of the hollow cone spray was influenced strongly by the air motion, although the influence of the air motion on the central part was weak. Hung et al. (2007) used high-speed imaging to visualize the spray pattern in a single cylinder DISI engine. Three spray patterns, i.e., a narrow 40° spray angle, a 60° spray angle with 5° offset angle, and a wide 80° spray angle with 10° offset angle, were studied by the authors. They concluded that for a given cylinder head, piston configuration and intake port flow characteristics, injector spray pattern plays a dominating role in how the fuel–air mixture is formed. It was suggested that if an appropriate injector spray pattern is chosen, the in-cylinder fuel–air mixing can be enhanced by minimizing the fuel impingement on in-cylinder surfaces, thus producing a more homogeneous fuel–air mixture prior to ignition. Aleiferis et al. (2008) studied the spray development of gasoline, iso-octane, and ethanol in a spark-ignition engine. They found that the spray characteristics of fuels differ between hot and cold engine operation to a large extent. Few differences were noticed in the spray development among all fuels at an engine head temperature of 20°C. However, the spray cone angle was reduced considerably at an engine head temperature of 80°C. Gasoline and a blend of 85% ethanol and 15% iso-octane appeared to demonstrate a classical spray collapse at this engine head temperature of 80°C. Iso-octane was found to be the least sensitive to the elevation of engine head temperature in terms of spray collapse. Yoo et al. (1998) discussed the effects of injector cone angle, location and injection timing on wall impingement inside a DISI engine. The authors found that the extent of wall impingement varies significantly with the injector mounting position and the spray cone angle; however, its effects can be reduced by optimizing the injection timing. Park et al. (1999) studied the direct-injection gasoline spray-wall interaction inside a pressurized chamber at both cold and elevated temperature conditions. Two hollow cone highpressure swirl injectors with different cone angles (20° and 60°) were used. The authors found that the impingement intensity is more forceful and focused for the narrow cone injector compared to the wide-cone injector. More rapid impingement was also observed at elevated temperature conditions. Hennessey et al. (2001) studied the effect of piston temperature on fuel impingement in a DISI engine. The authors observed the fuel impingement on piston surface for all the surface temperatures investigated. The objective of this paper is to investigate the crank angle resolved fuel impingement analysis on cylinder walls and piston top inside a DISI engine. Novel image processing algorithms are developed for this purpose. Experiments are performed with different fuel injection pressures, injection timing and number of injections. In the following sections, the experimental setup is outlined, with procedural details of fuel impingement analysis on in-cylinder surfaces. Results of various experiments are then presented. Finally, concluding remarks are summarized from this work. 2 Experimental setup and procedure 2.1 Experimental setup The engine used in the present work is a four-valve engine with two intake and two exhaust valves. It is a 0.4 l single cylinder spark-ignition engine, although not fired (Mittal et al. 2010). A quartz cylinder is used to provide the optical access to the cylinder. Engine specifications are listed in Table 1 with the bore Fuel spray visualization and its impingement analysis 151 Table 1 Engine specifications Specifications Bore (mm) Stroke (mm) Rod length (mm) Compression ratio Intake valve timing Exhaust valve timing Measure 83 73.9 123.2 13.5:1 Opens 355 CAD Closes -122 CAD Opens 124 CAD Closes -332 CAD Fig. 1 Experimental rig diameter of 83 mm and the stroke length of 73.9 mm. It should be noted that in this paper 0° crank angle corresponds to the top dead center (TDC) of the compression, and therefore -180 and 180 crank angle degrees (CADs) correspond to the bottom dead center of the intake, i.e., 180° BTDC (before TDC) and power strokes, i.e., 180° ATDC (after TDC), respectively. Figure 1 shows the experimental rig with the high-speed camera and laser in place for spray visualization in the cylinder. The laser light is introduced into the cylinder through the optical cylinder. In order to be able to fit with multiple configurations of the directinjection injectors, the cylinder head was designed to accommodate either the low-pressure or high-pressure direct-injection fuel injectors. To study the ionization combustion feedback, either 8 or 14 mm spark plugs may be fitted to the single cylinder head. The head also accommodates a pressure transducer to record the in-cylinder pressure data. A view of the top of the combustion chamber geometry showing intake and exhaust valves, direct-injection fuel injector, spark plug and the pressure transducer is illustrated in Fig. 2. The custom-designed piston was used, which allowed the geometrical compression ratio (knock-limited) from a baseline of 9.75:1 to increase up to 13.5:1. The increased compression ratio was largely made possible by a re-design of the piston bowl with deeper bowl contour in the mid section and four cavities matching the valve profiles. 2.2 Fuel injection system When developing combustion systems for DISI engines, it is important to achieve optimal fuel–air mixture prior to ignition. It has been shown previously (Hung et al. 2007) that multi-hole fuel injectors offer flexible spray pattern which can be tailored for a specific cylinder configuration. As fuel is injected directly into the engine cylinder, spray may mix well with the intake air or impinge on the in-cylinder surfaces such as piston top and cylinder wall. The goodness of fuel air mixture formation depends on several key factors such as the injection duration, injection crank angle timing, and injector axis orientation. If the injectors can be designed to offer spray tailoring flexibility, engine designers may utilize the injectors to deliver the specific flow and spray requirements without major compromises and limitations 152 M. Mittal et al. Fig. 2 Cylinder head configuration of the combustion chamber in the optical engine Table 2 Test parameters for spray visualization Parameter Fuel Fuel injection pressure Engine speed Engine load Injection timing Number of injection Description E85 3 (LPDI injector), 5 and 10 (HPDI injector) MPa 1500 rpm Full-load (WOT) 240°, 210°, and 180° BTDC Single and dual when running the engine at its optimized configuration. For this engine, the side-mounting option for installing the fuel injector is not feasible due to the packaging constraints around the coolant channels. In addition, the tilted angles for the inlet and exhaust valves are quite large, at 21.5° and 24.1°, respectively. The dimensions and their physical constraints of the engine cylinder head only allows the injector to be placed on the top side at about 12° along the piston axis in the vicinity of the spark plug (i.e., a centrally mounted injector configuration). Two types of multi-hole fuel injectors are considered; (i) low-pressure direct-injection (LPDI) 9-hole fuel injector with the fuel pressure of 3 MPa (Hung et al. 2007) and (ii) highpressure direct-injection (HPDI) 7-hole fuel injector with the fuel pressures of 5 and 10 MPa. The advantage of the multi-hole injector is that the hole pattern, hole orientation, internal flow cavity, and number of holes of a nozzle can all be designed to control individual spray plumes and the overall spray pattern, providing enhanced droplet dispersion in the cylinder with reduced fuel impingement on cylinder walls. For this series of optical engine tests, a laboratory type fuel supply system is used which consists of a fuel bladder, pressure regulator, and compressed nitrogen bottle. 2.3 Experimental procedure and operating conditions A Mie scattering technique is used to visualize the liquid phase of the fuel dispersion inside the combustion chamber. The fuel spray was imaged with a Photron APX-RS non-intensified high-speed CMOS camera and a Nikon 105 mm AF micro lens. The camera was set to operate at 10 kHz, which provided an image size of 512 9 512 pixels. A high repetition rate pulsed copper vapor laser, synchronized with the high-speed camera and the fuel injection timing logic, is used to illuminate the liquid fuel dispersion. A fiber optics cable is used to direct the laser pulse inside the engine cylinder through the quartz cylinder. The 20 W laser provides the high intensity short pulse duration of about 25 ns for visualization purpose. E85, which represents a blend of 85% ethanol and 15% gasoline (by volume) is used. Experiments are performed at 1500 rpm engine speed with full-load wide open throttle (WOT) condition. Different fuel injection timings (240°, 210°, and 180° BTDC) are studied to optimize the injection timing that minimizes the fuel impingement on in-cylinder surfaces. In addition, the effects of split (dual) injection are also studied by maintaining the same amount of fuel used in a single injection. Fuel injection duration (or pulse width) at each test point is defined to achieve a stoichiometric air–fuel ratio based on gasoline. This was selected for direct comparison between different fuels (gasoline and ethanol–gasoline blended fuels) (Mittal et al. 2010). The test matrix in Table 2 summarizes the key parameters studied. For each test condition, the engine was first motored to reach the desired rpm, i.e., 1500 rpm. Once the engine was stabilized, a pulse signal Fuel spray visualization and its impingement analysis 153 generated by the Cosworth engine controller was sent out to the fuel injector to trigger the start of injection (SOI) at a specific crank angle position as well as to trigger the camera to start recording the specified number of images in consecutive cycles. 3 Image processing technique for fuel impingement analysis Novel image processing algorithms are developed to analyze the spray images for fuel impingement on incylinder surfaces, i.e., cylinder walls and piston top, based on CAD. This technique is useful to optimize the fuel injection pressure, injection timing, and the number of injections for improved engine performance. To analyze the fuel impingement on cylinder walls (left and right) and on piston top based on CAD, first the cylinder boundaries are identified in each frame. Note that the boundaries are defined in the first (or any) frame with respect to the point of interest (POI). This is to be done only once for a given experimental setup. The boundaries are then updated based on the new identified location of this POI due to piston motion. In general, the POI is selected so that the most possible details are captured within the specified block size. The POI used in this work is shown in Fig. 3 (left) at 240° BTDC and the cylinder boundaries (at 186.9° BTDC) in Fig. 4 (left) at 1500 rpm engine speed. A normalized dot product procedure is used to identify the best match for the piston motion detection or to locate the new position of the POI in the processed frame. Mittal and Schock (2010) used a spatial correlation technique to identify the grid displacement in the molecular tagging procedure. Let u and v be any two nonzero vectors (two-dimensional sub-images of 31 9 31 pixel matrix in this work), then the normalized dot product of u and v is defined as follows, Shapiro and Stockman (2001): u v ; kuk kvk Note that for any two nonzero vectors u and v; 1  kuukkvvk  þ 1: A block size of 31 9 31 pixel matrix from the first frame, u, which is centered at the POI (of the first frame) is considered as a template. To identify its best match in the processed frame, a larger roam window is considered in the processed frame. Note that this roam window is centered at the previous frame’s POI location, and it is large enough to encompass the POI displacement (DX, DY) between the two frames, i.e., the processed frame and its previous frame. The roam window size of 61 9 61 pixel matrix is considered in this work. Note that it is not required to consider the whole image of the processed frame for the best match when there is prior information about the maximum possible piston displacement between any two consecutive frames. This is to reduce the computational time. The template is then moved in the larger roam window to identify its best match, i.e., the maximum absolute value of the normalized dot product. Figure 3 shows the location of POI in the first frame (left) with a circular mark and this identified location in the Fig. 3 Piston motion detection between first (left) and second (right) frames 154 M. Mittal et al. Fig. 4 Cylinder boundaries and interior pixels for impingement analysis second frame (right) with a square mark (red color) at 1500 rpm engine speed when the SOI is at 240° BTDC. The observed displacement is 2 pixels in the X direction and 0 pixels in the Y direction. For the analysis of fuel impingement on cylinder walls and piston top, a thin area band of 4 pixels (*0.75 mm) inside the cylinder is considered in this work; see Fig. 4 (right). Green-channel data of RGB spray image (background subtracted) is considered in the processing due to its broad range of histogram distribution compared to the red and blue channels data. Figure 5 shows the normalized histograms of red, green, and blue channels (lower) of the spray image (upper left). Fuel index, fi, which shows the presence of liquid fuel on respective in-cylinder surfaces, is then defined based on the weighted-mean-normalized intensity. For example, the fuel index at frame i for the left wall impingement is: PNplwi ! 1 Alwi j¼1 Ij   filwi ¼ ð1Þ L  1 Alw180BTDC Nplwi where i is the frame number and j is the processed pixel for left wall impingement. Nplwi represents the number of processed pixels in the measurement area of left wall impingement with intensity value, I, greater than zero. Each gray scale image considered in this processing is an 8 bit image and therefore, L = 28, and A represents the total considered area, where, Ntlwi Alwi ¼ X ð2Þ 1 j¼1 and Ntlw180BTDC Alw180BTDC ¼ X 1 ð3Þ j¼1 where Ntlwi and Ntlw180BTDC represent the total number of pixels in the measurement area of left wall impingement for the ith and 180° BTDC frames, respectively. Similarly, the fuel index (at each frame) for the right wall impingement is: PNprwi ! 1 Arwi j¼1 Ij ð4Þ   firwi ¼ L  1 Arw180BTDC Nprwi where each term is self explanatory, as discussed earlier in Eq. 1. Note that rw (in Eq. 4) is used to represent the right wall, whereas lw represented the left wall in Eq. 1. However, for the piston top impingement note that Apti ¼ Apt180BTDC ; and hence the fuel index (at each frame) is as follows: Fuel spray visualization and its impingement analysis 155 Fig. 5 Red, green, and blue channel contents (upper right) of the spray image (upper left) and their respective normalized histograms (lower graphs) fipti ¼ PNppti ! 1 j¼1 Ij  Nppti L1 ð5Þ Figure 6 shows the fuel index values evaluated for each frame for the piston top impingement with E85 when fuel injection pressure is 5 MPa and the SOI is at 180° BTDC. Note that the light reflection by the internal surfaces of the transparent wall in each spray frame is corrected by subtracting the corresponding background frame, i.e., without spray. The start of impingement on the piston top is calculated using the second derivative of the fuel index data. Due to the fact that derivative operation is very sensitive to noise, a two-way low pass filter is used (Zhu et al. 2007). Mittal et al. (2009) used this to filter the in-cylinder pressure data in a diesel engine. The upper graph in Fig. 6 shows the comparison between the experimental (image processing) and filtered fuel index data. The peak of the second derivative (lower graph of Fig. 6) is at 160.2° BTDC and shows the location where fuel impingement starts on the piston top. Due to this and to compare the different cases based on overall fuel index (see Eq. 6), the fuel index value of the previous frame (161.1° BTDC in this case) with respect to the frame where impingement starts is subtracted from each frame’s fuel index value. This processed data is also shown in the lower graph of Fig. 6. fipt ¼ Zhf fipti dh ð6Þ hSOI Figure 7 shows the effect of cycle-to-cycle variations considering five consecutive cycles on piston top impingement with E85 when fuel injection pressure is 5 MPa and the SOI is at 180° BTDC. Note that the cycle 1 results of this condition are already discussed in Fig. 6 and plotted again in Fig. 7 for the comparison 156 M. Mittal et al. Fig. 6 An example of the image processing procedure for piston top impingement with E85 (fuel pressure: 5 MPa and SOI: 180° BTDC) Fig. 7 An example of cycle-to-cycle variations on piston top impingement with E85 (fuel pressure: 5 MPa and SOI: 180° BTDC) purpose. The overall piston top impingement (using Eq. 6) is 10.5, 9.9, 9.7, 10.6, and 10.0 for cycles 1, 2, 3, 4, and 5, respectively. With the variations in peak impingement levels and the overall impingements, it is apparent that the cycle-to-cycle variations exist. However, it is to be noticed that only five consecutive spray cycles are recorded in this work (for each test point), which is insufficient to extract the statistical information, e.g., average fuel impingement levels on in-cylinder surfaces. It is expected that 25 (or more) consecutive cycles are required to obtain the average statistics of impingement. Therefore, only first cycle results are presented in this paper to compare the trends of different test conditions. 4 Results and discussion The effect of fuel spray on in-cylinder mixture preparation and its impingement on cylinder walls and piston top inside a DISI engine are presented. Figure 8 shows the spray development of E85 with LPDI injector at 3 MPa (left) and HPDI injector at 5 (middle) and 10 (right) MPa. The SOI is at 240° BTDC for each case and the engine is operated at 1500 rpm with full-load condition. At this engine speed each frame corresponds to 0.9 CAD. It should be pointed out here that the images presented are a two-dimensional representation of the three-dimensional spray development inside the engine cylinder. Note that the intake valves are located towards the left side of the images. As expected, the spray is first observed when the injection pressure is high, e.g., at 233.7° BTDC for 10 MPa. Similarly, it is first observed at 232.8° and 229.2° BTDC for 5 and 3 MPa, respectively. The LPDI injector (at 3 MPa) shows a hollow cone spray pattern and a strong symmetry along the injector axis. This is observed in the movie file of the crank angle resolved spray frames, however, not shown here. It is also observed that the overall spray angle (with LPDI injector) at 3 MPa is wider when compared to the HPDI injector at 5 and 10 MPa. It is evident from the images that the Fuel spray visualization and its impingement analysis 157 Fig. 8 Spray development of E85 (SOI 240° BTDC at 1500 rpm engine speed and full-load) with LPDI at 3 MPa (left) and HPDI at 5 (middle) and 10 (right) MPa spray tip penetration is faster with increased injection pressure. Thus piston top impingement starts early with higher injection pressure. Also, noticed is the fuel impingement on the right wall at 209.4° BTDC with LPDI injector. This is not observed at this point (209.4° BTDC) when HPDI injector is used. However, there is a vortex formation towards the lower right half of the image with 10 MPa which eventually ends up with some fuel impingement on the right wall at later CADs. There is no right wall impingement when fuel is injected at 5 MPa. The crank angle resolved spray images also show that the left wall impingement is significantly less with LPDI injector at 3 MPa compared to the HPDI injector, which is higher with 10 MPa compared to 5 MPa. Figure 9 shows the effect of fuel injection pressure on left (upper) and right (middle) wall impingement and on piston top (lower graphs) with E85 when the engine is operated at 1500 rpm with full-load condition. The SOI is at 240° BTDC for all the cases. The results confirm the observations of spray images (Fig. 8) that the left wall impingement is significantly less with LPDI injector at 3 MPa compared to the HPDI injector at 5 and 10 MPa. The impingement starts at 209.4°, 204.9°, and 193.2° BTDC for 10, 5, and 3 MPa, respectively. However, right wall impingement is significantly high with LPDI injector than that of HPDI injector. It is to be noticed that the piston top impingement starts early with higher injection pressure. However, the peak values and the overall impingement are higher with 3 MPa. The overall fuel impingement index (using Eq. 6) for piston top is 10.5 with HPDI injector (at 5 and 10 MPa) and 12.3 with LPDI injector at 3 MPa. Overall, results show that HPDI injector at 5 MPa reduces the fuel impingement on incylinder surfaces compared to LPDI injector at 3 MPa and HPDI injector at 10 MPa. It is to be noticed at this point that the fuel index values can be directly compared and provide relative tendency of the fuel 158 M. Mittal et al. Fig. 9 Effect of fuel injection pressure on left (upper graphs) and right (middle graphs) wall and piston top (lower graphs) impingement, SOI 240° BTDC with E85 Fig. 10 Effect of fuel injection timing on left (upper) and right (middle) wall and piston top (lower) impingement; injection pressure 5 MPa with E85 impingement among different test points for a given experimental setup. However, change in experimental setup may result in overall lower (or higher) values based on optical path and camera position; e.g., see Cronhjort and Wahlin (2004) for different illumination arrangements. Nonetheless, the experimental setup is not changed in this study. Hence the results are directly comparable. Figure 10 shows the effect of fuel injection timing at 240°, 210°, and 180° BTDC on left (upper) and right (middle) wall impingement and on piston top (lower graphs) with E85 when the engine is operated at 1500 rpm with full-load condition. HPDI injector is used in each case with 5 MPa of injection pressure. The results show that left wall impingement increases with the advancement of injection timing. However, right wall impingement is higher when SOI is at 180° BTDC compared to 240° and 210° BTDC. Note that the right wall impingement is almost negligible with SOI at 210° BTDC. The peak impingement on piston top reduces from 240° to 210° BTDC and then again increases from 210° to 180° BTDC. Overall piston top impingement reduces slightly when SOI is at 180° BTDC compared to the case when SOI is at 240° BTDC. However, due to the fact that a more homogeneous mixture is achieved when fuel is injected during the Fuel spray visualization and its impingement analysis 159 Fig. 11 Effect of number of injections on left wall (upper) and piston top (lower) impingement; injection pressure 5 MPa, SOI 210° BTDC with E85 intake stroke (by allowing more mixing time), 210° BTDC shows less overall impingement effects on incylinder surfaces. Figure 11 shows the effect of number of injections, i.e., single and dual injection, on left wall (upper) and piston top (lower) impingement with E85 when engine is operated at 1500 rpm with full-load condition. Note that there is no right wall impingement in both the cases (hence not shown in the plots). The SOI is at 210° BTDC for both the cases with the injection pressure of 5 MPa. Note that with split (or dual) injection the start of second injection is at 120° BTDC (90 CADs apart from the first injection) and the total amount of fuel is same as was used in single injection. It is evident that the left wall impingement reduces with the split injection. The overall reduction is about 50% with split injection compared to the single injection. Also, the peak and overall impingement on the piston top is less (about 22%) with split injection compared to the single injection. 5 Conclusion An experimental study is performed to investigate the fuel impingement on in-cylinder surfaces inside a DISI engine. Experiments are performed at 1500 rpm engine speed with full-load conditions. Two types of fuel injectors are used; (i) low-pressure production-intent injector with fuel pressure of 3 MPa, and (ii) highpressure production injector with fuel pressures of 5 and 10 MPa. In addition, the effects of split injection are also presented by maintaining the same amount of fuel used in single injection. Novel image processing algorithms are developed to analyze the fuel impingement quantitatively on cylinder walls and piston top inside the engine cylinder. The technique is useful to optimize the fuel pressure, injection timing and number of injections to minimize the fuel impingement on in-cylinder surfaces. Results show that the split injection is an effective way to reduce the overall fuel impingement on in-cylinder surfaces while maintaining a reasonably well air–fuel mixture in the cylinder. Acknowledgments This work was partially supported by the U.S. Department of Energy under Grant DE-FC26-07NT43275. References Aleiferis PG, Malcolm JS, Todd AR, Cairns A, Hoffmann H (2008) An optical study of spray development and combustion of ethanol, iso-octane and gasoline blends in a DISI engine. SAE Paper 2008-01-0073 Cronhjort A, Wahlin F (2004) Segmentation algorithm for diesel spray image analysis. Appl Opt 43(32):5971–5980. doi: 10.1364/AO.43.005971 Grimaldi CN, Postriot L, Stan C, Troger R (2000) Analysis method for the spray characteristics of a GDI system with high pressure modulation. SAE Paper 2000-01-1043 160 M. Mittal et al. Hennessey R, Fuentes A, Wicker R (2001) Effect of injection timing on piston surface fuel impingement and vaporization in direct-injection spark-ignition engines. SAE Paper 2001-01-2025 Hung DLS, Chmiel DM, Markle LE (2003) Application of an imaging-based diagnostic technique to quantify the fuel spray variations in a direct-injection spark-ignition engine. SAE Paper 2003-01-0062 Hung DLS, Zhu G, Winkelman JR, Stuecken T, Schock H, Fedewa A (2007) A high speed flow visualization study of fuel spray pattern effect on mixture formation in a low pressure direct injection gasoline engine. SAE Paper 2007-01-1411 Kawajiri K, Yonezawa T, Ohuchi H, Sumida M, Katashiba H (2002) Study of interaction between spray and air motion, and spray wall impingement. SAE Paper 2002-01-0836 Mittal M, Schock HJ (2010) A study of cycle-to-cycle variations and the influence of charge motion control on in-cylinder flow in an I.C. engine. ASME J Fluids Eng, 132(5):1–8, 051107 Mittal M, Zhu G, Stuecken T, Schock H (2009) Effects of pre-injection on combustion characteristics of a single cylinder diesel engine. In: Proceedings of the ASME international mechanical engineering congress and exposition, Paper IMECE200910493 Mittal M, Hung DLS, Zhu G, Schock H (2010) A study of fuel impingement analysis on in-cylinder surfaces in a directinjection spark-ignition engine with gasoline and ethanol-gasoline blended fuels. SAE Powertrains, Fuels and Lubricants Meeting, Paper No. 2010-01-2153 Park J, Xie X, Im KS, Kim H, Lai MC, Yang J, Han Z, Anderson RW (1999) Characteristics of direct-injection gasoline spray wall impingement at elevated temperature conditions. SAE Paper 1999-01-3662 Pereira JS, Aleiferis PG, Richardson D, Wallace S (2007) Spray development, flow interactions and wall impingement in a direct-injection spark-ignition engine. SAE Paper 2007-01-2712 Shapiro LG, Stockman GC (2001) Computer vision. London, Prentice-Hall Yoo JH, Kim SK, Zhao FQ, Lai MC (1998) Visulization of direct-injection gasoline spray and wall- impingement inside a motoring engine. SAE Paper 982702 Zhu G, Haskara I, Winkelman J (2007) Closed loop ignition timing control using ionization current feedback. IEEE Trans Control Syst Technol 15(3):416–427. doi:10.1109/TCST.2007.894634