Comparing the Cycle-to-Cycle Variations of Pulsing Spray Characteristics by Means of Ensemble Image and Probability Presence Image Analysis Techniques

This paper presents an investigation to reveal the cycle-to-cycle variations of pulsing spray characteristics of Spark-Ignition Direct-Injection (SIDI) fuel injectors. The objective is to quantify the spray’s cycle-to-cycle variation such that further insight of the operating principles in the fuel injection system could be developed to enhance the combustion efficiency and reduce emissions of SIDI engines. The experiments were carried out using a multi-hole SIDI fuel injector under an extended range of test conditions in a spray chamber. Using a strobe light as an illumination source, images of the spray structure were taken by a CCD camera. The analysis approach of the cycle-to-cycle variation was based on constructing two types of images, namely, 1) Ensemble Image, and 2) Probability Presence Image (PPI). The analysis of the ensemble image and PPI reveals that both approaches can be used to extract the variations of the spray structure. While an ensemble image is useful for determining the quantitative variation of the spray characteristics, such as the penetration and spray angle, in terms of average, maximum and minimum limits, a PPI provides a new way to examine the spray variation in terms of a probability defined for the presence of the liquid region. Not only a PPI is able to illustrate the magnitude of cycle-to-cycle variation in penetration and spray angle values quantitatively, it also displays the variations qualitatively in a two-dimensional manner in terms of the liquid presence probability. Introduction Spark-Ignition Direct-Injection (SIDI) engines are designed to inject fuel directly into the combustion chamber and possess the potential benefits of enhanced fuel economy, faster transient response, and lower coldstart hydrocarbon emission, among others [1]. In a SIDI engine, the quality of fuel mixture determines the successful ignition and proper burn rate for high quality engine performance. However, large cycle-to-cycle variations in the fuel mixture may increase the covariance of indicated mean effective pressure [2], thereby leading to combustion instabilities that may result in passenger-perceived engine roughness and transient vehicle vibration [3]. The fuel injector is considered as a critical component in the SIDI fuel system. Generally, most of the requirements of the SIDI injector are similar to that of port fuel injector, such as they both require the small pulseto-pulse variation in fuel quantity and spray characteristics. In certain areas, the specification criteria of a SIDI injector are more stringent. For example, SIDI injectors focus more on the spray penetration control with smaller flow variance under large thermal gradients. They should also withstand higher temperature at the injector body and tip [1]. Since the mixture formation process in a SIDI engine is very important, the study of spray characteristics is of great significance [4]. In this paper, we emphasize the study of the cycle-to-cycle variation in pulsing spray characteristics of SIDI fuel injectors through the use of optical diagnostic (image-based) analysis techniques. Optical diagnostic techniques have evolved as a versatile measurement tool in recent years. Programmable hardware and automated data post-processing algorithms have enhanced the efficiency in data reduction and analysis [5]. Different optical techniques, such as stroboscopic technique, light extinction method, laser-induced florescence, phase Doppler interferometry, laser Doppler velocimetry and photographic imaging [6-11] have been developed and applied in a variety of flow measurements relating to engine flow and combustion applications. Among these techniques, photographic imaging is especially useful for qualifying and quantifying the global characteristics of fuel spray over the injection period. By using optical diagnostic techniques, many researchers have investigated the influence of various factors such as fuel properties, fuel injection pressure, back pressure, and fuel temperature that affect the spray characteristics [12-20]. This work extends to a wider range from the previous work of fuel spray cycle-to-cycle characterization [21, 22] to explore various factors which may affect the spray cycle-to-cycle variation at a deeper understanding by using a direct photographic imaging method. The experiments of this study were carried out using a multi-hole  Corresponding author: [email protected] 12 ICLASS 2012 Abbreviated Paper Title Abbreviated Paper Title Abbreviated Paper Title Abbreviated Paper Title 2 SIDI fuel injector in a spray chamber: three types of fuel (n-heptane, ethanol and gasoline), two fuel injection pressures (5 MPa and 10 MPa), three levels of back pressure (50 kPa, 101 kPa and 500 kPa absolute pressure) and three fuel injection-temperatures (-10 C, 25 C and 80 C). Mie-scattered images of the spray structure were taken by a CCD camera using a strobing white light. The analysis approach of this work to reveal the cycle-to-cycle variation is based on image processing and analysis. Experimental Methods & Analysis Procedure 1) Experimental Apparatus & Test Procedure The spray behavior is investigated by means of a strobe light illumination to obtain the geometrical characteristics of spray in terms of spray angle and penetration. Figure 1 demonstrates the schematic of the experimental apparatus, (which consists of a constant volume chamber, a fuel supply system, a fuel temperature control system, a chamber pressurization system, a vacuum system and a photographic imaging system). An eighthole injector (the outside hole diameter was estimated to be about 0.2 mm) was installed vertically at the top of the constant volume chamber. Four quartz windows surrounding the chamber provided full optical access, and back pressure in the chamber from 20kPa (absolute) to 2MPa was regulated by a nitrogen supply and a vacuum pump. A piston accumulator with a separate nitrogen supply system was used to provide the required fuel pressure. In addition, a specially-fabricated injector with a thermocouple embedded near its tip was used to correlate the fuel temperature inside the injector with the temperature of water running through a coolant path surrounding the injector. Therefore, the fuel temperature in the injector could be regulated by controlling the temperature of the running water near the injector. Spray images were captured by the LaVision imaging system which consisted of a CCD camera which located at an angle of 150 from the light source. A timing device to control the image acquisition timing and the injection event was synchronized by the Davis® software. The experimental matrix was selected to represent various operating conditions of the injection system as shown in Table 1, where the ambient temperature was fixed at 25C. The test fuels, namely, n-heptane, ethanol and gasoline were chosen because they exhibited different physical properties, as shown in Table 2. The injection pressure, the back pressure in the chamber and the fuel temperature were varied to investigate their influence on the spray cycle-to-cycle variation. Table 1 Experimental conditions. Table 2 Physical properties of the test fuels. The spray characteristics illustrated in Fig. 2 are defined according to the SAE J2715 recommended practice on gasoline fuel injection spray measurements and reporting [23]. The spray penetration is defined as the lowest vertical distance to the injector tip. The spray angle is defined as the widest angle including all plumes in the image. For each test condition of Table 1, a series of spray images as well as background images were recorded. The background image was subtracted from the spray image during image processing. Fifty images were taken at 1.3 ms after the start of fuel (ASOF). The images were processed using the Matlab® image processing toolbox. Figure 1 Schematic Layout of Experimental Apparatus Figure 2 Definitions of Spray Characteristics. 2) Image Analysis Procedure and Methodology The development of the image processing technique to reveal the cycle-to-cycle variation of pulsing spray is a major objective of this work. The strobe light image of every single fuel injection provides detailed spatial and ICLASS 2012, 12 th Triennial International Conference on Liquid Atomization and Spray Systems, Heidelberg, Germany, September 2-6, 2012 3 temporal information into a peculiar spray event. The analysis approach includes constructing two types of images, i.e., 1) Ensemble Image, and 2) Probability Presence Image (PPI). The Ensemble Image is obtained by averaging the intensity value of each image from a series of test replicates at the same time delay after the injection logic pulse. This conventional image processing method is used to extract the overall characteristics of fuel spray, such as spray penetration and spray angle on a statistically valid basis. The intensity (grayscale value) of the Ensemble Image indicates the average information of the liquid presence. The Probability Presence Image (PPI) is constructed by the following steps. First, the region of the liquid fuel presence in each image is identified by a thresholding method and binarized images are obtained. According to the ‘Image Processing Handbook’ by John C. Russ [24], the ‘the simplest method’ of threshing methods is used in our analysis by locating the peaks in the histogram of images and setting the thresholds midway between them. By observing the histograms, the peaks are located in 12 and 37 on the full 256 levels (8 bit). Using this thresholding method, the threshold is set to 25 (about 10% of the 8 bit greyscale) so that any pixel with grey level lower than 25 are set to 0, and those with grey levels above the threshold are set to 1. Then, the binarized values of all images are added pixel-by-pixel in the same test replicates (i.e., the same ASOF images taken from the injection repetitions), and then the value is divided by the total number of images. Third, the value of these images at each pixel location is multiplied by the maximum luminance intensity of the image format (for our work, the maximum luminance value of an 8-bit image is 255), and then the ensemble PPI was formed as a grayscale image. For an 8-bit image, the pixels with a grayscale value of 255 on the PPI imply the 100% probability of liquid presence, while those with value of 0 represent no liquid presence and 0% probability. It is worth mentioning that the PPI, formed by using a set of binarized images, is also an ensemble image but illustrated with the probability of liquid presence. It provides another approach to examine the cycle-to-cycle variation. The numerical value of the probability represents the core region where the liquid is always present over the cycles considered. For instance, the outermost periphery indicates where the liquid may be least likely to exist beyond this boundary, which is of a low probability value. However, since the original liquid image is binarized, there is no account for the different cycles of liquid presence. It is simply a measure of whether liquid is present at the designated SOI time delay. In accords with constructing two types of images, the data analysis procedure applies these two indexes, namely, 1) the coefficient of variation (COV), and 2) the Probability Variation Index (PVI), to indicate the variation of each method. The coefficient of variation (COV) of the spray characteristics obtained from repetitive images that are formed as the Ensemble Image is a useful dimensionless number to indicate the dispersion from the average response and can be used to make comparison between data sets of different units. The standard equations to calculate the variations of the spray characteristics are shown below: