Introduction
Homogenous charge compression ignition (HCCI) mode is a process in which combustion is accomplished throughout charge auto-ignition. Fundamentally according to HCCI technological assumptions, the charge (air-fuel mixture) should be premixed (homogenous) or nearly premixed and the whole charge traversing identical in-cylinder pressure-temperature. The combustion initiation point is conditioned by fuel reactivity (chemical kinetics) and actual operating conditions-in-cylinder temperature, pressure, fuel-air ratio, etc [
1]. This constitutes difficulties in combustion control and thus limits the adaptability and application of this technology to, for example, automotive engines.
The widespread optical diagnostics techniques applicable to engine research field provide substantial, valuable and detailed insight into flow, auto-ignition or combustion activities patterns, and governing mechanisms by chemical species identification or eventual visualization of the aforementioned processes reflected by images [
2,
3]. The autoignition initiation and combustion development study using high-imaging speed imaging (CDD camera) technique in pure and spark assisted HCCI were reported by Aleiferis et al. [
4]. Their investigation objectives concerned different identifications of the engine operating conditions such as charge stratification, residuals and spark discharge causes to pure
n-heptanes autoignition initiation and further combustion development.
Persson et al. [
5] characterized differences and similarities between pure and spark assisted HCCI, contrasting those two combustion modes, basing on image processing of chemiluminescence images quantifying inherent image information. Conducting spectroscopic measurements along with chemiluminescence imaging for a number of different single components fuel and their blends (one and two-step ignition fuels) followed by a detailed exhaustive analysis, Hwang et al. [
6] identified the HCCI autoignition and combustion phases. They proved and reported that in pure HCCI operated engines, there are three phases for single-stage ignition fuel: intermediate temperature heat release (ITHR), high temperature heat release (HTHR) (also called main combustion), and burned out phase. The two-stage ignition (fuel) process is preceded by cool flame activities-low temperature heat release rate (LTHR).
The current work focuses on autoignition initiation and combustion progress investigation and understanding in pure PFI HCCI operated engine based on image derived information analysis of images captured and their correlation with pressure-derived data when needed (profitable). The engine operating conditions set, intake air temperature (no air heating to elevated temperatures), valve timing event (negative calve overlap (NVO) strategy) and fuel used (commercial gasoline) emulated natural, realistic engine operational conditions corresponding to low load HCCI envelope regime operation. The natural light images were collected by a high speed colour camera covering a range of light wavelengths from the visible to near infrared regions (above 400 nm and below 1000 nm). The flame profiles and reaction spreading velocity was analyzed based on digital image processing.
Experimental system and methods
Experiment setup
The experimental study presented in this paper was conducted utilizing a Jaguar single cylinder optically accessed research engine, designed at Ford Motor Company. The geometrical configuration of the optical engine derives from production scale automotive spark ignition Jaguar V8 engine. A port fuel injection (PFI) system, a flat top piston and an intake camshaft designed for PFI-HCCI were installed for the combustion experiment. The optical engine specifications are given in Table 1.
A transparent disc was inserted in the piston crown (see Fig. 1(a)) to create an optical access, which was manufactured from fused silica-full spectrum grade synthetic fused silica. The window diameter is 64 mm. This accessible (imaged) part of the combustion chamber orientation in reference to combustion chamber artefacts is illustrated in Fig. 1(b). When conducting a spray experiment, the metal cylinder could be replaced by a quartz ring to realize a full-stroke optical access, and there was also a side window in the cylinder head, offering an optical path into the pent roof combustion chamber. A sectioned view of the research engine and the optical path constructed are demonstrated in Fig. 2.
The optical engine can be operated with negative valve overlap strategy. Fuel can be injected utilizing the PFI system or Direct Injection (DI) system. Cylinder pressure measurements were made with a Kistler® 6051A piezoelectric pressure transducer mounted in the cylinder head, and precise referenced pressure which reflects directly cylinder pressure at corresponding crank angle around piston BDC instead of intake manifold pressure was given by a Druck PMP 1400 pressure transducer installed in the cylinder barrel. Volumetric air flow was measured using a positive displacement flow meter Romet® G40 certified gas meter, and mixture strength was monitored by broadband universal exhaust gas oxygen (UEGO) sensor incorporated in ETAS LA3 system. All the sensors’ data is recorded by an acquisition system developed in Labview.
The internal exhaust gas recirculation mass fractions presented in the combustion studies can be estimated utilizing an ideal-gas law. The trapped residuals temperature was assumed to correspond (being identical) to measured exhaust gases temperature. The inducted fresh charge mass of a cycle was calculated, since the volumetric air-flow rate along with the intake air temperature was recorded. Residual gas constant used in iEGR fraction estimation were taken from Heywood [
2] and correspond to isooctane combustion data (combustion products-burned gases) with equivalent conditions corresponding to conditions studied in this paper i.e. fuel-air ratios. The heat release rate (HRR) and the mass fraction burned (MFB) can be computed consistently from recorded in-cylinder pressure traces of engine cycle and cylinder volume changes according to the Rassweiler and Withrow method [
2] and differentiating recorded in-cylinder pressure and cylinder volume changes with respect to elapsed time (with time). A Vision Research CMOS Phantom v7.1 colour camera was used for high speed imaging at 7207 frame/s during combustion experiments, triggered by the TTL signal form the encoder.
Commercial gasoline ULG 95 was used as the fuel in this study. The engine works in PFI mode in the HCCI combustion experiment and the operating conditions are listed in Table 2.
Image processing
Image pre- as well as post-processing was conducted using Matlab® programming language. The first step in image pre-processing of raw test images was to subtract the image noise content by using standard technique-the dark frame subtraction. The image post-processing, reaction front spreading velocity was accomplished based on procedure that corresponds to the numerical differentiation of the distance which the individual reaction front contour has moved in each node with respect to elapsed time of two consecutive considered images-image pair. For this purpose, Matlab®bwdist function, which was widely used by previous researchers<FootNote>
Matlab Image Processing Toolbox Tutorials and Help, Matlab 7.1 Version, 2005
</FootNote> [
2,
7-
9], was applied in the image processing and analysis to investigate reaction front spreading velocity. Euclidean metrics which correspond to real distance measurements between image points was selected. Detailed expressions and equations are omitted here due to space limitations.
Results and discussion
Figure 3 presents autoignition and combustion image sequence at λ =1, stoichiometric operating conditions. The first autoignition traces (sites) are marked by a collective activities image at 358°-2° CA before combustion TDC. The area occupied by those autoignition zones is significantly undersized in comparison to the whole imaged combustion chamber area, and consequently barely seen. As correlated to corresponding heat release rates profiles plotted in Fig. 4, their mean and standard deviation of CA5 indications -360.7° CA and 1.02° CA respectively. The average cardinal points of heat release rate profile-respectively 5% burn point (CA5), 10% burn point (CA10) and 50% burned point (CA50) along with standard deviations values are plotted. Gradually higher kurtosis values of pixel intensity’s distribution (calculated, but not given here) of images 364°-366° CA potentially points out more peaked date (with higher occurrence probability) around also progressively increased observed means values. This suggested that higher total enflamed area fraction becomes more dominated with matching single cycles’ images areas.
A discussion of autoignition and combustion nature at lean (λ =1.2) operating conditions is based on process visualization illustrated in Fig. 5 and statistical parameters determined. Starting from 364° CA, the first initial autoignition sites are observable. At this point, noteworthy is the fact that no any preceding ensemble images before that aforementioned one has shown autoignition activities and the scrutinized 364° CA ensemble image therefore constitute a first image of image sequence in Fig. 5. Thus the first observation made comparing the λ =1 and λ =1.2 autoignition and combustion nature is the retarded combustion onset of lean conditions. However, this was expected and it is not new, being an effect of reaction rate and mixture strength dependency. The autoignition activities contained by 364° CA ensemble image, similarly to corresponding stage of λ =1 conditions show weak likelihood of repeatability tendency.
The later images presented ensemble images depicting combustion activities of 374°-378° CA indicate a gradual propensity of single cycle image to enflame whole accessible combustion chamber area. This occurs with the observed combustion progress pattern, from preferential identified autoignition sites. Figure 6 exhibits the heat-release rate profiles of individual cycles corresponding to imaged consecutive cycles. The autoignition initiation (combustion onset) expressed as CA5 is associated with great standard deviation value of CA5 as high as 2.3° CA as seen in Fig. 6. This could and presumably does contribute to decreased value of observed mean, joining images of early and late autoignition activities initiation cycles into one ensemble image at this point of cycle (stage).
The reactions fronts spreading velocities distributions along the reacting structures contours (local velocities) with mean values computed are presented in Figs. 7 and 8 for stoichiometric (λ =1) and lean (λ =1.2) operating conditions studied respectively. The left column shows the reacting structures (their contours) of image pair taken into analysis and the right column illustrates the velocities distribution along the reacting structure contour(s). Because the comprehensive illustration requires an image pair of consecutive crank angles, with highlighted reacting structures and also the distribution along the reacting structure contour to be presented, only one cycle of each operating condition case investigated is shown and discussed in this paper.
The insight into local values and their trend (distribution) during the whole cycle period is gained with limited comparison provided regarding to mixture strength dependency on reaction fronts spreading velocities magnitudes as only one cycle of each operating conditions studied is illustrated. The emphasis of single burning islands (with image containing multiple islands) is given in order to show how the velocities were estimated, more precisely speaking, to show which initial image reacting structure contour corresponds to successive image contour. Each of these single structure velocities along the contours was computed separately, as otherwise it could result in underestimation of values computed.
For both cases studied, the analysis clearly indicated that the low mean values for the initial combustion (autoignition) process reach as high as approximately 16 m/s at λ =1 (images 360°-361° CA), and 14.5 at λ =1.2 (images 364°-365° CA) operating conditions with maximum local values of 35 and 55 m/s for corresponding conditions respectively. This is observed in comparison with subsequent combustion stages. As correlated to heat release rates points of these corresponding single cycles aforementioned velocities magnitudes obtained are representations of around (slightly further) CA5 and CA10 points for stoichiometric and lean conditions respectively, since each image accounts for activities of additional 0.72° CA duration.
Further, as combustion progresses, reactions fronts spreading velocities become gradually higher, up to a given point and then start to decline, and this is noticeable for both cycles velocities examined. The greatest average (140 m/s) with local (up to 140 m/s) reactions front spreading velocities are seen for an image pair of 362°-363° CA at λ =1 operating conditions. Before this happens, the preceding image pair 361°-362° CA reveals maximum local values up to 100 m/s and mean value of approximately 38.8 m/s, for the same operating conditions considered. For this individual cycle burned points this is a region located somewhere between CA10 and CA50 points on heat release rate profile, significantly closer to CA10 than to CA50. Analyzing the single cycle reactions fronts spreading velocities at lean conditions, the highest values are observable for 367°-368° CA image pair reaching a mean value of approximately 31.2 m/s and a maximum local value of 110 m/s. Roughly speaking, this period corresponds to CA50 of the heat release rate. The further combustion stages beyond those reported above are characterized by the gradual drop in reactions fronts spreading velocities for both cycles shown in this study.
The analysis conducted and covered here produced three main observations. First, the initial stage of combustion (autoignition) exhibits significantly lower local and mean velocities magnitudes of reacting structures. Second, at λ=1 operating conditions, the highest magnitudes encountered in the whole cycle analysis are for the region CA10–CA50 of heat release rate, while at λ=1.2, these scale well with approximately CA50 point region. Third, the maximum reacting structures fronts spreading velocities, both local and mean values, were shown to be higher at λ=1 than at λ=1.2 operating conditions, although this is stated based on a single representative cycle of each case studied. The characterized nature of autoignition and combustion in this section is analogous to the previous section of this study investigating the similarities and differences of combustion growth rates distributions at stoichiometric and lean conditions studied. Reasons clarifying each behaviour observed of this section could be found in the preceding section’s corresponding justification.
Hultqvist et al. [
7] noticed reaction front spreading velocities values of approximately 82 m/s for developed combustion at
λ =2.9-4 lean operating conditions. These velocities magnitudes correspond rather to mean values observed and were the first ones reported in HCCI technology, to the best of the authors’ knowledge. Subsequently, Persson et al. [
5] reported reacting structures spreading speeds reaching up to 110 m/s at
λ =1.03 operating conditions. Apparently those magnitudes also constitute mean values of one direction, since these values were estimated using the analogous technique as in Ref. [
7]. Despite the fact that different techniques (assumptions) were utilized to estimate reacting structures fronts spreading velocities for current study and cited aforementioned studies (regardless of other differences in investigational approaches e.g. fuels characteristics) and direct comparison could or even should not be performed, it is noticed that local values of current studies scale well with their reactions fronts spreading values. Comparison of mean values of reacting structures fronts spreading velocities of current study with reported results in Refs. [
6] and [
7] would suggest underestimation of current results.
Conclusions
An optical autoignition and combustion processes diagnostics-characterization of PFI (premixed operation), HCCI at stoichiometric (λ =1) and lean (λ =1.2) operating conditions reflecting realistic HCCI scenario was conducted in this study. The HCCI characteristics were analyzed under the condition of NVO strategy, trapping significant amount of residuals, running on commercial gasoline (single-stage ignition fuel), and with a normal compression ratio and many other engine features very similar to a conventional SI production-scale engine of the same type.
Autoignition and combustion image sequences at λ =1 and λ =1.2 were investigated to analyze the HCCI combustion process. The reacting structures fronts spreading velocities were estimated illustrating local velocities along the burning island(s) contour(s). Besides, mean magnitudes were estimated for one representative cycle of each operating conditions studied.
The initial combustion stage is characterized by a maximum local reaction spreading velocity in the range of 40-55 m/s. The later combustion stage reveals values as high as 140 m/s in the case of stoichiometric combustion. The mixture as well as combustion stages effects are pronounced in these observed analytical results.
Higher Education Press and Springer-Verlag Berlin Heidelberg
PDF
(684KB)
