Corona Ignition – Assessment of Physical Effects by Pressure Chamber, Rapid Compression Machine, and Single Cylinder Engine Testing Dr. John Burrows, Kristapher Mixell, Paul-Benjamin Reinicke, Michael Riess, Marc Sens Abstract Federal-Mogul’s Advanced Corona Ignition System (ACIS) is presently under development internally, at customers, and fundamental experiments at IAV. The purpose of these activities is to determine the efÞcacy of the technology, characteristics of performance, and the ultimate industrial feasibility of the technology. Pressure vessel testing has shown consistency with earlier works regarding corona formation/pressure relationships, while also providing evidence of preferential formation with tip sharpness. Rapid compression machine (RCM) work has shown potential correlation with simulation work demonstrating radical formation patterns – yet still has demonstrable beneÞt for reduction in ignition delay in a variety of fuel-air mixtures at various ignition timings. Engine testing has shown signiÞcant reduction in ignition delay with a corresponding improvement in lean and dilute limits. Testing has shown notable improvement in BSFC. BeneÞt has also been shown in engine performance at the knock limit.
1.
Introduction
The concept of utilizing corona for the purpose of igniting fuel-air mixtures has been around for many decades. For example, patents have been granted for such concepts in the early 1960s in the U.S. Although interesting, corona ignition lacked the existence of essential market drivers and complementary technologies to make industrialization neither technically nor economically feasible.
87
In the decades since, numerous researchers in various organizational entities have continued to pursue corona ignition based on the assumptions that (1) a large, volumetric ignition source will be superior to a single point ignition source, and (2) initiation in a spatial model will not suffer from initial heat losses due to proximity effects related to the ignition source, and therefore have less ignition delay. The phenomena of corona inception and its use as an ignition source are well documented and will not be re-examined here. However, researchers have expanded the body of work to include the mechanics of ßame formation, as well as the formation of radicals relative to streamer length and hypotheses from said work. The combination of this body of work, emissions regulations, and the implementation of technologies such as direct injection, electronic engine control, and exhaust gas recirculation have made an environment where the proper execution of corona ignition can be both technically and commercially feasible. Federal-Mogul Powertrain has developed a corona ignition system known as the Advanced Corona Ignition System (or ACIS), based on principles well documented in public research, and a thorough analysis of the technical and execution risks associated with such a technology. This assessment has yielded a set of guiding principles (or pillars) regarding approach and execution. The system has been developed internally, and tested with multiple OE partners, who have generously shared time and data to enable the development that will be presented. Additionally, work has been conducted with IAV on single cylinder engines, and a rapid compression machine, to further analyze and understand various mechanisms related to the beneÞts of corona ignition. The following paper will present the approach and results, as well as some commentary regarding public works.
2.
Testing and results
To date, the ACIS system has been applied to a wide variety of engines. Bore sizes tested have ranged from automotive gasoline (80-90 mm bore class), stationary and transport CNG (90-150 mm bore class) and large bore CNG (> 300 mm bore) power generation. Engine loads up to 30 bar IMEP, and cylinder pressure approaching 100 bar at time of ignition, and over 200 bar Pmax, were tested. Engine speeds have exceeded 6,500 rpm at full load. Customer strategies have included lean, dilute (with and without boost), homogeneous, stratiÞed, and “alternative” combustion.
88
The objective of this paper is not intended to be an exhaustive review of all types of testing and strategies encountered. Many of these details are considered conÞdential. However, details regarding the following are discussed: 1.
Certain variables related to corona performance such as pressure dependency
2.
Effects on combustion initiation
3.
Part load performance enhancement through lean and dilute limit extensions
4.
Combustion performance at the knock limit
2.1. Pressure chamber analysis As known from the physics of electrical discharges, corona formation is strongly dependent upon many factors, including electrode geometry and gas pressure, or density. In order to gain insights into the speciÞc behavior of the ACIS ignition, basic analyses under different temperature and pressure conditions have been conducted on a pressure chamber. An ACIS igniter was placed into a nitrogen atmosphere in a simple pressure vessel at Federal-Mogul with an optical access parallel to the igniter tip plane, and perpendicular to the igniter axis. Select igniter tips were sharpened to demonstrate corona inception voltage preference related to electrical Þeld strength. A starting point of 50 V input and 3.9 kg/m3 was used to demonstrate a relatively large, stable corona without arc discharge, and corona formation was photographed. Pressure is increased isothermally until the discharge is “extinguished” with results recorded. Voltage is then increased 10 V, and the procedure is repeated with images recorded for each successive voltage setting until 90 V is achieved (see Þgure 1). First observations are that the corona size follows a linear behavior at lower voltage settings and gas densities. Densities range four times from stable onset to corona extinguishing density at 50 V. The difference between onset and extinguishing densities is reduced to about 33 % at 90 V. The trend is noted for corona inception voltage (CIV) and corona extinguishing voltage (CEV) along the vertical columns of the chart. Also noted is the preference for formation at the sharper tips, which remains preferential across tested densities and voltage settings. 89
Figure 1: Density and voltage effects on corona formation Even allowing for some igniter inefÞciencies, the results tend to suggest a narrowing of the stable range between CIV, arc onset, and CEV across the range of relevant densities. Furthermore, it tends to suggest the need for a strategy to maximize corona effectiveness at elevated densities and voltage input levels. Additional testing at IAV was conducted with a pressure chamber, which was Þlled with a nitrogen atmosphere, and operated within a range of 0.5 to 150 bar and 0 to 450 °C. For the observation of the corona streamer formation under different chamber conditions, the pressure vessel was equipped with two CCD cameras with top and side views of the igniter tip. In addition to the standard CCD imaging, Schlieren measurements have been conducted as well, in order to yield detailed information about the effects of the corona discharge to the surrounding ßuid. The experimental setup used for the Schlieren visualization is shown in Þgure 2.
90
Figure 2: Experimental setup for Schlieren experiments at the pressure vessel By controlling temperature in this second experiment, a better understanding of the corona discharge phenomenon may be realized. Testing over a series of pressures and temperatures conÞrms the principal dependence of corona formation on gas density and corona voltage. Figure 3 shows exemplary results for one voltage and one density variation. Photographic techniques relate that streamer distribution along its length is stable, as density is varied. A short streamer at higher density has a similar distribution of streamer counts as a long streamer at low density. However, the distribution of high streamer count is condensed closer to the tip. This gives some insight into the formation of radicals that are essential to the initiation of combustion.
91
Figure 3: Density and voltage effects on corona formation (root mean square images calculated from 100 single images) In addition to the beneÞcial volume ignition effect of the corona, it is known from literature that corona discharge generates a gas ßow (corona wind) characterized by signiÞcant turbulence [1], which in turn positively inßuences the initial ßame kernel development and propagation. To examine this phenomenon for ACIS ignition, Schlieren measurements of the corona discharge have been conducted in the pressure vessel. Figure 4 shows the ignition process for ambient conditions and increased gas density at 20 bar/200 °C. At the beginning of the corona discharge, no change in gas density along the streamers is visible (cases 1A and 2A). With increasing corona duration, the generation of local turbulences can be observed in the area covered by the streamers. The maximum magnitude is reached at corona switch-off (cases 1B and 2B). The distribution of the turbulence area is linked to the corona penetration, and thus depends on the gas density; highest intensity seems to occur in the vicinity of the electrodes. With the available measurement setup it was not possible to attribute the formation of turbulences unequivocally to ion motion or gas heating effects 92
caused by the corona discharge. Nevertheless, it is conceivable that the corona-induced turbulence can contribute to the initial ßame development – at least for quiescent mixtures, as prevailing in the RCM investigations described in subchapter 2.2. In a real engine application, it is assumed that the gas dynamic structure produced by corona discharge is insigniÞcant when compared to the level of charge motion and turbulence inside the engine.
Figure 4: Schlieren imaging of corona discharge under (1) ambient conditions and, (2) 20 bar/200 °C directly after (a) corona onset and, (b) corona switch-off
2.2. Rapid compression machine (RCM) In order to better understand the effects of the corona ignition to the combustion initiation, detailed investigations have been conducted by IAV on a rapid compression machine (RCM), in comparison with a standard spark plug ignition. In contrast to the pressure vessel, the RCM allows the examination of the ignition system in a combustible mixture close to real engine conditions. The RCM is a single-stroke device that disposes of a combustion chamber, in which an air-fuel charge is compressed by a piston, and combusted in a subsequent expansion stroke. In this process, the RCM offers various degrees of freedom, as compression ratio, pressure, temperature, as well as air-fuel composition and charge motion. In a Þrst step, the measurements have been conducted with a fully homogenized methane-air mixture, to examine the basic mechanisms of the
93
ßame initialization. Additionally, charge motion has been deactivated to minimize the side effects of turbulence. Figure 5 shows the examined operating points (chamber conditions at ignition timing). For each point, a variation of the air-fuel ratio for representation of stoichiometric, lean, and rich combustion has been performed. In a next step, there are planned experiments with direct gasoline injection during the compression stroke, for an evaluation of the beneÞts of corona ignition under stratiÞed mixture conditions. Chamber conditions at ignition timing Pressure in bar
Temperature in K
Density in kg/m³
Corresponding real engine OP
OP1
4.1
450
3.2
2,000 rpm/2.8 bar IMEP hom.
OP2
8
480
5.8
2,000 rpm/4.8 bar IMEP strat.
OP3
19
630
10.5
1,500 rpm/7 bar IMEP hom.
OP4
33
630
18.3
1,500 rpm/7 bar IMEP strat.
Figure 5: RCM operating conditions at ignition timing For selection of an appropriate optical measurement method, a spectroscopic analysis of the combustion process inside the RCM was done. It is known from existing literature that OH-radiation (wavelength 310 nm) is a very sensitive indicator for the start phase of the combustion, and in good correlation with the thermodynamic combustion rate. However, the detection of OH-radiation requires a complex optical setup with UV Þlter, image intensiÞer and – in case of laser induced ßuorescence (LIF) diagnostics – also a laser device. The spectroscopic measurement of the premixed methane combustion showed a characteristic OH band and a quasi-continuous spectrum in the visible wavelength range with lower intensity. The course of radiation intensity over time is almost completely synchronous for the entire spectrum (see Þgure 6). Therefore, all following ßame experiments have been conducted with a high-speed imaging of the combustion luminosity in the visible spectral range – which enables a much simpler measurement setup, in conjunction with a higher image quality.
94
1.6
normalized intensity in -
flame intensity
1.4 1.2
OH* chemiluminescence natural luminocity rate of heat release
1.0 0.8 0.6 0.4 0.2 0.0
TDC
-0.2
-10
-5
0 5 time in ms
10
15
Figure 6: Representative combustion spectrum (left), and correlation of ßame radiation to rate of heat release (right) The setup used for the experiments is shown in Þgure 7.
Figure 7: Setup for optical experiments on the RCM By means of the high-speed CMOS camera (20 kHz), the development of the corona streamers, as well as the ßame initialization and propagation can be visualized. Figure 8 shows exemplary results for a low load operating point with stoichiometric mixture, which are qualitatively applicable to all examined chamber and mixture conditions. It can be observed that
95
the corona streamers do not start simultaneously on all electrodes. This is probably caused by tolerances in the electrode tip geometry (tip sharpness) – and thus slight differences in the generated electric Þeld. 0.33 ms after ignition onset, the streamers reach their maximum protrusion, and lead to a large-area ßame initialization along their surface. In direct vicinity of the prong tips, where the electrical Þeld is highest, the inßammation appears more intensive. The ßame front propagation is faster than on the streamer heads. Reasons for this behavior can be seen in a higher radical density due to the locally higher electric Þeld strength [2], and an increased heat transfer by adjacent reaction zones in direct proximity. Furthermore, the faster ßame propagation could be promoted by local turbulences induced to the mixture by the corona discharge, as observed in the pressure vessel. With increasing combustion development, the ßame front transitions more and more from the characteristic star-shape into a spherical shape, as present at spark plug ignition. From this point, the ßame propagation is only determined by the chamber and mixture properties.
+ 0.1 ms
+ 0.2 ms
+ 0.4 ms
+ 0.6 ms
+ 1.0 ms
+ 2.0 ms
+ 4.0 ms
+ 6.0 ms
Figure 8: Ignition and ßame core formation for ACIS ignition (time after ignition timing) Comparing the results with a standard spark plug ignition, the beneÞt of the higher ignition volume of the corona becomes clearly evident in a signiÞcantly reduced ignition delay (see Þgure 9). In addition, the propagation of the initial ßame front of the corona ignition is with 4 to 5 ms approxi96
mately 1.5 times higher than with spark plug ignition. In addition to the very high radical density and the local turbulence induced by the corona discharge, the turbulent ßame character of the multiple ignition sites combustion also promotes faster ßame propagation. However, the advance in ßame initialization with the corona ignition cannot be maintained for the entire combustion process. Once the single ßame fronts of ACIS ignition merge to one spherical ßame front (around 6 ms after ignition timing), the combustion conditions are identical with spark ignition. Temporal and spatial distance from corona discharge is too high to still contribute to mixture inßammation. At this point less than 5 % of fuel mass is converted. As the main heat release proceeds under same conditions for both ignition systems, the overall combustion duration cannot be improved by the corona ignition. This becomes particularly obvious when comparing the heat release for spark and ACIS ignition with aligned combustion phasing.
IGN 2
+ 8 ms
50
TDC
180
40
140
30
100
20
60
10
20 -20
0
-15 stroke pressure heat release
-10
-5
0 time in ms
5
10
15
20
1.2 1.0 0.8 0.6 0.4 0.2 0.0
heat release in -
IGN 1
pressure in bar
stroke in mm
220
-0.2
Spark Ignition (IGN1) ACIS, same ignition timing (IGN1) ACIS, same HR50 timing (IGN2)
Figure 9: Comparison of ßame core formation for ACIS vs. spark ignition
97
As mentioned above, all the other examined variants (chamber conditions and air-fuel ratio) yield qualitatively similar results, so that an exhaustive explanation of those is disregarded in this paper.
2.3. Part load engine testing Pressure vessel testing conÞrms previous experiments with respect to the relationship between pressure (density) of the gas and corona formation; while high-speed photography in the rapid compression machine supports theories proposed regarding radical density at the base of streamers, raising questions as to the potential efÞcacy of corona ignition with respect to the ignition of fuel-air mixtures. However, subsequent testing with corona ignition shows a demonstrable improvement in ignition delay and combustion phasing in rich, stoichiometric, and lean conditions in these experiments. Attention turns to in-engine combustion testing that was conducted with several different customer engines – the results of which are presented here. Details of the engines are omitted to protect customer conÞdentiality, but it is given that each of the engines are production-based, with certain modiÞcations reßecting future direction.
2.4. Lean, part load combustion Lean limit testing at part load conditions was conducted to demonstrate the potential of the ignition system as an enabler to future fuel economy strategies. From rapid compression machine work, it follows that a significant reduction should be seen in the 5 % MFB indicating a reduction in the ignition delay. Correspondingly, the 50 % MFB (or CA 50) should show a “shift” in combustion phasing to a slightly earlier position. No effect is expected in the overall burn duration. The ability to measure extension of the lean limit may also be explored using COV of IMEP as indication of overall combustion stability. In the Þrst series of engine tests, the ACIS system is benchmarked against a high-energy conventional ignition system (> 120 mJ). The fuel-air mixture is reduced until combustion stability is considered unacceptable. As shown in Þgure 10, there is little difference in combustion stability from lambda 1 to 1.4. However, a considerable reduction in ignition delay is measured resulting in the ability to retard ignition timing by about 5 de98
grees. As the lambda ratio is increased beyond 1.5, the conventional system begins to exhibit instability far in excess of the 3 % COV limit. In contrast the ACIS ignited engine continued on to nearly lambda = 1.8 before COV exceeded the 3 % limit. The disparity in ignition delay is also increased at these levels.
Figure 10: Lean limit testing @ 2,000 rpm/3 bar IMEP In-cylinder video from a second customer engine test of the ignition and combustion events demonstrates the effect (see Þgures 11 and 12). At time of ignition, combustion initiates at arc in the conventional system, while multiple initiation sites are recorded along streamers. This initial phase of combustion can explain both the reduction in ignition delay and the extension of lean limit.
Figure 11: ACIS ignition at 2,000 rpm/3 bar IMEP
99
Figure 12: Comparison of ACIS (left) vs. spark ignition (right) 23 °CA after ignition timing at 800 rpm/1.5 bar IMEP Recall from the images of combustion in the RCM initiation of combustion appears orderly along the streamers, with higher concentration of radicals and combustion close to the base of the streamer. This was predicted in previous research. In contrast, the propagation of the corona initiated combustion moves very quickly into a large but less organized pattern, demonstrating the effect of turbulence even at the relatively low speed and load. Both examples show similar reduction in ignition delay and timing. This consistent behavior indicates that size and characteristics of the initial inßammation and low heat loss tend to be dominant in the event. The organization and distribution of radicals in the initiation predicted analytically and witnessed in the RCM have lesser effects.
2.5. Dilute, part load combustion Testing on the Þrst customer single-cylinder engine was performed with a similar method to measure the tolerance to externally cooled EGR. As expected from lean limit testing, there was a signiÞcant increase in the amount of EGR that could be introduced before instability was encountered. In the graphs below (please compare Þgures 13 and 14, respectively), it is demonstrated that the ACIS system could ignite mixtures in excess of 30 % EGR. It is also shown that there is a signiÞcant reduction in ignition delay of 28 % at the 25 % EGR point.
100
Spark Ignition ACIS Ignition
Figure 13: EGR testing @ 2,000rpm/3 bar IMEP The potential beneÞts of this extended dilution tolerance can be seen in a third customer’s engine testing. This testing was conducted in a four-cylinder research engine equipped with three spark plugs-per-cylinder, a turbocharger, and an external EGR loop. Cooled EGR was added until instability, indicated by a COV of IMEP > 3 %, was exceeded while BSFC was measured. Similar to the Þrst customer test, EGR could be pushed over 30 %. This test showed a reduction of nearly 5 % in BSFC when compared to the initial undiluted BSFC (see Þgure 14).
Figure 14: BSFC vs. EGR rate for ACIS system on four-cylinder engine Additional tests were conducted comparing a single plug, three-plug, and single ACIS igniter ignition systems. Each ignition system was run to its 101
EGR limit, and the best fuel consumption Þgures were recorded at loads of 2, 5, 8 and 10 bar. In Þgure 15, the baseline (or zero point) is the single plug ignition system at its best dilute fuel economy at the indicated load. At 5, 8 and 10 bar, ACIS performs signiÞcantly better than a single plug system, demonstrating between 2 and 5 % improvement, and about 1 % over the three-plug system. The single exception is at the 2 bar point, where the three-plug system has a slight advantage over ACIS. A possible reason for the anomaly is related to lower turbulence at time of ignition when compared to the higher loads, reducing the advantages of the highly distributed radical formation. By geometry, the three-plug system has initiation points spaced further apart, and could possibly have a better spatial effect at lower turbulence levels. More study and optimization needs to be applied here.
Figure 15: Fuel consumption comparison of single and three-plug systems vs. ACIS
2.6. Full load combustion Testing of ACIS at loads approaching 24 bar IMEP shows an increase in maximum cylinder pressure (Pmax), reduced standard deviation of Pmax, ignition delay, and standard deviation of 5 % MFB when compared to the reference high-energy ignition system (see Þgure 16). Combustion parameters at the knock limit indicate a strong increase in performance is pos102
sible through the ability to optimize timing much closer to the knock limit, without compromising operating safety margins. Furthermore, a reduction in the variation in maximum cylinder pressure can reduce stresses in the engine associated with this variability.
Spark Ignition ACIS Ignition
Figure 16: Combustion parameters at knock limit (2,000 rpm/full load) The precise and voluminous nature of the ignition source is responsible for the improvement. In general, corona size is positively correlated to combustion performance. However, next generation engines with a combination of high compression ratios (small head-to-piston clearance volumes), boosting (high cylinder pressure), turbulence and mixtures (lean, dilute, stratiÞed) present numerous challenges to ACIS. Previously presented pressure vessel testing shows an increase in voltage requirements with increasing cylinder pressure. A dilutant – such as cooled EGR – can increase voltage requirements. As ignition timing approaches top dead center (TDC), clearance between igniter tips and piston crown is reduced. This close condition increases the local electrical Þeld. The combination of voltage demand plus high electrical Þeld can create conditions where an arc discharge is formed, instead of the desired corona discharge. Due to a variety of mechanisms, this arc discharge is less desirable than a properly controlled corona discharge. Depending upon the ignition requirements at the particular speed and load condition, this arcing condition can have effects ranging from none to misÞre. For instance, a lambda = 1, full load condition (high breakdown voltage, turbulent, low ignition quality requirements) may suffer little or no negative effect resulting from an occasional arc discharge; whereas a highly diluted, medium load point with very high ignition quality requirements may suffer a misÞre. The ACIS driver supports arcing mode and a Federal-Mogul patented concept referred to the Ignition Optimization System (IOS). Op-
103
eration with IOS helps to mitigate the effects of arcing by controlling the corona/arc transition, and provides for increased tolerance to cycle-to-cycle variation.
Conclusion Pressure vessel and RCM testing conÞrms relationships well established in previous voltage/pressure/geometrical relationships. New Þndings in the RCM demonstrate and conÞrm analytical work predicting radical distribution along streamer length, as well as the orderly and instantaneous nature of inßammation. RCM testing demonstrates a signiÞcant improvement in the 5 % MFB characteristics compared to spark ignition. This is likely due to both the volume of the ignition source and turbulent nature of the ßame initiated by corona ignition previously documented. However, no signiÞcant change is quantiÞed in the 10-90 % burn duration. Engine testing and in-cylinder videos conÞrm the effect of the voluminous and turbulent ßame from RCM and Schlieren imaging, but show that mixture turbulence has an effect on the pattern of initiation. Although this effect was not directly studied for this paper, it may be inferred that turbulence at time of ignition has limited impact on combustion, as shown in the symmetry of ignition delay reduction between RCM and engine data across a variety of geometries, loads, and fuel types. The beneÞt of ACIS as an enabling technology is seen in its ability to greatly extend both lean and dilute limits. ACIS can also work at high engine loads improving combustion at knock limits, and has been shown to reduce cyclic variability across a wide range of operating conditions.
Outlook Beside the potential in terms of extending lean and EGR dilute limits, the voluminous ignition characteristics of ACIS is expected to be beneÞcial particularly for stratiÞed combustion robustness. In further investigations it is planned to evaluate the effects of the corona ignition to stratiÞed charge inßammation in detail on the rapid compression machine and on a single cylinder engine. First tests at IAV show promising results. Also, it is intended to deeper examine the mechanism of radical induced ignition by continuing RCM investigations, for example radical supported auto ignition.
104
Acknowledgements The authors would like to thank Eike Rindßeisch for supporting this paper with the conduction of the RCM testing work during his diploma thesis period at IAV. Further thanks to Dr. Jochen Mass for his support in terms of optical measurement techniques.
References [1] Akishev, Y., Goossens, O., Callebaut, T., Leys, C., Napartovich, A. & Trushkin, N. (2001). The Inßuence of Electrode Geometry and Gas Flow on Corona-to-Glow and Glow-to-Spark Threshold Currents in Air. Journal of Physics D: Applied Physics, 34, 18. [2] Breden, D., Raja, L. L., Idicheria, C. A., Najt, P. M. & Mahadevan, S. (2013). A Numerical Study of High-Pressure Non-Equilibrum Streamers for Combustion Ignition Application. Journal of Applied Physics, 114, 8.
105
The Authors Dr. John Burrows Electronics and Analysis, ACIS Federal-Mogul Corporation Manchester Suite 14, Styal Road Manchester, M22 5TN United Kingdom E-mail:
[email protected] Paul-Benjamin Reinicke Thermodynamics/Boost Systems, DA-M11 Division Development Advanced IAV GmbH Carnotstraße 1 10587 Berlin Germany E-mail:
[email protected] Kristapher Mixell ACIS and Ignition Coils Federal-Mogul Powertrain 47001 Port Street Plymouth, Michigan 48170 USA E-mail:
[email protected] Marc Sens Thermodynamics/Boost Systems, DA-M1 Division Development Advanced IAV GmbH Carnotstraße 1 10587 Berlin Germany E-mail:
[email protected]
106
Michael Riess Thermodynamics/Boost Systems, DA-M11 Division Development Advanced IAV GmbH Carnotstraße 1 10587 Berlin Germany E-mail:
[email protected]
107