- Email: [email protected]
Split fuel injection and Miller cycle in a large-bore engine
Applied Energy 162 (2016) 289–297
Contents lists available at ScienceDirect
Applied Energy journal homepage: www.elsevier.com/locate/apenergy
Split fuel injection and Miller cycle in a large-bore engine Matteo Imperato ⇑, Ossi Kaario, Teemu Sarjovaara, Martti Larmi Aalto University, Finland
h i g h l i g h t s High premixed combustion peak is reduced with pilot injection. Split injection in large bore engines increases engine efficiency. Pilot injection with high Miller rate can reduce NOx up to 60%. NOx reduction of 42% could be obtained without decrease in engine efficiency.
a r t i c l e
i n f o
Article history: Received 25 June 2015 Received in revised form 25 September 2015 Accepted 5 October 2015
Keywords: Large-bore engine Split injection NOx reduction Miller cycle Pilot injection Diesel combustion
a b s t r a c t The upcoming emission legislation for sea-going vessels issued by the international marine organization requires drastic reduction in nitric oxides. A well-known approach for meeting these requirements is to reduce the in-cylinder temperature prior to combustion by using the so-called Miller cycle. However, the mere use of this technique presents the actual limits due to long ignition delay, which occurs when the compression temperature is very low. As a consequence, premixed combustion develops quickly, increasing the local temperature in the combustion chamber and favoring NOx formation. Splitting the fuel injection into a small pilot and a main injection can reduce the magnitude of the premixed combustion and the local in-cylinder temperatures. The work presented here is divided in two parts and is novel by being the first systematic study of split injection combined with Miller cycle in large-bore engines. In its first stage, an extensive study of the injection dwell with two intake valve closings and three timings of the main injection are analyzed. In the second stage, both injection events are shifted later in the power stroke with fixed injection dwell. Overall, the pilot injection reduced the ignition delay but dropped the peak of the premixed combustion only with the most advanced intake valve closing. This improved fuel economy, but provided no advantages as far as emissions are concerned. In addition, while increasing injection dwell reduced NOx emissions, it also increased fuel consumption. The highest achieved NOx reduction was close to 60%, with a small drawback in fuel economy. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction The emission limits for ships’ diesel engines are becoming constantly stricter. In 2016, a new regulation, which reduces nitric oxi-
Abbreviations: °CA, crank angle degree; ATDC, after top dead center; BBDC, before bottom dead center; BDC, bottom dead center; BMEP, brake mean effective pressure; BTDC, Before Top Dead Center; CI, compression–ignition; CO, carbon oxide; HR, heat release; HRR, heat release rate; IMEP, indicated mean effective pressure; IVC, intake valve closing; LTO, low temperature oxidation; MI, main injection; NDIR, non-dispersive infrared; NOx, nitrogen oxides; NTC, negative temperature coefficient; PI, pilot injection; SFC, specific fuel consumption; SOMI, start of main injection; SOPI, start of pilot injection; TDC, top dead center. ⇑ Corresponding author at: Department of Mechanical Engineering, Aalto University, 5 Puumiehenkuja, Aalto 00076, Espoo, Finland. E-mail address: [email protected] (M. Imperato). http://dx.doi.org/10.1016/j.apenergy.2015.10.041 0306-2619/Ó 2015 Elsevier Ltd. All rights reserved.
des (NOx) emission by 80% – compared to 2001 level – for largebore compression–ignition (CI) engines [1], comes into effect. Therefore, the development of new solutions and technologies for reducing emissions is mandatory, in order to fulfil the stringent demands of the legislation. However, the engine performance should not be negatively affected by the new adopted configurations. NOx formation occurs in the combustion chamber when the local gas temperature is high [2]. An effective way to reduce the combustion temperature in CI engines is the so-called Miller cycle [3]. This can be realized by closing the intake valve before bottom dead center (BBDC), so that the effective compression ratio is lower than the effective expansion ratio. Thus, the compression incylinder temperature and the eventual combustion temperature decrease. Miller cycle has been theoretically studied [4], and it is
290
M. Imperato et al. / Applied Energy 162 (2016) 289–297
widely used in automotive [5,6], heavy-duty [7,8] and large-bore engines [9]. The Miller cycle approach is based on the correlation between NOx formation and the flame temperature. The approach states that, with reduced flame temperature, NOx emissions are expected to be lower [2,10]. As mentioned, the NOx formation kinetics is strongly temperature-dependent. However, recent investigation has reported several cases where the flame temperature correlation fails to capture the trend of NOx emissions for diesel combustion. A study demonstrated that at low load [11] increasing the Miller rate increases NOx close to the reference value with standard valve timing. In addition, it was shown that ignition delay becomes very long with advanced Miller rate, and the consequent high in-cylinder pressure fluctuations result in increased NOx values, despite the low in-cylinder temperature at ignition [12]. In addition, because of the lower compression temperature a higher Miller rate increases the ignition delay. Kyrtatos et al. [13,14] tested very advanced Miller timing in a large-bore mediumspeed engine and observed that a longer ignition delay – due to lower in-cylinder temperature – allowed more fuel being mixed prior to ignition, leading to a greater magnitude of premixed combustion. Moreover, the long ignition delay let the spray penetrate further, enhancing air entrainment and resulting in an overall leaner premixed combustion that could favor NOx formation. This might also result in unstable combustion and high cycle-to-cycle variation of the in-cylinder pressure, which could compromise the engine’s proper function and performance [12]. The amount of the premixed combustion may be correlated with the formation of NOx. Generally, a high peak of premixed combustion results in high NOx outcomes. Musculus [15] realized optical measurements in a diesel engine and found contradicting trends. It was reported that, reducing the intake temperature, NOx initially decreased and then increased monotonically with premixed combustion. Consistently, it was found that, for operating conditions with large premixed ratios, retardation of injection eventually led to an increase of NOx after reaching a minimum. It was concluded that large portions of premixed combustion generally led to faster combustion, with high increase in cylinder pressure early in the engine cycle. Thus, the reactants entering the diffusion flame are compressed to higher temperatures, leading to higher flame temperatures (compression heating). Split injection has been widely used in diesel engines for reducing both NOx and soot [16–18]. Injecting a small pilot dose with advanced intake valve closing (IVC) can trigger the combustion earlier before top dead center (BTDC), reducing also the ignition delay of the main injection. This is expected to reduce the magnitude of the premixed combustion and NOx. Brückner et al. [19] studied the influence of fuel pilot injection and Miller cycle in a heavy-duty diesel engine running at 1000 rpm and demonstrated that the pilot injection decreased the peak of the premixed combustion, reducing further NOx only in conditions of very low incylinder temperature. Pilot injection has not been extensively studied in marine engines. On a two-stroke marine engine, Andreadis et al. [20] tested pilot injection, achieving a NOx reduction of 15% and also a slight improvement in the engine economy. However, the study did not investigate any influence of Miller timing combined with split injection. Nonetheless, systematic studies of the use of the Miller cycle and split injection in large-bore engines cannot be found in the literature. The present research was carried out with a single-cylinder large-bore four-stroke mediumspeed research engine [21,22]. Former achievements include 40% NOx reduction at partial load, adopting Miller cycle [23]. Recently, split injection and Miller cycle running at partial load with 1500 bar injection pressure and with a pilot injection fuel fraction of 25% were attempted for the first time [11]. In those runs, pilot injection
timing was tested either very close or very far from the main injection. Hence, the influence of the split injection on the combustion characteristics was not completely understood. The novelty of this research consists of testing split injection and Miller, running with low fuel injection pressure, and investigating the combustion behavior at regular injection dwell steps. The main aim was to reduce the ignition delay when running with advanced Miller rate and to study the effect of the injection strategy on the combustion process and performance outcomes, especially that of NOx. The present study is divided in two parts. In the first part, different injection dwells are tested. Then, with a fixed dwell, the injection events are retarded to reduce the NOx amount, while attempting to fulfil the upcoming legislation requirements. 2. Material and methods 2.1. Experimental setup A single-cylinder large-bore CI medium speed engine was employed in this research: its main components were designed to withstand in-cylinder pressure of 30 MPa running and mean piston speed of 12 m/s [22,24]. The engine was connected to an electric motor, which also allowed running in the motored mode. In addition, the gas exchange valves were controlled by electrohydraulic valve actuators (EHVA) instead of the traditional mechanical camshaft, providing high flexibility in valve timing [25] for the gas exchange phase. Moreover, the fuel system permitted changing of injection pressure, duration and timing. An external charge-air supply plant and an exhaust gas throttle valve controlled the boundary conditions: the maximum allowed intake air pressure was 10 bar. The parameters of injection, valve timing and ancillary systems could be freely set and remotely monitored [21]. A schematic of the engine is shown in Fig. 1. The charge air is processed by the compressor (C); then, it is cooled, dried and heated before going through the flow meter. The air reservoirs reduce the pressure fluctuations and the intake pressure is regulated by a control valve. The engine (E) is connected to the electric motor (M). The exhaust gas pressure is controlled by the throttle valve, and exhaust gas samples are diverted to the emission analyzers and to the smoke meter (AVL 415). The engine was equipped with Kistler in-cylinder pressure sensor 6045AU20 connected to the Kistler charge amplifier 5064. An Emerson Micromotion ELITE CMF100 flow meter was employed for charge air measurement, and an Emerson Micromotion ELITE CMF025 flow meter for the measurement of fuel mass flow before the engine. The emission measurement system was composed of different analyzers for each gas. NOx emissions were measured with an ECO Physics CLD822Sh analyzer. Soot emissions were measured as filter smoke number (FSN) by an AVL 415 smoke meter. A non-dispersive infrared (NDIR) absorption analyzer by Sick was employed to measure carbon oxide (CO). The emission probes were located in the exhaust pipe five meters from the cylinder head
Fig. 1. Schematic of the engine test bed.
291
M. Imperato et al. / Applied Energy 162 (2016) 289–297 Table 1 Main dimensions and engine configuration. Displaced volume Stroke Bore Connecting rod length Compression ratio Injection nozzle tip
(cm3) (mm) (mm) (mm) – (#Holes £ mm)
8796 280 200 614.2 16.7:1 9 0.30
outlet, and the gases were transported in a heated line (191 °C) to avoid condensation. Table 1 presents the main dimensions of the engine. The compression ratio was high to reduce fuel consumption. The injector nozzle tip holes were reasonably small, to obtain good spray properties [2] without exceeding the injection duration. 2.2. Calculation methods The specific fuel consumption (SFC) and the NOx were normalized, with 100% corresponding to the reference point. The reference point was run with a single injection, and in particular the specific NOx outcome was close to the IMO Tier II value at 900 rpm. The heat release rate (HRR) was calculated using the in-cylinder pressure and the cylinder volume along °CA. The apparent HRR equation [2] below was adopted:
dQ c dV 1 dp þ ¼ p V d CA c 1 d CA d CA c 1
ð1Þ
where Q is the heat released, p the in-cylinder pressure, V the incylinder volume and c the specific heat ratio. The variable c is temperature-dependent and an exponential correspondence was chosen for determining its value in the calculation [26]. The HRR calculation started at 40°CA before top dead center (BTDC) and terminated at 50°CA ATDC in the power stroke. The ignition delay could be evaluated from the start of injection (SOI) and the instant when the HRR curve departs from zero [2]. However, because of measurement errors and unstable combustion with some setups, the ignition delay was determined as the time between the SOI and the time when the heat release (HR) reached the value of 0.5 kJ, corresponding to energy content of 12 mg fuel (<2% of the total injected fuel). Although the results might be overestimated, clear trends of the different strategies chosen in this research were expected. In the two-injection cases, the ignition delay was considered along with the pilot injection. A simulation model was employed to attain the values of incylinder temperature at the injection start. Temperature is a crucial parameter for evaluating the ignition delay [2], but it is not possible to calculate it directly from the test results. Thus, a simulation model was developed with GT-Power, 1-D fluid dynamic code powered by Gamma Technologies. GT-Power has been extensively used in diesel engine research [27,28], and a detailed model had been already applied to this engine [23]. In particular, the Redlich-Kwong [2] real gas law was employed for the combustion chamber. Since the simulation was used for the evaluation of the in-cylinder temperature prior to combustion, the model was calibrated with the engine inlet and outlet conditions obtained from the test runs. Only motored cycles were run.
the injection dwell was constant and both the injections were retarded. Four valve timings were tested in this study, and the details can be seen in Fig. 2. The IVC of the reference case was 30°CA BBDC, and this parameter was advanced until 110°CA BBDC. When changing the valve timing, only IVC was modified, whilst the exhaust valve timing and the intake valve opening were kept unchanged. As shown in Fig. 2, the maximum valve lift was reduced as the IVC was advanced. Hence, the closing slope of the intake valve phase was augmented with the configuration Mil100 and Mil110, to avoid undue reduction of the maximum lift value. The injection parameters are defined in Fig. 3. It must be pointed out that the injection duration referred to the electric signal of the injector, since the injection valve was not instrumented. In particular, the injection dwell is intended as the time between the start of the pilot injection (SOPI) and the SOMI. Because of the mechanics of the fuel injection valve, the smallest stable injection opening time was set to 0.93 ms, which corresponded to 5°CA. Hence, the minimum injection dwell was set to 7°CA, with 2°CA tolerance to permit good and steady operation of the injection system. The pilot injection duration was 33% of the total injection time, and this value should be a good approximation of the relative pilot fuel ratio over the global injection quantity. Since the study was focused on valve timing and injection timing, some parameters were constant in each test point, and they are reported in Table 2. All the tests were carried out at partial load running at 900 rpm. Injected fuel mass flow and air-to-fuel ratio were also constant in all the tests. In particular, the injected fuel
Fig. 2. Gas exchange valve lift curves.
2.3. Test settings The research had two parts, each with different gas exchange valve timing. In the first stage of the research, the injection dwell was tested by modifying the time between the injections, together with three starts of the main injection (SOMI). In the second stage,
Fig. 3. Injection nomenclature.
292
M. Imperato et al. / Applied Energy 162 (2016) 289–297
Table 2 Constant engine parameters. Engine power
Engine speed Engine load (BMEP)
(rpm) (bar)
900 11 ± 0.4
Injection parameters
Total injection quantity Injection duration (single) Pilot duration Injection pressure
(mg/cyc) (°CA) (°CA) (bar)
700 ± 14 15 5 1200
In-cylinder conditions
Density at TDC Total k
(kg/m3) (#)
40 2.1
mass per cycle amounted to 700 mg (with a 2% error range), and the total k was 2.1, resulting in 40 kg/m3 charge density at the end of the compression stroke. Since the intake valve opening time was reduced in Miller cases, it was necessary to increase the boost pressure values in order to achieve the same global air-to-fuel ratio. The main injection signal was of 15°CA, and the fuel injection pressure was 1200 bar, reasonably lower than in the former study. The complete test matrix is shown in Table 3. The engine setup name was chosen along the IVC. In the first part of the work, a detailed study of the injection dwell was carried out with Mil80 and Mil100. The second part of the work aimed to reduce the NOx by retarding both the injection events and increasing the Miller rate. The injection dwell was constant, and the comparison with the single injection cases was also carried out. Table 3 shows also the simulation results at top dead center (TDC). In the reference point, the IVC was set close to the bottom dead center (BDC) and the motored temperature from simulation was close to 1000 K. Temperature reduction of 100 K was obtained by advancing IVC of 50°CA, and, with the most advanced timing, the temperature reduction at TDC was almost 200 K compared to the reference. Although these values show the influence of the Miller timing on the in-cylinder temperature, it must be considered that in most of the test points the combustion starts before the TDC. Hence, the in-cylinder temperature at the injection instant will be taken into account later in the paper.
Fig. 4. Influence of Miller and pilot injection on the combustion (early SOMI).
pressure increases without remarkable changes in the magnitude of the premixed combustion, which occurs earlier because of earlier injection timing. Controversially, very early pilot injection results in high premixed combustion and two-fold increase of its maximum value. The fuel injected during the pilot injection completed its combustion and, as a consequence, the diffusive combustion has a lower magnitude. It must be pointed out that the overall injected fuel mass is the same in all the run cases (Table 2). Fig. 5 shows the influence of all four tested injection dwells with timing Mil80 and late SOMI. The premixed combustion increases with earlier pilot injection timing, and the diffusive combustion has lower magnitude. Here, the peak pressure occurs later
3. Test results 3.1. Effect of the injection dwell Fig. 4 shows the effect of Miller timing and pilot injection on the combustion process. The in-cylinder compression pressure is lower with Miller timing, because the same air mass quantity is compressed at a lower temperature. Miller cycle increases ignition delay and premixed combustion magnitude. In addition to the cases with single injection, also two cases with pilot are plotted in the same figure. With short injection dwell, the in-cylinder peak
Fig. 5. Influence of injection dwell with Mil80 on the combustion (late SOMI).
Table 3 Test matrix. Engine setup
Injection strategy
IVC (°CA bBDC)
Simulated motored temp. at TDC (K)
SOMI (°CA aTDC)
Inj. dwell (°CA)
Relative pilot quantity (%)
Reference Mil80 Mil100
Single injection
30 80 100
998 904 855
2
–
0
Mil80 Mil100
Split injection
80 100
904 855
2, 0, 2
7, 10, 13, 16
33
Mil80 Mil100 Mil110
Single injection
80 100 110
904 855 807
3
–
0
Mil80 Mil100 Mil110
Split injection
80 100 110
904 855 807
3, 4, 5
10
33
M. Imperato et al. / Applied Energy 162 (2016) 289–297
and it is lower. In comparison, the tests with earlier SOMI (Fig. 4) display similar in-cylinder peak pressure regardless of the pilot timing: with earlier timing, the pilot combustion takes place mostly during the compression stroke, augmenting to a greater extent the in-cylinder pressure. In Fig. 6, the cumulative heat release is plotted. It can be seen that the late part of combustion was similar regardless of the pilot injection timing. In fact, all the curves match close to 9 kJ, which is slightly lower than 50% of the total HR. Fig. 7 depicts the effect of Miller timing and pilot injection. The increased Miller rate lengthens the ignition delay and enhances the premixed combustion. Its peak is decreased using late pilot injection (dwell = 7°CA), although it is apparent that the overall combustion develops mainly around the TDC, increasing the incylinder pressure level. Similarly to Fig. 4, a large dwell resulted in high peak of the premixed combustion, and the maximum HRR value is 3 kJ/°CA. In this case, the pilot and main injection combustion are clearly distinct. With a more advanced Miller rate, the ignition delay increases and, accordingly, the peak of premixed combustion. Fig. 8 represents the combustion curve of the test points with Mil100 and SOMI = 2°CA ATDC. Here, the premixed phase is of a higher magnitude with the injection setup similar to that of Mil80 (Fig. 5). Because of colder in-cylinder temperature, the pressure fluctuations are larger, especially with the earlier pilot. As a summary of the combustion analysis, in Fig. 9 the peak heat release rate of the premixed and diffusive combustion are plotted as injection dwell changes. High heat release rate might imply high
Fig. 6. Influence of injection dwell with Mil80 on the cumulative HR (late SOMI).
Fig. 7. Influence of Miller and injection dwell on the combustion.
293
Fig. 8. Influence of injection dwell with Mil100 on the combustion (late SOMI).
in-cylinder temperature, and this could be correlated with NOx formation [2]. As shown previously, earlier pilot injection – i.e. larger dwell – increases the maximum value of the heat release rate in the premixed phase. In particular, earlier IVC implies a large increase in the HRR, and the major differences occur with larger dwell. In addition with Mil100, the peak of the HRR in the premixed combustion converges to the same value, as the dwell is reduced. On the other hand, with Mil80 the values are different with injection timing for each tested dwell. However, the values with Mil100 are in all cases higher than Mil80, due to the longer ignition delay. In the diffusive combustion, the maximum value of the HRR increases as the dwell is reduced. This can be explained as an effect of the short injection dwell: the combustion of the pilot dose is not completed at SOMI. While in the premixed combustion the maximum HRR increases as SOPI is advanced, the trend of the maximum value of diffusive combustion as SOMI changes is not well-defined. In fact, the differences of the absolute values are small and mainly within the measurement error range. As stated in the literature [2], ignition delay generally increases with lower air temperature. Fig. 10 shows the measured ignition delay plotted with the simulated in-cylinder temperature at the injection. The pilot ignition occurs before SOMI in all the cases. Generally, early pilot injection shortens the ignition delay compared with single injection cases. In addition, one can notice that different trends are measured with the different valve timings. Indeed, with Mil80 all the points lie on the same curve regardless of the injection timing. One can observe that the same values of ignition delay are measured with in-cylinder temperature at injection in range 860–890 K. In comparison, with Mil100 the incylinder temperature is cooler and the injection timing affects more the ignition delay. For instance, its value changes from 0.9 to 1.2 ms at temperature of 800 K. At low temperature the hydrocarbon oxidation, and eventually the ignition delay, is governed earlier by a low temperature oxidation (LTO) range and later by a negative temperature coefficient (NTC) range [29,30]. In the latter, the ignition delay increases with increasing temperature. With Mil100, the pilot event is injected between 750 and 840 K, which is reasonable in LTO-NTC range. However, it must be kept in mind the experiments above mentioned [29,30] were achieved in conditions of perfect air–fuel premixing, and hence the results might differ from these experiments. The influence of the injection dwell on the NOx outcomes is plotted in Fig. 11. As expected, in single injection cases high Miller rate brings a drastic reduction of NOx, over 50% with Mil100 timing. Split injection does not yield further NOx reduction, when compared with the single injection cases. As shown in Figs. 5 and 8, early pilot injection creates large premixed combustion, but the main combustion starts when the pilot fuel is not completely
294
M. Imperato et al. / Applied Energy 162 (2016) 289–297
Fig. 9. The peak heat release rate in premixed and diffusion combustion.
Fig. 10. The in-cylinder simulated temperature at injection instant and the ignition delay.
Fig. 11. Influence of injection dwell on NOx emission. Percentile normalization along the reference point.
burnt. In addition, the effect of the injection dwell changes with the Miller rate. Indeed, with Mil80 timing, larger dwell implies lower NOx. Nonetheless, with Mil100 the injection dwell has minor effects on NOx outcomes, registering the lowest value with dwell = 10°CA. As stated by Musculus [15], both a large premixed combustion peak and great heat released in diffusive combustion increase the
in-cylinder temperature, which may form high NOx. With these assumptions and according to the NOx outcomes, the runs with dwell = 10°CA indicate the best trade-off between these two phenomena. Herfatmanesh [31] achieved similar results in a study of dwell angle with an optical engine at 1500 rpm. Although the application was different from the one here, the NOx trend was not monotonous and minimum values were attained similarly with dwell = 10°CA. The results concerning soot and CO are presented in Fig. 12. In all the cases, seen objectively, the absolute values are low, considering that the tests were carried out at partial load. Generally, soot (as FSN) increases with CO. Split injection with large dwell increases both, especially when running with Mil100. Lower CO is also obtained with a small dwell. As shown in Figs. 5 and 8, with a large injection dwell the pilot combustion is extinguished before SOMI; with a narrow dwell, on the contrary, the main combustion commences before the pilot combustion is completed. Thus, the evaluation of the start of the injection of the main combustion is not possible with sufficient accuracy. From the FSN results, it can be seen that later injection – i.e. shorter dwell – reduces the amount of soot in the exhaust gases, and this is in accordance with previous experiences with diesel engines [16–18]. Fig. 13 shows SFC, normalized to the reference point. Split injection results in lower SFC, when compared with the single injection cases, because of the remarkably higher in-cylinder pressure around TDC (Figs. 4 and 7). In addition, larger dwell angle denotes higher fuel consumption for all the tested configurations. This might be due to the high amount of heat released in the compression stroke, which represents negative work of the engine cycle. A narrower dwell angle shifts the combustion phase toward the expansion stroke, improving the engine economy. Considering the same injection dwell, earlier injection is more favorable for decreasing fuel consumption. Summing up the results shown in Figs. 11 and 13, the general trend is such that the larger dwell brought to lower NOx values but higher SFC. However, the configuration dwell = 10°CA was chosen for the second part of the study, since there was a certain fuel economy decrease with a reasonable NOx reduction. Shifting the injection events later was expected to decrease the amount of NOx when keeping the SFC value at the same level of the reference point.
3.2. Effect of the injection timing with constant dwell As shown in Table 3, three different Miller rates were tested with late fuel injection and constant dwell = 10°CA. A more
M. Imperato et al. / Applied Energy 162 (2016) 289–297
295
Fig. 12. The soot and CO values along the injection dwell.
Fig. 15. Influence of injection timing with Mil100 on the combustion. Fig. 13. Influence of injection dwell on SFC variation. Percentile normalization along the reference point.
Fig. 14. Influence of injection timing with Mil80 on the combustion.
extended study was carried out with Mil80 and Mil100, and the latest injection timings were tested also with Mil110. Figs. 14 and 15 display the combustion curves of the cases with Mil80 and Mil100. Later timing decreases the in-cylinder maximum pressure and the premixed combustion magnitude. The diffusive part shifts later in the expansion stroke without any remarkable modi-
fication in shape and maximum value. In addition, in Mil100 cases (Fig. 15) with retarded fuel injection the in-cylinder pressure fluctuations diminish, reducing also the cycle-to-cycle variation [14]. Mil110 was tested with late injection timing. Figs. 16 and 17 show the combustion with two injection timings and the three Miller rates, considered in this paper. It is apparent that Mil100 and Mil110 provide similar results. Indeed, the combustion curves nearly overlap, especially in the premixed phase. In addition, the combustion starts at the same instant with Mil100 and Mil110, despite the different in-cylinder conditions at injection. Highmagnitude in-cylinder pressure fluctuations (hence in HRR) occur with Mil110, although later injection reduces these phenomena. Fig. 18 shows the NOx with each IVC. The cases with single injection (solid markers) demonstrate that increasing IVC from 80 to 110°CA BBDC (with late fuel injection) implies negligible NOx differences: compared to the reference point, 50% reduction is achieved. However, the values are higher than with the earlier injection timing and Mil100 (Fig. 11). As expected, later fuel injection results in lower NOx, but the trend is different depending on the Miller timing. In fact, with Mil80 the use of pilot injection does not bring to lower NOx compared to the single injection case. With the more advanced Miller rate, NOx is reduced by 5% with the use of split injection and by retarding the main event. There are unimportant differences in the results between Mil100 and Mil110, with 57% as the maximum achieved NOx reduction. The soot values are shown in Fig. 19 along the injection timing, indicated as SOPI. Since the dwell is constant, the SOMI occurs at
296
M. Imperato et al. / Applied Energy 162 (2016) 289–297
Fig. 16. Influence of Miller timing on the combustion (SOMI = 3°CA ATDC).
Fig. 19. The soot values along the SOPI.
Fig. 17. Influence of Miller timing on the combustion (SOMI = 5°CA ATDC). Fig. 20. Influence of injection timing on SFC. Single injection and split injection comparison. Percentile normalization along the reference point.
is 0.83FSN, which represents a reasonable result for these engine applications. The analysis of the fuel consumption variation is plotted in Fig. 20. Generally, later injection diminishes the fuel economy. The cases with a single injection and very advanced IVC present substantial drawbacks, up to 10% increase compared to the reference value. Although the NOx outcome is comparable, a remarkable difference in fuel consumption can be found between Mil80 and the other valve timings. In this respect, it is apparent that it is not convenient to use a very early IVC with a single late injection. Similarly to Fig. 13, the use of the pilot improves the engine fuel economy. With Mil80, the fuel consumption is nearly same as the reference (SOPI = 5°CA ATDC) with 42% NOx reduction. Fig. 18. Influence of injection timing on NOx emission. Single injection and split injection comparison. Percentile normalization along the reference point.
10°CA after SOPI (ex: SOPI = 8°CA ATDC, SOMI = 2°CA ATDC). Generally, pilot injection produces more soot than the single injection cases. Higher Miller rate generates more soot, and with Mil100 the trend is not monotonous, reaching its maximum outcome with SOPI = 7°CA ATDC. Comparing Figs. 14 with 15, it is apparent that the combustion is more regular with Mil80, indicating a linear trend. Unlike with Mil100, high fluctuations can be seen in the in-cylinder pressure and HRR curves, which characterize more unstable combustion, promoting conditions for soot formation. Indeed, with SOPI = 6°CA ATDC the combustion is more regular, inverting the FSN trend in Fig. 19. However, the maximum value
4. Conclusions and outlook The purpose of this research consisted in studying the effect of the injection strategy on the combustion process and engine performance. An experimental study with a single-cylinder largebore four-stroke medium-speed CI engine was carried out. Very advanced Miller rate, resulting in low in-cylinder temperature, was tested. This may lengthen the ignition delay and create enhanced premixed combustion. Hence, a fast in-cylinder temperature increase takes place, promoting NOx formation. A pilot dose, injected during the compression stroke, was employed to reduce the ignition delay and the magnitude of the premixed combustion. Specifically, three advanced IVCs (80, 100 and 110°CA BBDC) were tested with different pilot injection. The work was carried
M. Imperato et al. / Applied Energy 162 (2016) 289–297
out at partial load (BMEP ffi 11 bar) with the engine speed of 900 rpm. Also the injection pressure and injected mass were constant at each point. The boost pressure was increased as the IVC was advanced, in order to maintain the same air-to-fuel ratio. The study was divided in two parts. In the first stage of the study, the injection dwell between SOPI and SOMI was studied with two IVCs and three SOMIs. The dwell was modified within a range of 7–16°CA and the main injection started around TDC. In the second stage, the dwell was kept constant, retarding the whole injection phase later in order to reduce NOx. The reference point was set to obtain the specific NOx outcome close to the Tier II value. Hence, an evaluation of the effectiveness of the chosen strategies could be immediately performed. The conclusions of the study are: 1. Split injection generally increased engine efficiency. On the other hand, compared to single injection, no advantage in NOx emissions was observed. 2. The peak of the premixed combustion increases with larger dwell, but this is not stringently correlated to NOx outcomes. Indeed, the injection timing had important influence on emissions, when the same values of premixed combustion peaks were measured. 3. The optimum injection dwell was 10°CA in this study. With longer dwell and Mil100, both fuel consumption and NOx emissions increase. 4. NOx reduction of 42% could be obtained without a decrease in engine efficiency. When 4% SFC increase was allowed, 57% NOx reduction was possible. 5. Soot values (measured as FSN) were generally low, and shorter dwell resulted in lower FSN. The trend was in accordance with previous experiences with diesel engines, and it is probably related to the local in-cylinder temperature and oxygen content during the main combustion. The present study pointed out that it is not straightforward to use split injection together with Miller cycle. It was shown that the pilot timing, dwell length, and the main injection timing have to be carefully optimized in order to realize remarkable benefits. In addition, smaller pilot quantity may improve fuel consumption and achieve a good reduction in NOx. According to this work, the IMO Tier III legislation cannot be met by using Miller cycle and split injection alone. However, taken the considerable NOx reduction obtained (42%) without efficiency loss, the present techniques can be applied together with e.g. exhaust gas recirculation in order to meet the IMO III limits and keep the fuel consumption in control. Acknowledgements This study was funded by the European Commission within the Hercules C project. The authors want to thank all the framework partners for their active collaboration during the research. References [1] International Standard, ISO 8178-4. Reciprocating internal combustion engines – exhaust emission measurement, Genève; 1996. [2] Heywood JB. Internal combustion engine fundamentals. New York: McGrawHill; 1988. [3] Miller R, Lieberherr HU. The Miller supercharging system for diesel and gas engines operating characteristics. In: 4th CIMAC congress, Zurich, Switzerland; 1957. [4] Al-Sarkhi, Jaber JO, Probert SD. Efficiency of a Miller engine. Appl Energy 2006;83(4):343–51.
297
[5] Wang Y, Lin L, Zeng S, Huang J, Roskilly AP, He Y, et al. Application of the Miller cycle to reduce NOx emissions from petrol engines. Appl Energy 2008;85 (6):463–74. [6] Rinaldini CA, Mattarelli E, Golovitchev VI. Potential of the Miller cycle on a HSDI diesel automotive engine. Appl Energy 2013;112:102–19. [7] Wang Y, Zeng S, Huang J, He Y, Huang X, Lin L, et al. Experimental investigation of applying Miller cycle to reduce NOx emission from diesel engine. Proc Inst Mech Eng, Part A: J Power Energy 2005;219(A8):631–8. [8] Gonca G, Sahin B, Parlak A, Ust Y, Ayhan V, Cesur I, et al. Theoretical and experimental investigation of the Miller cycle diesel engine in terms of performance and emission parameters. Appl Energy 2015;138:11–20. [9] Codan E, Vlaskos I. Turbocharging medium speed diesel engines with extreme Miller timing. In: 9th Turbocharging conference, Dresden; 2004. [10] Plee SL, Ahmad T, Myers JP. Diesel NOx emissions – a simple correlation technique for intake air effects. In: 19th Symposium (international) on combustion, vol. 19, issue 1; 1982. pp. 1495–1502 [http://dx.doi.org/10. 1016/S0082-0784(82)80326-3]. [11] Imperato M, Nurmiranta J, Sarjovaara T, Larmi M, Wik C. Multi-injection and advanced Miller timing in large-bore CI engine. In: 27th CIMAC congress, paper no 157, Shanghai, China; 2013. [12] Kyrtatos P, Hoyer K, Obrecht P, Boulouchos K. Apparent effects of in-cylinder pressure oscillations and cycle-to-cycle variability on heat release rate and soot concentration under long ignition delay conditions in diesel engines. Int J Eng Res 2014;15(3):325–37. [13] Kyrtatos P, Hoyer K, Obrecht P, Boulouchos K. Recent developments in the understanding of the potential of in-cylinder NOx reduction through extreme Miller valve timing. In: 26th CIMAC congress, paper no 225, Shanghai, China; 2013. [14] Kyrtatos P. The effects of prolonged ignition delay due to charge air temperature reduction on combustion in a diesel engine. Dissertation, Swiss Federal Institute of Technology Zurich, no. 21064; 2013. http://dx.doi.org/10. 3929/ethz-a-009933871. [15] Musculus, M. On the correlation between NOx emissions and the diesel premixed burn. SAE 2004–01-1401; 2004. http://dx.doi.org/10.4271/2004-011401. [16] Liu H, Ma S, Zhang Z, Zheng Z, Yao M. Study of the control strategies on soot reduction under early-injection conditions on a diesel engine. Fuel 2015;139:472–81. [17] Park SH, Yoon SH. Injection strategy for simultaneous reduction of NOx and soot emissions using two-stage injection in DME fueled engine. Appl Energy 2015;143:262–70. [18] Zheng Z, Tian X, Zhang X. Effects of split injection proportion and the second injection time on the mixture formation in a GDI engine under catalyst heating mode using stratified charge strategy. Appl Therm Eng 2015;84:237–45. [19] Brückner C, Kyrtatos P, Boulouchos K. Extending the NOx reduction potential with Miller valve timing using pilot fuel injection on a heavy-duty diesel engine. SAE Int J Engine 2014;7(4):1838–50. http://dx.doi.org/10.4271/201401-2632. [20] Andreadis P, Zompanakis A, Chryssakis C, Kaiktsis L. Effect of the fuel injection parameters and emissions formation in a large-bore marine diesel engine. Int J Engine Res 2011;12:14–29. [21] Kallio I, Rantanen P, Imperato M, Antila E, Sarjovaara T, Larmi M, Huhtala K, Liljenfeldt G. The design and operation of the fully controllable medium-speed research engine EVE. In: 25th CIMAC congress, paper no. 163, Wien, Austria; 2007. [22] Kaario O, Imperato M, Tilli A, Lehto K, Ranta O, Antila E, et al. The design of a new generation medium-speed research engine. In: 26th CIMAC congress, paper no. 145, Bergen, Norway; 2010. [23] Imperato M, Antila E, Sarjovaara T, Kaario O, Larmi M, Kallio I, et al. NOx reduction in a medium-speed single-cylinder diesel engine using Miller cycle with very advanced valve timing. SAE 2009-240112; 2009. [24] Imperato M, Kaario O, Sarjovaara T, Larmi M. Influence of the in-cylinder gas density and fuel injection pressure on the combustion characteristics in a large-bore diesel engine. Int J Eng Res 2015;7(1):1–9. http://dx.doi.org/ 10.1177/1468087415589043. [25] Herranen M, Huhtala K, Villenius M, Liljenfeldt G. The electro-hydraulic valve actuation for medium speed diesel engines – development steps with simulation and measurements. SAE 2007-011289; 2007. [26] Egnell R. Combustion diagnostics by means of multizone heat release analysis and NO calculation. SAE 981424; 1998. [27] Sridharan S, Rutland C. Model-based diesel HCCI combustion phasing controller in integrated system level modeling. SAE 2010–01-0886; 2010. [28] Harrison J, Aihara R, Eshraghi M, Dmitrieva I. Modeling engine oil variable displacement vane pumps in 1D to predict performance, pulsations, and friction. SAE 2014–01-1086; 2014. [29] Warnatz J, Maas U, Dibble RW. Combustion. 4th ed. New York: Springer; 2006, ISBN 978-3-540-67751-2. [30] Ando H, Sakai Y, Kuwahara K. Universal rule of hydrocarbon oxidation. SAE 2009–01-0948; 2009. [31] Herfatmanesh MR. Investigation of single and split injection strategies in an optical diesel engine. Dissertation, Brunel University School of Engineering and Design; 2010 [http://bura.brunel.ac.uk/handle/2438/4776].
Contents lists available at ScienceDirect
Applied Energy journal homepage: www.elsevier.com/locate/apenergy
Split fuel injection and Miller cycle in a large-bore engine Matteo Imperato ⇑, Ossi Kaario, Teemu Sarjovaara, Martti Larmi Aalto University, Finland
h i g h l i g h t s High premixed combustion peak is reduced with pilot injection. Split injection in large bore engines increases engine efficiency. Pilot injection with high Miller rate can reduce NOx up to 60%. NOx reduction of 42% could be obtained without decrease in engine efficiency.
a r t i c l e
i n f o
Article history: Received 25 June 2015 Received in revised form 25 September 2015 Accepted 5 October 2015
Keywords: Large-bore engine Split injection NOx reduction Miller cycle Pilot injection Diesel combustion
a b s t r a c t The upcoming emission legislation for sea-going vessels issued by the international marine organization requires drastic reduction in nitric oxides. A well-known approach for meeting these requirements is to reduce the in-cylinder temperature prior to combustion by using the so-called Miller cycle. However, the mere use of this technique presents the actual limits due to long ignition delay, which occurs when the compression temperature is very low. As a consequence, premixed combustion develops quickly, increasing the local temperature in the combustion chamber and favoring NOx formation. Splitting the fuel injection into a small pilot and a main injection can reduce the magnitude of the premixed combustion and the local in-cylinder temperatures. The work presented here is divided in two parts and is novel by being the first systematic study of split injection combined with Miller cycle in large-bore engines. In its first stage, an extensive study of the injection dwell with two intake valve closings and three timings of the main injection are analyzed. In the second stage, both injection events are shifted later in the power stroke with fixed injection dwell. Overall, the pilot injection reduced the ignition delay but dropped the peak of the premixed combustion only with the most advanced intake valve closing. This improved fuel economy, but provided no advantages as far as emissions are concerned. In addition, while increasing injection dwell reduced NOx emissions, it also increased fuel consumption. The highest achieved NOx reduction was close to 60%, with a small drawback in fuel economy. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction The emission limits for ships’ diesel engines are becoming constantly stricter. In 2016, a new regulation, which reduces nitric oxi-
Abbreviations: °CA, crank angle degree; ATDC, after top dead center; BBDC, before bottom dead center; BDC, bottom dead center; BMEP, brake mean effective pressure; BTDC, Before Top Dead Center; CI, compression–ignition; CO, carbon oxide; HR, heat release; HRR, heat release rate; IMEP, indicated mean effective pressure; IVC, intake valve closing; LTO, low temperature oxidation; MI, main injection; NDIR, non-dispersive infrared; NOx, nitrogen oxides; NTC, negative temperature coefficient; PI, pilot injection; SFC, specific fuel consumption; SOMI, start of main injection; SOPI, start of pilot injection; TDC, top dead center. ⇑ Corresponding author at: Department of Mechanical Engineering, Aalto University, 5 Puumiehenkuja, Aalto 00076, Espoo, Finland. E-mail address: [email protected] (M. Imperato). http://dx.doi.org/10.1016/j.apenergy.2015.10.041 0306-2619/Ó 2015 Elsevier Ltd. All rights reserved.
des (NOx) emission by 80% – compared to 2001 level – for largebore compression–ignition (CI) engines [1], comes into effect. Therefore, the development of new solutions and technologies for reducing emissions is mandatory, in order to fulfil the stringent demands of the legislation. However, the engine performance should not be negatively affected by the new adopted configurations. NOx formation occurs in the combustion chamber when the local gas temperature is high [2]. An effective way to reduce the combustion temperature in CI engines is the so-called Miller cycle [3]. This can be realized by closing the intake valve before bottom dead center (BBDC), so that the effective compression ratio is lower than the effective expansion ratio. Thus, the compression incylinder temperature and the eventual combustion temperature decrease. Miller cycle has been theoretically studied [4], and it is
290
M. Imperato et al. / Applied Energy 162 (2016) 289–297
widely used in automotive [5,6], heavy-duty [7,8] and large-bore engines [9]. The Miller cycle approach is based on the correlation between NOx formation and the flame temperature. The approach states that, with reduced flame temperature, NOx emissions are expected to be lower [2,10]. As mentioned, the NOx formation kinetics is strongly temperature-dependent. However, recent investigation has reported several cases where the flame temperature correlation fails to capture the trend of NOx emissions for diesel combustion. A study demonstrated that at low load [11] increasing the Miller rate increases NOx close to the reference value with standard valve timing. In addition, it was shown that ignition delay becomes very long with advanced Miller rate, and the consequent high in-cylinder pressure fluctuations result in increased NOx values, despite the low in-cylinder temperature at ignition [12]. In addition, because of the lower compression temperature a higher Miller rate increases the ignition delay. Kyrtatos et al. [13,14] tested very advanced Miller timing in a large-bore mediumspeed engine and observed that a longer ignition delay – due to lower in-cylinder temperature – allowed more fuel being mixed prior to ignition, leading to a greater magnitude of premixed combustion. Moreover, the long ignition delay let the spray penetrate further, enhancing air entrainment and resulting in an overall leaner premixed combustion that could favor NOx formation. This might also result in unstable combustion and high cycle-to-cycle variation of the in-cylinder pressure, which could compromise the engine’s proper function and performance [12]. The amount of the premixed combustion may be correlated with the formation of NOx. Generally, a high peak of premixed combustion results in high NOx outcomes. Musculus [15] realized optical measurements in a diesel engine and found contradicting trends. It was reported that, reducing the intake temperature, NOx initially decreased and then increased monotonically with premixed combustion. Consistently, it was found that, for operating conditions with large premixed ratios, retardation of injection eventually led to an increase of NOx after reaching a minimum. It was concluded that large portions of premixed combustion generally led to faster combustion, with high increase in cylinder pressure early in the engine cycle. Thus, the reactants entering the diffusion flame are compressed to higher temperatures, leading to higher flame temperatures (compression heating). Split injection has been widely used in diesel engines for reducing both NOx and soot [16–18]. Injecting a small pilot dose with advanced intake valve closing (IVC) can trigger the combustion earlier before top dead center (BTDC), reducing also the ignition delay of the main injection. This is expected to reduce the magnitude of the premixed combustion and NOx. Brückner et al. [19] studied the influence of fuel pilot injection and Miller cycle in a heavy-duty diesel engine running at 1000 rpm and demonstrated that the pilot injection decreased the peak of the premixed combustion, reducing further NOx only in conditions of very low incylinder temperature. Pilot injection has not been extensively studied in marine engines. On a two-stroke marine engine, Andreadis et al. [20] tested pilot injection, achieving a NOx reduction of 15% and also a slight improvement in the engine economy. However, the study did not investigate any influence of Miller timing combined with split injection. Nonetheless, systematic studies of the use of the Miller cycle and split injection in large-bore engines cannot be found in the literature. The present research was carried out with a single-cylinder large-bore four-stroke mediumspeed research engine [21,22]. Former achievements include 40% NOx reduction at partial load, adopting Miller cycle [23]. Recently, split injection and Miller cycle running at partial load with 1500 bar injection pressure and with a pilot injection fuel fraction of 25% were attempted for the first time [11]. In those runs, pilot injection
timing was tested either very close or very far from the main injection. Hence, the influence of the split injection on the combustion characteristics was not completely understood. The novelty of this research consists of testing split injection and Miller, running with low fuel injection pressure, and investigating the combustion behavior at regular injection dwell steps. The main aim was to reduce the ignition delay when running with advanced Miller rate and to study the effect of the injection strategy on the combustion process and performance outcomes, especially that of NOx. The present study is divided in two parts. In the first part, different injection dwells are tested. Then, with a fixed dwell, the injection events are retarded to reduce the NOx amount, while attempting to fulfil the upcoming legislation requirements. 2. Material and methods 2.1. Experimental setup A single-cylinder large-bore CI medium speed engine was employed in this research: its main components were designed to withstand in-cylinder pressure of 30 MPa running and mean piston speed of 12 m/s [22,24]. The engine was connected to an electric motor, which also allowed running in the motored mode. In addition, the gas exchange valves were controlled by electrohydraulic valve actuators (EHVA) instead of the traditional mechanical camshaft, providing high flexibility in valve timing [25] for the gas exchange phase. Moreover, the fuel system permitted changing of injection pressure, duration and timing. An external charge-air supply plant and an exhaust gas throttle valve controlled the boundary conditions: the maximum allowed intake air pressure was 10 bar. The parameters of injection, valve timing and ancillary systems could be freely set and remotely monitored [21]. A schematic of the engine is shown in Fig. 1. The charge air is processed by the compressor (C); then, it is cooled, dried and heated before going through the flow meter. The air reservoirs reduce the pressure fluctuations and the intake pressure is regulated by a control valve. The engine (E) is connected to the electric motor (M). The exhaust gas pressure is controlled by the throttle valve, and exhaust gas samples are diverted to the emission analyzers and to the smoke meter (AVL 415). The engine was equipped with Kistler in-cylinder pressure sensor 6045AU20 connected to the Kistler charge amplifier 5064. An Emerson Micromotion ELITE CMF100 flow meter was employed for charge air measurement, and an Emerson Micromotion ELITE CMF025 flow meter for the measurement of fuel mass flow before the engine. The emission measurement system was composed of different analyzers for each gas. NOx emissions were measured with an ECO Physics CLD822Sh analyzer. Soot emissions were measured as filter smoke number (FSN) by an AVL 415 smoke meter. A non-dispersive infrared (NDIR) absorption analyzer by Sick was employed to measure carbon oxide (CO). The emission probes were located in the exhaust pipe five meters from the cylinder head
Fig. 1. Schematic of the engine test bed.
291
M. Imperato et al. / Applied Energy 162 (2016) 289–297 Table 1 Main dimensions and engine configuration. Displaced volume Stroke Bore Connecting rod length Compression ratio Injection nozzle tip
(cm3) (mm) (mm) (mm) – (#Holes £ mm)
8796 280 200 614.2 16.7:1 9 0.30
outlet, and the gases were transported in a heated line (191 °C) to avoid condensation. Table 1 presents the main dimensions of the engine. The compression ratio was high to reduce fuel consumption. The injector nozzle tip holes were reasonably small, to obtain good spray properties [2] without exceeding the injection duration. 2.2. Calculation methods The specific fuel consumption (SFC) and the NOx were normalized, with 100% corresponding to the reference point. The reference point was run with a single injection, and in particular the specific NOx outcome was close to the IMO Tier II value at 900 rpm. The heat release rate (HRR) was calculated using the in-cylinder pressure and the cylinder volume along °CA. The apparent HRR equation [2] below was adopted:
dQ c dV 1 dp þ ¼ p V d CA c 1 d CA d CA c 1
ð1Þ
where Q is the heat released, p the in-cylinder pressure, V the incylinder volume and c the specific heat ratio. The variable c is temperature-dependent and an exponential correspondence was chosen for determining its value in the calculation [26]. The HRR calculation started at 40°CA before top dead center (BTDC) and terminated at 50°CA ATDC in the power stroke. The ignition delay could be evaluated from the start of injection (SOI) and the instant when the HRR curve departs from zero [2]. However, because of measurement errors and unstable combustion with some setups, the ignition delay was determined as the time between the SOI and the time when the heat release (HR) reached the value of 0.5 kJ, corresponding to energy content of 12 mg fuel (<2% of the total injected fuel). Although the results might be overestimated, clear trends of the different strategies chosen in this research were expected. In the two-injection cases, the ignition delay was considered along with the pilot injection. A simulation model was employed to attain the values of incylinder temperature at the injection start. Temperature is a crucial parameter for evaluating the ignition delay [2], but it is not possible to calculate it directly from the test results. Thus, a simulation model was developed with GT-Power, 1-D fluid dynamic code powered by Gamma Technologies. GT-Power has been extensively used in diesel engine research [27,28], and a detailed model had been already applied to this engine [23]. In particular, the Redlich-Kwong [2] real gas law was employed for the combustion chamber. Since the simulation was used for the evaluation of the in-cylinder temperature prior to combustion, the model was calibrated with the engine inlet and outlet conditions obtained from the test runs. Only motored cycles were run.
the injection dwell was constant and both the injections were retarded. Four valve timings were tested in this study, and the details can be seen in Fig. 2. The IVC of the reference case was 30°CA BBDC, and this parameter was advanced until 110°CA BBDC. When changing the valve timing, only IVC was modified, whilst the exhaust valve timing and the intake valve opening were kept unchanged. As shown in Fig. 2, the maximum valve lift was reduced as the IVC was advanced. Hence, the closing slope of the intake valve phase was augmented with the configuration Mil100 and Mil110, to avoid undue reduction of the maximum lift value. The injection parameters are defined in Fig. 3. It must be pointed out that the injection duration referred to the electric signal of the injector, since the injection valve was not instrumented. In particular, the injection dwell is intended as the time between the start of the pilot injection (SOPI) and the SOMI. Because of the mechanics of the fuel injection valve, the smallest stable injection opening time was set to 0.93 ms, which corresponded to 5°CA. Hence, the minimum injection dwell was set to 7°CA, with 2°CA tolerance to permit good and steady operation of the injection system. The pilot injection duration was 33% of the total injection time, and this value should be a good approximation of the relative pilot fuel ratio over the global injection quantity. Since the study was focused on valve timing and injection timing, some parameters were constant in each test point, and they are reported in Table 2. All the tests were carried out at partial load running at 900 rpm. Injected fuel mass flow and air-to-fuel ratio were also constant in all the tests. In particular, the injected fuel
Fig. 2. Gas exchange valve lift curves.
2.3. Test settings The research had two parts, each with different gas exchange valve timing. In the first stage of the research, the injection dwell was tested by modifying the time between the injections, together with three starts of the main injection (SOMI). In the second stage,
Fig. 3. Injection nomenclature.
292
M. Imperato et al. / Applied Energy 162 (2016) 289–297
Table 2 Constant engine parameters. Engine power
Engine speed Engine load (BMEP)
(rpm) (bar)
900 11 ± 0.4
Injection parameters
Total injection quantity Injection duration (single) Pilot duration Injection pressure
(mg/cyc) (°CA) (°CA) (bar)
700 ± 14 15 5 1200
In-cylinder conditions
Density at TDC Total k
(kg/m3) (#)
40 2.1
mass per cycle amounted to 700 mg (with a 2% error range), and the total k was 2.1, resulting in 40 kg/m3 charge density at the end of the compression stroke. Since the intake valve opening time was reduced in Miller cases, it was necessary to increase the boost pressure values in order to achieve the same global air-to-fuel ratio. The main injection signal was of 15°CA, and the fuel injection pressure was 1200 bar, reasonably lower than in the former study. The complete test matrix is shown in Table 3. The engine setup name was chosen along the IVC. In the first part of the work, a detailed study of the injection dwell was carried out with Mil80 and Mil100. The second part of the work aimed to reduce the NOx by retarding both the injection events and increasing the Miller rate. The injection dwell was constant, and the comparison with the single injection cases was also carried out. Table 3 shows also the simulation results at top dead center (TDC). In the reference point, the IVC was set close to the bottom dead center (BDC) and the motored temperature from simulation was close to 1000 K. Temperature reduction of 100 K was obtained by advancing IVC of 50°CA, and, with the most advanced timing, the temperature reduction at TDC was almost 200 K compared to the reference. Although these values show the influence of the Miller timing on the in-cylinder temperature, it must be considered that in most of the test points the combustion starts before the TDC. Hence, the in-cylinder temperature at the injection instant will be taken into account later in the paper.
Fig. 4. Influence of Miller and pilot injection on the combustion (early SOMI).
pressure increases without remarkable changes in the magnitude of the premixed combustion, which occurs earlier because of earlier injection timing. Controversially, very early pilot injection results in high premixed combustion and two-fold increase of its maximum value. The fuel injected during the pilot injection completed its combustion and, as a consequence, the diffusive combustion has a lower magnitude. It must be pointed out that the overall injected fuel mass is the same in all the run cases (Table 2). Fig. 5 shows the influence of all four tested injection dwells with timing Mil80 and late SOMI. The premixed combustion increases with earlier pilot injection timing, and the diffusive combustion has lower magnitude. Here, the peak pressure occurs later
3. Test results 3.1. Effect of the injection dwell Fig. 4 shows the effect of Miller timing and pilot injection on the combustion process. The in-cylinder compression pressure is lower with Miller timing, because the same air mass quantity is compressed at a lower temperature. Miller cycle increases ignition delay and premixed combustion magnitude. In addition to the cases with single injection, also two cases with pilot are plotted in the same figure. With short injection dwell, the in-cylinder peak
Fig. 5. Influence of injection dwell with Mil80 on the combustion (late SOMI).
Table 3 Test matrix. Engine setup
Injection strategy
IVC (°CA bBDC)
Simulated motored temp. at TDC (K)
SOMI (°CA aTDC)
Inj. dwell (°CA)
Relative pilot quantity (%)
Reference Mil80 Mil100
Single injection
30 80 100
998 904 855
2
–
0
Mil80 Mil100
Split injection
80 100
904 855
2, 0, 2
7, 10, 13, 16
33
Mil80 Mil100 Mil110
Single injection
80 100 110
904 855 807
3
–
0
Mil80 Mil100 Mil110
Split injection
80 100 110
904 855 807
3, 4, 5
10
33
M. Imperato et al. / Applied Energy 162 (2016) 289–297
and it is lower. In comparison, the tests with earlier SOMI (Fig. 4) display similar in-cylinder peak pressure regardless of the pilot timing: with earlier timing, the pilot combustion takes place mostly during the compression stroke, augmenting to a greater extent the in-cylinder pressure. In Fig. 6, the cumulative heat release is plotted. It can be seen that the late part of combustion was similar regardless of the pilot injection timing. In fact, all the curves match close to 9 kJ, which is slightly lower than 50% of the total HR. Fig. 7 depicts the effect of Miller timing and pilot injection. The increased Miller rate lengthens the ignition delay and enhances the premixed combustion. Its peak is decreased using late pilot injection (dwell = 7°CA), although it is apparent that the overall combustion develops mainly around the TDC, increasing the incylinder pressure level. Similarly to Fig. 4, a large dwell resulted in high peak of the premixed combustion, and the maximum HRR value is 3 kJ/°CA. In this case, the pilot and main injection combustion are clearly distinct. With a more advanced Miller rate, the ignition delay increases and, accordingly, the peak of premixed combustion. Fig. 8 represents the combustion curve of the test points with Mil100 and SOMI = 2°CA ATDC. Here, the premixed phase is of a higher magnitude with the injection setup similar to that of Mil80 (Fig. 5). Because of colder in-cylinder temperature, the pressure fluctuations are larger, especially with the earlier pilot. As a summary of the combustion analysis, in Fig. 9 the peak heat release rate of the premixed and diffusive combustion are plotted as injection dwell changes. High heat release rate might imply high
Fig. 6. Influence of injection dwell with Mil80 on the cumulative HR (late SOMI).
Fig. 7. Influence of Miller and injection dwell on the combustion.
293
Fig. 8. Influence of injection dwell with Mil100 on the combustion (late SOMI).
in-cylinder temperature, and this could be correlated with NOx formation [2]. As shown previously, earlier pilot injection – i.e. larger dwell – increases the maximum value of the heat release rate in the premixed phase. In particular, earlier IVC implies a large increase in the HRR, and the major differences occur with larger dwell. In addition with Mil100, the peak of the HRR in the premixed combustion converges to the same value, as the dwell is reduced. On the other hand, with Mil80 the values are different with injection timing for each tested dwell. However, the values with Mil100 are in all cases higher than Mil80, due to the longer ignition delay. In the diffusive combustion, the maximum value of the HRR increases as the dwell is reduced. This can be explained as an effect of the short injection dwell: the combustion of the pilot dose is not completed at SOMI. While in the premixed combustion the maximum HRR increases as SOPI is advanced, the trend of the maximum value of diffusive combustion as SOMI changes is not well-defined. In fact, the differences of the absolute values are small and mainly within the measurement error range. As stated in the literature [2], ignition delay generally increases with lower air temperature. Fig. 10 shows the measured ignition delay plotted with the simulated in-cylinder temperature at the injection. The pilot ignition occurs before SOMI in all the cases. Generally, early pilot injection shortens the ignition delay compared with single injection cases. In addition, one can notice that different trends are measured with the different valve timings. Indeed, with Mil80 all the points lie on the same curve regardless of the injection timing. One can observe that the same values of ignition delay are measured with in-cylinder temperature at injection in range 860–890 K. In comparison, with Mil100 the incylinder temperature is cooler and the injection timing affects more the ignition delay. For instance, its value changes from 0.9 to 1.2 ms at temperature of 800 K. At low temperature the hydrocarbon oxidation, and eventually the ignition delay, is governed earlier by a low temperature oxidation (LTO) range and later by a negative temperature coefficient (NTC) range [29,30]. In the latter, the ignition delay increases with increasing temperature. With Mil100, the pilot event is injected between 750 and 840 K, which is reasonable in LTO-NTC range. However, it must be kept in mind the experiments above mentioned [29,30] were achieved in conditions of perfect air–fuel premixing, and hence the results might differ from these experiments. The influence of the injection dwell on the NOx outcomes is plotted in Fig. 11. As expected, in single injection cases high Miller rate brings a drastic reduction of NOx, over 50% with Mil100 timing. Split injection does not yield further NOx reduction, when compared with the single injection cases. As shown in Figs. 5 and 8, early pilot injection creates large premixed combustion, but the main combustion starts when the pilot fuel is not completely
294
M. Imperato et al. / Applied Energy 162 (2016) 289–297
Fig. 9. The peak heat release rate in premixed and diffusion combustion.
Fig. 10. The in-cylinder simulated temperature at injection instant and the ignition delay.
Fig. 11. Influence of injection dwell on NOx emission. Percentile normalization along the reference point.
burnt. In addition, the effect of the injection dwell changes with the Miller rate. Indeed, with Mil80 timing, larger dwell implies lower NOx. Nonetheless, with Mil100 the injection dwell has minor effects on NOx outcomes, registering the lowest value with dwell = 10°CA. As stated by Musculus [15], both a large premixed combustion peak and great heat released in diffusive combustion increase the
in-cylinder temperature, which may form high NOx. With these assumptions and according to the NOx outcomes, the runs with dwell = 10°CA indicate the best trade-off between these two phenomena. Herfatmanesh [31] achieved similar results in a study of dwell angle with an optical engine at 1500 rpm. Although the application was different from the one here, the NOx trend was not monotonous and minimum values were attained similarly with dwell = 10°CA. The results concerning soot and CO are presented in Fig. 12. In all the cases, seen objectively, the absolute values are low, considering that the tests were carried out at partial load. Generally, soot (as FSN) increases with CO. Split injection with large dwell increases both, especially when running with Mil100. Lower CO is also obtained with a small dwell. As shown in Figs. 5 and 8, with a large injection dwell the pilot combustion is extinguished before SOMI; with a narrow dwell, on the contrary, the main combustion commences before the pilot combustion is completed. Thus, the evaluation of the start of the injection of the main combustion is not possible with sufficient accuracy. From the FSN results, it can be seen that later injection – i.e. shorter dwell – reduces the amount of soot in the exhaust gases, and this is in accordance with previous experiences with diesel engines [16–18]. Fig. 13 shows SFC, normalized to the reference point. Split injection results in lower SFC, when compared with the single injection cases, because of the remarkably higher in-cylinder pressure around TDC (Figs. 4 and 7). In addition, larger dwell angle denotes higher fuel consumption for all the tested configurations. This might be due to the high amount of heat released in the compression stroke, which represents negative work of the engine cycle. A narrower dwell angle shifts the combustion phase toward the expansion stroke, improving the engine economy. Considering the same injection dwell, earlier injection is more favorable for decreasing fuel consumption. Summing up the results shown in Figs. 11 and 13, the general trend is such that the larger dwell brought to lower NOx values but higher SFC. However, the configuration dwell = 10°CA was chosen for the second part of the study, since there was a certain fuel economy decrease with a reasonable NOx reduction. Shifting the injection events later was expected to decrease the amount of NOx when keeping the SFC value at the same level of the reference point.
3.2. Effect of the injection timing with constant dwell As shown in Table 3, three different Miller rates were tested with late fuel injection and constant dwell = 10°CA. A more
M. Imperato et al. / Applied Energy 162 (2016) 289–297
295
Fig. 12. The soot and CO values along the injection dwell.
Fig. 15. Influence of injection timing with Mil100 on the combustion. Fig. 13. Influence of injection dwell on SFC variation. Percentile normalization along the reference point.
Fig. 14. Influence of injection timing with Mil80 on the combustion.
extended study was carried out with Mil80 and Mil100, and the latest injection timings were tested also with Mil110. Figs. 14 and 15 display the combustion curves of the cases with Mil80 and Mil100. Later timing decreases the in-cylinder maximum pressure and the premixed combustion magnitude. The diffusive part shifts later in the expansion stroke without any remarkable modi-
fication in shape and maximum value. In addition, in Mil100 cases (Fig. 15) with retarded fuel injection the in-cylinder pressure fluctuations diminish, reducing also the cycle-to-cycle variation [14]. Mil110 was tested with late injection timing. Figs. 16 and 17 show the combustion with two injection timings and the three Miller rates, considered in this paper. It is apparent that Mil100 and Mil110 provide similar results. Indeed, the combustion curves nearly overlap, especially in the premixed phase. In addition, the combustion starts at the same instant with Mil100 and Mil110, despite the different in-cylinder conditions at injection. Highmagnitude in-cylinder pressure fluctuations (hence in HRR) occur with Mil110, although later injection reduces these phenomena. Fig. 18 shows the NOx with each IVC. The cases with single injection (solid markers) demonstrate that increasing IVC from 80 to 110°CA BBDC (with late fuel injection) implies negligible NOx differences: compared to the reference point, 50% reduction is achieved. However, the values are higher than with the earlier injection timing and Mil100 (Fig. 11). As expected, later fuel injection results in lower NOx, but the trend is different depending on the Miller timing. In fact, with Mil80 the use of pilot injection does not bring to lower NOx compared to the single injection case. With the more advanced Miller rate, NOx is reduced by 5% with the use of split injection and by retarding the main event. There are unimportant differences in the results between Mil100 and Mil110, with 57% as the maximum achieved NOx reduction. The soot values are shown in Fig. 19 along the injection timing, indicated as SOPI. Since the dwell is constant, the SOMI occurs at
296
M. Imperato et al. / Applied Energy 162 (2016) 289–297
Fig. 16. Influence of Miller timing on the combustion (SOMI = 3°CA ATDC).
Fig. 19. The soot values along the SOPI.
Fig. 17. Influence of Miller timing on the combustion (SOMI = 5°CA ATDC). Fig. 20. Influence of injection timing on SFC. Single injection and split injection comparison. Percentile normalization along the reference point.
is 0.83FSN, which represents a reasonable result for these engine applications. The analysis of the fuel consumption variation is plotted in Fig. 20. Generally, later injection diminishes the fuel economy. The cases with a single injection and very advanced IVC present substantial drawbacks, up to 10% increase compared to the reference value. Although the NOx outcome is comparable, a remarkable difference in fuel consumption can be found between Mil80 and the other valve timings. In this respect, it is apparent that it is not convenient to use a very early IVC with a single late injection. Similarly to Fig. 13, the use of the pilot improves the engine fuel economy. With Mil80, the fuel consumption is nearly same as the reference (SOPI = 5°CA ATDC) with 42% NOx reduction. Fig. 18. Influence of injection timing on NOx emission. Single injection and split injection comparison. Percentile normalization along the reference point.
10°CA after SOPI (ex: SOPI = 8°CA ATDC, SOMI = 2°CA ATDC). Generally, pilot injection produces more soot than the single injection cases. Higher Miller rate generates more soot, and with Mil100 the trend is not monotonous, reaching its maximum outcome with SOPI = 7°CA ATDC. Comparing Figs. 14 with 15, it is apparent that the combustion is more regular with Mil80, indicating a linear trend. Unlike with Mil100, high fluctuations can be seen in the in-cylinder pressure and HRR curves, which characterize more unstable combustion, promoting conditions for soot formation. Indeed, with SOPI = 6°CA ATDC the combustion is more regular, inverting the FSN trend in Fig. 19. However, the maximum value
4. Conclusions and outlook The purpose of this research consisted in studying the effect of the injection strategy on the combustion process and engine performance. An experimental study with a single-cylinder largebore four-stroke medium-speed CI engine was carried out. Very advanced Miller rate, resulting in low in-cylinder temperature, was tested. This may lengthen the ignition delay and create enhanced premixed combustion. Hence, a fast in-cylinder temperature increase takes place, promoting NOx formation. A pilot dose, injected during the compression stroke, was employed to reduce the ignition delay and the magnitude of the premixed combustion. Specifically, three advanced IVCs (80, 100 and 110°CA BBDC) were tested with different pilot injection. The work was carried
M. Imperato et al. / Applied Energy 162 (2016) 289–297
out at partial load (BMEP ffi 11 bar) with the engine speed of 900 rpm. Also the injection pressure and injected mass were constant at each point. The boost pressure was increased as the IVC was advanced, in order to maintain the same air-to-fuel ratio. The study was divided in two parts. In the first stage of the study, the injection dwell between SOPI and SOMI was studied with two IVCs and three SOMIs. The dwell was modified within a range of 7–16°CA and the main injection started around TDC. In the second stage, the dwell was kept constant, retarding the whole injection phase later in order to reduce NOx. The reference point was set to obtain the specific NOx outcome close to the Tier II value. Hence, an evaluation of the effectiveness of the chosen strategies could be immediately performed. The conclusions of the study are: 1. Split injection generally increased engine efficiency. On the other hand, compared to single injection, no advantage in NOx emissions was observed. 2. The peak of the premixed combustion increases with larger dwell, but this is not stringently correlated to NOx outcomes. Indeed, the injection timing had important influence on emissions, when the same values of premixed combustion peaks were measured. 3. The optimum injection dwell was 10°CA in this study. With longer dwell and Mil100, both fuel consumption and NOx emissions increase. 4. NOx reduction of 42% could be obtained without a decrease in engine efficiency. When 4% SFC increase was allowed, 57% NOx reduction was possible. 5. Soot values (measured as FSN) were generally low, and shorter dwell resulted in lower FSN. The trend was in accordance with previous experiences with diesel engines, and it is probably related to the local in-cylinder temperature and oxygen content during the main combustion. The present study pointed out that it is not straightforward to use split injection together with Miller cycle. It was shown that the pilot timing, dwell length, and the main injection timing have to be carefully optimized in order to realize remarkable benefits. In addition, smaller pilot quantity may improve fuel consumption and achieve a good reduction in NOx. According to this work, the IMO Tier III legislation cannot be met by using Miller cycle and split injection alone. However, taken the considerable NOx reduction obtained (42%) without efficiency loss, the present techniques can be applied together with e.g. exhaust gas recirculation in order to meet the IMO III limits and keep the fuel consumption in control. Acknowledgements This study was funded by the European Commission within the Hercules C project. The authors want to thank all the framework partners for their active collaboration during the research. References [1] International Standard, ISO 8178-4. Reciprocating internal combustion engines – exhaust emission measurement, Genève; 1996. [2] Heywood JB. Internal combustion engine fundamentals. New York: McGrawHill; 1988. [3] Miller R, Lieberherr HU. The Miller supercharging system for diesel and gas engines operating characteristics. In: 4th CIMAC congress, Zurich, Switzerland; 1957. [4] Al-Sarkhi, Jaber JO, Probert SD. Efficiency of a Miller engine. Appl Energy 2006;83(4):343–51.
297
[5] Wang Y, Lin L, Zeng S, Huang J, Roskilly AP, He Y, et al. Application of the Miller cycle to reduce NOx emissions from petrol engines. Appl Energy 2008;85 (6):463–74. [6] Rinaldini CA, Mattarelli E, Golovitchev VI. Potential of the Miller cycle on a HSDI diesel automotive engine. Appl Energy 2013;112:102–19. [7] Wang Y, Zeng S, Huang J, He Y, Huang X, Lin L, et al. Experimental investigation of applying Miller cycle to reduce NOx emission from diesel engine. Proc Inst Mech Eng, Part A: J Power Energy 2005;219(A8):631–8. [8] Gonca G, Sahin B, Parlak A, Ust Y, Ayhan V, Cesur I, et al. Theoretical and experimental investigation of the Miller cycle diesel engine in terms of performance and emission parameters. Appl Energy 2015;138:11–20. [9] Codan E, Vlaskos I. Turbocharging medium speed diesel engines with extreme Miller timing. In: 9th Turbocharging conference, Dresden; 2004. [10] Plee SL, Ahmad T, Myers JP. Diesel NOx emissions – a simple correlation technique for intake air effects. In: 19th Symposium (international) on combustion, vol. 19, issue 1; 1982. pp. 1495–1502 [http://dx.doi.org/10. 1016/S0082-0784(82)80326-3]. [11] Imperato M, Nurmiranta J, Sarjovaara T, Larmi M, Wik C. Multi-injection and advanced Miller timing in large-bore CI engine. In: 27th CIMAC congress, paper no 157, Shanghai, China; 2013. [12] Kyrtatos P, Hoyer K, Obrecht P, Boulouchos K. Apparent effects of in-cylinder pressure oscillations and cycle-to-cycle variability on heat release rate and soot concentration under long ignition delay conditions in diesel engines. Int J Eng Res 2014;15(3):325–37. [13] Kyrtatos P, Hoyer K, Obrecht P, Boulouchos K. Recent developments in the understanding of the potential of in-cylinder NOx reduction through extreme Miller valve timing. In: 26th CIMAC congress, paper no 225, Shanghai, China; 2013. [14] Kyrtatos P. The effects of prolonged ignition delay due to charge air temperature reduction on combustion in a diesel engine. Dissertation, Swiss Federal Institute of Technology Zurich, no. 21064; 2013. http://dx.doi.org/10. 3929/ethz-a-009933871. [15] Musculus, M. On the correlation between NOx emissions and the diesel premixed burn. SAE 2004–01-1401; 2004. http://dx.doi.org/10.4271/2004-011401. [16] Liu H, Ma S, Zhang Z, Zheng Z, Yao M. Study of the control strategies on soot reduction under early-injection conditions on a diesel engine. Fuel 2015;139:472–81. [17] Park SH, Yoon SH. Injection strategy for simultaneous reduction of NOx and soot emissions using two-stage injection in DME fueled engine. Appl Energy 2015;143:262–70. [18] Zheng Z, Tian X, Zhang X. Effects of split injection proportion and the second injection time on the mixture formation in a GDI engine under catalyst heating mode using stratified charge strategy. Appl Therm Eng 2015;84:237–45. [19] Brückner C, Kyrtatos P, Boulouchos K. Extending the NOx reduction potential with Miller valve timing using pilot fuel injection on a heavy-duty diesel engine. SAE Int J Engine 2014;7(4):1838–50. http://dx.doi.org/10.4271/201401-2632. [20] Andreadis P, Zompanakis A, Chryssakis C, Kaiktsis L. Effect of the fuel injection parameters and emissions formation in a large-bore marine diesel engine. Int J Engine Res 2011;12:14–29. [21] Kallio I, Rantanen P, Imperato M, Antila E, Sarjovaara T, Larmi M, Huhtala K, Liljenfeldt G. The design and operation of the fully controllable medium-speed research engine EVE. In: 25th CIMAC congress, paper no. 163, Wien, Austria; 2007. [22] Kaario O, Imperato M, Tilli A, Lehto K, Ranta O, Antila E, et al. The design of a new generation medium-speed research engine. In: 26th CIMAC congress, paper no. 145, Bergen, Norway; 2010. [23] Imperato M, Antila E, Sarjovaara T, Kaario O, Larmi M, Kallio I, et al. NOx reduction in a medium-speed single-cylinder diesel engine using Miller cycle with very advanced valve timing. SAE 2009-240112; 2009. [24] Imperato M, Kaario O, Sarjovaara T, Larmi M. Influence of the in-cylinder gas density and fuel injection pressure on the combustion characteristics in a large-bore diesel engine. Int J Eng Res 2015;7(1):1–9. http://dx.doi.org/ 10.1177/1468087415589043. [25] Herranen M, Huhtala K, Villenius M, Liljenfeldt G. The electro-hydraulic valve actuation for medium speed diesel engines – development steps with simulation and measurements. SAE 2007-011289; 2007. [26] Egnell R. Combustion diagnostics by means of multizone heat release analysis and NO calculation. SAE 981424; 1998. [27] Sridharan S, Rutland C. Model-based diesel HCCI combustion phasing controller in integrated system level modeling. SAE 2010–01-0886; 2010. [28] Harrison J, Aihara R, Eshraghi M, Dmitrieva I. Modeling engine oil variable displacement vane pumps in 1D to predict performance, pulsations, and friction. SAE 2014–01-1086; 2014. [29] Warnatz J, Maas U, Dibble RW. Combustion. 4th ed. New York: Springer; 2006, ISBN 978-3-540-67751-2. [30] Ando H, Sakai Y, Kuwahara K. Universal rule of hydrocarbon oxidation. SAE 2009–01-0948; 2009. [31] Herfatmanesh MR. Investigation of single and split injection strategies in an optical diesel engine. Dissertation, Brunel University School of Engineering and Design; 2010 [http://bura.brunel.ac.uk/handle/2438/4776].