The Influence of Charge Dilution and Injection Timing on Low-Temperature Diesel Combustion and Emissions

Transcription

1 Licensed to University o New South Wales SAE TECHNICAL PAPER SERIES The Inluence o Charge Dilution and Injection Timing on Low-Temperature Diesel Combustion and Emissions Sanghoon Kook and Choongsik Bae Korea Advanced Institute o Science and Technology Paul C. Miles, Dae Choi and Lyle M. Pickett Sandia National Laboratories Powertrain & Fluid Systems Conerence and Exhibition San Antonio, Texas USA October 4-7, Commonwealth Drive, Warrendale, PA U.S.A. Tel: (74) Fax: (74) Web:

2 Licensed to University o New South Wales By mandate o the Engineering Meetings Board, this paper has been approved or SAE publication upon completion o a peer review process by a minimum o three (3) industry experts under the supervision o the session organizer. All rights reserved. No part o this publication may be reproduced, stored in a retrieval system, or transmitted, in any orm or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission o SAE. For permission and licensing requests contact: SAE Permissions 400 Commonwealth Drive Warrendale, PA USA permissions@sae.org Fax: Tel: For multiple print copies contact: SAE Customer Service Tel: (inside USA and Canada) Tel: (outside USA) Fax: CustomerService@sae.org ISSN Copyright 005 SAE International Positions and opinions advanced in this paper are those o the author(s) and not necessarily those o SAE. The author is solely responsible or the content o the paper. A process is available by which discussions will be printed with the paper i it is published in SAE Transactions. Persons wishing to submit papers to be considered or presentation or publication by SAE should send the manuscript or a 300 word abstract o a proposed manuscript to: Secretary, Engineering Meetings Board, SAE. Printed in USA

3 Licensed to University o New South Wales The Inluence o Charge Dilution and Injection Timing on Low- Temperature Diesel Combustion and Emissions Sanghoon Kook and Choongsik Bae Korea Advanced Institute o Science and Technology Paul C. Miles, Dae Choi and Lyle M. Pickett Sandia National Laboratories Copyright 005 SAE International ABSTRACT INTRODUCTION The eects o charge dilution on low-temperature diesel combustion and emissions were investigated in a smallbore single-cylinder diesel engine over a wide range o injection timing. The resh air was diluted with additional N and CO, simulating 0 to 65% exhaust gas recirculation in an engine. Diluting the intake charge lowers the lame temperature T due to the reactant being replaced by inert gases with increased heat capacity. In addition, charge dilution is anticipated to inluence the local charge equivalence ratio φ prior to ignition due to the lower O concentration and longer ignition delay periods. By inluencing both φ and T, charge dilution impacts the path representing the progress o the combustion process in the φ-t plane, and oers the potential o avoiding both soot and NO x ormation. In-cylinder pressure measurements, exhaust-gas emissions, and imaging o combustion luminosity were perormed to clariy the path o the combustion process and the eects o charge dilution and injection timing on combustion and uel conversion eiciency. Based on the indings, a postulated combustion process in the φ-t plane is presented or dierent dilution levels and injection timings. Although the ignition delay increased with high dilution and early injection, the heat release analysis indicated that a large portion o the combustion and emissions ormation processes was still dominated by the mixing-controlled phase rather than the premixed phase. Because o the incomplete premixing, and the need to mix a greater volume o charge with unburned or partially-burned uel to complete combustion, the diluted mixtures increased CO emissions. Injecting the uel at earlier timings to extend the ignition delay helped alleviate this problem, but did not eliminate it. Fuel conversion eiciencies calculated or each dilution level and start o injection provide guidance as to the appropriate combustion phasing and practical levels o charge dilution or this low-temperature diesel combustion regime. Low-temperature diesel combustion systems limit combustion temperatures to levels at which NO x and soot ormation rates are low. These systems can be roughly divided into two categories: those in which the combustion phasing is largely decoupled rom injection timing and dominated by the kinetics o the chemical reactions, and those in which the control o the combustion phasing is closely coupled to the uel injection event. In the ormer category, the uel and air are typically thoroughly premixed, such that at the start o combustion the mixture is near homogeneous and characterized by an equivalence ratio φ that is everywhere less than. Such systems are generally termed homogeneous charge compression ignition (HCCI) systems. Low combustion temperatures are achieved by either pre-mixing to very lean equivalence ratios (φ 0.5) or by employing exhaust gas recirculation (EGR) to reduce the combustion temperatures o mixtures with higher equivalence ratio (and to control the combustion phasing, e.g. [, ]). In contrast, or the second category, the short times between the uel injection event and the start o combustion preclude thorough pre-mixing, and signiicant regions exist where φ > at the start o combustion. Accordingly, a large raction o the heat is released in a mixing-controlled process and peak combustion temperatures near the adiabatic lame temperature o a stoichiometric mixture could be realized. Various strategies are used to keep the lame temperatures low, including low compression ratio, large amounts o cooled EGR, and use o retarded injection timing. This latter category o low-temperature combustion systems is the ocus o this paper. Charge dilution through use o EGR is a widely used strategy employed or maintaining low lame temperatures and low NO x emissions in diesel engines [3, 4]. Soot emissions are typically ound to increase with

4 Equivalence Ratio Flame temperature with % ambient O at 000 K Conventional Diesel Adiabatic Mixing Premixed burning Re [9] 5% 0% 5% 0% 5% % Premixed burning Mixing-controlled burning Soot 500 ppm 5000 ppm NOx Temperature [K] Fig. Depiction o conventional and two variants o lowtemperature combustion on the φ-t plane increasing dilution, however. Soot emissions rom diesel engines are a result o a competition between soot ormation and oxidation processes. Traditionally, it has been thought that charge dilution aects the oxidation processes more strongly, and soot emissions increase due to a reduction in oxidation rates associated with the low lame temperatures o dilute mixtures. A recent study [6], however, suggests that low soot ormation rates may also be achievable with very high levels o charge dilution, as the lame temperature is limited to levels at which the soot ormation rate is low. Understanding o the soot and NO x ormation processes can be aided through examination o the combustion process in a φ-t parameter space [7], which has proven to be a useul tool or visualizing how the combustion and emission ormation/destruction processes progress. An example is shown in Fig.. The igure is an equivalence ratio versus temperature plot with contours indicating the locations where soot and NO x ormation occur or a diesel-like uel. Here, the zones o soot and NO x ormation are sketched according to the numerical results o Kitamura et al. [8], or n-heptane uel at a pressure o 6 MPa and a residence time o ms. Also a solid line shown in the igure is the maximum lame temperature achieved or uel at 373 K reacting with ambient gas at temperature o 000 K. The open arrows are the path ollowed by a typical uel element in the φ-t plane or a conventional diesel combustion process. Injection o uel into the combustion chamber initiates an adiabatic mixing process which, ater the ignition delay, is ollowed by rapid heat release that brings the mixture to the lame temperature. During this premixed burning Licensed to University o New South Wales process, the heat release is rapid and little additional mixing takes place. Subsequently, additional mixing with oxidant or with the products o more complete combustion causes the path o the uel element to ollow the lame temperature curve. Beyond φ =, no signiicant chemical heat release occurs and mixing serves to lower the temperature o the uel element. Two additional, contrasting paths are also shown on Fig.. The irst, denoted by the black arrows, depicts a typical HCCI combustion process, in which mixing is substantially complete beore the onset o combustion. The second path, denoted by the gray arrows, is the path one might expect to ollow i complete mixing to a lean equivalence ratio is achieved even ater the time o premixed burn, made possible by high rates o mixing beore the lit-o length using small injector holes and high injection pressures [9]. Note that these latter paths are expected to produce very little soot or NO x emissions. It is evident rom the above description that the ormation o both soot and NO x depends strongly on the path ollowed during the combustion process. A major objective o this study is thus to clariy the eect o EGR rate and start o injection (SOI) on the path o a diesel combustion process. The EGR rate inluences the path not only through changes in the lame temperature, but also in ignition delay and the amount o ambient luid that must be mixed with the uel to attain a given equivalence ratio. Similarly, the SOI inluences the temperature (and density) during the ignition delay period, the peak lame temperature reached, and the cooling o the in-cylinder charge during the latter part o the combustion process as volume expansion takes place. In previous work [0], we have begun to clariy the path ollowed or a subset o low-temperature combustion systems characterized by moderate EGR rates and retarded injection timings (examples o these systems include MK (Modulated Kinetics) combustion [], HCLI (Homogeneous Charge Late Injection), and HPLI (Highly Premixed Late Injection) combustion []). Here, we vary both the EGR rate and start o injection over a broad range encompassing not only the aorementioned combustion systems, but also conventional diesel combustion and early injection systems such as the Toyota smokeless system [6] or the DCCS (Dilution Controlled Combustion System) []. Subsequently, we contrast the various parameters inluencing the φ-t plane path o the combustion process, the emission ormation/destruction rates along the path, and (qualitatively) sketch the path itsel. The various actors inluencing the uel conversion eiciency o these combustion systems are also considered.

5 Licensed to University o New South Wales Table Engine speciications Camera EPERIMENT Injector tip Spray Centerline Trajectory Photodiode Neutral density ilter High Speed Mirror Fig. Optical single-cylinder diesel engine RESEARCH ENGINE & DIAGNOSTICS Measurements were obtained in a single-cylinder optical diesel engine with typical characteristics o small-bore engines intended or automotive applications, i.e.: our valves, a central, vertical injector, 6-hole minisac nozzle, and a concentric, re-entrant bowl. A schematic o the engine is shown in Fig., and its speciications are shown in Table. Fuel injection parameters including injection pressure, injected mass and injection timing were controlled with a common-rail uel system, capable o a maximum injection pressure o 350 bar. In-cylinder pressure was acquired using a water-cooled (KISTLER 6043A60) piezoelectric pressure transducer. The in-cylinder pressures acquired over 50 engine cycles were averaged to calculate the indicated mean eective pressure (IMEP), the apparent heat release rate, and the cumulative heat release. In-cylinder pressure data, and the photodiode luminosity data described below, were acquired with a resolution o 0.5 crank angle degrees (CAD). Engine-out NO x was measured using a Caliornia Analytical Instruments chemiluminescent NO x analyzer (Model 600-HCLD). Samples were taken rom the exhaust plenum and transerred to the analyzer through a heated sample line. Moisture and condensable hydrocarbons were removed rom the sample by a condenser prior to the analyzer. Samples were also analyzed or CO content using a Caliornia Analytical Basic Geometry Bore: 79.5 [mm] Stroke: 85.0 [mm] Disp. Vol.: 4 [cm 3 ] Comp. Ratio: 8.7 Valve Events IVO: -360 CAD ATC IVC: -40 CAD ATC EVO: 45 CAD ATC EVC: -350 CAD ATC Fuel Injection Equipment Bosch Flow No.: 30 Included Angle: 45 Number o Holes: 6 Hydro-erosion: Nozzle Style: Cylindrical Minisac (Bosch DLLA) Table Operating Parameters Speed [rpm] 500 T Coolant [ C] 88 P Injection [bar] 800 Swirl Ratio [R s ] 3.77 P Intake [bar]. T Intake [ C] 90 P Exhaust [bar].3 Start O Injection (SOI) [CAD ATDC] to 7.75 Instruments NDIR (Non-Dispersive Inra-Red) CO analyzer (Model 300). The ormation and spatial distribution o in-cylinder soot was monitored with two optical diagnostics. First, a spatially-integrated measure o natural combustion luminosity was obtained with a high-sensitivity photodiode viewing the combustion chamber through the liner window. Because the luminosity rom hot soot is considerably stronger than the luminosity rom natural chemiluminescence [4], the photodiode data is dominated by soot luminosity at all but the highest EGR rates and extremes o injection timing. Neutral density ilters were selected to maintain the signal level within the range o the photodiode signal saturation. Second, images o natural combustion luminosity were obtained to provide qualitative inormation on the spatial distribution o luminous soot within the combustion chamber. Both bottom-view images acquired through the extended piston assembly (as illustrated in Fig. ) and side-view images acquired through the liner side windows were obtained. A high speed digital video camera (Integrated Design Tools Inc.; -Stream Vision) was employed to obtain these images. The rame rate was set to 9000 rames per second, corresponding to CAD (Crank Angle Degree) at,500 rpm. The image resolution is 80 x 80 pixels and an exposure time o 30 µs (0.97 CAD) was employed.

6 ENGINE OPERATION The engine operating parameters are described in Table. All data were acquired at a ixed engine speed o 500 rpm. The coolant temperature was set to 88 C. Recirculated exhaust gas was simulated by diluting the intake air stream with N and CO. The proportions o air, CO, N were selected to obtain the desired O concentration. The relative proportions o CO and N were chosen to match the mixture molar speciic heat o real engine exhaust gas at the selected load and O concentration. The selected O concentrations, the corresponding EGR rates, stoichiometric air-uel ratio, and the calculated average in-cylinder equivalence ratios are listed in Table 3. Prior to obtaining in-cylinder pressure and exhaust emissions, the engine was motored or a minimum o 90 s in order to pre-heat the combustion chamber walls and to allow the intake plenum pressure to stabilize. Subsequently, the engine was skip-ired or 70 s, with uel injection occurring on only one o every our engine cycles. This skip-iring period allowed the emission analyzers to stabilize. Finally, data were acquired over 50 additional skip-ired cycles or the pressure, emissions, and spatially-integrated luminosity data. For imaging o Table 3 Charge Dilution O [%] N [%] CO [%] EGR [%] (A/F) st Avg. φ Cylinder Pressure [kpa] Motoring Peak Pressure (Cylinder Temperature) Motored: 4875 kpa (89 K) Fired: SOI = -4.5 CAD ATDC Fired: SOI = -6.5 CAD ATDC Motored: 490 kpa (900 K) Fired: SOI = -4.5 CAD ATDC Fired: SOI = -6.5 CAD ATDC Motored: 4970 kpa (909 K) Fired: SOI = -4.5 CAD ATDC Fired: SOI = -6.5 CAD ATDC CAD ATDC Fig. 3 Change o combustion phasing with dierent peak motoring pressure (bulk temperature) Licensed to University o New South Wales the spatial distribution o soot luminosity, or acquisition o spatially-integrated luminosity at high O concentration (9 and %), window ouling considerations dictate that the data be acquired immediately upon starting skip-ired operation. Accordingly, beore acquiring these data, the engine was motored or an extended period to raise cylinder wall temperatures to levels commensurate with those existing ater the 70 s skip-iring period. While perorming injection timing sweeps, the in-cylinder density was ixed by maintaining a ixed intake mass low rate. Accordingly, to maintain the same average incylinder temperature ( bulk temperature) or all tests, the same in-cylinder pressure history must be attained. With changing SOI, this pressure history varies slightly, or reasons thought to be associated with changes in the cylinder wall temperatures. Thereore, or some tests, the motored period was extended rom the minimum o 90 s to match the motored pressure traces beore the injection event. Figure 3 shows how changes in the bulk temperature, maniested as small changes in the pressure trace beore injection, can cause signiicant dierences in iring pressure traces and combustion phasing. The in-cylinder pressure data obtained is repeatable to within ±0 kpa, corresponding to a bulk incylinder temperature tolerance o approximately ± K. For the EGR sweeps (ixed SOI), the pressure traces beore injection were also careully controlled to match the bulk in-cylinder temperature as shown in Fig. 4. This was accomplished by ixing the total intake mass low rate at all EGR rates, such that the bulk gas density remained ixed, and by adjusting the motored period so that the in-cylinder pressure decreases as much as the gas constant R decreases or diluted gases ( Pv = RT ). Following this procedure, slightly lower TDC pressures are observed as the EGR rate is increased. The ueling rate was controlled to maintain 3 bar IMEP (indicated mean eective pressure) or dierent SOI and EGR cases. I the ueling rate was over ± o the reerence case, the SOI and EGR sweeps were stopped. The reerence was 7.7 mg/stroke injected or O concentration and SOI o -4.5 CAD ATDC. DATA ANALYSIS HEAT RELEASE ANALYSIS The in-cylinder pressure data were analyzed or the apparent heat release rate using an iterative, two-zone heat release code, in which gas properties are dependent on both mixture composition and temperature. The heat transer losses are partially accounted by subtracting the apparent heat release calculated or a corresponding motored engine cycle, a procedure which is also expected to compensate to a large degree or enthalpy losses associated with crevice lows.

7 Consequently, the net apparent heat release rate Q app is approximately equal to the rate o chemical heat release Q chem less the dierence between the heat transer losses Q hl with combustion Q ht, comb and the motored losses : Q Q ht, motored Q ( Q ht comb Q ht motored ) = Q chem Q hl app = chem,, () The cumulative apparent heat release history is obtained by integrating Q app rom SOI until the integral reaches a maximum Q total just prior to exhaust valve opening. Normalized (ractional) heat release histories are calculated by normalizing the cumulative heat release by Q total. These normalized histories are used to determine burn periods and to estimate equivalence ratios at ignition, as described below. ADIABATIC FLAME TEMPERATURE To clariy the eects o lame temperature on the combustion process and engine emissions, several parameters are correlated below with the maximum adiabatic lame temperature T. This methodology has been proven as a reasonable approach to analyze engine-out emissions by Plee et al. [3] and Musculus [5] as well. The adiabatic lame temperature was calculated or a stoichiometric mixture o diesel uel and diluted intake gas using the STANJAN chemical equilibrium code [5]. The peak adiabatic lame temperature is here deined as the adiabatic lame temperature achieved by combustion o a stoichiometric mixture at the peak gas temperature ( core temperature). The peak gas temperatures were estimated rom the initial intake gas temperature and pressure assuming that the core gases are compressed isentropically. The estimation accounted or the gas composition and or temperature dependent speciic heats. For the lame temperature calculation, a single heat o combustion was used or the multicomponent diesel uel, calculated rom the net heat o combustion and H/C ratio as given in Table 4. RESULTS AND DISCUSSION IN-CYLINDER PRESSURE At ixed SOI, charge dilution decreases the peak incylinder pressure due to the higher heat capacity o the diluent gases as well as slower reaction rates during the premixed combustion. Figure 4 shows the variation in the in-cylinder pressure traces or the various inlet gas O concentrations at a ixed injection timing o -6.5 CAD ATDC. Licensed to University o New South Wales Table 4 Fuel Properties Cylinder Pressure [kpa] 007 Emission Certiication Diesel Fuel Cetane Number 47. H/C Ratio.85 Net Heat o Combustion 4.98 [kj/kg] Speciic Gravity Viscosity at 40 C. Flash Point 77 [ C] Particulate Matter.6 [mg/ l ] Sulur 8. [ppm] Carbon wt% 86.7 Hydrogen, wt% Same Bulk Temperature Low-T Reaction O Concentration CAD ATDC Fig. 4 Eect o charge dilution on cylinder pressure (SOI= -6.5 CAD ATDC) Cylinder Pressure [kpa] Start O Injection [CAD ATDC] Motoring CAD ATDC Fig. 5 Eect o injection timing on cylinder pressure (0% O concentration) Despite the above mentioned reduction in peak pressure at ixed SOI, the highest in-cylinder pressures are oten observed at the lower O concentration. Under these circumstances, the long ignition delay and slow premixed combustion permitted very advanced injection timing. As a result, the portion o uel burned during the latter part o the compression stroke increased, and the peak in-cylinder pressure increased when the timing is advanced, as shown in Fig. 5.

8 Licensed to University o New South Wales Adiabatic Flame Temperature [K] SOI [CAD ATDC] Fig. 6 Peak adiabatic lame temperature or a stoichiometric lame 4 Skip-Fired NO x Emission [ppm] SOI [CAD ATDC] Fig. 7 Eect o O concentration and injection timing on NO x emission ADIABATIC FLAME TEMPERATURE The peak adiabatic lame temperature calculated or each EGR rate and SOI is shown in Fig. 6. Because the peak adiabatic lame temperature is based on the peak in-cylinder temperature, the additional compression associated with early heat release, discussed in conjunction with Fig. 5, generally results in an increase in adiabatic lame temperature as SOI is advanced at any ixed O concentration. This behavior can be clearly seen in Fig. 6. Although the peak pressures or low O concentration are oten higher than those seen at greater O concentrations due to the advanced injection timing, overall the peak adiabatic lame temperatures are lower. More speciically, at ixed SOI the peak lame temperature is always observed to decrease with decreasing O concentration. NO x EMISSIONS The NO x emissions measured or an injection timing sweep at each o the eight O concentrations are shown in Fig. 7. As noted in the Experiment section, emissions were measured operating the engine in a skip-ired mode. Had the engine been continuously ired, the emissions would be 4 times the value shown. Figure 7 indicates that, or all but the lowest O concentration, NO x emissions are reduced as injection timing is retarded. Also note that, at ixed injection timing, NO x decreases with decreasing O concentration behavior which is consistent with the calculated maximum lame temperatures shown in Fig. 6. From the extended Zeldovich mechanism o NO ormation, adopting a partial equilibrium assumption [6], an approximate NO ormation rate can be written as 4 Skip-Fired NO x Emission [ppm] 00 0 d / / E RT NO = A T O N e a / 3 [ ] [ ] [ ] mol / cm s () dt where A is a constant, T is the local gas temperature, R is the universal gas constant, and E a is an overall activation energy. Figure 8 illustrates the lame temperature correlation obtained when NO x emissions are plotted against T on a semi-log plot. Plee et al. [3] suggest that the eect o charge dilution on NO x emissions is primarily due to a lowering o lame temperature, and that the exhaust gas NO x mole raction can be correlated with an expression similar to Eq.(): NOx E/R = K Ea RT NOx = A e /T (Plee et al. [3]) T = 60 K /T [/K] O E/R = K Fig. 8 Correlation o NO x emission with the stoichiometric lame temperature, obtained with reactants initially at the maximum gas temperature ( core temperature) = A e (3)

9 where T is the stoichiometric adiabatic lame temperature. The theoretical basis or Eq.(3) is the premise that engine-out NO x emissions will be proportional to the peak NO x ormation rate. Plee et al. obtain E a R = K rom their experimental data. NO x emissions are thus decreased exponentially with decreasing lame temperature. Figure 8 indicates that only a partial correlation o the NO x emissions measured here with peak lame temperature is obtained using Eq.(3). Two signiicant deviations are observed. First, the slope o the correlation (activation energy E a /R) decreases in magnitude at lower temperatures (higher T ). Second, the activation energies obtained dier rom the value obtained by Plee and co-workers. In the irst region, where correlated by where NOx E a R = 5845 K. T 60 > K, the data are well Ea RT = e (4) In the second, low-temperature region ( T < 60 K ), the apparent activation energy is decreased in magnitude and the correlation is better expressed as: with NOx E a R = 3047 K. 4 Ea RT = e (5) The two regions o the NO x correlation with lame temperature have also been observed by Plee et al., although their low-temperature measurements were not as extensive as is seen in Fig. 8 due to the higher O concentrations considered. The cause o the dierent behavior in the low-temperature region has not yet been identiied, but several possibilities exist:. Because NO x emissions are lower than ppm (8 ppm had the engine been continuously ired), it is possible that lame temperatures are low enough that thermal NO ormation no longer dominates, and prompt NO ormation becomes signiicant [3].. Due to the low temperatures, ormation reactions may be quenched earlier [3]. 3. At the lowest O concentrations, less uel may be ound in stoichiometric or leaner mixtures prior to ignition, and the peak lame temperature or the remaining, rich uel mixtures is not reached until later in the cycle, ater additional mixing has taken place. By this time, cylinder volume expansion has cooled the gases. Estimates o the equivalence ratio at ignition and a simple adiabatic mixing analysis presented below support this suggestion. Licensed to University o New South Wales Regardless o the cause, the trends observed in Figs. 7 and 8 are clear: at the lowest O concentrations NO x emissions are less sensitive to SOI despite the signiicant variation in calculated peak lame temperature (Fig. 6). Although NO x emissions correlate reasonably well with peak adiabatic lame temperature alone, examination o Eq.() suggests that NO x emissions might be aected by other, additional parameters. That is, or the constant density tests conducted here, NOx / / / Ea RT O N M T e (6) For various O concentrations, both the mixture molecular weight M and the N mole raction N vary little (rom 8.96 to 9.95 and rom 79 to 80.6 (see Table 3), respectively). Consequently, they should not aect the correlation considerably but rather modiy the constants o Eq (4) and (5). The O, however, were changed rom to a change that can not be easily ignored. To examine the inluence o O concentration, the adiabatic lame temperatures were re-calculated using the bulk (average) in-cylinder temperature as the reactant temperature at SOC, in contrast to our earlier use o the maximum core temperature. Dec [7] suggests that the periphery o a burning diesel jet is a location where signiicant NO x ormation occurs. It is conceivable that, in the periphery o the diesel jet, the local gas temperature is between the hot core area and relatively cold outer area. Accordingly, the maximum bulk in-cylinder temperature estimated by the ideal gas equation might be a more appropriate initial temperature rom which to calculate the peak adiabatic lame temperature. The results obtained with this second approach are shown in Fig. 9, which indicates that 4 Skip-Fired NO x Emission [ppm] 00 0 NOx = A e /T (Plee et al. [3]) O /T [/K] Fig. 9 Correlation o NO x emission with stoichiometric lame temperature, calculated with the reactants at an initial bulk in-cylinder temperature

10 -4.6 N - M / T / NOx O E/R = K T = 70 K O E/R = -508 K /T [/K] Fig. 0 Correlation o NO x emission with stoichiometric lame temperature calculated with the reactants at an initial bulk in-cylinder temperature, ater compensation or the additional parameters suggested by the orm o Eq.() Licensed to University o New South Wales Soot Luminosity [a.u.] more accurate estimation o actual emission levels, multi-zone models (e.g. [8],[9]), are a more appropriate choice. SOOT LUMINOSITY SOI [CAD ATDC] Fig. Eect o O concentration and injection timing on soot luminosity Eq. (3) no longer successully correlates the NO x emissions and lame temperature. There is an apparent residual dependency on the O concentration. From Fig. 0, we see a dependency on O concentration to the 4.6 power collapses the data to a single curve rather than 0.5 rom Eq. (4). Most o the data where T > 70 K, are well correlated by NOx 4.6 N / / T Ea RT = M e (7) where E a R = 3700 K. Note the reasonable agreement with the value o E a R obtained by Plee and coworkers ( E a R = K ). Two regions o diering slope remain, as was observed previously in Fig. 8. In the low-temperature region ( T < 70 K ), the apparent activation energy is reduced considerably in magnitude and the correlation is better expressed as: NOx 4.6 N / / T Ea RT 7 =.7 0 M e (8) with. E a R = 508 K. It is clear rom the above discussion that it is diicult to accurately predict NO x emissions rom a simple correlation o the orm o Eqs.(3) or (6) over the wide range o O concentrations and injection timing considered here. Choice o a single, suitable temperature to employ introduces urther complications. Accordingly, simple correlations o the orm employed here are best employed to gain an understanding o the important variables inluencing NO x emissions. For a As noted in the introduction, engine-out soot emissions typically increase with increasing dilution level. However, beyond a critical dilution level, corresponding to an airuel ratio o about 5 [6, ] or an EGR rate o approximately 55% [3], soot levels are again ound to decrease. Because soot oxidation rates are generally thought to decrease with dilution, the decrease in soot at high EGR rates is likely due to reduced ormation. Multi-dimensional modeling studies o soot ormation in engines [8] show that or equivalence ratios less than approximately, no soot will be ormed regardless o the local temperature (see Fig. ). These numerical results are supported by both experimental engine studies using an oxygenated uel [0], as well as studies o diesel jets in a constant volume combustion chamber [9]. Moreover, or local temperatures lower than the value required or soot inception ( K rom Fig. ), no soot is ormed whatever the equivalence ratio is. Maintaining low lame temperatures thus has the potential or simultaneously low soot and NO x emissions. Figure shows the peak level o spatially-integrated soot luminosity observed during the cycle, measured using the high-sensitivity photodiode shown in Fig.. The results were normalized with the maximum value, which occurred at % O concentration. The integrated soot luminosity is a complicated unction o temperature, the spatial distribution o the soot within the cylinder, and the overall soot volume raction [4]. A quantitative interpretation o the measured soot luminosity in terms o in-cylinder soot mass is thus not possible. However, airly simple observations rom Fig. 0 can reveal inormation regarding the soot ormation process as dilution or SOI is varied.

11 A irst observation to make rom Fig. is that soot luminosity is seen at all O concentrations. Consequently, or our engine compression ratio and inlet conditions, dilution alone does not suiciently lower lame temperatures to impede the ormation o soot. However, we also observe that retarding injection suiciently at high to moderate EGR rates results in a signiicant reduction in soot luminosity, suggesting that soot ormation can be urther suppressed by the additional premixing and the lower lame temperatures associated with the cooling eect o cylinder volume expansion. On the other hand, at 0% O concentration, low soot luminosity can also be achieved at the earliest injection timings. With early injection, however, Fig. 6 demonstrates that the lame temperature is substantially equal (or higher) than those seen at later injection timings. Accordingly, in this case premixing alone is the likely mechanism responsible or reduced soot ormation. IMAGES OF NATURAL COMBUSTION LUMINOSITY The spatial distribution o soot luminosity, and its temporal evolution, was investigated using the high speed digital video camera. Figures, 3 and 4 show pairs o bottom- and side-view images o natural combustion luminosity or, 5, and % O concentrations, respectively. The apparent heat release rate and the normalized cumulative heat release are shown above the images, with dots indicating the time o image acquisition. The acquisition time, as well as the time o each image ater start o injection (ASOI), is superimposed on each image pair. Similarly, the transmissivity associated with the neutral density ilter used during the acquisition o each image set is shown on the upper-let image pair. Several observations are common to the images obtained at all three O concentrations. First, the initial soot luminosity is detected in the latter hal o the premixed burn period. This behavior is consistent with the path o conventional diesel combustion discussed in the context o Fig., in which the premixed combustion phase is required to raise the temperature o the mixture to above the soot inception temperature. Second, although some discrete jet structure can be observed in the % O soot luminosity distributions, the soot quickly spreads and is more or less uniormly distributed azimuthally by roughly CAD ASOI. Finally, between approximately and 0 CAD ATDC, the bulk o the luminous soot appears to be located within the bowl, but away rom the outer periphery. This suggests that the phasing and magnitude o the premixed burning does not signiicantly aect the low structures transporting unburned uel and soot during the mixing-controlled phase o combustion. Licensed to University o New South Wales In addition to these common eatures, dierences were also observed. As was seen in the integrated luminosity data o Fig. 0, the level o soot luminosity dramatically decreased as the O concentration was decreased (note the increasing ilter transmissivity). Additionally, or % Fig. Images o natural combustion luminosity: -6.5 CAD ATDC SOI, % O, 30µs exposure time

12 Licensed to University o New South Wales Fig. 3 Images o natural combustion luminosity: -6.5 CAD ATDC SOI, 5% O, 30µs exposure time Fig. 4 Images o natural combustion luminosity: -6.5 CAD ATDC SOI, % O 30µs exposure time

13 O, soot luminosity was detected in the squish area ( B - C ) at early crank angles (-0.5 and.75 CAD ATDC). However, soot luminosity in the squish area appeared later (.75 CAD) at O concentration and did not appear at 0% O concentration. These dierences in the squish area soot luminosity suggest dierences in the soot ormation rates in this region or the dierent O concentrations. CO EMISSIONS CO emissions are typically small or conventional, hightemperature diesel combustion strategies. With high combustion temperatures, CO is easily oxidized even though large amounts o CO are ormed during the premixed burning o uel-rich regions []. On the other hand, low-temperature combustion systems oten exhibit high CO emissions, particularly at high dilution levels. There remains some question as to the source o these CO emissions. Due to the extended ignition delay encountered in low-temperature systems, signiicant portions o the uel could mix to very lean equivalence ratios ( T < 500 K ). Under these over-mixed conditions, combustion temperatures are too low or the oxidation o CO to be completed on engine time scales []. Over-mixed uel has been shown to be the dominant source o CO emissions or conventional diesel combustion []. On the other hand, the presence o soot (discussed above) implies that some raction o the premixed combustion occurs under uel-rich conditions. Combustion o this under-mixed uel will produce copious amounts o CO (and H ). Completion o combustion (burn-out o the CO ormed) is conditional on additional mixing o the products o rich combustion with oxidant. This mixing process is more diicult to complete quickly with low O concentrations, due to the need to mix with a greater volume o ambient luid. Additionally, the lower lame temperatures o diluted mixtures imply Licensed to University o New South Wales that less time is available beore expansion drops the temperature below a threshold level required or suiciently rapid CO oxidation. Measured CO emissions are shown in Fig. 5. At low-tomoderate dilution levels ( O 7 % ), the CO emissions correlate strongly with the ignition delay results presented below (see Fig. 6), increasing with both increased diluent and retarded SOI. This observation, coupled with the numerical results o [] and the general inverse correlation o the CO emissions with the peak lame temperature (Fig. 6), suggests that the incomplete combustion o over-mixed uel is the dominant mechanism responsible or the CO emissions at low EGR rates. However, the inverse lame temperature correlation is also consistent with CO emissions rom the products o combustion o undermixed uel. Furthermore, as O concentrations decrease, the CO emissions begin to correlate inversely with ignition delay. At high dilution rates it thus appears more probable that CO emissions originating rom undermixed uel dominate. A correlation o CO emissions with the average in-cylinder temperature late in the combustion process, discussed below, urther reinorces this view. At low-to-moderate dilution levels and retarded injection timing, charge cooling and low lame temperatures may impede the burn-out o the CO ormed in the rich premixed combustion, and additional numerical simulations and/or in-cylinder measurements will be required to clariy the dominant source o the CO emissions IGNITION DELAY Ignition delay is inluenced by both physical and chemical actors, and is thus aected by characteristics o the uel injection equipment (injection pressure, nozzle hole diameter), injection timing (ambient temperature and density) and charge dilution (EGR rate) [, 3]. Here, 4 Skip-Fired CO Emission [ppm] SOI [CAD ATDC] Fig. 5 Eect o O concentration and injection timing on CO emission ignition delay τ id [CAD] Average Temperature T a O SOI [CAD ATDC] Fig. 6 Eect o O concentration and injection timing on ignition delay together with the average ambient temperature estimated or an O concentration o 0% Average Temperature Ta

14 ignition delay is deined as the time rom SOI to start-ocombustion (SOC), as determined by the crank angle at which 0% o the cumulative heat release occurred. This crank-angle correlates well with the beginning o rapid, high-temperature heat release. Figure 6 shows that, at the same injection timing, the ignition delay increases with decreasing O concentration. To isolate the eect o O concentration rom the eects o changing ambient temperature or density as the dilution level is increased, the temperature and density o the ambient gas were maintained at the same value regardless o O concentrations (see Fig. 4 and the accompanying discussion). As the injection timing is varied, the ignition delay varies inversely with the estimated average in-cylinder temperature Ta Licensed to University o New South Wales bulk in-cylinder temperature T SOC was estimated rom the cylinder pressure at the reerence state assuming an isentropic compression or expansion process. This observed eect o ambient temperature on ignition delay is well-known an increase in temperature shortens the ignition delay [3]. To examine the inluence o T a more closely, the ignition delay τ id is plotted on a logarithmic axis against T a in Fig. 7. Here it is seen that τ id increases exponentially with T a. Two additional observations are noteworthy. First, an apparent dependency o ignition delay on O concentration clearly remains. Second, at ixed O concentration, the same τ id is measured or two dierent T a. The latter observation shows that the ignition delay cannot be expressed with T a alone, even at constant. However, ignition delay is also known to be O TSOI + TSOC Ta = (9) where T SOI is the bulk in-cylinder temperature at SOI and T SOC is bulk in-cylinder temperature at SOC. T SOI was estimated rom the ideal gas law using the known pressure and volume history, an estimated reerence temperature at TDC (without combustion), and the assumption o a ixed mass. The reerence temperature at TDC was taken to be the typical bulk gas temperature, and was estimated based on the value calculated rom multi-dimensional numerical calculation [0]. With this procedure or estimating temperature, the peak temperature occurs earlier than the peak pressure, due to heat loss during compression. For example, the peak in-cylinder temperature at 0% O concentration was around - CAD ATDC, which was near the crank angle where the shortest ignition delay was measured. The 3 τ id p [ms kpa] Single Line or each /T [/K] a Fig. 8 Correlation o ignition delay with average ambient pressure and temperature 700 τ id [ms] Dierent T a /T [/K] a τ id ( ). p [ms kpa] E a /R = 34.4 K O /T [/K] a Fig. 9 Correlation o ignition delay with average ambient Fig 7. Correlation o ignition delay with average ambient pressure, temperature and O concentration temperature

15 inluenced by ambient pressure, and correlations or ignition delay measured in constant pressure and temperature environments are typically o the orm [4]: n Ea / RTa τ = A p e ms (0) id A and n are constants dependent on uel and airlow characteristics, p is the ambient pressure, R is the universal gas constant, and E a /R is an apparent activation energy or the auto-ignition process. The constant n is typically close to. The inluence o pressure (with n = ) on ignition delay is n illustrated in Fig. 8, where τ id p is plotted against T a. Now, or ixed O concentration, a single ignition delay is predicted at each temperature. However, a signiicant dependency on O remains. Consequently, we have included the inluence o O concentration on ignition delay by considering the ollowing expression: Licensed to University o New South Wales Normailed Cummulative Heat Release Mixing-Controlled Start o High-Temperature Heat Release CAD ATDC Fig. 0 Estimation o the percentage o premixed O Apparant Heat Release Rate [J/CAD]. Ea / RTa τ =.54 p e ms, E a / R = K () id As can be seen in Fig. 9, Eq.() eectively predicts the ignition delay at all O concentrations. From this equation, we see that ignition delay is decreased not only by uel injection into high ambient pressures and temperatures, but also by high O concentrations. The latter dependency is to be expected due to the slowing o chemical reactions in dilute mixtures, as well as the decreased temperature rise rom early low-temperature reactions due to the heat capacity o the diluent molecules. EQUIVALENCE RATIO AT IGNITION As can be inerred rom Fig., the equivalence ratio at ignition will inluence the typical path ollowed by a uel element in the φ - T plane as the combustion process progresses, thereby aecting soot and NO x ormation. By changing the dilution level, we anticipate that the equivalence ratio at ignition will change due to the ollowing two opposing actors:. Ignition delay: As shown in the previous section, decreased O concentrations resulted in increased ignition delay, allowing a greater time period or premixing to occur. For example, at the same injection timing, the ignition delay observed with 0% O concentration can be over twice as great as is seen or O concentrations o 7% and above.. Low ambient O concentration: Due to the decreased O concentration, a greater amount o ambient luid must be mixed with the uel to achieve the same uel-air equivalence ratio. For example, at 0% O concentration, over twice the volume o ambient luid must be mixed with the uel to achieve the same uel/o ratio as or % O. J/Q LHV O Fig. The raction o the energy released as a unction o the equivalence ratio (SENKIN calculation ater high-temperature ignition, ambient temperature = 900K, Fuel temperature = 373K) Studies o lame lit-o in ree jets [5] show that as O concentrations are reduced, the lame lit-o increases such that the average mixture equivalence ratio at lit-o remains approximately constant. Assuming that lame lit-o is dominated by ignition-like processes, we might suppose that the opposing eects o increased ignition delay and reduced O concentration identiied above would be largely sel-canceling. However, steady-state mixing rates in a ree jet and the average mixing rates in our short injection-duration, transient engine experiments are likely to be signiicantly dierent, and such a supposition cannot be justiied. An estimate o the equivalence ratio at ignition can be made, however, through examination o the cumulative heat release observed with % O. With % O, the rates o chemical reaction are ast and beyond the φ

16 Ignition Delay [CAD] Intake O % Fig. Ignition delay and the raction o premixed O present at SOC premixed burning period the cumulative heat release can be correlated with the raction o O needed or complete combustion which has been mixed with the uel. We also assume that, given the same SOI and bulk temperature, the uel/ambient gas volumetric mixing rates are substantially independent o O concentration or the presence o heat release. With these approximations, a measure o the average equivalence ratio at ignition or the various dilute mixtures can be obtained by examination o the % O cumulative heat release (measured at the same SOI) at the crank angle that the dilute mixture ignites. The estimated equivalence ratio at ignition or the lower O concentrations is obtained rom Figures 0 and using the ollowing procedure: First, the crank angle at which high temperature heat release begins is determined or the dilute combustion case being considered. At this crank angle, the normalized cumulative heart release or the % O case is determined rom Fig. 0, and the corresponding equivalence ratio (or raction o needed O ) is then ound rom Fig.. It is notable that the ractions o energy released at a given equivalence ratio are identical or dierent O concentrations. Finally, the raction o needed O is multiplied by the ratio o stoichiometric airuel ratios to account or the lower O concentration o the dilute case. This procedure could not be ollowed or the case o 9% O, as ignition occurred beore the onset o mixing-limited combustion or % O. For example, at 0% O concentration, ignition occurred at 7.3 CAD ATDC. At this crank angle, Fig. 0 indicates that 70% o the cumulative heat release is observed in the % O case, corresponding to (rom Fig. ) an equivalence ratio o.3 or about 8 % o the needed O. In the dilute case, however, the raction o needed O Licensed to University o New South Wales % o Needed O at SOC will be considerably lower, and it is necessary to multiply by the ratio o stoichiometric uel-air ratios given in Table 3 (4.5/3.6). Consequently, only about 37% o the needed is present at ignition or 0% O. Figure shows the results o this estimation and the ignition delay or data obtained at -6.5 CAD ATDC SOI. Note that the raction o O needed or complete combustion does not increase as the O concentration decreases, despite the increasing ignition delay. On the contrary, we observe an increase in the equivalence ratio at ignition or more dilute mixtures. Consequently, i a leaner equivalence ratio at ignition is desired, modiying the ambient gas temperature (through modiied injection timing, compression ratio, or intake charge cooling) to increase ignition delay appears a better strategy than employing dilute mixtures. As will be seen below, a simple adiabatic mixing argument supports the above observations o increased φ at ignition or dilute mixtures, as well as the eectiveness o ambient temperature in modiying φ at ignition. APPARENT COMBUSTION DURATION Figure 3 shows the time rom the start o rapid heat release until one-hal o the total apparent heat release has occurred the 0-50% burn time. This parameter is o interest or two reasons. First, it provides inormation on the path o the combustion process in the φ-t plane. For rapid premixed burning, very little additional mixing takes place during the premixed burning portion o the combustion path shown in Fig.. However, the early combustion duration can increase 5-old or high dilution levels and/or retarded injection timing. Under these circumstances, it is anticipated that signiicant additional mixing can occur between SOC and the time the peak combustion temperatures are achieved at the end o premixed burning. Second, the 0 50% burn period is typically ound to vary inversely with the peak heat 0-50 % Burn Time [CAD] SOI [CAD ATDC] Fig. 3 Eect o O concentration and injection timing on 0-50 % burn time