Diesel Engine Combustion of Biomass Pyrolysis Oils - American


Experimental ignition delay and heat release rates were interpreted using a ... rate, the model showed that the longer ignition delays of the bio-oils...
0 downloads 0 Views 401KB Size


260

Energy & Fuels 2000, 14, 260-274

Articles Diesel Engine Combustion of Biomass Pyrolysis Oils Alan Shihadeh and Simone Hochgreb* Massachusetts Institute of Technology, Department of Mechanical Engineering, Cambridge, Massachusetts 02139

Downloaded via UNIV OF CALIFORNIA SANTA BARBARA on June 23, 2018 at 20:27:05 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Received March 23, 1999

To investigate their ignition delay and combustion behavior, experiments with two biomass pyrolysis oils and No. 2 diesel fuel were performed in a direct injection diesel engine. It was found that while the indicated thermal efficiency of both pyrolysis oils equaled that of the diesel fuel, they exhibited excessive ignition delays and required a moderate degree of combustion air preheating to ignite reliably. Despite the longer ignition delays associated with the pyrolysis oils, the cylinder pressure rise rates were significantly less than with No. 2 diesel fuel. Experimental ignition delay and heat release rates were interpreted using a phenomenological spray combustion model. Using a three parameter fit for vaporization, ignition, and combustion rate, the model showed that the longer ignition delays of the bio-oils result from slow chemistry relative to diesel fuel. The model also showed that the heat release profiles of the bio-oils are consistent with slow combustion chemistry and rapid mixing relative to diesel fuel. As a result, whereas diesel combustion is predominantly mixing limited, pyrolysis oil combustion is predominantly limited by chemistry through much of the process.

Introduction With the advancement of high-recovery flash pyrolysis processes, interest in biomass pyrolysis oils as a potentially renewablesand CO2 recyclingsenergy source for conventional power systems has increased greatly in the past few years. Biomass pyrolysis oils are manufactured through a moderate-temperature process (∼500 °C) in which the biomass feedstock is subjected to rapid heating in the absence of air, where it vaporizes, cracks, and is condensed after a short residence time (∼500 ms) into a dark brown liquid composed of a complex mixture of oxygenated hydrocarbons whose heating value is approximately half that of No. 2 diesel fuel. They contain significant quantities of moisture (∼15-25 wt. %), particulates, nitrogen, alkali, and tar, and are more dense and viscous than No. 2 diesel fuel. Whereas petroleum distillates are composed primarily of pure hydrocarbons (paraffins, aromatics, naphthalenes), pyrolysis oils are typically composed of carbohydrates, organic acids, aldehydes, and other oxygenated organics.1,2 Their physical properties and molecular composition vary with the pyrolysis process parameters such * Author to whom correspondence should be addressed at Sandia National Labs, P.O. Box 969, MS 9053, Livermore, CA 94550. Telephone: (925) 294-4724. E-mail: [email protected] (1) Radlein, D.; Piskorz, J.; Scott, D. S. Lignin derived oils from the fast pyrolysis of poplar wood. J. Anal. Appl. Pyrolysis 1987, 12, 5159. (2) Evans, R. J.; Milne, T. A. Molecular characterization of the pyrolysis of biomass. 1. Fundamentals. Energy Fuels 1987, 1 (2).

as reactor temperature and residence time, as well as the feedstock moisture content and particle size.3 While there is a considerable body of literature addressing pyrolysis chemistry and bio-oil production spanning two decades of work, only recently has a systematic effort begun to study the combustion characteristics of biomass pyrolysis oils. To date, combustion research with biomass pyrolysis oils has focused on single droplet combustion behavior in atmospheric pressure laminar flow reactors,4-7 on ignition delay times in a combustion bomb,8 and on combustion behavior of bio-oil in furnace applications.9 In addition, (3) Evans, R. J.; Milne, T. A. Molecular characterization of the pyrolysis of biomass. 2. Applications. Energy Fuels 1987, 1, (2). (4) Wornat, M. J.; Porter, B. G.; Yang, Y. C. Single Droplet Combustion of Biomass Pyrolysis Oils. Energy Fuels 1994, 8, 11311142. (5) Shaddix, C. R.; Huey, S. P. Combustion Characteristics of Pyrolysis Oils Derived from Hybrid Poplar. Developments in Thermochemical Biomass Conversion, Conference proceedings, Banff, Canada, 1996. (6) Shaddix, C. R.; Huey, S. P.; Wornat, M. J.; Davis, K. A. Fundamental Aspects of Combustion of Biomass Pyrolysis Oils. Biomass Usage for Utility and Industrial Power, Conference proceedings, Snowbird, UT, 1996. (7) Shaddix, C. R.; Tennison, P. J. Effects of Char Content and Simple Additives on Biomass Pyrolysis Oil Droplet Combustion. 27th Symposium (International) on Combustion, The Combustion Institute, 1998. (8) Suppes, G. J.; Rui, Y.; Regehr, E. V. Hydrophilic Diesel Fuels Ignition Delay Times of Several Different Blends. SAE Paper 971686, 1997. (9) Shihadeh, A.; Lewis, P.; Manurung, R.; Bee´r, J. Combustion Characterization of Wood-Derived Flash Pyrolysis Oils in IndustrialScale Turbulent Diffusion Flames. Biomass Pyrolysis Oil Properties and Combustion Meeting, Proceedings, U.S. National Renewable Energy Lab, Estes Park, CO, September 1994.

10.1021/ef990044x CCC: $19.00 © 2000 American Chemical Society Published on Web 02/15/2000

Diesel Engine Combustion of Biomass Pyrolysis Oils

Energy & Fuels, Vol. 14, No. 2, 2000 261

Table 1. Fuel Properties manufacturer/batch number water content [wt %] viscosity at 30 C [cSt] surface tension at 20 °C [mN/m] specific gravity at 20 °C LHV [MJ/kg] LHV stoichiometric mixture [MJ/kg] A/F stoichiometric whole oil ultimate analysis [wt % dry] C H N O (by difference)

Table 2. Engine Specifications

NREL ENSYN diesel M2-10 RTP 15TPD fuel 16.9 80 34.7 1.2 17.0 2.28 6.45

26.3 96 40.0 1.2 16.3 2.46 5.62

58.25 7.40 1.52 32.83

57.95 7.23 1.64 33.19

29.3 0.84 44 2.79 14.5 87 13

one study was published on the techno-economic feasibility of bio-oil utilization in a high-speed diesel engine.10 Taken together, this young body of literature reveals a picture of pyrolysis oil combustion which is complicated by long ignition delays and varying microexplosive behavior across oils with varying impacts of water, char, and volatiles affecting relative burn rates. Furthermore, it shows that considerable practical problems, including corrosivity, coking, and high viscosity, make pyrolysis oil utilization in a diesel engine difficult at present. In the work reported here, the combustion of wood pyrolysis oil to that of No. 2 diesel fuel in a high-speed diesel engine is characterized and compared. The primary diagnostics are crank-angle resolved cylinder pressure and ignition delay, which are interpreted using a phenomenological model of diesel spray combustion to give information about characteristic vaporization, ignition, and combustion rates relative to diesel fuel. Experimental Setup Two wood pyrolysis oils were used in this study: one manufactured from hardwoods using the ENSYN Rapid Thermal Process, and the other from Poplar using the NREL Vortex Ablative Pyrolysis. The ENSYN process represents the most widely implemented commercial pyrolysis technology at present. It utilizes a thermal mixer in which a high-temperature solid particulate contacts and rapidly heats the biomass feed, after which both are fed into a residence time and temperature-controlled tubular transport reactor.11 The NREL process utilizes an ablative vortex reactor in which the biomass feed is forced to slide in a helical path on the hot cylindrical wall of the reactor. The sliding contact of the biomass particles results in an ablative process in which char buildup on the particle surface is constantly scraped, continually exposing fresh biomass material to the reducing atmosphere.11 Furthermore, the NREL process is distinguished by a hot-gas filtration stage upstream of the condenser. Thus the NREL process represents a more complex experimental system which produces a low char content oil, but with lower liquid yields, while the ENSYN process is representative of a commercial product. Properties of both pyrolysis oils and diesel fuel are shown in Table 1. The viscosity, density, and surface tension of the pyrolysis oils are significantly greater than those of diesel fuel, suggesting that poor atomization might hamper efficient (10) Solantausta, Y.; Nylund, N.; Westerholm, M.; Koljonen, T.; Oasmaa, A. Wood Pyrolysis Oil as Fuel in a Diesel-Power Plant. Bioresour. Technol. 1993, 46, 177-188. (11) Elliot, D. C.; Beckman, D.; Bridgwater, A. V.; Diebold, J. P.; Gevert, S. B.; Solantausta, Y. Developments in Direct Thermochemical Liquefaction of Biomass: 1983-1990. Energy Fuels 1991, 5, 399-410.

model cylinders bore [mm] stroke [mm] swept volume [liters] compression ratio aspiration rated speed [rpm] water outlet temperature [°C] oil outlet temperature [°C] IVO/IVC [DATC] EVO/EVC [DATC] injector nozzle opening pressure [bar]

Ricardo Hydra Mark 4 1 80.26 88.9 0.4498 19.8 natural 4500 85 85 -10/221 122/11 4 hole × 0.21 mm dia × 155 deg cone angle 250

ignition and combustion. Other important distinguishing features are the high water contents (which can affect local temperature and vaporization rates via the high heat of vaporization and specific heat), and the low heating values of the pyrolysis oils. It should be noted that the relatively low stoichiometric ratio offsets the low heating value of the pyrolysis oils, resulting in an overall heating value close to that of diesel fuel on a stoichiometric fuel-air charge basis. This means that for a given diesel engine displacement, a similar power output can in theory be achieved when operating on pyrolysis oil. Table 2 shows the relevant characteristics of the single cylinder, direct injection diesel engine used in this work. The engine employs a toroidal bowl piston to achieve rapid fuel/ air mixing and is naturally aspirated. Its bore/stroke ratio and operating speed range are typical of modern light-duty diesel engines. The fuel injection system utilizes a Bosch Type A inline fuel pump (maximum pressure ) 600 bar) and a holetype injector nozzle which opens at 250 bar; the injection timing is varied by rotating the fuel pump relative to the crank shaft. The combustion air inlet temperature can be preheated up to 130 °C through the use of an in-line electric heater, which allows engine operation with fuels that have long ignition delay, without relying on any ignition additives. Cylinder pressure was measured using a flush-mounted Kistler 6125 piezo-electric transducer located near the center of the head. The transducer was connected to a Kistler model 504 dual mode charge amplifier and the voltage output sampled using a PC-based data acquisition system at a rate of 2.5 points per crank-angle degree. Injector needle lift was measured using a Hall-effect proximity sensor which was designed and machined into the fuel injector by Wolff Controls Corp. Average cylinder temperature at the start of injection was calculated from the cylinder pressure measurement using the ideal gas law, while instantaneous heat release was calculated from cylinder pressure using a standard single-zone First Law analysis of the cylinder contents. Ignition delay was calculated as the time elapsed from start of injection (SOI) to the time at which the instantaneous heat release rate reached a value of 5% of the peak release. All experiments were performed at a load of 5.0 bar indicated mean effective pressure (IMEP) (φ ∼ 0.4) and 2400 rpm, while the combustion air temperature was varied from 30 to 100 °C. The load condition was chosen such that experiments with low heating value fuels could operate at the same load (and approximately the same chemical energy input per cycle) as diesel fuel given the existing fuel delivery system capacity. This is realistic for an application where pyrolysis oil is being substituted for diesel fuel. Several special measures were necessary to burn the pyrolysis oils in the diesel engine; they are described in Appendix A.

262

Energy & Fuels, Vol. 14, No. 2, 2000

Figure 1. NREL oil. Cylinder pressure and needle lift traces. Needle lift in arbitrary units. Average of 100 cycles. 5.0 bar IMEP, 2400 RPM, TSOI ) 920 K.

Figure 2. ENSYN oil. Cylinder pressure and needle lift traces. Needle lift in arbitrary units. Average of 100 cycles. 5.0 bar IMEP, 2400 RPM, TSOI ) 920 K. SOI ) -7 DATC, ID ) 7.9 CAD.

Figure 3. Cylinder pressure and needle lift traces. Needle lift in arbitrary units. Average of 100 cycles. 5.0 bar IMEP, 2400 RPM, TSOI ) 920 K. SOI ) -3 DATC, ID ) 6.3 CAD.

Experimental Results General Observations. Some observations regarding the differences of pyrolysis oil and diesel fuel combustion can be made directly from the (typical) cylinder pressure traces shown in Figures 1-3, for NREL, ENSYN, and diesel fuels. First, while the three cases exhibit the same peak pressure timing of 9.5 degrees after top center (DATC), the peak pressure of the pyrolysis oils is approximately 15% greater than

Shihadeh and Hochgreb

Figure 4. NREL and diesel fuels average rate of pressure rise rate during rapid rise period following SOC. Ignition delay shown in gray. 100 cycle average, 2400 RPM, 5.0 bar IMEP.

that of diesel fuel. Second, the qualitative shape of the pressure trace during the bulk of the combustion is significantly differentsthe pyrolysis oils do not exhibit the period of almost constant pressure volume expansion following the initial rapid pressure rise, as exhibited by the diesel case. Third, as shown in Figure 4, the rate of cylinder pressure rise of the diesel fuel is some 30-70% greater than that of the pyrolysis oils, even when the ignition delay of the pyrolysis oils is significantly greater. Normally, fuels with long ignition delay (low Cetane number) give high-pressure rise rates because there is more time for fuel-air premixing prior to ignition, resulting in diesel knock in extreme cases. As will be shown in the Analysis section, these qualities together result from the fact that pyrolysis oil combustion is predominantly kinetically controlled (even when the ignition delay is short), a fundamental difference from the predominantly mixing-controlled diesel fuel combustion. It is noteworthy that despite these differences, the gross indicated thermal efficiency was approximately constant at ∼36% for the pyrolysis oils and diesel fuel. Finally, the engine was found to operate as smoothly with pyrolysis oils as it did with diesel fuel, even when the ignition delay was long; the IMEP coefficient of variation remained within an acceptable range in all cases from 1 to 3%, and did not correlate with fuel type. Ignition Delay. Ignition delays for the pyrolysis oils are plotted versus charge temperature at the start of injection in Figures 5 and 6. The data fall within typical ignition delay ranges for high-speed DI diesel engines of 0.4-1 ms, corresponding to 6-14 crank angle degrees (CAD) at 2400 rpm.12 Corroborating the earlier cited work,13 we found that both pyrolysis oils exhibited excessive ignition delays and would not auto-ignite at cylinder charge temperatures below 870 K, which corresponded to an air inlet temperature of 55 °C for the experimental engine. In contrast, the diesel fuel ignited under all test conditions. The natural logarithm of the ignition delay is plotted against 1/T, where T is calculated average charge (12) Plee, S. L.; Ahmad, T. Relative Roles of Premixed and Diffusion Burning in Diesel Combustion. SAE paper 831733, SAE Trans., Vol. 92, 1983. (13) Solantausta, Y.; Nylund, N.; Westerholm, M.; Koljonen, T.; Oasmaa, A. Wood Pyrolysis Oil as Fuel in a Diesel-Power Plant. Bioresour. Technol. 1993, 46, 177-188.

Diesel Engine Combustion of Biomass Pyrolysis Oils

Figure 5. NREL oil ignition delay versus charge temperature at SOI. 2400 RPM, 5.0 bar IMEP, 100 cycle average.

Figure 6. ENSYN oil ignition delay versus charge temperature at SOI. 2400 RPM, 5.0 bar IMEP, 100 cycle average.

Figure 7. Comparison of apparent activation energies for ENSYN, NREL, and diesel fuels. 2400 RPM, 5.0 bar IMEP, 100 cycle average. (R2 > 0.9 in all cases.)

temperature at start of injection (TSOI), in Figure 7 for the pyrolysis and diesel fuels. When plotted this way the apparent activation temperature (EA/R) is approximated by the slope of the logarithm of time and inverse temperature. Greater apparent activation energy implies that chemistry is more important relative to mixing and evaporation processes, as chemical kinetic rates vary exponentially with temperature whereas mixing processes vary at most linearly with temperature. As shown in the figure the apparent activation energies of the pyrolysis oils are more than two times greater than that of the diesel fuel. It should be noted that when calculated this way the apparent activation

Energy & Fuels, Vol. 14, No. 2, 2000 263

Figure 8. Diesel fuel. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5 bar IMEP. TSOI ) 870 K.

energy is not a true chemical property of the fuel, as it lumps physical (vaporization and mixing) and chemical processes together. It can also be seen that the NREL oil ignition delay is significantly lower than that of the ENSYN oil, though both exhibit approximately the same temperature dependence. Moreover, at the higher temperatures the ignition delay of the pyrolysis oils approaches that of diesel fuel, which is consistent with the explanation that the longer ignition delay is more likely due to slow chemical kinetics, than poor atomization, vaporization, or mixing. This hypothesis is developed further in the Analysis section. Heat Release. Instantaneous heat release data for three conditions corresponding to TSOI ) 870, 920, and 950 K are plotted for the NREL, ENSYN, and diesel fuels in Figures 8 through 16. We found that the diesel cases consistently exhibited the greatest peak heat release, followed by the NREL and ENSYN oils, respectively. The intense fluctuations characteristic of the diesel cases are due to pressure oscillations recorded by the pressure transducer, and whose frequency (ca. 4000 Hz) was found to correspond to the characteristic acoustic frequency of the cylinder chamber. This acoustic phenomenon was reported by the engine manufacturer as inherent to the design, and has been widely reported in the literature. The intensity of the oscillations can be seen to attenuate as the charge temperature was increased, likely due to the fact that the ignition delay decreased, resulting in a less steep pressure rise and smaller heat release spike. In the case of Figure 10 (TSOI ) 950 K) the oscillations are less intense, and one can see the two phases usually associated with diesel combustion: a strong initial heat release rate, usually attributed to the “premixed” combustion of fuel after the ignition delay, followed by a “diffusion-controlled” stage of slower heat release, where the rate of combustion is governed by the rate of fuel injection. In contrast, the pyrolysis oil heat release traces shown in Figures 11-16 do not exhibit any discernible qualitative shift from one combustion mode to another. Furthermore, with the pyrolysis oils the peak heat release rates occur more than 50% later relative to start of combustion (SOC) than with the diesel fuel, even when the ignition delay is the same. As a result, by the time the peak heat release is reached, more than 40% of the

264

Energy & Fuels, Vol. 14, No. 2, 2000

Shihadeh and Hochgreb

Figure 9. Diesel fuel. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5 bar IMEP. TSOI ) 920 K.

Figure 12. NREL oil. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5.0 bar IMEP. TSOI ) 920 K.

Figure 10. Diesel fuel. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5 bar IMEP. TSOI ) 950 K.

Figure 13. NREL oil. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5.0 bar IMEP. TSOI ) 950 K.

Figure 11. NREL oil. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5.0 bar IMEP. TSOI ) 870 K.

fuel has been consumed, in contrast to the diesel cases in which less than 20% has burned. Thus the combustion process differs markedly between pyrolysis oil and diesel fuel, quite apart from ignition delay. The pyrolysis oil heat release profiles are qualitatively consistent with those obtained with the rapid compression machine experiments by Colella et al.14 of diesel fuel combustion, in which long ignition delay and slow combustion chemistry induced by low charge temperatures yielded purely premixed combustion with relatively low heat release rates. Their data showed that in addition to reducing the peak heat release magnitude, slower chemical kinetics resulted in reduced rate of instantaneous heat release (the slope of the instanta-

Figure 14. ENSYN oil. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5 bar IMEP. TSOI ) 870 K.

neous heat release curve) in the early combustion phase, resulting in delayed peak heat release timing relative to SOC. Figure 17 shows the characteristic burn duration for the three fuels versus charge temperature at SOI. The characteristic burn duration is defined as the time elapsed from SOC to the point at which the cumulative heat release fraction (normalized by the lower heating value of the injected fuel) reaches a value of (1 - e-1) ) 0.632, corresponding to the characteristic time of the (14) Colella, K. J.; Balles, E. N.; Ekchian, J. A.; Cheng, W. K.; Heywood, J. B. A Rapid Compression Machine Study of the Influence of Charge Temperature on Diesel Combustion. SAE Paper 870587, 1987.

Diesel Engine Combustion of Biomass Pyrolysis Oils

Figure 15. ENSYN oil. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5 bar IMEP. TSOI ) 920 K.

Energy & Fuels, Vol. 14, No. 2, 2000 265

oils exhibited indicated thermal efficiencies similar to those with diesel fuel, the combustion with the former was characterized by much greater apparent ignition activation energies and longer ignition delays. Despite the longer time for mixing before ignition, operation with pyrolysis oil showed lower peak heat release and pressure rise rates. The oils would not ignite without combustion air preheating. Furthermore, whereas the diesel heat release profiles were consistent with the classic two-stage (kinetic and mixing control) combustion description, the pyrolysis heat release traces did not exhibit any perceptible qualitative shift from kineticto mixing-controlled phases, and also gave 50% later peak heat release timing relative to SOC. As a result, nearly 40% of the fuel burned prior to the peak, in comparison to less than 20% for diesel cases, even when the ignition delays were equal. The heat release profiles were consistent with previous research on diesel combustion in which chemical kinetics had been suppressed by low temperatures, suggesting that slow chemistry may underlie the observed pyrolysis oil heat release profiles, andsbecause of the high apparent activation energies of ignitionsmay also underlie the long ignition delays. In summary, the pyrolysis oils exhibited behavior symptomatic of slow ignition and combustion chemistry in comparison to diesel fuel. Analysis

Figure 16. ENSYN oil. Instantaneous heat release normalized by heat input per cycle, averaged over 100 cycles. 2400 RPM, 5 bar IMEP. TSOI ) 950 K.

Figure 17. Characteristic burn duration versus charge temperature. 2400 RPM, 5.0 bar IMEP, 100 cycle average.

integral of an exponential decay. As shown in the figure, the burn duration is generally shorter for the pyrolysis oils than for diesel fuel, which is partly explained by the fact that they also have longer ignition delay, allowing more fuel and air to mix prior to SOC; when the ignition delays are equalsabove approximately 920 K TSOIsthere is no significant difference in burn duration. This is consistent with published data which show that premixed burn fraction is proportional to ignition delay.15 Summary of Experimental Results. While the operation of a compression-ignition engine on pyrolysis (15) Watson, N.; Pilley, A. D.; Marzouk, M. A Combustion Correlation for Diesel Engine Simulation. SAE Paper 800029, 1980.

Heat release profiles contain information about the mixing and chemical processes prior to and during combustion. For example, a relatively large heat release spike in the early combustion stage suggests that a significant amount of fuel has premixed with air prior to the start of combustion. If this condition is observed even when there is a relatively short ignition delay, it would suggest a relatively rapid fuel-air mixing process during the ignition delay period. On the other hand, if the ignition delay is relatively long, one would expect a greater peak heat release rate due simply to the fact that the fuel had more time to mix with air. Thus with a combustion model relating fuel properties, ignition delay, and heat release, one could predict what sort of heat release profile would be expected, for example, in a case where long ignition delay results from slow fuelair mixing (as opposed to slow chemistry), and compare this to experimental data to determine whether the observed heat release and ignition delay are consistent with this picture. The model employed in this work was formulated by Hiroyasu et al.16 and updated and commercially packaged by GTI, Inc. It accounts for the in-cylinder processes of fuel-jet formation, its breakup into droplets, entrainment of air, evaporation of fuel droplets, mixing of air with fuel vapor, ignition, and combustion.17 The model and its verification against experimental data are presented in more detail in Appendices B and C, respectively. Vaporization, Ignition, Combustion Rates. Experimental heat release and ignition delay data were (16) Hiroyasu, H.; Kadota, T.; Arai, M. Development and Use of Spray Combustion Modeling to Predict Diesel Engine Efficiency and Pollutant Emission. Bull. JSME 1983, 26 (214), 569-575. (17) Morel, T.; Wahiduzzaman, S. Modeling of Diesel Combustion and Emissions, Gamma Technologies Report, 1996.

266

Energy & Fuels, Vol. 14, No. 2, 2000

Shihadeh and Hochgreb

Figure 18. Experimental and computed heat release and cylinder pressure for NREL Case II. Mv ) 8.0, Mi ) 1.4, Mc ) 0.29.

interpreted in terms of three major fuel-side processes: vaporization, ignition, and combustion. In particular, simulations were made using default diesel fuel values for all constants appearing in equations B.1-B.6 and adjusting the governing rate equations using multipliers for ignition delay (Mi), combustion (Mc), and vaporization (Mv) until the predicted and experimental ignition delay and heat release data agreed. The algorithm for searching for the unique set of Mi, Mc, and Mv which gives agreement between predictions and experimental data relies on matching the predicted and experimental ignition delay and peak heat release magnitude and timing to within a specified tolerance (see Appendix D). The method yields good agreement between experimental and predicted cylinder pressure and heat release, as illustrated in Figure 18 for NREL oil case II, for which Mi, Mc, and Mv are equal to 1.4, 0.29, and 8.0, respectively. The results in this case mean that relative to diesel fuel at these operating conditions, the NREL oil vaporizes more rapidly and has slower combustion and ignition chemistry. While a parametric study of ignition, combustion, and vaporization multipliers is presented elsewhere,18 it is sufficient here to point to the impact of independently varying the combustion rate multiplier. As shown in Figure 19, the combustion rate multiplier affects the (18) Shihadeh, A. Rural electrification from local resources: Biomass pyrolysis oil combustion in a direct injection diesel engine. ScD Thesis, MIT Department of Mechanical Engineering, September, 1998.

Figure 19. Impact of combustion multiplier on cylinder pressure (a), heat release (b), liquid fuel (c), and unburned fuel profiles (d). Mc varied from 0.2 to 1.0. Mi, Mv ) 1.

Diesel Engine Combustion of Biomass Pyrolysis Oils

Energy & Fuels, Vol. 14, No. 2, 2000 267

Table 3. Pyrolysis Oil Cases Modeleda fuel/case

NREL I

Tcyl at SOI [K] fueling [mg/cyc] SOI [DATC] ignition delay [CAD]

870 28 -9 9.9

II

III

920 950 32 34 -4 -3 6.2 4.9

ENSYN I 870 33 -13 12.1

II

III

920 950 35 37 -7 -4 7.9 6.2

a Air inlet temperature varied from 55 to 89 °C, SOI varied from -3 to -8 DATC. Engine run at 2400 rpm at constant load. SOI temperature calculated from cylinder pressure.

Table 4. Combustion, Vaporization, and Ignition Multipliers Used to Match Experimental Data for Each Case fuel/case

NREL I

combustion, Mc vaporization, Mv ignition, Mi

0.25 1.0 2.1

II

III

0.29 0.39 8.0 10.0 1.4 1.0

ENSYN I 0.23 0.2 2.0

II

III

0.18 0.17 2.0 2.0 1.5 1.2

Table 5. Experimental and Predicted Instantaneous Heat Release Magnitude and Timing Obtained Using Multipliers Given in Table 4a fuel/case

NREL I

peak magnitude [1/deg] model [1/deg] peak timing [DASI] model [DASI]

0.15

0.12 0.11

II

III ENSYN I 0.13

0.15 14.7 15.0

0.12 0.11 11.4 8.4 11.3 8.0

0.13 17.3 17.3

II

III

0.11

0.08

0.11 0.08 14.1 12.8 13.8 12.8

a Heat release magnitude normalized by fuel energy input per cycle.

heat release profile by shifting the magnitude and timing of the peak (at constant ignition delay), with slower combustion yielding later and lower peak heat release and cylinder pressure. While the in-cylinder liquid fuel mass profile is not significantly affected by changing the combustion rate, the amount of in-cylinder unburned fuel increases as the combustion rate multiplier is reduced, meaning that a greater fuel-vapor buildup occurs. Thus as the combustion chemical kinetics are slowed, the heat release profile essentially loses its “mixing controlled phase” and the entire combustion process becomes “premixed.” This can be seen by comparing the heat release of the 0.4 and 1.0 combustion multiplier cases. In the latter, and consistent with the classical DI combustion model, a rapid heat release phase (caused by the buildup of a fuel-air charge prior to SOC) is followed by a second combustion phase with a lower characteristic heat release rate determined by the rate at which fuel-air charge is prepared. In the former case, however, because the chemical kinetics are slow, there is an abundance of fuel-air charge present during the bulk of the combustion process, and as a result the heat release profile never switches to a “mixing controlled” curve, as in the pyrolysis oil cases. Modeling Results. For each pyrolysis oil, representative 100-cycle ensemble averaged pressure and needle lift traces from each of the three experimental operating conditions listed in Table 3 were used for the simulations. The resulting rate equation multipliers are given in Table 4, and the corresponding peak heat release magnitudes and timings are listed in Table 5. It can be seen that the pyrolysis oil combustion kinetics are slower (i.e., it requires a lower multiplier), the vaporization as fast or faster (except in one case), and the ignition delay chemistry slower than would be predicted for diesel fuel properties. It should be noted that the

parameters are interdependent; for example, in NREL case II, the ignition delay is 1.4 times longer than predicted given that the vaporization multiplier is 8.0. With respect to the longer experimentally observed ignition delays of the pyrolysis oils at the lower cylinder temperatures, the model shows that this could not result from slow vaporization alone. In fact, with the simulation case I for the NREL oil, the model suggests that the longer delay is entirely due to slow ignition chemistry, as the experimental heat release profile is consistent with the same vaporization rate that would be predicted for diesel fuel (Mv ) 1.0), while the ignition chemistry is a more than a factor of 2 slower (Mi ) 2.1). Even with the simulation case I for the ENSYN oil, where the vaporization rate multiplier is 0.2 (the vaporization rate being one-fifth that predicted for diesel fuel), an ignition delay multiplier of 2 is required to match the experimental data, which means here that slow vaporization alone is insufficient to explain the long experimental delay; if it were, the ignition delay multiplier would be equal to unity. It should be noted that ENSYN I is the only case whose heat release profile implied slower vaporization than would be predicted for diesel fuel under the same conditions. The fact that slow vaporization cannot alone explain the longer ignition delays can also be shown directly with the model by reducing the vaporization rate multiplier (leaving the ignition delay multiplier equal to unity) such that the ignition delay of the pyrolysis oils is replicated. When this is done, it is found that the predictions yield excessively low heat release rates. In the simulation case I for the NREL fuel (Figure 20), the vaporization rate multiplier had to be reduced to a value of 0.07 in order to reproduce the experimental ignition delay of 9.9 CAD. It can be seen in the figure that the calculated peak heat release is approximately half the experimental peak heat release. This remains true even if the combustion multiplier is increased because the combustion is now overwhelmingly mixing limited. In sum, the experimentally observed heat release profiles are characteristic of much greater vaporization rates than would be required to reproduce the longer experimentally observed ignition delays through slow vaporization alone. This is found to be true with both pyrolysis oils under all conditions tested. As shown in Table 4, the ignition delay multipliers decrease with increasing cylinder temperature (from Case I to Case III), meaning that the ignition kinetics of both oils are more sensitive to temperature than is diesel fuel; i.e., that they have higher apparent global activation energies of ignitions4200 to 4300 K for the pyrolysis oils versus 3500 K for diesel fuel, as shown in Table 6. The ignition delay constants given in Table 6 were fit to the experimental data for each pyrolysis oil by using the input activation energy to match the slope of the experimental ignition delay curve (as a function of TSOI), and then using the preexponential term (Ai in eq A.4) to shift the entire curve up or down to match the data. It can also be seen in Table 4 that for the NREL oil, the combustion multiplier increases by more than 50% from Case I to Case III, while with the ENSYN oil decreases by 15% for the same range. This suggests that the apparent activation energy of combustion of the

268

Energy & Fuels, Vol. 14, No. 2, 2000

Shihadeh and Hochgreb

Figure 21. In-cylinder fuel. Case (a), all multipliers equal to one, representative of diesel combustion with experimental setup. All curves normalized by total fuel injected per cycle.

Figure 20. Predicted and measured cylinder pressure (a) and heat release rate (b) when correct ignition delay is obtained by reducing vaporization rate. NREL Case I; Mv ) 0.07; Mc, Mi ) 1.

Figure 22. In-cylinder fuel. Case (b), Mc ) 1, Mv ) 8.0, Mi ) 1.5. Represents NREL II hypothetical case where combustion rate equals that of diesel.

Table 6. Ignition Properties Implied by Matching Experimental Ignition Delay Data for Cases I-III When Mi ) 1.0 and Mv Is Varied According to Table 4a ignition Ai* ignition Ci (Ea/R), K

NREL

ENSYN

diesel

0.7 4300

1 4200

1 3500

a Pre-exponential term A is normalized by diesel fuel value. (See i Equation B.4.).

NREL oil is greater than, and that of the ENSYN oil less than that of diesel. In all cases the NREL oil combustion remains more rapid than ENSYN oil combustion, though both burn more slowly than diesel fuel. This slower chemistry could in principle result from lower local temperatures (due to a high heat of vaporization resulting from the high water content), and/or from the chemical structure of the oils. These two possibilities are currently under study in an effort to elucidate relationships between combustion characteristics and the pyrolysis oil production process. Biomass Oil Combustion Characterization. Figures 21-24 show the normalized in-cylinder injected, unburned, and liquid fuel mass calculated using NREL case II operating conditions and three different sets of rate equation multipliers: (a) all multipliers set to unity (default diesel), (b) high evaporation rate (Mv ) 8.0, Mi ) 1.4, and Mc ) 1), (c) high evaporation rate and slow combustion chemistry (Mc ) 0.29, Mv ) 8.0, and Mi ) 1.4). Case (c) represents the best fit of the NREL

Figure 23. In-cylinder fuel. Case (c) calculated in-cylinder fuel distribution versus crank angle for NREL Case II. Mc ) 0.29, Mv ) 8.0, Mi ) 1.5.

fuel data, while case (a) represents a diesel fuel simulation. All have the same ignition delay of 6.2 CAD. These plots capture the essential differences of diesel and pyrolysis oil combustion from a macro-perspective. For this purpose, NREL case II can be considered typical. First, the effect of rapid vaporization can be seen by comparing cases (a) and (b) in Figures 21 and 22. The liquid fuel mass curves illustrate the more rapid vaporization of case (b), as well as a greater buildup of fuel vapor up to and during the premixed combustion phase. In both cases, following the early premixed phase, the

Diesel Engine Combustion of Biomass Pyrolysis Oils

Figure 24. Fuel vapor fraction of unburned fuel versus burned fraction. Burned fraction of 0+ corresponds to SOC, 1 corresponds to EOC.

unburned fuel mass follows the liquid fuel mass profile, consistent with a mixing controlled process. As expected, the overall burn rate is greater with the greater evaporation rate of case (b). Second, considering only the impact of slow combustion kinetics, Figures 22 and 23 show that the greater buildup of fuel vapor in case (c) results from the fact that a combustible charge is prepared faster than it can be consumed by chemical processes. In contrast, the case (b) simulation shows that after the initial premixed combustion phase, the unburned fuel closely follows the liquid fuel curve, indicating that the combustion proceeds according to the availability of fuel-air charge. Fuel vapor as a mass fraction of unburned fuel is plotted versus fuel burn fraction for the three cases in Figure 24. Whereas in case (c), for the best fit of NREL fuel data, the average fuel-vapor fraction throughout the combustion event remains near 40%, in case (a), the diesel simulation, the average vapor fraction is near 20%. The fact that the curve (b) approaches curve (a) following the early premixed combustion phase indicates that the air entrainment and reaction rate is sufficient to allow most of the excess fuel vapor which results from the large vaporization multiplier to burn. In fact, under these conditions the combustion is limited only by the rate at which fuel vaporizes, i.e., the combustion is mixing controlled. Most importantly, this means that the excess fuel vapor built up in case (c) is actually mixed with enough air to burn, and that the combustion in this case is always kinetically limited, a result that holds for all the pyrolysis oil cases. Apart from the longer ignition delay, this is the major characterizing difference between pyrolysis oil and diesel fuel combustion. Conclusions The purpose of this work was to characterize how the combustion of biomass pyrolysis oils in a high-speed diesel engine differs from that of No. 2 fuel oil. This was done by comparing the ignition delay and heat release data of diesel fuel and two wood pyrolysis oils made using the ENSYN Rapid Thermal Process, and NREL Vortex Ablative Pyrolysis. Combustion experiments demonstrated that while the pyrolysis oils exhibited similar indicated thermal efficiencies to diesel fuel, they were characterized by much greater apparent ignition

Energy & Fuels, Vol. 14, No. 2, 2000 269

activation energies, longer delays, shorter burn duration, and lower peak heat release and pressure rise rates. Most problematic was that neither pyrolysis oil could ignite without combustion air preheating. We undertook a modeling study to interpret the experimental heat release and ignition delay data in terms of characteristic vaporization rates, and ignition and combustion kinetics relative to diesel fuel. In summary, the results showed the following: (1) The indicated thermal efficiency of both the pyrolysis oils equaled that of the diesel fuel, though they exhibited excessively long ignition delay and required a moderate degree of combustion air preheating (55 °C) to ignite reliably. (2) The longer ignition delay time observed with the pyrolysis oils cannot be attributed solely to poor vaporization or atomization characteristics, but is rather associated with the chemical composition of the fuel. The apparent global activation energy of ignition (EA/R) for the pyrolysis oils was found to be 4200-4300 K, in comparison to 3500 K for the diesel fuel. In general, the heat release profiles were consistent with faster fuel vaporization rates than would be predicted for diesel fuel, though this may not be the case at the lower cylinder temperatures in which the oils did not ignite. The slow reaction rates implied by the long ignition times are also reflected in the slow combustion rate of pyrolysis fuels. As a result, whereas diesel combustion is predominantly limited by mixing after ignition, combustion of pyrolysis oils is apparently limited by the chemical reaction throughout much of the process. Acknowledgment. The authors thank John Heywood, Brian Corkum, Peter Menard, Nancy Cook, James Diebold, Stefan Czernik, Wai Cheng, Janos Bee´r, Don Huffman, and Sayed Wahiduzzaman for their contributions in formulating or executing this work. Financial support was provided by the US Department of Energy and ENEL Societa′ per Azioni (Italy). Appendix A: Special Considerations for Burning Pyrolysis Oil in a Diesel Engine Because the pyrolysis oils contain significant amounts of char and are known to polymerize when exposed to high temperatures,19 several modifications to the fuel injection system were necessary to obtain reliable engine operation during the experimental runs. The primary hindrance to continuous operation was the abrasive wear of the fuel pump plunger-barrel assembly and injector nozzle caused by the char, as well as the apparent in situ growth of solid particles larger than could be traced to the parent fuel. As is typical with inline diesel injection pumps, the fuel was continuously cycled through the pump to provide cooling, while a small fraction of the internal flow was delivered to the injector. Stainless steel mesh filters (40 µm) were installed in the circulation loop to capture particles formed within the system, in addition to pre-filtering the oils in a batch method with a separate pressurized filtration rig employing a 10 µm paper element automo(19) Czernik, S. Storage of Biomass Pyrolysis Oils. Biomass Pyrolysis Oil Properties and Combustion Meeting, Proceedings; U.S. National Renewable Energy Lab: Estes Park, CO, September 1994.

270

Energy & Fuels, Vol. 14, No. 2, 2000

tive type oil filter. As a further measure, the fuel injector spill return, which is normally reintroduced to the fuel pump, was diverted to an external reservoir, since it was thought that much of the particulate formation occurred within the hot environment of the fuel injector body. In addition to these modifications, the fuel system was configured to allow on-line switching between diesel, nitrate-enriched ethanol, and the pyrolysis oils, so that the fuel pump and injector could be flushed with ethanol between data collection (typically every 15 min) to prevent gumming, and also to reduce scaling of the combustion chamber surfaces, since alcohol is an effective detergent for pyrolysis oil. Furthermore, since diesel fuel and pyrolysis oil are not miscible, ethanol, which is miscible in both, was used as an intermediary when switching from diesel fuel (which was used to warm the engine) to pyrolysis oil. At the end of each experiment, the engine was run for 30 min with ethanol, and then switched to diesel fuel for an additional 20 min before shut-down. Because ethanol has poor lubrication and ignition properties, Lubrizol 9520A additive was used (0.1 vol %) for lubricity, and di-ethylhexylnitrate was added (15 vol %) to ensure good ignition at all operating conditions. The Lubrizol additive was also used with the pyrolysis oils. This operational procedure and fuel filtering scheme was chosen after several design iterations and pump failures. Even with these measures however, engine operation could only be sustained with reasonable performance for approximately 6 h, after which the pump required over-hauling, and the nozzle required replacement. Thus to ensure consistent data and engine performance across tests, the pump was overhauled and the nozzle replaced after each day’s runs (typically less than 3 h on pyrolysis oils). Compounding the fuel injection problems was the buildup of carbon deposits in the combustion chamber and exhaust valve and port which required that the head, valves, and piston be removed and cleaned, the cylinder wall scraped clean, and the valves re-seated at the end of each day’s experiments. Clearly, this is not a desirable mode of operation in any practical situation, though it did allow repeatable combustion measurements to be taken. Appendix B: Model Description The model treats the spray issuing from the fuel injector nozzle as a series of discrete packets in the axial and radial directions. For simplicity, it is assumed that no intermixing among the packets occurs. Each packet entrains air as it travels through the combustion chamber, at a rate determined from the momentum conservation equation which is solved using an empirically prescribed ordinary differential equation for fuel jet penetration as a function of time (i.e., spray velocity as a function of radial and axial distance from the nozzle). As each packet (initially containing only liquid fuel) travels through the cylinder entraining air, fuel is assumed to vaporize according to single droplet evaporation equations (after an initial jet column breakup time), and the vapor mixes instantly with the air in the packet. The local equivalence ratio is calculated from the fuel vapor and air in each package. When a fuel-

Shihadeh and Hochgreb

air mixture in a packet is within the combustion limits, the prepared fuel-air charge is allowed to burnsat a rate depending on the local temperature and fuel equivalence ratiosafter some ignition delay period which is calculated using an Arrhenius rate expression. (Therefore as the combustion proceeds, and the temperature increases, the significance of the chemical delay for each packet decreases, and the combustion approaches a “mixing-controlled” process.) Thus each package consists of 3 subzones: liquid droplets, vaporized fuel plus entrained air, and products. At every time step, the heat released from each package is summed, and the overall cylinder temperature and pressure are updated according to the energy conservation equation, which is solved at each time step for the individual subzones and the cylinder as a whole. Each package has a different temperature, which varies with time and space, and is different from the mass average temperature calculated from the cylinder pressure. The thermodynamic properties of the air and eleven individual species (N2, O2, CO2, H2O, CO, OH, NO, H2, O, H, and N) are described by JANAF polynomial fits, and during combustion the mixture is treated on the basis of kinetic equilibrium between them. Key model inputs are listed in Table B1. The detailed equations governing the various processes can be found in refs 16 and 17. It is useful for the ensuing discussion, however, to present the rate-governing equations for fuel vaporization, ignition delay, and combustion below. Droplet Vaporization. The droplets in each fuel package are assumed to vaporize according to the 1-D single droplet diffusion and continuity equations. Thus the evaporation rate per droplet is given by

dmv/dt ) MvGπD2d

(B.1)

where Mv is a user-input “vaporization multiplier,” and Dd is the instantaneous droplet diameter. In the case of diffusion-limited evaporation, the mass flux, G, is given by

G ) F Sh R ln(1 + β)/Dd

(B.2)

where F is the droplet density, R is the mass diffusion coefficient, and β is the mass transfer number, defined as β ) (Yo - Y∞)/(1 - Y∞). The Sherwood number is given by Sh ) 2 + 0.57Sc1/3 Re1/2. In the case of boiling-limited evaporation, the vaporization rate is proportional to the heat transfer rate and thus G becomes

G ) Q/hfg

(B.3)

where Q is the heat flux to the droplet and hfg is the latent heat of vaporization (which is treated as a constant during the droplet lifetime). The heat flux is calculated assuming convective transfer only, and follows the well-known Nusselt number correlation of Ranz and Marshall.20 The evaporation regime is determined by calculating the fuel vapor pressure at the droplet surface. If it exceeds the cylinder pressure, the evaporation is switched to the boiling-limited mode. Ignition. The ignition delay is calculated separately for each packet using an Arrhenius type expression (20) Ranz, W. E.; Marshall, W. R. Evaporation of Drops. Chem. Eng. Prog. 1952, 48, 141.

Diesel Engine Combustion of Biomass Pyrolysis Oils

τ ) MiAi/[φ(3 - φ)2]pBi exp(Ci/T)

Energy & Fuels, Vol. 14, No. 2, 2000 271

(B.4)

where Mi is a user-input “ignition delay multiplier,” φ is the local vapor-phase fuel equivalence ratio, and T and p are the local temperature and pressure. The constants Ai, Bi, and Ci are input to the program, and assume default values based on correlation with published data. Note that Ci is the activation energy divided by the gas constant, yielding an “activation temperature.” Because cylinder and local conditions (p, T, φ) are changing continually, the ignition delay calculated at each time step from eq B.4 for a given fuel packet is different. To circumvent this problem, this model calculates the ignition delay by integrating the kinetic rate expression and assumes that ignition occurs when the value of the ignition integral, I, equals one:

I)

∫1τ dt

Figure C1. Experimental and predicted ignition delay for Cases 1 through 5. 4.8-4.9 bar IMEP, 55-89 deg C air inlet, -3 to -8 DATC SOI. Experimental delay determined by averaging data over 100 cycles.

(B.5)

conditions. A sample of the more important inputs are listed in Table B1.

Combustion. Under normal diesel engine conditions, combustion proceeds at a rate determined by the availability of fuel vapor and air mixed at the molecular level to combustible proportions, and the combustion kinetics provide no additional limit on the rate of combustion. However, when the local temperature is sufficiently low, or the mixture sufficiently fuel lean, chemical kinetics can become the rate-limiting step. In this case the rate of combustion is determined by Equation B.6

Appendix C: Model Calibration The model was calibrated using measured engine data for a number of test conditions in which the injection timing and air inlet temperature were varied over a range of typical experimental conditions. Ignition delay, cylinder pressure, and cumulative heat release predictions were compared to the experimental data, and found to be in good agreement. The cases modeled are described in Table C1 and the assumed diesel fuel properties (defaults in the model) are given in Table C2. The primary diagnostics used for comparing the model to the data were the ignition delay, cylinder pressure, and heat release traces. Ignition Delay and Cylinder Pressure. As shown in Figure C1, the predicted delay agrees very well with the experimental data (with all multipliers equal to unity). The peak cylinder pressure and timing for the 5 cases tested are given in Figure C2, and, as shown, are slightly over-predicted by the model. The average error in peak cylinder pressure and timing relative to SOI over the five cases is 6.6 and 2.0%, respectively. The error partly derives from the fact that the experimental

dmk/dt ) McAcφ(3.0 - φ)2pBc exp(-Cc/T) (B.6) where mk is the mass of fuel in a given packet. Mc can be input as a “combustion rate multiplier.” The constants Bc and Cc are fixed at values of 2.5 and 400 K, respectively. The model transfers the calculated mass of fuel which burns in a given time step along with a mass of air proportional to the local fuel/air ratio from the unburned to the burned subzone of a given packet. Energy will continue to be released due to the oxidation of partial combustion products from the burned zone as it entrains air. Inputs to the program include engine geometry, fuel injection profile, fuel properties, and cylinder initial

Table B1. Model Inputs engine data

fuel injection

fuel properties

compression ratio bore stroke connecting rod length piston cup diameter and depth piston TDC clearance height wall, head temperatures cylinder P and T at IVC

mass of fuel injected/cycle number of nozzle holes nozzle diameter injection timing, duration

LHV C/H/O content liquid density heat of vaporization 10, 50, 75% distillation temp ignition activation energy

Table C1. Diesel Fuel Cases Used to Verify Modela case

IMEP [bar]

fueling [mg/cyc]

SOI [DATC]

EOI [DATC]

air flow [g/s]

air temp [K]

[email protected] [bar]

[email protected] [K]

1 2 3 4 5

4.8 4.9 4.8 4.8 4.8

12 12.3 12.8 12.5 12.5

-3 -3 -4 -5 -8

8 9 8 7 3

9.5 9.2 8.7 9.2 9.2

55 67 89 67 67

1.14 1.12 1.13 1.16 1.14

361 373 395 373 373

a Air inlet temperature varied from 55 to 89 °C, SOI varied from -3 to -8 DATC. Engine run at 2400 rpm at constant load. IVC temperature calculated from cylinder pressure using ideal gas law.

272

Energy & Fuels, Vol. 14, No. 2, 2000

Shihadeh and Hochgreb

Figure C2. Experimental and predicted peak cylinder pressure and timing relative to SOI for the 5 cases.

Figure C3. Measured and predicted cylinder pressure for Case 1.

Figure C4. Measured and predicted normalized heat release rate for Case 1.

Figure C5. Characteristic burn duration defined.

Table C2. Default Diesel Fuel Properties in Model lower heating value heat of vaporization liquid density ignition delay activation temperature ignition delay pressure exponent distillation temperature

42.5 0.25 830 3500 -1.25 490 K 542 K 562 K

MJ/kg MJ/kg kg/m3 K 10 wt % 50 wt % 75 wt %

cylinder pressure oscillations cannot be predicted by the model, though this does not diminish its utility for the present purposes. Figures C3 and C4 show the predicted and experimental cylinder pressure and heat release profiles for Case 1. Heat Release. While the predicted ignition delay and cylinder pressure agree closely with the experimental data, the heat release is more difficult to compare due to the experimentally recorded pressure oscillations during combustion, which are reflected in the dp/dθ term in the heat release calculation by the oscillations shown in Figure C4. In addition, because the instantaneous slope of the recorded cylinder pressure trace during these oscillations is very different from that of the mean cylinder pressure, the resulting heat release rate calculation can yield unrealistically large positive and negative values. Thus rather than attempting to compare particular features (e.g., the peak heat release and timing) of the experimental and calculated heat release rate, the characteristic burn duration (the integrated, or cumulative heat release) were compared. Because the heat release profile resembles an exponen-

Figure C6. Experimental and predicted burn duration for Cases 1 through 5.

tial decay, the characteristic burn duration for the comparison was specified as the time elapsed from SOI to the point at which the cumulative normalized heat release reached a value of (1 - e-1) ) 0.632. This is shown schematically in Figure C5. When compared this way the predicted and experimental data agree well (2% average error for the 5 cases, max error ) 4.5%), as shown in Figure C6. Taken together, the ignition delay, cylinder pressure, and heat release data demonstrate that the model provides a good representation of the mean physical processes leading up to and during combustion.

Diesel Engine Combustion of Biomass Pyrolysis Oils

Energy & Fuels, Vol. 14, No. 2, 2000 273

Figure D1. Impact of vaporization and combustion rate multipliers on peak heat release timing relative to SOC. (Case 1 operating conditions.)

Figure D3. Obtain correct peak heat release timing. Mc ) 0.29.

Figure D2. Calculated and experimental cylinder pressure (a) and apparent heat release (b) for NREL Case II when Mc, Mi, Mv ) 1.

Appendix D: Multiplier Methodology As noted, it is possible to interpret the experimental pyrolysis oil heat release data through the manipulation of the vaporization, ignition delay, and combustion rate multipliers. In particular, it is possible to search for a combination of the three multipliers (ignition delay, vaporization, combustion) which would produce the same heat release profile (using diesel fuel properties) as the experimental data for each experimental operat-

ing condition. In this way the experimental results can be interpreted using physical arguments, i.e., that in relation to diesel fuel, the pyrolysis oils vaporize faster, burn slower, and so on. The procedure for searching for the three multipliers begins with obtaining the correct ignition delay for the experimental case under question. Once the correct ignition delay is obtained with the ignition delay multiplier, the combustion multiplier is adjusted so that the peak heat release occurs at the proper timing with respect to SOI. Finally, the vaporization multiplier is adjusted to provide the correct peak heat release magnitude. Since the vaporization multiplier affects the ignition delay, it is sometimes necessary to iterate by readjusting the ignition delay multiplier. The tolerances employed in the procedure were that the ignition delay, peak heat release magnitude, and peak heat release timing would agree to within 2, 5, and 5% of the experimental data, respectively. This method converges to a unique solution because only the combustion multiplier can significantly affect the location of the peak heat release relative to SOC. As shown in Figure D1, the impact of vaporization multiplier on peak heat release timing relative to SOC is negligible over a 3 orders of magnitude variation. In contrast, varying the combustion rate multiplier from 1.0 to 0.2 increased the time to peak heat release by a factor of 2.5. Thus once the ignition delay and correct peak heat release timing are obtained, only the vaporization rate manipulation can yield the correct magnitude of the peak release. Physically, this is explained by the fact that the positive slope of the early combus-

274

Energy & Fuels, Vol. 14, No. 2, 2000

Shihadeh and Hochgreb

ignition delay, vaporization, and combustion multipliers are equal to one (the baseline diesel case), the resulting heat release profile does not match the experimental data (Figure D2). The peak heat release is approximately 20% less than, and the peak heat release timing relative to SOC 30% earlier than the experimental data. To match the experimental peak heat release timing and magnitude, the combustion rate multiplier was reduced to a value of 0.29, and the vaporization multiplier was increased to 8.0, respectively (Figure D3). This greater vaporization rate required that the ignition delay multiplier be increased from 1 to 1.4 to maintain the correct ignition delay of 5.5 CAD. As shown in Figure D4, the calculated heat release and pressure traces produced by this method agree well with the experimental data. Physically, the results mean that relative to diesel fuel at these cylinder conditions, the NREL oil vaporizes more rapidly, burns more slowly, and takes longer to ignite.

Nomenclature

Figure D4. Obtain experimental peak heat release magnitude and ignition delay (a). Resulting predicted and experimental cylinder pressure b. Mv ) 8.0, Mi ) 1.4, Mc ) 0.29.

tion phase is dictated by chemical kinetics, since the fuel is already premixed, while the maximum heat release is determined by the amount of fuel which has premixed; it reaches a maximum and decreases as the premixed charge is consumed. Figures D2-D4 illustrate the application of the multiplier method to NREL oil case II. When the

BMEP CAD DATC DASOC DSOI EOC EOI φ ID IMEP IVC LHV PM SOC SOI TDC TSOI

brake mean effective pressure crank angle degrees crank angle degrees after top center crank angle degrees after start of combustion crank angle degrees after start of injection end of combustion end of injection fuel equivalence ratio ignition delay indicated mean effective pressure inlet valve closed timing lower heating value premixed start of combustion start of injection top dead center average cylinder charge temperature at start of combustion EF990044X