Stock and Optimized Performance and Emissions ... - ACS Publications

Dec 21, 2009 - Stock and Optimized Performance and Emissions with 5 and 20% Soy Biodiesel Blends in a Modern Common Rail Turbo-Diesel Engine ...... Ag...
3 downloads 13 Views 5MB Size
Energy Fuels 2010, 24, 928–939 Published on Web 12/21/2009

: DOI:10.1021/ef9011033

Stock and Optimized Performance and Emissions with 5 and 20% Soy Biodiesel Blends in a Modern Common Rail Turbo-Diesel Engine Michael Bunce,† David Snyder,† Gayatri Adi,† Carrie Hall,† Jeremy Koehler,† Bernabe Davila,† Shankar Kumar,‡ Phanindra Garimella,‡ Donald Stanton,‡ and Gregory Shaver*,† †

Energy Center, Herrick Laboratories, School of Mechanical Engineering, Purdue University, 585 Purdue Mall, West Lafayette, Indiana 47907 and ‡Cummins, Incorporated, 500 Jackson Street, Columbus, Indiana 47201 Received September 29, 2009

Stock engine design and decision making target optimal performance with conventional diesel fuel, leading to suboptimal results for biodiesel. The main result of this study is the determination of the appropriate engine decision making for the air/fuel ratio (AFR), exhaust gas recirculation (EGR) fraction, injection (rail) pressure, and start of main fuel injection (SOI) in a modern common rail diesel engine using variablegeometry turbo charging and operating with 5% (B5) and 20% (B20) soy-based biodiesel fuel mixtures to minimize brake-specific fuel consumption (BSFC) while adhering to strict combustion noise, NOx and PM emission constraints. In so doing, this effort determined to what extent the optimal AFR, EGR fraction, rail pressure, and SOI settings can (1) overcome the well-known “biodiesel-NOx effect” and (2) mitigate the impact of lower biodiesel energy density on BSFC for B5 and B20 biodiesel blends. Study findings indicate that lower AFR, higher EGR fraction, and earlier start of main injection can completely eliminate biodiesel NOx increases with blends of soy-based biodiesel in a modern diesel engine. While the BSNOx reductions were achieved with acceptable BSPM and noise, it was not possible to reduce the BSFC for B5 and B20 to conventional diesel levels.

particulate matter (PM) emissions.2-6 Much of the reduction in these emissions is credited to the 11% oxygen content by mass of biodiesel.7,2-4 There is also a net carbon dioxide (CO2) reduction largely attributed to the consumption of CO2 in the atmosphere by the crops used to form biodiesel.8,9 Observed trends in biodiesel emissions across a range of fuel blends are displayed in Figure 1. The fuel blends range from conventional diesel (B0) to 100% biodiesel (B100), and the findings are based on experiments involving a large number of pre-1998 model year production diesel engines.3

1. Introduction 1.1. Background. Petroleum-based fuels, especially in the transportation sector, are used to meet a significant portion of modern energy demands. The United States is increasingly being forced to meet this demand with foreign sources of petroleum, a situation that will become exacerbated by an expected 50% rise in demand between 2005 and 2030, with no appreciable increase in domestic supply.1 Suitable alternatives or supplements to petroleum-based fuels are therefore beneficial. Because it is produced from renewable sources and can be produced domestically, biodiesel has emerged as an attractive alternative to petroleum diesel fuel. Biodiesel is formed by reacting triglycerides from vegetable oils or animal fats with an alcohol in the presence of a catalyst to form fatty acid esters and glycerin. The fatty acid esters are what are commonly referred to as biodiesel. This process can use a number of different feedstocks, including soybean oil, rapeseed oil, palm oil, waste vegetable oils, and animal fat byproducts from the meat processing industry, such as lard and beef tallow. Biodiesel is an attractive alternative fuel for diesel engines because it can be used in its pure form or in any combination with diesel fuel due to its miscibility with diesel. When used in diesel engines, biodiesel also usually exhibits several combustion-related advantages, including reductions in carbon monoxide (CO), unburned hydrocarbon (UHC), and

(2) McCormick, R.; Tennant, C.; Hayes, R.; Black, S.; Ireland, J.; McDaniel, T.; Williams, A.; Frailey, M.; Sharp, C. Regulated emissions from biodiesel tested in heavy-duty engines meeting 2004 emission standards. SAE Tech. Pap. 2005-01-2200, 2005. (3) United States Environmental Protection Agency (EPA). A comprehensive analysis of biodiesel impacts on exhaust emissions. Technical Report, 2002. (4) Wang, W.; Lyons, D.; Clark, N.; Gautam, M.; Norton, P. Emissions from nine heavy trucks fuel by diesel and biodiesel blend without modification. Environ. Sci. Technol. 2000, 34, 933–939. (5) Kegl, B. Numerical analysis of injection characteristics using biodiesel fuel. Fuel 2006, 85, 2377–2387. (6) Postrioti, L.; Battistoni, M.; Grimaldi, C.; Millo, F. Injection strategies tuning for the use of bio-derived fuels in a common rail HSDI diesel engine. SAE Tech. Pap. 2003-01-0768, 2003. (7) Yuan, W. Computational modeling of NOx emissions from biodiesel combustion based on accurate fuel properties. Ph.D. Thesis, Iowa State University, Ames, IA, 2003. (8) Sheehan, J.; Camobreco, V.; Duffield, J.; Graboski, M.; Shapouri, H. Life cycle inventory of biodiesel and petroleum diesel for use in an urban bus. Technical Report, National Renewable Energy Laboratory, Golden, CO, 1998. (9) Hill, J.; Nelson, E.; Tilman, D.; Polasky, S.; Tiffany, D. Environmental, economic, and energetic costs and benefits of biodiesel and ethanol biofuels. Proc. Natl. Acad. Sci. U.S.A. 2006, 103 (30), 11206– 11210.

*To whom correspondence should be addressed. E-mail: gshaver@ purdue.edu. (1) Energy Information and Administration (EIA). International energy outlook 2008 DOE/EIA-0484(2008). Technical Report, 2008. r 2009 American Chemical Society

928

pubs.acs.org/EF

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al. Table 1. Fuel Properties Provided by BP property molecular formula density cloud point pour point cetane number heat of vaporization lower heating value

units kg/m3 °C °C J/kg MJ/kg

diesel

soy biodiesel

C13.5H23.6 830 -12 -23 49 250000 43.25

C18.8H34.5O2 877.25 3 -3 55 357000 37.5

diesel, including molecular composition and structure, cetane number, higher and lower heating values, and heat of vaporization,7,15,18 as displayed in Table 1. 1.2. Previous Studies. Substantial mitigation or elimination of the negative performance- and combustion-related effects of biodiesel is a key step to making biodiesel a viable alternative/supplemental transportation fuel. Prior efforts have indicated that higher biodiesel fuel consumption and biodiesel-NOx emissions can be mitigated to some extent via modulation of the following four engine parameters, alone or in concert: air/fuel ratio (AFR),6 EGR fraction,19,20 fuel injection pressure,21,5 and start of injection (SOI) timing.22-25 Simultaneous modulation of fuel injection pressure and timing was considered in ref 20, while ref 26 also incorporated EGR fraction optimization. Many of these efforts were based on experiments performed at a single operating location using single-cylinder engines with mechanical fuel injection or were based solely on combustion model results. 1.3. Current Effort. The effort described in this paper differs from the aforementioned studies in that it considers simultaneous modulation of four parameters (AFR, EGR fraction, rail pressure, start of main injection pulse) across their effective ranges on a modern six-cylinder engine with common rail injection, a variable geometry turbocharger (VGT), and EGR at three different operating locations to determine optimal engine decision making. In this study, the optimization problem is characterized by eq 1. This states that, at any given operating location with any given fuel, the optimal settings for AFR, EGR fraction, rail pressure (RP), and SOI will produce the lowest BSFC that is possible without exceeding the stock diesel levels of BSNOx, BSPM, and peak dP/dt (a surrogate for acoustic combustion noise). In optimization terminology, BSFC is the cost function,

Figure 1. Average emission impacts of biodiesel for heavy-duty highway engines according to a 2002 EPA report.

There are also several combustion-related disadvantages to using biodiesel in modern diesel engines. Biodiesel has a calorific value that is approximately 13% lower than diesel (43 MJ/kg for diesel versus 37.5 MJ/kg for soybased biodiesel), meaning that biodiesel has a lower energy density. This manifests itself in a larger biodiesel fuel requirement to achieve the same power level as diesel,7 resulting in increased brake-specific fuel consumption (BSFC). Also for many operating conditions, biodiesel combustion emits more nitrogen oxides (NOx) than conventional diesel combustion.10,11 The Environmental Protection Agency (EPA) study results in Figure 1 show a 10% increase in NOx emissions with B100. This NOx increase with biocontent is known as the “biodiesel-NOx effect” and appears to be even more troublesome in more modern engines [i.e., those with exhaust gas recirculation (EGR) and common rail fuel injection].2,12 There are many theories that attempt to explain the biodiesel-NOx effect, but there is little consensus.10,11 Effects of cetane number differences,13 soot radiation,11,14 bulk modulus differences,15 engine control module (ECM) decision-making effects,16 and adiabatic flame temperature differences17,11,16 have all been considered. The performance- and emission-related differences occur as a result of fuel property differences between biodiesel and

(18) Szybist, J.; Song, J.; Alam, M.; Boehman, A. Biodiesel combustion, emissions, and emission control. Fuel Process. Technol. 2007, 88, 679–691. (19) Agarwal, D.; Sinha, S.; Agarwal, A. Experimental investigation of control of NOx emissions in biodiesel-fueld compression ignition engine. Renewable Energy 2006, 31, 2356–2369. (20) Yuan, W.; Hansen, A.; Tan, Z. Combustion optimization of biodiesel for diesel engines with the aid of kiva-3 code. ASAE 026083, 2002. (21) Basavaraja, T.; Reddy, R.; Swamy, V. Effect of injection pressure on emission performance of bio-diesel and its blends. SAE Tech. Pap. 2005-26-030, 2005. (22) Szybist, J.; Kirby, S.; Boehman, A. NOx emissions of alternative diesel fuels: A comparative analysis of biodiesel and FT diesel. Energy Fuels 2005, 19, 1484–1492. (23) Yuan, W.; Hansen, A.; Tan, Z. Modeling of NOx emissions of biodiesel fuels. ASAE 056116, 2005. (24) Kegl, B.; Kegl, M.; Pehan, S. Optimization of a fuel injection system for diesel and biodiesel usage. Energy Fuels 2008, 22, 1046–1054. (25) Choi, C.; Bower, G.; Retiz, R. Mechanisms of emissions reduction using biodiesel fuels;Final report for the national biodiesel board. Technical Report, National Biodiesel Board, Jefferson City, MO, 1997. (26) Tennison, P.; Reitz, R. An experimental investigation of the effects of common-rail injection system parameters on emmisions and performance in a high-speed directinjection diesel engine. J. Eng. Gas Turbines Power 2001, 123, 167–174.

(10) McCormick, R.; Graboski, M.; Alleman, T.; Herring, A.; Tyson, K. Impact of biodiesel source material and chemical structure on emissions of criteria pollutants from a heavy-duty engine. Environ. Sci. Technol. 2001, 35, 1742–1747. (11) Ban-Weiss, G.; Chen, J.; Buchholtz, B.; Dibble, R. A numerical investigation into the anomalous slight NOx increase when burning biodiesel; a new (old) theory. Fuel Process. Technol. 2007, 88, 659– 667. (12) Bunce, M.; Snyder, D.; Adi, G.; Hall, C.; Koehler, J.; Davila, B.; Garimella, P.; Kumar, S.; Stanton, D.; Shaver, G. Optimization of soybiodiesel combustion in a modern diesel engine. Fuel, 2009, manuscript in review. (13) Stone, R. Introduction to Internal Combustion Engines, 3rd ed.; Society of Automotive Engineers: Warrendale, PA, 1999. (14) Musculus, M. Measurements of the influence of soot radiation on in-cylinder temperatures and exhaust NOx in a heavy-duty DI diesel engine. SAE Tech. Pap. 2005-01-0925, 2005. (15) Fuel property effects on biodiesl NOx emissions. Agricultural Products Utilization Forum, Kansas City, MO, Nov 2002. (16) Eckerle, W.; Lyford-Pike, E.; Stanton, D.; LaPointe, L.; Whitacre, S.; Wall, J. Effects of methyl ester biodiesel blends on NOx emissions. SAE Tech. Pap. 2008-01-0078, 2008. (17) Jha, S.; Fernando, S.; Filip To, S. Flame temperature analysis of biodiesel blends and components. Fuel 2007, 87, 1982–1988.

929

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al. Table 2. Engine Parameters value

rating displacement volume number of cylinders valves per cylinder bore diameter stroke compression ratio

325 hp at 2500 rpm 6.7 6 4 107 124 17.3

units hp L mm mm

multiplexer A/D I/O boards, a DS2103 D/A I/O board, and a DS4002 timing and digital I/O board. Each board has a corresponding breakout box for the external signal connections. Fuel is injected using a common rail fuel injection system which can distribute fuel in up to three separate pulses: pilot, main, and post injections. The ECM determines and dictates the number of pulses and the time delay between each. For this study, the engine was held at a specific operating location using the speed mode function of the dynamometer controller and by dictating the total fueling quantity to the ECM via a control area network (CAN) interface. The dynamometer accuracy is 0.005%, indicating negligible drift from the desired speed set point. Factory-installed on-engine sensors measure quantities including intake manifold temperature and pressure, EGR valve position, pressure differential across the EGR valve, EGR temperature after cooling, engine speed, VGT turbine speed, fuel rail pressure, crankcase pressure, and coolant temperature. These quantities are measured by Calterm III (Cummins proprietary software that monitors and controls the ECM) and logged at a rate of 50 Hz by the dSPACE system via the CAN interface. The ECM also uses virtual measurements (ECM estimates) of torque, air flow, fuel flow, and exhaust flow. Additional temperatures and pressures are measured using external lab-grade sensors. The engine head has been modified to allow for the insertion of a pressure transducer in each of the six cylinders. For this study, one Kistler model 607C pressure transducer was used in cylinder 4. The transducer operates with a piezoelectric quartz crystal and can withstand pressures up to 70 000 psi. Pressure data is conditioned through a charge amplifier and then logged through dSPACE at a sampling frequency of 50 kHz. CO2, NOx, and PM emissions were monitored using exhaust analyzers. The CO2 analyzer is a Cambustion NDIR500 CO/ CO2 two-channel fast response analyzer that measures carbon monoxide (CO) and carbon dioxide on a wet basis. It operates via the nondispersive infrared (NDIR) principle and has a response time of 7 ms. For the experiments detailed in this study, one NDIR channel measures the CO2 concentration in the intake manifold, while the other channel measures the CO2 concentration in the exhaust pipe downstream of the turbocharger. Dividing intake CO2 by exhaust CO2 provides real-time determination of the EGR fraction. This calculation is integrated into a closed loop controller, allowing for a continuous, accurate command of the EGR fraction. The NOx analyzer is a Cambustion fNOx400 CLD two-channel fast analyzer that measures the nitric oxide concentration on a wet basis. It operates on the chemiluminescent (CLD) principle. Both fNOx400 channels are used to enhance data accuracy. The PM concentration is measured with an AVL 483 Microsoot sensor, operating on the photoacoustic measurement principle. 2.2. Experimental Procedure. Tests were conducted using conventional petroleum diesel (B0) and blends of 20% biodiesel/80% diesel (B20) and 5% biodiesel/95% diesel (B5). All biodiesel and blends used in this study are soy-based. The fuels were tested at three very different operating locations on the speed-torque curve. These locations are commonly known as A100, B50, and C100 and are located in the not-to-exceed (NTE) region, where tailpipe emissions are stringently regulated by the U.S. EPA, as shown by the shaded region in

Figure 2. Possible control diagram incorporating biocontent estimation.

while BSNOx, BSPM, and peak dP/dt are three inequality constraints. minimize : BSFC ¼ functionðAFR, EGR, RP, SOIÞ subject to : BSNOx ¼ functionðAFR, EGR, RP, SOIÞ eðBSNOx ÞB0, nominal BSPM ¼ functionðAFR, EGR, RP, SOIÞ eðBSPMÞB0, nominal peak dP=dt ¼ functionðAFR, EGR, RP, SOIÞ eðpeak dP=dtÞB0, nominal

parameter

ð1Þ

This study acts as a counterpart to a previous study12 that used a similar approach to compare nominal and optimal performances of conventional diesel with 100% soy-based biodiesel. While some of the results of that study are mentioned here, the present study focuses on the stock and optimal performance of two intermediate biodiesel blends, B20 (20% biodiesel/80% diesel) and B5 (5% biodiesel/95% diesel). This study also incorporates a regression model with optimal selection to the data sets to effectively improve the determination of optimal decision making, performance, and emissions for B5 and B20 blends. In addition, this study outlines how stock and optimized performance and emissions trend with increasing biocontent. When the optimal ECM decision making is known, one possible approach would be to apply it along with an onboard real-time biocontent estimation27 or direct sensing scheme. This would allow the ECM to detect the specific fuel blend that is being combusted and adjust its decision making to optimize combustion accordingly, thus providing true fuel-flexible combustion as illustrated in Figure 2. 2. Experimental Arrangement and Procedure 2.1. Experimental Arrangement. The engine considered is a 2007 Cummins ISB 6.7 L inline six-cylinder turbo-diesel engine outfitted with common rail fuel injection, externally cooled EGR, and VGT. Engine specifications are provided in Table 2. The engine is fitted to an eddy current dynamometer. Quantities including exhaust emissions, temperature, pressure, speed, torque, and mass air flow are measured using external sensors. Data from all external and factory-installed engine sensors are collected with a dSPACE data acquisition system, which also has the capability to override ECM actuator commands. The system consists of a DS1005 processor board, DS2002/2003 (27) Snyder, D.; Washington, E.; Indrajuana, A.; Shaver, G. Biodiesel blend estimation via a wideband oxygen sensor. American Controls Conference, 2008.

930

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al.

Figure 3. Engine torque-speed curve and operating locations (note: 280 hp curve).

Figure 4. Heat release rate and fuel mass flow rate versus the crank angle at C100.

Table 3. Operating Locations Considered in This Study operating point A100 B50 C100

speed (rpm)

torque (lb. ft) for diesel fuel

1576 1944 2311

660 330 637

Table 4. Constant and Varied Parameters constant parameters total fueling quantity pilot fueling quantity post fueling quantity pilot-main fuel injection separation time main-post fuel injection separation time intake manifold temperature

Figure 3.28 Table 3 lists these operating locations and the corresponding speed and torque values. The NTE region is an area of the torque-speed curve where “mid-range” diesel engines spend most of their operating lives. The EPA has specified exhaust emission levels throughout this region that are not to be exceeded by production engines, including the engine used in this study.28 The EPA can regulate emission levels at any location in the NTE region, regardless of whether or not the location corresponds to a delineated operating condition. The A100, B50, and C100 operating locations in the NTE region exhibit mixing-controlled (i.e., diffusion) combustion. In this combustion mode, the combustion event largely occurs as the fuel is injected into the cylinder, as illustrated in Figure 4. This is in contrast to other operating locations, typically at lower speeds and loads, which exhibit kinetic-controlled (i.e., premixed) combustion. At any given speed and torque, the stock ECM makes a series of decisions in an effort to bring exhaust emissions (BSNOx and BSPM) and noise within acceptable levels while simultaneously attempting to minimize BSFC. These decisions include the fuel injection profile (rail pressure and injector on-times) and “gas exchange” parameters, namely, AFR and EGR fraction. The ECM translates these predetermined decisions into actuator commands (e.g., EGR and VGT valve positions) based on a series of look-up tables and control loops developed from extensive testing and simulation by the engine manufacturer. As expected, these stock look-up tables and controllers were developed for conventional diesel fuel and are not intended to be optimal for combustion of biodiesel blends. 2.2.1. B5 and B20 Performance and Emissions with Stock ECM Decision Making. For the first sets of tests performed at the three chosen operating locations in this study, ECM decisions were dictated to correspond to decisions that the ECM would normally make at each given point while burning conventional diesel fuel to ensure that all inputs to the engine, except fuel properties, would initially remain constant when introducing biodiesel blends. These stock ECM decisions will be referred to as “nominal settings” in this paper. The ECM settings that were dictated include pilot, post, and total fueling quantities (mg/stroke), rail pressure (bar), timing

varied parameters charge flow EGR fraction fuel rail pressure start of main fuel injection pulse

for start of main fuel pulse (degrees BTDC), pilot-main and main-post injection timing separation (μs), charge flow (kg/ min), and EGR fraction (the latter two dictated in tandem to effectively dictate AFR), as shown in Table 4. Because of the lower torque associated with lower energy content for biodiesel, operating locations on the speed-torque curve were defined as “speed-fueling” locations for this study to establish a more valid comparison between fuels. 2.2.2. B5 and B20 Performance and Emissions with Random ECM Decision Making. A series of experiments was performed to determine appropriate ranges for the four modulated parameters (AFR, EGR fraction, rail pressure, and SOI) at each of the three operating locations. Each parameter was modulated individually using a wide, predetermined range that comprehensively spanned nearly the entire range of values that the ECM could reasonably dictate for all operating locations. These initial ranges were then constrained given mechanical (e.g., EGR and VGT position) limitations and reasonable ranges for NOx and PM emissions and BSFC. Closed-loop control of AFR, EGR fraction, and total fueling using lab-grade measurements was implemented to achieve high-fidelity manipulation of these engine decision parameters. With the ranges finalized, 150-185 random combinations of AFR, EGR fraction, rail pressure, and main injection timing that spanned these ranges were generated for each of the three operating locations. The remaining parameter overrides including fuel quantities and pulse-pulse separations were kept constant for the entire set of combinations, as summarized in Table 4. Data were collected for each parameter combination at each operating location using conventional diesel, B20, and B5. After recording data at 20 different parameter combinations, all engine overrides were returned to the nominal diesel settings for each particular operating location to check data repeatability. After recording data at 40 different parameter

(28) United States Environmental Protection Agency (EPA). Control of emissions of air pollution from new motor vehicles: In-use testing for heavy-duty diesel engines and vehicles. Technical Report, 2005.

931

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al.

combinations, each emission analyzer was recalibrated. These tasks were performed to ensure data quality. 2.3. Data Collection. All ECM decisions were dictated via a CAN-ECM interface to allow for rapid transitioning from ECM decision making to user-dictated decision making and vice versa. Any fueling or air handling parameter values could be specified, with the only limitations being hardware restrictions of the EGR valve and VGT. Total fueling, charge flow, and EGR fraction were dictated using closed-loop controllers. Once these values and the charge temperature stabilized, steady-state data were recorded from all of the production and lab-grade sensors. The on-engine sensor data are monitored by the ECM. These data, along with data from all of the external sensors, were integrated into the dSPACE data acquisition system via a CAN interface, and data were logged at 100 Hz for 30 s. Data from the PM analyzer were logged at 1 Hz for 30 s using an AVL software interface. Data from a Kistler in-cylinder pressure transducer were logged at 50 kHz to allow for sufficient in-cylinder pressure data resolution for calculating the heat release rate, indicated mean effective pressure (IMEP), centroid of heat release, and peak in-cylinder pressure. Experimental data were used to calculate numerous other quantities, including brake-specific NOx (BSNOx), brake-specific particulate matter (BSPM), peak rate of the change of the in-cylinder pressure (peak dP/dt), and BSFC.

Figure 5. B20 results with nominal ECM decision making.

3. Results 3.1. Observed Experimental Results at the Nominal ECM Settings. Figure 5 summarizes the behavior of B20 at the B0 nominal settings for the A100, B50, and C100 operating locations. As expected, all three operating locations exhibit BSFC increases. There are also BSNOx increases of 8-34%. The average increase is substantially larger than what the EPA study suggests (Figure 1) but is consistent with recent studies suggesting that the biodiesel-NOx increase is typically more pronounced with MY2004 and newer engines.2,12 BSPM emissions are reduced by an average of 50% among the three operating locations, again exhibiting a more pronounced trend than expected (Figure 1). Figure 6 displays the B5 performance at the B0 nominal settings. The three locations all experience BSNOx increases of over 10%. The average BSFC increase is roughly 4%, down slightly from the B20 levels. The PM reduction is still largely retained, except at the C100 location, where the reduction has interestingly shrunk to only 4% below B0 nominal levels. In general, the goal of engine decision making optimization pursued in this study is to (1) bring BSNOx and peak dP/dt to or below the nominal B0 engine results (the x axis in Figures 5 and 6), (2) maintain BSPM to or below the B0 results, and (3) minimize BSFC as much as possible while meeting the other requirements. It seems reasonable that it may be possible to leverage the significant BSPM reductions to reduce the BSNOx, peak dP/dt, and BSFC as desired. An attempt to achieve this experimentally is discussed in the following sections. 3.2. Observed Experimental Optimal Performance. Experimental data points for B5 and B20 were determined to be optimal candidates if their corresponding BSNOx, BSPM, and peak dP/dt levels were at or below the associated B0 levels. For a particular blend and operating condition, the three experimental results (out of the 150-185 random combinations of AFR, EGR fraction, rail pressure, and main timing) that adhered to these constraints and had the lowest BSFC values at each

Figure 6. B5 results with nominal ECM decision making.

Figure 7. B20 performance at A100 nominal settings, experimental optimal settings, and “best BSFC” settings.

operating location with each fuel were defined as the “experimental optimal points”. Each of the six fuel/operating location combinations therefore have three associated “experimental optimal points”. 3.2.1. Optimal Experimental Results for 20% Biodiesel. Stock (“nominal”), optimal (“best three constrained cases”), and “best BSFC” results with B20 are displayed in Figures 7-9 for the A100, B50, and C100 operating conditions. The results show that it is possible to eliminate the 932

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al.

Figure 8. B20 performance at B50 nominal settings, experimental optimal settings, and “best BSFC” settings.

Figure 10. B5 performance at A100 nominal settings, experimental optimal settings, and “best BSFC” settings.

Figure 9. B20 performance at C100 nominal settings, experimental optimal settings, and “best BSFC” settings.

Figure 11. B5 performance at B50 nominal settings, experimental optimal settings, and “best BSFC” settings.

BSNOx increases for biodiesel blends while maintaining noise (except at the B50 condition) and BSPM at or below B0 results. At the C100 operating condition, BSNOx went from a 34% increase over B0 levels at the nominal ECM settings to as much as a 20% decrease at one of the three optimal points. The BSPM reductions experienced at the nominal settings were also largely retained. With the exception of a 5% reduction at one of the three B50 optimal points, all optimal points experienced BSPM reductions of at least 17%. Optimal peak dP/dt levels were at or below nominal B0 levels at A100 and C100 but not at B50, where the peak dP/dt constraint had to be relaxed by 4% to locate points with emissions adhering to strict constraints. At A100, the BSFC penalty was reduced slightly from 8% above B0 levels to 6%. At the other two operating locations, optimal BSFC was slightly higher than the nominal penalty with stock engine decision making. These data suggest difficulty in mitigating the BSFC penalty while meeting emissions and noise constraints, because of the difference in energy content. The point with the lowest BSFC regardless of emissions and noise levels was determined at each B20 operating location. There were no parameter combinations that resulted in BSFC levels comparable to or below B0 levels at A100 and C100, strongly suggesting that overcoming the lower energy content of biodiesel may not be possible, even when emissions and noise constraints are ignored. The BSFC

penalties were reduced but were accompanied by substantial BSNOx increases. A parameter combination does exist that results in a BSFC level slightly lower than the nominal B0 level at B50, but this combination carries an associated BSNOx increase of over 100%, indicating that this parameter combination is not practical from an emissions standpoint. 3.2.2. Optimal Experimental Results for 5% Biodiesel. Figures 10-12 show the experimental optimal performance with B5 fuel. BSNOx and BSPM could be brought in limits at all operating locations, but at the B50 and C100 locations, the peak dP/dt constraint had to be relaxed by 13% to do so. The optimal parameter combinations were not able to mitigate the BSFC penalty whatsoever at the B50 and C100 locations, although they were able to reduce BSFC to levels essentially even with B0 levels at A100 (1% increase over B0). Here again, the substantial PM reductions were largely maintained with the optimal settings. The “best BSFC” results shown in the figures show that it is possible to make B5 BSFC competitive with base diesel fuel but with substantial increases in noise, BSNOx, or BSPM. In summary, at all three operating locations, for both B20 and B5 fuel blends, the optimal parameter settings resulted in BSNOx levels that were at or below B0 levels while largely retaining the substantial BSPM reductions. However, these 933

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al. Table 5. Experimental Optimal Parameter Trendsa parameter AFR EGR Rail P SOI

B5 A100

B5 B50

B5 C100

B20 A100

B20 B50

B20 C100

0 þ 0 0

0 -

þ þ -

0 þ þ þ

þ þ 0

0 þ 0

a -, from -10 to -50%; 0, (10%; þ, from þ10 to þ50%, with percentages taken as a portion of the parameter range tested; þ SOI, early injection; - SOI, delayed timing.

reducing local in-cylinder temperatures.29,30 Biodiesel is 11% oxygen content by weight. This may lead to an equivalence ratio closer to 1 for biodiesel blends in the rich premixed region of the mixing-controlled combustion process, resulting in higher temperatures and increased NOx formation. Shifting to a lower AFR may raise the local premixed zone equivalence ratio, thereby reducing NOx.31,32 The B20 and B5 optimal settings also shift to higher EGR fractions at all but one operating location to reduce NOx. This is consistent with the fact that increasing the EGR fraction is an effective tool for mitigating NOx by lowering flame temperatures and in-cylinder oxygen concentrations.19,33-35 In summary, the optimal settings generally shift to lower AFRs and higher EGR fractions to mitigate the biodiesel-NOx effect. There is, however, a limit as to how far these parameters can be shifted. Overly aggressive EGR increases or AFR reductions result in less complete combustion, leading to increases in PM and fuel consumption.29,36 3.3.2. Start of Injection (SOI) Modulation. The A100 location exhibited a B20 optimal parameter shift to more advanced main injection time. This is consistent with data that show that advancing the start of the main injection pulse increases work output, thereby reducing BSFC.26,20,37 However, timing advancement generally results in a more vigorous combustion process with higher temperatures in a

Figure 12. B5 performance at C100 nominal settings, experimental optimal settings, and “best BSFC” settings.

settings tend to have little or no net impact on mitigating the effects of the lower energy density of biodiesel on BSFC and, in several cases, result in increased noise. In other words, these results demonstrate that the large BSNOx increases at the nominal settings can be mitigated completely with certain optimal parameter settings while keeping BSPM below nominal limits. However, difficulty remains in mitigating the BSFC penalty, regardless of the fuel blend. In fact, when emissions and noise limits were ignored, optimal parameter settings were able to bring BSFC to within comparable B0 nominal levels at just one operating location with a fuel blend that contained 5% biocontent. The engine decision making that leads to the more optimal performance outlined in this section and the reasons for the changes from the nominal settings are discussed in the next section. 3.3. Optimal Parameter Settings. Each of the experimental optimal data points determined in section 3.2 correspond to certain AFR, EGR fraction, rail pressure, and SOI settings. The ECM settings corresponding to the three experimental optimal points for each fuel/operating location were averaged and defined as the experimental optimal parameter settings. Table 5 shows the shifts in optimal parameter settings with B20. The optimal AFR settings are comparable to the nominal settings at two locations and shift to a lower AFR at the remaining location. The optimal EGR fraction is higher than the nominal settings at all three locations. Optimal SOI is advanced at one location and remains comparable to the nominal settings at B50 and C100. While the optimal parameter region shifts to a lower rail pressure at C100, the other two locations experience shifts to higher rail pressures. The optimal parameter shifts for B5 are also displayed in Table 5 and are mostly similar to those for B20. The optimal AFR is flat or shifted lower at two of the three operating locations and increases slightly above the nominal setting at C100. The EGR fraction is flat or shifted higher at all three locations. SOI remains flat at one location and shifts earlier at the remaining two locations. Rail pressure shifts lower at two locations and is flat at one location. 3.3.1. AFR and EGR Modulation. The shift to lower AFRs for NOx reduction is consistent with studies that have suggested that lower AFRs reduce BSNOx emissions by

(29) Park, S.W.; Reitz, R. D. Modeling the effect of injector nozzlehole layout on diesel engine fuel consumption and emissions. J. Eng. Gas Turbines Power 2008, 130, No. 032805. (30) Senatore, A.; Cardone, M.; Allocca, L.; Vitolo, S.; Rocco, V. Expirimental characterization of a common rail engine fuelled with different biodiesel. SAE Tech. Pap. 2005-01-2207, 2005. (31) Adi, G. An experimental and simulation study of fuel consumption and NOx emissions from biofueled diesel engines. Master’s Thesis, Purdue University, West Lafayette, IN, 2008. (32) Mueller, C. J.; Pitz, W. J.; Pickett, L. M.; Martin, G. C.; Siebers, D. L.; Westbrook, C. K. Effects of oxygenates on soot processes in DI diesel engines: Experiments and numerical simulations. SAE Tech. Pap. 20030193, 2003. (33) Barths, H.; Hasse, C.; Peters, N. Computational fluid dynamics modelling of non-premixed combustion in direct injection diesel engines. Int. J. Engine Res. 2000, 1 (3), 249–267. (34) Chan, M.; Das, S.; Reitz, R. Modeling multiple injection and EGR effects on diesel engine emissions. SAE Tech. Pap. 972864, 1997. (35) Senecal, P. K.; Reitz, R. D. Simultaneous reduction of engine emissions and fuel consumption using genetic algorithms and multidimensional spray and combustion modeling. SAE Tech. Pap. 2000-011890, 2000. (36) Aghav, Y. V.; Lakshminarayanan, P. A.; Babu, M. K. G.; Udin, A.; Dani, A. D. Validating the phenomenological smoke model at different operating conditions of DI diesel engines. J. Eng. Gas Turbines Power 2008, 130, No. 012803. (37) Yoshizaki, T.; Nishida, K.; Hiroyasu, H. Approach to low NOx and smoke emission engines by using phenomenological simulation. SAE Tech. Pap. 930612, 1993.

934

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al.

smaller in-cylinder volume, leading to increases in NOx and noise.38-42,35,43 While advancing the start of the main injection pulse can produce slight BSFC decreases from the nominal penalties with B100,12 all but the one aforementioned operating location exhibited optimal SOI unchanged or slightly reduced from the nominal settings with B20 and B5. This may be due to difficulty in identifying data points that meet the peak dP/dt constraint. In fact, the peak dP/dt constraint had to be relaxed at the C100 location with B5 and at the B50 location with both B5 and B20 because there were no data points that could meet strict constraints on BSNOx, BSPM, and peak dP/dt simultaneously. Because advancing SOI tends to increase noise and noise levels in the B20 and B5 data sets are much closer to nominal B0 levels than B100, further SOI advancement was prohibited. 3.3.3. Rail Pressure Modulation. Studies have shown that increasing rail pressure tends to decrease BSFC.37,20 Increasing injection pressure causes more fuel to enter the cylinder in a shorter amount of time. This generally increases the rate of heat release, resulting in increased peak pressures and temperatures. These in turn usually result in higher torque (lower BSFC), NOx, and noise.35,44,37 While the above indicates that increasing rail pressure should decrease BSFC, optimal rail pressures were shifted higher only at the A100 and B50 locations with B20 fuel. The remaining fuel/operating location combinations exhibited unchanged or decreased rail pressures, likely indicating that optimal rail pressure cannot be increased at these blend/ operating locations because of the increased noise that this shift would produce. The above indicates that advancing timing and increasing rail pressure is a potentially effective strategy for partially mitigating the energy density penalty of biodiesel, although its effectiveness is somewhat stunted by (1) the trade-off between BSFC and noise and (2) the inherent difficulty in overcoming a 12% reduction in calorific value for biodiesel. Point 1 was made manifest in this experiment by actually preventing most optimal rail pressures from being increased above the nominal; in half of those cases, rail pressure even had to be decreased to remain within limits on noise. Point 2 was demonstrated by the fact that the optimal settings were only able to produce a data point with BSFC, BSNOx, BSPM, and peak dP/dt levels comparable to or lower than the nominal B0 levels at one of the six fuel/operating location combinations, and that was with a fuel that had only 5% biocontent.

4. Development and Implementation of a Regression Model with Optimal Selection 4.1. Regression Model and Optimal Selection Purpose. Regression models were fitted to each data set at each operating location across all fuel blends. The purpose of these regression models is to effectively interpolate between data points in a given set. This essentially creates a “data set” that is one continuous sheet as opposed to a series of 150þ discrete points for each blend/operating condition. Using an optimal selection routine, optimal parameter combinations can now be evaluated at “data points” not specifically tested in the experiments but within the experimentally defined data boundaries. Specifically, the selection routine scanned the regression model corresponding to each fuel blend/operating location for minimal BSFC in regions corresponding to allowable noise, BSNOx, and BSPM according to eq 1. This increased data set resolution should theoretically result in “better” optimal points that have slightly lower BSFC values than the experimental optimal points in each set. This lower BSFC may come at the expense of one or all of the three constrained parameters (BSNOx, BSPM, and peak dP/dt), which may increase slightly from the experimental optimal levels but will still, by definition, remain at or below corresponding B0 nominal levels. 4.2. Regression Model Description. The regression model uses a second-order least-squares fit with 33 total terms. The independent variables are the parameters: AFR, EGR fraction, rail pressure, and SOI. The dependent variables are the monitored outputs: BSFC, BSNOx, BSPM, and peak dP/dt. An example of how the model output predictions generally agreed with the experimental outputs is presented in Figures 13-16 for B20 fuel at the B50 operating condition. All 150þ data points are in each window. The red lines correspond to (10% of the BSNOx, BSPM, and peak dP/dt experimental data and (2% of the BSFC experimental data. As expected, BSPM and peak dP/dt display the most variability between model and experimental results. The green points in each window are identified outliers, most of which originate from BSPM measurement errors at very low PM concentration levels. 4.3. Optimal Selection Routine Description. Using the regression models, an optimal selection routine is employed to determine the model optimal point in each data set. This is

(38) Zhou, P.; Zhou, S.; Clelland, D. A modified quasidimensional multi-zone combustion model for direct injection diesels. Int. J. Engine Res. 2006, 7, 335–345. (39) Yoshizaki, T.; Nishida, K.; Hiroyasu, H. Reduction of heavy duty diesel engine emission and fuel economy with multi-objective genetic algorithm and phenomenological model. SAE Tech. Pap. 2004-01-0531, 2004. (40) Jung, D.; Assanis, D. Multi-zone DI diesel spray combustion model for cycle simulation studies of engine performance and emissions. SAE Tech. Pap. 2001-01-1246, 2001. (41) Jung, D.; Assanis, D. Quasidimensional modeling of direct injection diesel engine nitric oxide, soot, and unburned hydrocarbon emissions. J. Eng. Gas Turbines Power 2006, 128 (2), 388–396. (42) Szybist, J. P.; Boehman, A. L.; Taylor, J. D.; McCormick, R. L. Evaluation of formulation strategies to eliminate the biodiesel NOx effect. Fuel Process. Technol. 2005, 86, 1109–1126. (43) Tat, M. Investigation of oxides of nitrogen emissions from biodiesel-fueled engines. Ph.D. Thesis, University of Illinois at UrbanaChampaign, Urbana and Champaign, IL, 2003. (44) Hampson, G. J.; Patterson, M. A. ; Kong, S. C.; Reitz, R. D. Modeling the effects of fuel injection characteristics on diesel engine soot and NOx emissions. SAE Tech. Pap. 940523, 1994.

Figure 13. Regression model of predicted BSFC versus experimental BSFC.

935

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al. Table 6. Average Experimental B0 Nominal versus Regression B0 Optimal output

A100

B50

C100

percent change in BSFC (%) percent change in BSNOx (%) percent change in BSPM (%) percent change in peak dP/dt (%)

-2.2 0.0 -0.2 0.0

-0.1 -0.4 -0.1 0.0

-1.5 -3.3 -1.1 -0.1

Table 7. Regression Model Optimal Parameter Trendsa parameter AFR EGR Rail P SOI

B5 A100

B5 B50

B5 C100

B20 A100

B20 B50

B20 C100

0 þ 0 þ

; 0

0 0 0 0

0 þþ þ þ

; þ þ þ

0 þþ þ

a ;, -50% and below; -. from -10 to -50%; 0, (10%; þ, from þ10 to þ50%; þþ, þ50% and above, with percentages taken as a portion of the parameter range tested; þ SOI, early injection; - SOI, delayed timing.

Figure 14. Regression model of predicted BSNOx versus experimental BSNOx.

4.4. Regression Model Optimal Comparison to B0 Experimental Data. The regression model optimal search routine validity was first examined by comparing the model optimal B0 results to the experimental B0 results with stock engine decision making. The stock ECM parameter settings have been optimized for B0 combustion for a set of BSNOx, BSPM, and peak dP/dt constraints. As such, with these constraints in effect, an accurate regression model for B0 should predict optimal BSFC levels that are similar to the corresponding average experimental levels that occur at the nominal ECM settings at each of the three operating locations. Table 6 displays the differences between the model optimal B0 results and the average experimental B0 nominal results. There is excellent agreement between model optimals and experiments with nominal decision making at A100, B50, and C100, with the largest deviation being a 3.3% difference in BSNOx values at C100, meaning that the model that predicted the optimal B0 BSNOx value was 3.3% lower than the experimental average B0 nominal value at C100. The strong correlations in Table 6 and Figures 13-16 indicate validity of the regression model, optimal search routine, and stock engine decision making for B0. 4.5. Regression Model Optimal Parameter Results. Table 7 displays the regression model optimal parameter settings for B5 and B20 at each operating location. Compare this table to the experimental data in Table 5. As expected, the model optimal parameter trends generally correlate with the experimentally observed optimal parameter trends. In each case, optimal AFR is either comparable to or lower than the corresponding nominal setting. Likewise, optimal EGR fractions shift higher, except at the B50 and C100 locations with B5, where it appears that the lower optimal AFR can accommodate the NOx increase without the assistance of a higher EGR fraction. These trends exhibit excellent agreement with previously discussed experimental optimal results. The SOI trends are more consistent with intuition than some of the experimental results. Model optimal SOI is advanced or flat in all cases. The variability in the optimal rail pressure trend witnessed in the experimental results is repeated in the model results. B20 data sets tend to favor optimal shifts to higher rail pressures, whereas B5 data sets tend to favor flat or decreased rail pressures. The following section discusses the predicted performance that results from these optimal

Figure 15. Regression model of predicted BSPM versus experimental BSPM.

Figure 16. Regression model of predicted peak dP/dt versus experimental peak dP/dt.

performed by setting BSNOx, BSPM, and peak dP/dt limits for each data set corresponding to the average nominal B0 experimental values. The routine then scans the entirety of the regression fit to locate a model point with the lowest BSFC that has emissions and noise levels at or below the specified limits in the same manner that the experimental optimal points were determined in section 3.2. 936

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al.

Figure 17. B20 optimization results at A100.

Figure 18. Nominal and model optimal BSFC trends across blends.

Table 8. Nominal and Model Optimal Performance

regression models are summarized in Table 8. The first column in each set displays the experimental performance of each fuel blend at the nominal B0 settings. The second column in each set displays the predicted regression model performance of each fuel blend at the model optimal parameter settings. Values that are red are unfavorable (high BSFC, high BSPM, etc.), and values that are green are favorable (low BSFC, low BSPM, etc.). At each of the six fuel/operating location combinations, optimal parameter settings were able to produce BSNOx, BSPM, and peak dP/dt levels that were comparable to or below those of the corresponding B0 nominal settings. However, BSFC could not be improved to B0 levels for any of the blends or operating locations. The results in Table 8 demonstrate both the effectiveness of the biodiesel-NOx effect mitigation strategies employed in this study and the inherent difficulty in mitigating the effect of lower energy content. According to these data, the lower energy content effect can be mitigated only slightly by modulating AFR, EGR fraction, rail pressure, and SOI.

parameter settings and what performance is ultimately achievable with B20 and B5 combustion. 4.6. Regression Model Optimal Performance Results. To further demonstrate the effectiveness of using the regression model to produce optimal engine decision making for a particular fuel blend at a given operating location, the results for B20 at the A100 location is examined as an example. Figure 17 displays the changes in performance with B20 at the A100 location. The first set of bars is performance at the nominal settings, followed by performance at each of the three experimental optimal points, as discussed in section 3.2. The final set of bars in figure 7 is the predicted performance at the model optimal point. As expected, the model optimal point has a slightly lower (i.e., “more optimal”) BSFC than the three experimental optimal points. This comes at the expense of the peak dP/dt decrease and the BSPM decrease, with little impact on the BSNOx decrease. While these three parameters are still, by definition, comparable to or below the corresponding B0 levels, the peak dP/dt and BSPM decreases are now negligible compared to decreases between 9 and 13% for peak dP/dt and between 20 and 62% for BSPM at the experimental optimal points. The ability to simultaneously consider a large number of parameter combinations, including untested combinations, with the regression model allows for additional minimization of the BSFC penalty. The optimized performance and emission results for B5 and B20 for all operating locations according to the

5. Trends Across Biodiesel Blends 5.1. Nominal Performance Trends Across Biodiesel Blends. Performance, emissions, noise, and efficiency trends across fuel blends with the stock ECM settings are plotted as solid lines in Figures 18-22 using the B20 and B5 data from this study and the B100 data from a prior study.12 There appears to be no prevailing trend for either noise or thermal efficiency at the nominal settings. BSFC, BSNOx, and BSPM all display clear trends across blends. Figure 23 shows the relationship between the energy content and the BSFC penalties experienced by each fuel blend at the nominal ECM settings. Assuming that the B0 used for these experiments has a lower heating value (LHV) of 43 MJ/kg and B100 has a LHV of 37.5 MJ/kg,31 there appears to be a slightly nonlinear relationship between the LHV and BSFC penalty experienced by each of the fuel blends with the B0 nominal settings. This indicates that, while the LHV appears to be the main driver for BSFC increases, there may be other property differences that influence combustion efficiency, as shown in Figure 22. The dashed line in Figure 23 delineates the energy content difference. A fuel with a LHV of 43 MJ/kg achieves a certain thermal efficiency. A fuel with a lower LHV will exhibit an increase in BSFC as shown by the dashed line if it achieves the same thermal efficiency. BSFC increases that are lower 937

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al.

Figure 22. Nominal and model optimal thermal efficiency trends across blends.

Figure 19. Nominal and model optimal BSNOx trends across blends.

Figure 20. Nominal and model optimal BSPM trends across blends.

Figure 23. LHV versus BSFC penalty.

physical and chemical fuel property differences, less dominant than LHV, that lead to BSFC increases via thermal efficiency reductions. However, as shown in Figure 23, the LHV reductions increasing with biocontent appear to be the prevailing cause of BSFC increases with higher B0 blends. 5.2. Model Optimal Performance Across Biodiesel Blends. The trends in model optimal performance versus biocontent are also exhibited in Figures 18-22. As desired and previously demonstrated, BSNOx and peak dP/dt levels move below the x axis and BSPM levels generally shift upward, closer to but still below the x axis. BSFC levels only show substantial shifts downward at the A100 location. At the C100 location, the optimal curve actually shifts above the nominal curve, meaning that the model optimal settings produce higher BSFC levels than the nominal settings (at two of the three fuel blends). The regression model optimal parameters appear to be producing the same trends in optimal BSFC and BSNOx across the four operating locations, with the notable exception of BSNOx at the C100 location, which exhibits a substantially larger reduction. This particular fuel/operating location combination also retains a substantial BSPM reduction, although the optimization does appear to leverage the peak dP/dt reduction (as do all fuel/location combinations) because it is nearly flat with B0 levels at C100. The inability to produce significant BSFC reductions at the pure

Figure 21. Nominal and model optimal peak dP/dt trends across blends.

than the energy content-specific BSFC effect (i.e., below the dashed line) are consistent with thermal efficiency increases, whereas BSFC increases above this line correspond to efficiency decreases. At the nominal B0 settings, no operating locations with B20 or B5 showed thermal efficiency levels comparable to or above B0 levels. This again indicates that there may be other 938

Energy Fuels 2010, 24, 928–939

: DOI:10.1021/ef9011033

Bunce et al.

biodiesel/C100 combination may be indicative of an inherent limitation in leveraging either of the emission reductions. The average trends for BSFC, BSNOx, and BSPM across blends are displayed as bolded lines in Figures 18-20. With the optimal parameter settings (SOI, EGR fraction, rail pressure, and AFR), the average BSNOx shifts below the x axis while BSPM curves shift upward, closer to but still below the x axis. The nonlinearities, particularly in the average BSPM curve, indicate the ability of the approach to leverage emissions decreases differently with different fuel blends. As is evident, the average BSFC curve does shift downward but only very slightly, not enough to bring it below the x axis for any of the fuel blends. In summary, despite the aggressive leveraging strategies used with the regression models to identify optimal parameter settings that reduce emissions, noise, and BSFC to comparable B0 levels, the ability to mitigate the biodiesel BSFC penalty (primarily as a result of lower energy content) is severely limited.

A second-order regression model with optimal selection was also implemented to interpolate throughout the data sets and find more precise optimal parameter combinations. This model demonstrated excellent agreement with experimental data. As expected, optimal parameter settings with biodiesel blends were different than the optimal parameter settings with B0. For each blend at each operating location, optimal AFR generally shifted lower and optimal EGR fraction generally shifted higher. These trends serve to lower the in-cylinder oxygen content, likely lowering near-flame temperatures and local oxygen fractions, reducing BSNOx. Optimal SOI was generally advanced, while optimal rail pressure generally increased, although the latter parameter showed the most variability. These trends serve to increase work output, thereby reducing BSFC. With each fuel at each operating location, the optimal ECM parameter settings resulted in the reduction of BSNOx, BSPM, and peak dP/dt to levels comparable to or lower than the corresponding nominal B0 levels. The ability to produce emission and noise decreases were subsequently leveraged to produce BSFC reductions. Despite these efforts, BSFC could only be reduced slightly from the penalties observed with stock ECM settings. In summary, this study proves that, with engine parameter modulation including lower AFR, higher EGR fraction, and earlier start of main injection, it is possible to mitigate the biodiesel-NOx effect with blends of soy-based biodiesel in a modern diesel engine. While these BSNOx reductions are also accompanied by reductions in both BSPM and noise, these settings have a limited net effect on counter-acting the lower energy density effects of biodiesel.

6. Summary In this study, several blends of soy biodiesel (B5 and B20) were tested with stock and optimal ECM decision making in a modern diesel engine with common rail fuel injection, cooled EGR, and a VGT. Experiments with B0 showed that the stock ECM decision making was indeed optimal for B0 combustion at all three operating locations in terms of minimizing BSFC subject to contrainsts on BSNOx, BSPM, and noise. Biodiesel blend combustion was associated with increases in both BSFC and BSNOx and decreases in BSPM, as expected. Previous studies had concluded that AFR, EGR fraction, rail pressure, and SOI, acting alone or in concert, are likely major drivers for reducing BSNOx and BSFC. This study differs from previous studies by simultaneously modulating all four parameters extensively on a modern diesel engine with a VGT, common rail fuel injection, and EGR. Other notable features of this study include complete control over ECM decision making. Through modulation of AFR, EGR fraction, rail pressure, and SOI, this study attempted to identify optimal parameter combinations that resulted in BSNOx, BSPM, and peak dP/dt levels with B5 and B20 that were comparable to or lower than corresponding nominal B0 levels while minimizing BSFC, to the greatest extent possible.

Acknowledgment. Funding for this work was provided by Cummins, Inc., Office of Naval Research, and Energy Center at Purdue Discovery Park. Fuel was donated by BP. The authors thank the following researchers at the Cummins Technical Center in Columbus, IN: Donald Stanton, Tim Frazier, Wayne Eckerle, Shankar Kumar, and Phanindra Garimella. Also, special thanks to Cummins, Inc. for providing the engine and technical support. The authors are also grateful to the Ray W. Herrick Laboratories Technical Services staff of Fritz Peacock, Bob Brown, Gil Gordon, and Frank Lee.

939