Physical and chemical effects of low octane gasoline fuels on compression ignition combustion

Applied Energy 183 (2016) 1197–1208

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

Physical and chemical effects of low octane gasoline fuels on compression ignition combustion Jihad Badra a,⇑, Yoann Viollet a, Ahmed Elwardany b,c, Hong G. Im b, Junseok Chang a a

Fuel Technology Division, R&DC, Saudi Aramco, Dhahran, Saudi Arabia Clean Combustion Research Center, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia c Mechanical Engineering Department, Faculty of Engineering, Alexandria University, Alexandria 21544, Egypt b

h i g h l i g h t s
New experimental data SOI sweep in GCI engine running on naphtha fuels.
Despite the differences in the fuel’s properties, the combustion and emissions were similar.
CFD engine simulations successfully reproduced the experimental trends.
The chemical and physical effects were isolated numerically and detailed analysis was performed.

a r t i c l e

i n f o

Article history: Received 8 June 2016 Received in revised form 31 August 2016 Accepted 24 September 2016 Available online xxxx Keywords: Naphtha fuels Low-octane fuel Gasoline compression ignition engine Mixing Ignition delay times

a b s t r a c t Gasoline compression ignition (GCI) engines running on low octane gasoline fuels are considered an attractive alternative to traditional spark ignition engines. In this study, three fuels with different chemical and physical characteristics have been investigated in single cylinder engine running in GCI combustion mode at part-load conditions both experimentally and numerically. The studied fuels are: Saudi Aramco light naphtha (SALN) (Research octane number (RON) = 62 and final boiling point (FBP) = 91 °C), Haltermann straight run naphtha (HSRN) (RON = 60 and FBP = 140 °C) and a primary reference fuel (PRF65) (RON = 65 and FBP = 99 °C). Injection sweeps, where the start of injection (SOI) is changed between 60 and 11 CAD aTDC, have been performed for the three fuels. Full cycle computational fluid dynamics (CFD) simulations were executed using PRFs as chemical surrogates for the naphtha fuels. Physical surrogates based on the evaporation characteristics of the naphtha streams have been developed and their properties have been implemented in the engine simulations. It was found that the three fuels have similar combustion phasings and emissions at the conditions tested in this work with minor differences at SOI earlier than 30 CAD aTDC. These trends were successfully reproduced by the CFD calculations. The chemical and physical effects were further investigated numerically. It was found that the physical characteristics of the fuel significantly affect the combustion for injections earlier than 30 CAD aTDC because of the low evaporation rates of the fuel because of the higher boiling temperature of the fuel and the colder in-cylinder air during injection. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction Around 40% increase in global transport energy demand is expected by the year 2040 according to energy projections [1–4]. Even with this increase in energy demand, more than 90% of the global transport energy demand is currently and is expected to continue be supplied by the petroleum-based liquid fuels such as gasoline, diesel, jet and heavy fuel oil fuels [2]. Therefore, improv-

⇑ Corresponding author at: P.O. Box 62, Dhahran 31311, Saudi Arabia. E-mail address: [email protected] (J. Badra). http://dx.doi.org/10.1016/j.apenergy.2016.09.060 0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.

ing the fuel efficiency at low cost in transportation sector can help not only saving the global energy usage but also reducing the greenhouse gas (GHG) CO2 emissions. Gasoline compression ignition (GCI) engines have recently emerged as a promising technology for diesel-like thermal efficiencies with significantly reduced engine-out nitrogen oxides (NOx) and soot emissions by achieving low temperature combustion (LTC) conditions [5]. Recent studies [6–10] reported that GCI combustion occurs as a series of autoignition events with minor contributions from flame fronts. The autoignition timing is controlled by manipulating the mixture composition and temperature

1198

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

Nomenclature 3D three dimensional ASTM American Society for Testing and Materials aTDC after top dead center BiCGSTAB biconjugate gradient stabilized CAD crank angle degree CFD computational fluid dynamics CO carbon monoxide COV coefficient of variation CR compression ratio DCN derived cetane number EGR exhaust gas recirculation ETC/ED effective thermal conductivity/effective diffusivity EVO exhaust valve opening FBP final boiling point FSN filter smoke number FTP federal test procedure GCI gasoline compression ignition GDI gasoline direct injection GHG greenhouse gas HC hydrocarbon HCCI homogeneous charge compression ignition HSRN Haltermann straight-run naphtha

stratification within the cylinder through late injection in the compression stroke. This is in contrast to homogeneous charge compression ignition (HCCI) [11–15] and premixed ignition [16,17] engines where the fuel and air are fully mixed prior to entering the combustion chamber. Recently, a new engine combustion concept named ‘‘multiple premixed compression ignition” (MPCI) has been investigated as an option to increase efficiency and reduce pollutants [18–21] by manipulating the stratification of the charge to control the pressure rise rates. The enabling fuels for such GCI, partially premixed compression ignition (PPCI) and MPCI combustion modes are generally in the gasoline boiling range. Compared to commercial gasoline and diesel fuels, refinery streams such as petroleum naphtha with low octane numbers (RON) in the 50–80 range have recently been considered attractive alternatives to provide suitable chemical characteristics (longer ignition delay than diesel) in GCI engines at lower production cost and well-to-tank CO2 emissions. Hao et al. [22] found that compared with the conventional pathway, the lowoctane gasoline-GCI pathway leads to a 24.6% reduction in energy consumption and a 22.8% reduction in GHG emissions. A naphtha stream with a RON of 65 was utilized in [18–21] to investigate the MPCI mode. They have also tested diesel and gasoline blends with the aim of reducing the ON to the 50–80 range. They have reported that the naphtha fuel can be used to cover wide operating load points while resulting in higher thermal efficiencies and lower emissions. Han et al. [10] also tested the combustion of diesel and gasoline blends in compression ignition mode. It is clear from literature that the properties (RON, MON, density, boiling range and . . .) of the fuel that is suitable for PPCI or MPCI are not fully determined yet. Therefore, it is of interest to understand the effects of the fuel’s physical and chemical properties on the combustion and emission behavior of petroleum naphtha in GCI conditions. Kim et al. [23] studied the effects of some physical properties (density, vapor pressure, viscosity, surface tension, heat of vaporization and specific heat capacity) in a reaction fuel sprays. They found that density, viscosity, heat of vaporization and specific heat had significant impact on liquid penetration length. They also reported

IMEP IQT KH-RT LHV LTC MON MPCI NMEP NOx NTC OI PISO PPCI PRF RON RPM SALN SOI SOR TDC TPRF

indicated mean effective pressure ignition quality tester Kelvin-Helmholtz and Rayleigh-Taylor lower heating value low temperature combustion motor octane number multiple premixed compression ignition net mean effective pressure nitric oxides negative temperature coefficient octane index Pressure Implicit with Splitting of Operators partially premixed compression ignition primary reference fuel research octane number revolutions per minute Saudi Aramco light naphtha start of injection successive over-relaxation top dead center toluene primary reference fuel

that specific heat and density significantly affected the ignition delay of the system. Lacey et al. [24] also studied the effects of the properties of gasoline refinery streams on the auto-ignition quality of a fuel and the HCCI combustion. They have observed that the fuel composition significantly affects the combustion phasings of gasoline fuels with the same RON and MON. They have proposed a new octane index (OI) correlation that accounts for the aromatics, olefins, saturates and ethanol contents in the gasoline. Full cycle computational fluid dynamics (CFD) simulations with detailed chemical kinetics and turbulent transport can also provide fundamental understanding of the spray development and stratification and their effects on the combustion process. Ra et al. [25] performed extensive numerical studies on GCI combustion to investigate the effects of injection parameters, gas temperatures, boost pressure and exhaust gas recirculation (EGR) on combustion characteristics such as combustion phasing and important emissions. Recently, Badra et al. [26] reported a numerical study on the optimization of the spray models for outwardly opening hollow cone spray and the effects of primary reference fuel (PRF) and toluene primary reference fuel (TPRF) chemical surrogates on the combustion phasing of GCI engine running on naphtha fuel. Another study [27] also investigated the mixing effects of a light naphtha stream on the combustion phasing of a compression ignition engine. In this work, we investigate the effects of mixing and spray/piston interactions on the combustion phasing and emissions over a range of injection parameters using three carefully chosen low octane fuels. We first report new experimental data that investigate the effect of start of injection (SOI) on the combustion phasing and emissions in a GCI engine running on three fuels with different chemical and physical characteristics. Full-cycle GCI engine 3D CFD simulations using different chemical and physical surrogate fuels are then presented and validated with experimental data. Subsequently, the simulation data are thoroughly investigated through detailed equivalence ratio-temperature (U-T) for carefully chosen fuels with different chemical and physical properties, in order to provide insights into the observed engine combustion behavior.

1199

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

To our knowledge, this is the first time a detailed investigation of the fuel’ chemical and physical properties effects on the combustion characteristics and emissions of GCI engines running on low octane fuels is carried on. 2. Experimental setup A single cylinder 4 valve engine with a 14:1 geometric compression ratio and an outwardly-opening piezo-electric hollow-cone gasoline direct injection (GDI) injector is used in this investigation. Table 1 lists the engine specifications utilized in the presented experimental work. More details about the engine, combustion chamber, emission and pressure measurements are provided in [7,8,18,28,29]. The combustion chamber is originally designed to accommodate stratified charge spark ignition combustion. An outwardly opening piezo-electric hollow cone GDI (Siemens) injector is centrally mounted adjacent to the spark plug which is disabled in the subsequent GCI combustion tests. As shown in Fig. 1a, the injector is located between two intake valves and slightly skewed from the vertical direction where the fuel is introduced as a hollow cone spray through the outwardly opening piezo-electric injector. The spray is more widely distributed with relatively short penetration compared to spray-jet style multi-hole gasoline direct injector. The operating fuel injection pressure range is 50–150 bar, which is about 10 times lower than a conventional diesel injector. As such, spray atomization and penetration characteristics are expected to be significantly different. For the purpose of this study, a new set of pistons was designed and manufactured. To improve fuel containment, a diesel bowl-like feature was added. It also has a reentry feature to ensure that spray interaction occurs inside the bowl. The squish height is 1.5 mm. Since the combustion chamber is originally manufactured for a gasoline engine (pent-roof style), the piston height is relatively large in order to match the CR and compensate for the bowl volume, resulting in a smaller clearance height than a conventional gasoline piston. As shown in Fig. 1b (side view), the spark plug is still installed although it was not utilized in this study. The spray pathway is also highlighted in Fig. 1b. GCI testing was performed with fixed intake and exhaust valve events and symmetric lift profiles (8 mm peak lift and 207 CAD duration), as shown in Fig. 2. This is similar to conventional production diesel valve lift profiles where no variable valve timing with minimum valve overlap was used in the subsequent experiments. The fuels used in all the experiments in this work are the straight-run Saudi Aramco light naphtha (SALN), Haltermann straight-run naphtha (HSRN) and PRF65. These fuels have been chosen to have close cetane and octane numbers and different physical characteristics as shown in Table 2. The temperature ranges of the boiling characteristics for these fuels are shown in Fig. 3. SALN is highly paraffinic (>90%) and have low sensitivity

Table 1 Single cylinder engine specifications. Cylinders Number of valves Displacement Bore Stroke Connecting rod Compression ratio Fuel injector Fuel injection pressure Intake system Exhaust system

1 4 499 cm3 84 mm 90 mm 159 mm 14:1 Outwardly opening piezo-electric hollow cone injector 130 bar Conditioned air with external boost Back pressure valve

Fig. 1. Schematics of (a) base engine Combustion Chamber Shape: Pent-Roof Style 4 Valve Head and (b) side view of compression ratio 14 diesel bowl reentry piston.

Fig. 2. GCI valve lift profiles.

Table 2 Properties of the tested fuels. Fuel

SALN

HSRN

PRF65

DCN from IQT RON MON Sensitivity Density (kg/m3) at 15 °C Lower heating value (LHV) (MJ/kg) Hydrogen to carbon (H/C) ratio Normal-paraffins vol.% Iso-paraffins vol.% Aromatics vol.% Naphthenes vol.% Olefins vol.% Oxygenates vol.% Sulfur (ppm) Final boiling point (FBP (°C))

34.7 62 60 2 661 45.1 2.175 53.3 38.7 1.14 6.87 0 0 <1 90.7

34.2 60 58.3 1.7 705 44.6 2.146 36.0 40.8 8.94 14.24 0 0 14 138

34.2 65 65 0 695 44.4 2.262 35.0 65.0 0 0 0 0 0 99

1200

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

Fig. 3. Evaporation curves of the tested fuels using American Society for Testing and Materials (ASTM) D86 method.

(S = RON – MON) of about 2. On the other hand, HSRN has more than 20% aromatics and naphthenes, which are reflected as a lower sensitivity of 1.7. The boiling range of HSRN is wider than SALN, while PRF65 has a very narrow boiling range as expected from the similar boiling temperatures of n-heptane (98.4 °C) and isooctane (99 °C).

3. Experimental results An injection timing sweep was performed at 1500 RPM, 1.2 bar (abs) intake pressure and intake temperature of 32 ± 0.5 °C to achieve a target low-load net mean effective pressure (NMEP) of 3.6 bar using the three fuels. This particular operating speed and load condition is heavily weighted in a Federal Test Procedure 75 (FTP75) city cycle using a 1590 kg vehicle equipped with a 2 L engine [8]. Around 10 mg/cycle of fuel was injected to achieve a targeted global equivalence ratio of 0.225. No EGR was utilized in these experiments. Only the injection durations and subsequently the injected fuel masses were slightly adjusted to match the targeted NMEP while performing the SOI sweep. The injection timings ranged from 60 to 11 CAD aTDC. This injection sweep allows studying the effects of the physical properties of the fuel on the combustion performance and emissions of GCI engines. The physical property effects are investigated under different levels of mixture and temperature stratification. Less stratification is expected at SOI of 60 CAD aTDC while higher stratification is anticipated at SOI of 11 CAD aTDC. Moreover, GCI engines mostly operate within this SOI range at this particular load and speed. Therefore, the findings of this work will provide very useful insights on the impact of the fuel’s physical properties on the operation of this engine technology. The experimental CA10 and CA50 for the SOI sweep of the three tested fuels are presented in Fig. 4a. The CA10 (hollow symbols) and CA50 (solid symbols) are surprisingly similar for all the fuels at SOI later than 30 CAD aTDC. For earlier SOIs, SALN results in slightly earlier CA10 and CA50 than HSRN and PRF65. The combustion phasings for HSRN and PRF65 are similar throughout the SOI sweep despite their RON being moderately different by 5. Our hypothesis is that, for the conditions under study, the expected combustion phasing changes due to RON differences were offset by one or a combination of the following factors:

Fig. 4. Experimental CA10 (hollow symbols), CA50 (solid symbols), NOx, HC, smoke number and CO for the SOI sweep of the three tested fuels.


The differences in physical properties mainly the boiling range and the density. These properties affect the spray velocities and breakup and hence the mixing fields (temperature and equivalence ratio). This might subsequently influence the start of ignition and burn durations.
The slight variances in the injected mass which was required to match the NMEP in the experiment. The injected mass also affects the temperature/equivalence ratio distribution within the combustion chamber.
The differences in boundary conditions such as wall temperatures and the amount of trapped residual gases due to the slightly different burn durations and maximum pressure rise. These all can affect the autoignition of the mixture which is proportional to the temperature, residual gases and equivalence ratio. Fig. 4a also shows an important trend that combustion is most advanced at SOI of 30 and 40 CAD aTDC for all fuels. The NOx, hydrocarbon (HC), the filter smoke number (FSN) and CO emissions for all the tested SOI of the studied fuels are shown in Fig. 4. Similar to the combustion characteristics of the tested fuels,

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

the emissions are also nearly identical, with NOx (solid symbols in Fig. 4b) peaking at 30 CAD aTDC SOI, HC (hollow symbols in Fig. 4b) and CO emissions are lowest at 18 CAD aTDC SOI, and FSN are very low for all tested fuels at all SOI conditions. However, HSRN and PRF65 produce more soot at late injections, which is attributed to slower evaporation resulting in higher stratification. The emissions are mainly dictated by the combustion temperatures because all fuels have similar H/C ratio. The burn duration (CA90CA10), ignition delay (CA10SOI), maximum pressure rise, combustion efficiency, indicated mean effective pressure (IMEP) coefficient of variation (COV) and fuel injected mass are presented in Fig. 5. The burn durations (Fig. 5a) are similar for the SALN and HSRN, and they are slightly faster than that of the PRF65, resulting in slightly higher pressure rises as seen in Fig. 5c. The ignition delay times defined as CA10SOI (Fig. 5b) and the combustion efficiencies (Fig. 5d) are similar for all fuels at all SOI. The engine combustion is stable as shown from the IMEP COV (Fig. 5e) being lower than 3% for all cases except for the 60 CAD aTDC SOI where it rises up to around 5%. The fuel injected mass is around 0.45 ± 0.05 kg/h as shown in Fig. 5f with HSRN having consistently more mass compared to SALN and PRF65 to match the targeted NMEP.

1201

The studied fuels have differences in the ignition characteristics, indicated by their ON and in the physical properties such as the distillation curves, densities, and LHV. Yet, they present very similar combustion and emission features for almost all the conditions covered in these experiments. Further analysis and explanations of these differences/similarities will be performed in the subsequent computational investigation.

4. Computational model description and validation 4.1. Numerical setup CONVERGE CFD [30] was used to perform the engine simulations. Grid was initialized with the real engine geometry and was subsequently generated during run-time by both fixed embedding of cells and adaptive mesh refinement as will be elaborated more in Section 4.3. The pressure-velocity coupling in CONVERGE was achieved using a modified Pressure Implicit with Splitting of Operators (PISO) method of Issa [31]. The equations for momentum, pressure, density, energy, species, turbulence and heat transfer have been solved throughout the engine simulations presented

Fig. 5. Burn duration (CA90-CA10), ignition delay (CA10-SOI), ignition delay (CA10-SOI), combustion efficiency, maximum pressure rise, IMEP COV, and fuel injected mass for the SOI sweep of the three tested fuels.

1202

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

here. The momentum, density, energy, species, turbulence and heat transfer governing equations were solved using the pointwise successive over-relaxation (SOR) algorithm and the pressure was solved using the biconjugate gradient stabilized (BiCGSTAB) method. The convergence tolerances for density, energy, species and heat transfer are 1e4, for momentum is 1e5, for pressure is 1e8 and for turbulence is 1e3. These are the default values in CONVERGE for engine setup. The variable time-step algorithm was adopted here with a minimum and maximum time steps of 1e9 s and 5e4 s, respectively. For the spray modeling, a Lagrangian method is used by introducing discrete parcels (groups of droplets) of liquid into the gas phase computational domain. The spray from the outwardly opening piezo-injector is represented by a hollow-cone spray with string-like structures at the nozzle exit [32]. The mechanism of string-like structure has not yet been understood, and no appropriate models to predict such a string formation and accurate droplet sizes exist in the literature. As such, the modified Kelvin-Helmholtz and Rayleigh-Taylor (KH-RT) breakup model is implemented by feeding small cylindrical liquid blobs along the circular liquid sheet of a hollow-cone spray [26,27,33]. The spray models properly account for different in-cylinder pressure at different SOI and the increase in the in-cylinder pressure during injection. A sample of the resulting spray exit velocities from the engine simulations running with PRF65 is shown in Fig. 6. The engine simulations were performed for a full cycle starting before the exhaust valve opening (EVO) (160 crank angle degree (CAD)) and ending at 880 CAD. Prior to modeling the combustion cases, a motored run with a compression ratio (CR) of 14 and 1500 RPM was simulated from 160 CAD to 880 CAD, using the experimental in-cylinder, intake and exhaust port pressures for direct comparison. The default model parameters in CONVERGE were adopted, including the RNG k-e turbulent mixing model and the wall heat transfer model by O’Rourke and Amsden [34]. With the default simulations, the in-cylinder pressure at top dead center (TDC) was slightly over-predicted compared to experiments. This difference is attributed to multiple factors such as the uncertainties in the actual CR, the wall temperatures, and most importantly, the blow-by effect which cannot be measured accurately. These uncertainties were accounted for by adjusting the effective CR to 13.85 in the simulations to yield good agreement in the peak pressure. The multi-zone approach based on the SAGE chemistry solver [35] was used as a combustion sub-model. The NOx emissions were computed using the extended Zel’dovich mechanism [36].

Fig. 6. Spray exit velocities from the CFD engine simulations with different SOI.

Table 3 GCI numerical operating and boundary conditions. Effective CR Injector SOI (CAD aTDC) Injection duration (CAD) Injected mass (mg/cycle) Injection pressure (bar) Intake pressure (bar) Intake temperature (K) Fuel temperature (K) EGR level (%) Start of simulations (CAD) Tliner (K) Thead (K) Tpiston (K)

13.85 Hollow cone spray 60 to 11 4 (experimental values) 10 (experimental values) 130 1.2 304 363 0 160 403 403 423

The engine cases were setup according to the experimental boundary conditions (intake/exhaust temperature and pressure, injected fuel mass, injection duration and valve profiles) and setting the initial temperatures for the fuel (363 K), piston (423 K), liner (403 K) and head (403 K). The injected fuel mass (Fig. 5f) and injection duration for each simulated SOI case are taken from the experimental measurements. The boundary conditions utilized in the subsequent engine simulations are listed in Table 3. 4.2. Surrogate formulation For simulations of GCI engines, accurate description of ignition chemistry is critical. Since the real naphtha fuels contain a large number of components, they are often represented by a surrogate fuel consisting of a finite number of well-defined components. Different methodologies to formulate multi-component surrogates for gasoline fuels while matching various physical and chemical properties have been proposed [7,37–40]. Our recent work [26,27] showed that PRF and TPRF are good autoignition surrogates for the low sensitivity naphtha fuels. These surrogate fuel formulations were incorporated in the present modeling study. For the description of the complex chemical reactions, the reduced kinetic mechanism developed by Liu et al. [41] were used for all cases. In determining the fuel formulations, the PRF65 experiments were simulated using PRF65 as chemical surrogate. For the SALN and the HSRN, both have low sensitivities of around 2 and contain a large number of components. Therefore, a PRF62 and PRF60 have been adopted to represent SALN and HSRN, respectively, in order to match the RON and H/C ratios closely (2.264). For accurate description of mixing and combustion for a wide range of injection timings, the droplet breakup and evaporation are also important factors. For this, the PRF65 fuel was represented by the physical properties of the PRF65 components. For SALN, the physical properties were represented by a three-component surrogate containing n-pentane, n-hexane and 2,3-dimethyl-butane, with mass fractions of 0.493, 0.266 and 0.241, respectively. Furthermore, the lower heating value (LHV) of the PRF62 surrogate is forced to be that of SALN (45.1 MJ/kg). In other words, the SALN experiments were simulated with PRF62 as a chemical surrogate but with the liquid physical properties of the three-component surrogate. The vapor physical properties are those of the multicomponent chemical surrogate (PRF62). The exact formulation of the physical surrogate is determined to match the evaporation of a single droplet by employing the effective thermal conductivity/effective diffusivity (ETC/ED) model [42–44], which accounts for the heat and mass transport inside a spherical droplet. The results are shown in Fig. 7a, where it is shown that the three-component surrogate has evaporative characteristics similar to SALN but different from PRF62.

1203

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

Fig. 7. Plots of predicted droplet surface temperatures and radii versus time for different droplet mixtures for (a) SALN and its surrogates and (b) HSRN and its surrogates.

A similar methodology was adopted to formulate a physical surrogate for the HSRN as seen from Fig. 7b. PRF60 has a shorter droplet lifetime and different droplet surface temperature than the HSRN. Therefore, a three-component surrogate was formulated to mimic the evaporative characteristics of the HSRN, with 0.445 of n-hexane, 0.461 of 2,2-dimethyl-hexane and 0.093 of m-xylene in mass fractions. Similarly, the physical properties of the surrogate are used for the PRF60 engine simulations and LHV is set to be that of HSRN (44.63 MJ/kg). 4.3. Grid independent solution and model validation CONVERGE CFD [30] code automatically generates a high quality orthogonal mesh at run-time thus eliminating all user meshing time. CONVERGE automatically adds mesh resolution when and where it is needed based upon field variables to maximize accuracy while minimizing the run-time. The base mesh size used in this work was varied to study the grid resolution on the numerical solution. Four different base grid sizes (2 mm, 4 mm, 8 mm and 16 mm) were tested here. Few fixed and adaptive mesh refinement (AMR) embeddings were adopted here for finer grid resolutions at

areas of interest. For example, the mesh in the 4 mm base grid case was refined to 2 mm in a geometrical cylinder that contains the cylinder region and a part of the intake and exhaust ports. The cylinder region had additional embedding levels resulting in a grid size of 1 mm throughout the simulations. The mesh in contact with the intake and exhaust valve seats is refined to 0.5 mm. The mesh near the nozzle exit is refined to 0.25 mm during the injection process. In addition to these fixed embedding refinements, temperature and velocity AMR with an embedding level of 3 for each, resulting in the smallest grid size of 0.5 mm in the domain. Table 4 shows the grid sizes in the different locations of the numerical domain for the various base grid sizes. The PRF65 with SOI of 30 CAD aTDC was simulated using the four different base grid sizes and the resulting errors in the CA50 (CA50 from the 2 mm case is taken as the basis) and computational times (logarithmic scale) are shown in Fig. 8. As can be shown from Fig. 8, the computational time exponentially increases while decreasing the base grid size. However, the errors in CA50 decrease while decreasing the base grid size as expected. An error of 0.38 CAD in CA50 is obtained for a base grid size of 4 mm. Therefore, a base grid size of 4 mm is considered sufficient for the subsequent engine

Table 4 Fixed and AMR embedding grid sizes for different base grids. Base grid size (mm)

Geometrical cylinder grid size (mm)

Cylinder region grid size (mm)

Valve seats grid size (mm)

Injector nozzle grid size (mm)

Velocity and temperature AMR grid sizes (mm)

2 4 8 16

1 2 4 8

0.5 1 2 4

0.25 0.5 1 2

0.125 0.25 0.5 1

0.25 0.5 1 2

1204

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

Fig. 8. Errors in CA50 and computational times for different base grid sizes. Fig. 10. Experimental and calculated CA10, CA50 and CA90 for PRF65 at all SOI.

Fig. 9. Measured and calculated CA50 of the various PRF65 cases with different SOI.

simulations considering the significant savings in computational times (480 CPU hours for 4 mm compared to 12,000 CPU hours for 2 mm base grid size). Starting with PRF65, the cases with various SOI were run from EVO (160 CAD) for a full cycle. The resulting CA50 from the PRF65 engine simulations are plotted against the experimental data in Fig. 9. The calculated CA50s over-predict the measurements for PRF65 at all SOI. However, the qualitative trends that the combustion is most advanced at 30 CAD aTDC SOI are successfully reproduced. To better represent the experimental combustion phasings and the resulting emissions, the SAGE reaction multiplier was set to 1.25 for all simulations in order to account for other uncertainties in the boundary and initial conditions, resulting in good agreement with the experiments as shown in Fig. 9. Note that the overall trends and conclusions remain unchanged without the reaction multiplier adjustment. 5. Results and discussion The numerical simulations for the PRF65 case were conducted and the resulting CA10, CA50 and CA90 are compared with the experimental data in Fig. 10. The trends in CAs are successfully reproduced by the simulations where CA10 levels off for SOI earlier

than 30 CAD aTDC, and CA50/CA90 continue to increase with reasonable quantitative agreements. Similar calculations were performed for SALN and HSRN using the experimental conditions presented earlier. Fig. 11 shows the simulation results of combustion phasings and emissions for the three fuel cases, to be compared with the experimental data shown in Fig. 2. The calculated CA10 and CA50 are similar for all three fuels for SOI later than 30 CAD aTDC in agreement with the measurements. However, they slightly deviate for earlier SOIs; the CA10 and CA50 for PRF65 are slightly longer than SALN and HSRN, which is attributed to the higher RON for PRF65. This was not observed experimentally where combustion phasings were closer for all fuels. Considering the number of uncertainties in both the experimental measurements and numerical calculations, it is concluded that the computational trends compare reasonably well with the experimental data. Note also that the modeling assumption that all three fuels have the same boundary (walls temperatures) and initial (trapped residual gases concentrations and temperatures) conditions may have an effect the combustion phasings as noted earlier. The NOx and HC emissions also show similar behavior as experiments with peak values at 30 and 18 CAD aTDC SOI, respectively. Note that the NOx and HC emissions are slightly under-predicted numerically and it is beyond the scope of the current work to achieve quantitative agreements with the experiments. To better understand the similarities and differences in the combustion characteristics of the three different fuels in GCI engines, a more systematic approach to isolate the chemical and physical effects was undertaken. First, the chemical effects of the fuels on engine combustion were examined for the PRF65 fuel case. Two sets of simulations were run with the same physical properties, boundary and initial conditions but with PRF60 and PRF65 chemical surrogates. Fig. 12 shows the results, indicating that the RON difference of 5 yields noticeable differences in combustion phasings because of the higher reactivity of the fuel. Thus the chemical effects are believed to be primarily responsible for the differences in CA10 and CA50 among different fuels observed in Fig. 11. Note that the discrepancies between the two simulations decrease as SOI is delayed, which is attributed to the lower sensitivity of the ignition delay times of PRF 60 and PRF65 at rich mixtures. The homogeneous ignition delay times for PRF60 and PRF65/air mixtures were calculated at 40 bar and a range of temperatures for different equivalence ratios, and the results are presented in Fig. 13. The combustion phasing advancement when using PRF60 instead

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

1205

Fig. 13. Calculated constant volume homogeneous ignition delay times for PRF60 (solid lines) and PRF65 (dashed lines)/air mixtures at 40 bar using Liu et al. mechanism and reaction multiplier of 1.25.

Fig. 11. Calculated CA10 (hollow), CA50 (solid), NOx (solid) and HC (hollow) for the three fuels at all SOI.

Next computational exercise was to isolate the physical properties effect on the combustion phasing in compression ignition mode. Three simulations were run for the SALN fuel case with the same PRF62 chemical surrogate but with different physical properties: the physical properties of SALN, those of HSRN, and finally the physical properties of n-hexadecane which has very high boiling point (559.8 K) were chosen as an extreme scenario in order to show the extent of the effects by different physical properties. The same boundary and initial conditions and fuel injected mass and duration were used. The calculated CA10 and CA50 for PRF62 with different physical properties as a function of SOI are shown in Fig. 14. While the simulations with SALN and HSRN physical properties yielded similar CA10 and CA50 at all SOI, the simulations with much heavier n-hexadecane physical properties show that the combustion is significantly delayed for early injections (40 CAD aTDC) and even leads to misfire for 50 CAD aTDC SOI. This exercise indicates that the CA10/CA50 predictions are more sensitive to chemical properties (RON) than to the physical properties; only an extreme condition of boiling point has a comparable level of effects.

Fig. 12. Calculated CA10 and CA50 for PRF65 and PRF60 with PRF65 physical properties (indicated in parentheses) at all SOI.

of PRF65 is purely due to the shorter ignition delay times as can be seen from Fig. 13. The differences in ignition delay times between PRF60 and PRF65 decrease as U increases. This explains the lower combustion phasing sensitivity at later SOI where the fuel is mixed at richer U as will be shown later.

Fig. 14. Calculated CA10 (hollow symbols) and CA50 (solid symbols) for PRF62 with different physical properties at all SOI.

1206

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

ducted. As a heuristic measure, density-weighted mean equivalence ratio and temperature are determined by:

R q dV U ¼ RV F ; q dV V F

R q TdV T ¼ RV F q dV V F

ð1Þ

where qF is the local fuel vapor density excluding the liquid phase which is calculated using the following expression:

qF ¼ Y F  qTot

ð2Þ

where qTot is the total local vapor density and Y F is the local fuel mass fraction. In other words, U and T represent the equivalent ratio and temperature in the region of higher fuel vapor concentration. The two quantities were calculated for the non-reacting cases at TDC position (0 CAD) to represent the start of combustion. Fig. 15 shows the two quantities for the three different physical properties as considered in Fig. 14. With SALN and HSRN physical properties, both quantities behave similarly, while with the physical properties of much heavier n-hexadecane, higher mean temperature and lower mean equivalence ratio are observed due to the lower evaporation rate. Note that the injected fuel will be mostly distributed around the reported Umean and Tmean. Having similar Umean

Fig. 15. Calculated U (solid symbols) and T (hollow symbols) for the cases presented in Fig. 14 from the non-reacting simulations at TDC position.

To further substantiate the main effects of the different physical properties on the combustion phasings, a detailed investigation of the equivalence ratio (U)-temperature (T) distribution is con-

n-Hexadecane

HSRN

SALN

Фmean

Фmean Фmean

Фmean

Фmean

Фmean

Фmean

Фmean Фmean

Fig. 16. Non-reacting contours of U-T maps for PRF62 with different physical properties with various SOI at TDC position.

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

Fig. 17. Calculated constant volume homogeneous ignition delay times for PRF62/ air mixtures at 40 bar using Liu et al. mechanism and reaction multiplier of 1.25 (solid lines) and reaction multiplier of 1 (dashed lines).

Fig. 18. Global equivalence ratio for the total evaporated mass with the air intake at three SOI and different fuels: solid lines (PRF62 (SALN)), dotted lines (PRF62 (HSRN) and dashed lines (PRF62 (n-hexadecane)).

and Tmean for PRF62 (SALN) and PRF62 (HSRN) does not necessarily mean they have the exact same U-T distribution. The detailed U-T maps colored by the fuel mass fraction for all three fuels at different SOI are presented in Fig. 16. As can be seen from Fig. 16, even though Umean and Tmean values for PRF62 with SALN and HSRN properties are similar, their U-T maps are slightly different due to their different physical properties. However, these small differences in U-T maps were not reflected on the combustion phasings because ignition is initiating at similar conditions (U and T). The homogenous constant volume ignition delay times for PRF62/air mixtures at 40 bar, U ranging from 0.25 to 3 and temperatures between 700 and 1200 K are presented in Fig. 17 with reaction multipliers of 1.25 and 1. The ignition delay times of PRF62/air mixtures decrease with increasing U and have a negative temperature coefficient (NTC) region between 750 and 950 K. Most of the Tmean values presented in Fig. 15 (right Y-axis) are within this NTC temperature range. The ignition delay times are sensitive to both the equivalence ratios and temperatures of the

1207

fuel/air mixtures and it is always a trade-off between these two parameters which can explain the differences in the combustion phasings at different SOI. The sensitivity of the ignition delay times to U and T is conservative when using a reaction multiplier of 1.25 compared to 1. Tmean values for SOI less than 30 CAD aTDC are well within the NTC region where the ignition delay times are more sensitive to the equivalence ratio. Even when Tmean for SOI of 50 CAD aTDC is higher than that of 30 CAD aTDC and having more residence time, the combustion phasing is more advanced for 30 CAD aTDC because of the higher Umean. For SOI later than 30 CAD aTDC, the ignition delay times might be shorter than those of earlier SOI because of the higher Umean but the combustion phasing is retarded because of the shorter residence time. As an alternative metric, the global equivalence ratio defined as the total mass of the evaporated fuel and the total mass of air intake to the engine is computed at three SOI (50, 30 and 11 CAD aTDC) and for the same three cases (PRF62 chemistry with SALN, HSRN and n-hexadecane physical properties). The results are presented in Fig. 18. For SOI of 50 CAD aTDC, it is clearly seen that less than half of the PRF62 with n-hexadecane physical properties is evaporated at TDC. For injections earlier than 30 CAD aTDC, the in-cylinder air temperature is relatively low (500– 600 K) at which the low volatility of the high boiling temperature of n-hexadecane (559.8 K) is pronounced. Moreover, the higher spray exit velocity at earlier SOI conditions results in increased fuel penetration and collision with the colder piston wall, making it harder to evaporate. This effect is seen as the plateau of the dashed line for the 50 SOI case. Such excessively low evaporation behavior with the n-hexadecane fuel is responsible for the observed misfire results shown in Fig. 14. Most of the PRF62 (n-hexadecane) fuel has evaporated at TDC for SOI of 30 CAD aTDC which explains the similar TDC Umean for all fuels at this SOI. For SOI later than 18 CAD aTDC, the PRF62 (n-hexadecane) completely evaporates after TDC (10 CAD at 11 CAD aTDC SOI). This explains the much lower TDC Umean for PRF62 (n-hexadecane) at SOI later than 18 CAD aTDC (see Fig. 15). The evaporated global equivalence ratio for PRF62 (SALN) and PRF62 (HSRN) are very similar with PRF62 (HSRN) being slightly slower right after SOI as can be seen from the solid and dotted lines in Fig. 18. This small difference in evaporative characteristics does not affect the combustion phasings of these two fuels because both fuels are fully evaporated close to TDC position where ignition occurs. The different distillation (Fig. 3) and evaporative (Fig. 7) curves of these two fuels did not reflect on the experimental and numerical combustion phasings of the cases studied here because of the high in-cylinder air temperatures during injection causing these fuels to evaporate quickly and mix similarly with the in-cylinder air. In summary, for the three fuels and engine operation conditions under study, the combustion phasings are found to be more sensitive to the auto-ignition chemistry (to the extent of the RON difference by 5) than the differences in the physical properties. Only an extremely low volatility (represented by the n-hexadecane case) was able to affect the combustion phasing at comparable magnitudes. The experimental measurements still did not show the same level of differences in CA10/CA50 as suggested by the simulation results shown in Fig. 12.

6. Conclusions In this study, an experimental and numerical investigation of the Saudi Aramco GCI engines with three fuels with different chemical and physical characteristics have been performed. We have investigated the effects of chemical and physical properties for different injection timings (SOI) at part load conditions using

1208

J. Badra et al. / Applied Energy 183 (2016) 1197–1208

two naphtha streams and one primary reference fuel in a single cylinder engine under the GCI combustion mode. The effect of fuel/air mixing on the combustion phasing of the engine was investigated in detail using equivalence ratio-temperature maps and evaporation characteristics. Key findings are summarized as follows: 1. The measurements and calculations showed that an SOI of 30° CAD aTDC has the most advanced combustion phasing and ignition is retarded for earlier and later SOI. 2. NOx emissions were highest for SOI of 30 CAD aTDC which corresponded to the lowest CO and HC emissions. 3. Experimental results showed that all three fuels behaved closely in terms of combustion phasings and emissions. 4. Computational parametric studies showed that some observable differences in combustion phasings may result from the differences in the ignition chemistry among the fuels (RON ranging from 60 to 65). 5. On the other hand, the realistic differences in the physical properties among the three fuels were found to be insufficient to lead to meaningful differences in the combustion phasing. 6. The similar combustion phasing behavior among the three fuels observed in the experiment is attributed to the fact that the RON differences were offset by the few factors that were not accounted for numerically. Acknowledgments This work was sponsored by the Fuel Technology Division at Saudi Aramco R&DC. The surrogate formulation work at King Abdullah University of Science and Technology (KAUST) was supported by KAUST and Saudi Aramco under the FUELCOM program. References [1] U.S. Energy Information Administration. International energy outlook 2013; 2013. [2] ExxonMobil. 2012 Energy Outlook; 2012. [3] Kalghatgi G. Fuel/engine interactions. Warrendale, PA: SAE International; 2013. [4] Kalghatgi GT. The outlook for fuels for internal combustion engines. Int J Engine Res 2014. http://dx.doi.org/10.1177/1468087414526189. [5] Kalghatgi G, Risberg P, Ångström H. Advantages of fuels with high resistance to autoignition in late-injection, low-temperature, compression ignition combustion. SAE Technical Paper 2006-01-3385; 2006. [6] Manente V, Johansson B, Cannella W. Gasoline partially premixed combustion, the future of internal combustion engines? Int J Engine Res 2011;12:194–208. [7] Li J, Yang WM, An H, Chou SK. Modeling on blend gasoline/diesel fuel combustion in a direct injection diesel engine. Appl Energy 2015;160:777–83. [8] Chang J, Kalghatgi G, Amer A, Viollet Y. Enabling high efficiency direct injection engine with naphtha fuel through partially premixed charge compression ignition combustion. SAE Technical Paper 2012-01-0677; 2012. [9] Manente V, Johansson B, Tunestal P. Characterization of partially premixed combustion with ethanol: EGR sweeps, low and maximum loads. J Eng Gas Turbines Power 2010;132. 082802-1-7. [10] Han D, Ickes AM, Bohac SV, Huang Z, Assanis DN. Premixed low-temperature combustion of blends of diesel and gasoline in a high speed compression ignition engine. Proc Combust Inst 2011;33:3039–46. [11] Najt PM, Foster DE. Compression-ignited homogeneous charge combustion. SAE Technical Paper 830264; 1983. [12] Onishi S, Jo SH, Shoda K, Jo PD, Kato S. Active Thermo-Atmosphere Combustion (ATAC) – a new combustion process for internal combustion engines. SAE Technical Paper 790501; 1979. [13] Noguchi M, Tanaka Y, Tanaka T, Takeuchi Y. A study on gasoline engine combustion by observation of intermediate reactive products during combustion. SAE Technical Paper 790840; 1979. [14] Thring RH. Homogeneous-Charge Compression-Ignition (HCCI) engines. SAE Technical Paper 892068; 1989. [15] Gan S, Ng HK, Pang KM. Homogeneous Charge Compression Ignition (HCCI) combustion: implementation and effects on pollutants in direct injection diesel engines. Appl Energy 2011;88:559–67.

[16] Iwabuchi Y, Kawai K, Shoji T, Takeda Y. Trial of new concept diesel combustion system - premixed compression-ignited combustion. SAE Technical Paper 1999-01-0185; 1999. [17] Kimura S, Aoki O, Ogawa H, Muranaka S, Enomoto Y. New combustion concept for ultra-clean and high-efficiency small DI diesel engines. SAE Technical Paper 1999-01-3681; 1999. [18] Wang B, Wang Z, Shuai S, Xu H. Combustion and emission characteristics of Multiple Premixed Compression Ignition (MPCI) mode fuelled with different low octane gasolines. Appl Energy 2015;160:769–76. [19] Wang B, Wang Z, Shuai S, Yang H, Wang J. Combustion and emission characteristics of Multiple Premixed Compression Ignition (MPCI) fuelled with naphtha and gasoline in wide load range. Energy Convers Manage 2014;88:79–87. [20] Yang H, Shuai S, Wang Z, Wang J. Fuel octane effects on gasoline multiple premixed compression ignition (MPCI) mode. Fuel 2013;103:373–9. [21] Kim K, Wang Z, Wang B, Shuai S, Yang H, Bae C. Load expansion of naphtha multiple premixed compression ignition (MPCI) and comparison with partially premixed compression ignition (PPCI) and conventional diesel combustion (CDC). Fuel 2014;136:1–9. [22] Hao H, Liu F, Liu Z, Zhao F. Compression ignition of low-octane gasoline: life cycle energy consumption and greenhouse gas emissions. Appl Energy 2016;181:391–8. [23] Kim D, Martz J, Violi A. Effects of fuel physical properties on direct injection spray and ignition behavior. Fuel 2016;180:481–96. [24] Lacey J, Kameshwaran K, Sathasivam S, Filipi Z, Cannella W, Fuentes-Afflick PA. Effects of refinery stream gasoline property variation on the auto-ignition quality of a fuel and homogeneous charge compression ignition combustion. Int J Engine Res 2016. http://dx.doi.org/10.1177/1468087416647646. [25] Ra Y, Yun JE, Reitz RD. Numerical parametric study of diesel engine operation with gasoline. Combust Sci Technol 2009;181:350–78. [26] Badra JA, Sim J, Elwardany A, Jaasim M, Viollet Y, Chang J, et al. Numerical simulations of hollow-cone injection and gasoline compression ignition combustion with naphtha fuels. J Energy Res Technol 2016;138:052202. [27] Badra J, Elwardany A, Sim J, Viollet Y, Im HG, Chang J. Effects of in-cylinder mixing on low octane gasoline compression ignition combustion. SAE Technical Paper 2016-01-0762; 2016. [28] Chang J, Viollet Y, Amer A, Kalghatgi G. Fuel economy potential of Partially Premixed Compression Ignition (PPCI) combustion with naphtha fuel. SAE Technical Paper 2013-01-2701; 2013. [29] Viollet Y, Chang J, Kalghatgi G. Compression ratio and derived cetane number effects on gasoline compression ignition engine running with naphtha fuels. SAE Int J Fuels Lubricants 2014;7:412–26. [30] Senecal P, Richards K, Pomraning E. CONVERGE (version 2.2.0) manual. Madison, WI: Convergent Science Inc.; 2014. [31] Issa RI. Solution of the implicitly discretised fluid flow equations by operatorsplitting. J Comput Phys 1986;62:40–65. [32] Martin D, Cardenas M, Pischke P, Kneer R. Experimental investigation of near nozzle spray structure and velocity for a GDI hollow cone spray. Atom Sprays 2010;20:1065–76. [33] Sim J, Badra J, Elwardany A, Im HG. Spray modeling for outwardly-opening hollow-cone injector. SAE Technical Paper 2016-01-0844; 2016. [34] Amsden AA. Los Alamos National Laboratory; 1997. p. Report No. LA-13313MS. [35] Senecal PK, Pomraning E, Richards KJ, Briggs TE, Choi CY, McDavid RM, et al. Multi-dimensional modeling of direct-injection diesel spray liquid length and flame lift-off length using CFD and parallel detailed chemistry. SAE Technical Paper 2003-01-1043; 2003. [36] Heywood JB. Internal combustion engine fundamentals. McGraw-Hill; 1988. [37] Ahmed A, Goteng G, Shankar VS, Al-Qurashi K, Roberts WL, Sarathy SM. A computational methodology for formulating gasoline surrogate fuels with accurate physical and chemical kinetic properties. Fuel 2015;143:290–300. [38] Su M, Chen CP. Heating and evaporation of a new gasoline surrogate fuel: a discrete multicomponent modeling study. Fuel 2015;161:215–21. [39] Yan J, Lee DJ, Chou SK, Desideri U, Li H, Shuai S, et al. Combustion and emission characteristics of Multiple Premixed Compression Ignition (MPCI) mode with low octane gasoline fuels. Energy Procedia 2014;61:2127–31. International Conference on Applied Energy, ICAE2014. [40] Kalghatgi GT, Babiker H, Badra J. A simple method to predict knock using toluene, n-heptane and iso-octane blends (TPRF) as gasoline surrogates. SAE Int J Eng 2015;8:505–19. [41] Liu Y-D, Jia M, Xie M-Z, Pang B. Development of a new skeletal chemical kinetic model of toluene reference fuel with application to gasoline surrogate fuels for computational fluid dynamics engine simulation. Energy Fuels 2013;27:4899–909. [42] Gusev IG, Krutitskii PA, Sazhin SS, Elwardany AE. New solutions to the species diffusion equation inside droplets in the presence of the moving boundary. Int J Heat Mass Transfer 2012;55:2014–21. [43] Sazhin S. Droplets and sprays. London: Springer; 2014. [44] Sazhin SS, Al Qubeissi M, Nasiri R, Gun’ko VM, Elwardany AE, Lemoine F, et al. A multi-dimensional quasi-discrete model for the analysis of Diesel fuel droplet heating and evaporation. Fuel 2014;129:238–66.