Applied Energy 111 (2013) 758770

Contents lists available at SciVerse ScienceDirect

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

An advanced combustion model coupled with detailed chemical reaction mechanism for D.I diesel engine simulation
Amin Maghbouli, Wenming Yang , Hui An, Jing Li, Siaw Kiang Chou, Kian Jon Chua
Department of Mechanical Engineering, National University of Singapore, 9 Engineering Drive 1, Singapore 117576, Singapore

h i g h l i g h t s
 An advanced multi-component fuel

g r a p h i c a l a b s t r a c t

combustion model was developed.

 Numerical simulations were

conducted for a diesel engine at full and mid loads/speeds.  Temporal/spatial variation of the emissions were extracted from numerical simulations.  CO2, CO and UHC emissions were reduced at the engine mid load condition.

a r t i c l e

i n f o

a b s t r a c t
A multi-dimensional computational uid dynamics (CFD) modeling was conducted on a direct injection turbo-charged diesel engine based on KIVA-4 code under full and mid engine loads. Multi-component fuel evaporation model of KIVA-4 was used and coupled with advanced combustion chemistry to generate a multi-component fuel combustion model by integrating CHEMKIN II into the KIVA-4 code. As the coding schema of KIVA-4 in the case of data/parameter allocation, etc. was different compared to previous version of KIVA-3V, a considerable amount of FORTRAN programming was performed in order to develop a multi-component fuel combustion model. The developed combustion model was capable of modeling combustion process of number of chemical species as the components of direct injected liquid fuel. Comparing to the single component fuel combustion model, new model is capable of comprehensive combustion modeling of blend fuel and heavy hydro-carbon fuels. Furthermore, spray breakup and collision models were changed to more advanced KelvinHelmholtz and RayleighTaylor (KHRT) and ORourke models, respectively. The model was used to simulate direct injected diesel engine under full and mid engine loads at three engine speed conditions. Extracted temporal and spatial results for equivalence ratio distribution inside the combustion chamber showed that under full load condition, a considerable amount of fuel was trapped in piston bowl after initiation of the injection process where such fuel rich local regions provide the potential for production of higher soot emission. Mean value of the fuel concentration history showed that the ignition delay was increased under mid engine load at all engine speeds producing higher amounts of unburned hydro carbons and carbon monoxide. By reducing engine load and speed, output power was decreased as well. However, same trend was not reported for the indicated thermal efciency as the middle engine speed in considered engine loads, had slightly higher efciency. 2013 Elsevier Ltd. All rights reserved.

Article history: Received 21 September 2012 Received in revised form 6 April 2013 Accepted 13 May 2013 Available online 15 June 2013 Keywords: Diesel engine Direct injection KIVA-4 Multi-component combustion model

Corresponding author. Tel.: +65 65166481.

E-mail address: mpeywm@nus.edu.sg (W. Yang). 0306-2619/$ - see front matter 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.apenergy.2013.05.031

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

759

Nomenclature A0 D J M P Nc _c Q _s Q t u W uid ow constant diffusion coefcient heat ux number of reactions in chemical kinetics mechanism pressure number of cylinders source term due to chemistry source term owing to spray time velocity vector molecular weight Abbreviations ATDC After Top Dead Center CAD Crank Angle Degree CFD computational uid dynamics CO Carbon Monoxide CO2 Carbon Dioxide D.I Direct Injection DVODE double-precision variable-coefcient ordinary differential equation solver EGR Exhaust Gas Recirculation EVO Exhaust Valve Opening HCCI Homogeneous Charge Compression Ignition HRR Heat Release Rate IC Internal Combustion IVC Inlet Valve Closure ITE indicated thermal efciency NOx Nitrogen Oxides SOI Start of Injection TDC Top Dead Center UHC Unburned Hydro Carbon

Subscripts M number of species in chemical kinetics mechanism n number of the fuel components Greek symbols dml dirac delta function Dh heat of formation at absolute zero f q density _c q density change rate due to chemistry m _s q density change rate owing to Spray m e dissipation rate of turbulence _m x molar production rate

1. Introduction Diesel engines have proved their reliability as a beating heart for the mass transportation and power generation units. The most important factor that has caused diesel engines to surpass gasoline engines, is their lower fuel consumption associated with high specic output power and torque. Despite the fact that diesel engines fulll the daily increasing power demand and seem to be a promising tool to achieve efcient fuel consumption, conventional diesel cycle suffers from high levels of NOx and Soot emissions at high engine loads, and low thermal efciency and high levels of Unburned Hydro Carbon (UHC) emissions at lower engine load conditions [1,2]. This is why there has been a signicant amount of researches worldwide to optimize and introduce new technologies in order to minimize emissions from diesel combustion while providing acceptable power output. These modern technologies have mostly concentrated on direct injecting of the fuel into the combustion chamber with precise timing using common-rail fuel injectors and the use of Exhaust Gas Recirculation (EGR) to reduce in-cylinder peak temperature, avoiding massive production of NOx emissions [36]. As an alternative approach to reduce emission levels and fossil fuel usage, application of several environmentally friendly fuels such as: ether, alcohol, vegetable/fat based bio-diesel or gaseous fuel in diesel cycle was investigated by many researchers. Ying et al. studied behavior of the diesel and di-methyl ether (DME) blends on a diesel engine at low and high engine loads [7]. They concluded that at the high load conditions, smoke emissions were reduced about 5868% for the blend fuel in comparison to the pure diesel. At low load conditions, smoke emissions for the blend fuel were almost comparable to the neat diesel. In addition, by increasing DME, NOx emission was gradually decreased, while CO and UHC emissions were increased at the most operating conditions. Papagiannakis and Hountalas conducted experiments on a diesel engine running under dual-fuel mode for various engine loads, where liquid diesel was partially replaced with natural gas [8].

The results showed that using gaseous fuel reduced the peak cylinder pressure. In the case of pollutant emissions, they concluded that addition of the gaseous fuel had a positive effect on reducing NO emission. However, at lower engine load, the CO and UHC emissions were higher compared to the normal diesel operation. Although experiments are essential tools to study, understand and optimize the typical IC engines, the range of their applicability in terms of economy, time and available infrastructures is the matter of debate. Numerical simulations of the IC engines through detailed mathematical framework have been shown to be a very useful tool for saving experimental resources. Nowadays, with the advent of powerful computers with high computational capacities, numerical modeling has proved its exibility and low cost. Generally, engine modeling mathematical framework can be classied into two main groups: thermodynamic and dimensional models. Thermodynamic models assume energy equation as the main conservation equation and are subdivided into single and multi-zone models [911]. Since physical processes such as turbulent ow effects are not directly modeled, some correlations with numerous constants are used instead. Kouremenos et al. used a multi-zone thermodynamic model to study a D.I diesel engine at different engine loads and speeds [12]. Results showed that fuel consumption was increased by increasing the engine load for all operating engine speeds. To simulate performance and emissions of a D.I diesel engine, Rakopoulos and Hountalas developed a 3D multi-zone thermodynamic model [13]. In their multi-zone model, engine uid ow was not taken into account directly in simulations; however, they achieved to a good agreement between simulations and experimental measurements for in-cylinder pressure, heat release rate and emission data. Hountalas et al. developed a multi-zone thermodynamic model to investigate characteristics of a diesel engine under full and partial loads [14]. They concluded that higher injection pressure signicantly increased NO emission especially at low speed and partial load, whereas positive effect was resulted on soot reduction.

760

A. Maghbouli et al. / Applied Energy 111 (2013) 758770 Table 1 Toyota 2KD-FTV engine specications. Engine aspiration Engine cycle Number of cylinders Injector orices Fuel injection system Bore stroke Displacement volume Compression ratio Max. output power Max. output torque Turbo-charged Diesel 4 Stroke 4 Inline 6 Centric Common rail, denso 92 93.8 (mm) 2.494 (lit) 18.5:1 75 (kW) @ 3600 (rpm) 260 (N m) @ 1600 (rpm)

By contrast, in dimensional models temporal and spatial variations of the ow eld, pressure, composition, temperature and turbulence inside the combustion chamber are taken into account [15]. Various sub-models are applied in dimensional models to simulate different aspects of in-cylinder processes. Besson et al. optimized HCCI dedicated diesel engine combustion chamber geometry under engine full load condition using KIVA II-Renault code [16]. They concluded that widest bowl geometry at full load provides the highest gross mean effective pressure and lowest soot emission. Ehleskog et al. used KIVA-3 code and diesel surrogate fuel chemical kinetics together with turbulence-chemistry interactions in multi-dimensional simulation of D.I diesel engine at a single engine speed and just for 25% load condition [17]. They studied effects of split injection and concluded that split injection using a common rail piezo-electric common rail injector could be used to reduce NOx emission and specic fuel consumption without increasing soot emissions at lower loads. However, CO and UHC emissions tended to increase in all cases including split injection. Helmantel and Golovitchev used 3D-CFD with chemical kinetics to simulate triple injection strategy to reduce soot and NOx emission levels in a single cylinder D.I. diesel engine at the medium load [18]. As they inferred, although split injection together with high levels of EGR could reduce NOx emission, soot levels would increase due to reduction in combustion chamber temperature. To overcome this, triple injection was introduced to oxidize the produced soot particles. Suggested amount of mass and timing for the third injection or so called post injection was 1730% of total fuel mass at 1012 crank angle ATDC, respectively. Shuai et al. used KIVA-3V coupled with chemical kinetics to simulate injection rate-shape effect on engine performance and emission under low load operating condition [19]. They concluded that the early injection timing gave lower soot, UHC and CO emissions but higher NOx emissions compared to the late injection timing. By the reference to extensive research on 3D-CFD modeling of the diesel engines, rare works have used KIVA-4 incorporated with detailed chemical reaction model to investigate the combustion and emission formation process in the terms of temporal and spatial variation. Especially, if the purpose of the diesel engine development and optimization is in the vehicle application, detailed studying of its behavior in different engine loads and speeds through numerical simulations is essential. The most important aim of this study is to generate a promising numerical workstation which would be able to accurately predict combustion and

emission characteristics of a typical diesel engine for variety of engine loads and speeds. In this paper, the conducted experiments will be briey discussed rst and then the developed mathematical simulation framework will be presented. After validation of the numerical results by the experiments variety of engine combustion and performance parameters will be analyzed. 2. Experimental procedure 2.1. Engine setup and specications The experiments were performed on a four cylinder Toyota 2KD-FTV common rail fuel injection diesel engine in the engine laboratory of National University of Singapore. A schematic diagram of the engine test bed is shown in Fig. 1 and the detailed engine specications are listed in Table 1. The engine was loaded with an AVL DP 160 water-cooled passive eddy current dynamometer which is able to provide a peak brake power of 160 kW and a maximum torque of 400 N-m with an accuracy of 0.3%. The air ow rate was measured using an AVL Sensyow P air ow meter with a resolution of 100 ms for sampling frequency. An AVL 733S.18 fuel balance was used to measure the fuel consumption rate with a sampling frequency of 500 ms and an accuracy of 1%. The cylinder pressure was measured at a resolution of 1 CAD by an AVL GH13P water-cooled pressure transducer which was mounted on the cylinder head, and it can sustain a peak pressure of 250 bars. 2.2. Experimental test cases Experimental cases including three engine speeds of 1200, 2400 and 3600 rpm under mid and full engine loads were considered to

Fig. 1. Schematic diagram of the diesel engine test bed.

A. Maghbouli et al. / Applied Energy 111 (2013) 758770 Table 2 Selected experimental cases (IVC = 149 CAD ATDC, EVO = 150 CAD ATDC). Engine speed (rpm) Engine Load Pressure at IVC (bar) Temperature at IVC (K) Wall temperature (K) Initial charge Diesel mass (gr/cycle) Air mass (gr/cycle) Total equivalence ratio SOI (CAD ATDC) 1200 50% 1.161 373 523 Air + 5% 0.026 0.844 2.16 7 100% 1.27 374 525 EGR 0.035 0.7 1.33 5 2400 50% 1.686 380 529 Air + 7% 0.022 0.957 1.72 9 100% 1.81 383 530 EGR 0.037 0.955 1.72 10 3600 50% 1.87 397 531 100% 1.89 398 539

761

1:5C14 H28 0:5O2 ) 2C7 H16 C7 H8 H2 O

Air + 10% EGR 0.025 0.044 0.967 0.957 2.58 1.45 16 17

perform numerical simulations. Input parameters including initial and boundary conditions for these cases are illustrated in Table 2. These parameters will be used on simulation process of the cases discussed above. 3. Multi-dimensional modeling 3.1. 3D-CFD modeling tool Accurate and more realistic simulation of the liquid fuel spray, particle collision, oscillation, atomization and evaporation encountered by highly turbulent reactive uid ow in the diesel engines requires utilization of a multi-dimensional CFD code. In this study, uid ow simulation was carried out using the latest version of Los Alamos National Laboratory (LANL) CFD code, KIVA-4 [20]. KIVA is a computer code for the numerical calculation of the transient, two and three dimensional chemically reactive ows with sprays. It uses a time-marching, nite-volume scheme which solves the conservation equations of mass, momentum, energy and accounts for turbulence using an Arbitrary Lagrangian Eulerian (ALE) method in three solution phases. Comparing to the earlier versions of the KIVA code, detailed evaporation model together with ability for modeling unstructured mesh were added to the KIVA-4. Nonetheless, some basic sub-models in the default KIVA-4 require further modications to maximize code predictability. In doing so, KelvinHelmholtz and RayleighTaylor (KH-RT) spray break up models [21] were implemented in KIVA-4 calculations for spray primary and secondary breakup stages, respectively. Moreover, ORourke collision model [22] was used for fuel particle collision and the RNG ke model was used as turbulence model. 3.2. Multi-component fuel combustion model 3.2.1. Insights to the applied chemical kinetics mechanism Developing comprehensive reduced chemical kinetics mechanisms for different fuels, researchers needed to include one or two global reactions at the beginning of the chemical kinetics mechanism of the heavy hydrocarbons such as diesel and biodiesel fuels [2325]. This is because either chemical pathways of molecular break down of the considered fuel surrogate was not available or number of intermediate chemical species and kinetics pathways of their production/destruction would be so large that the nal mechanism cannot be used in 3D CFD calculations. For instance: Bergman and Golovitchev [23] developed a Diesel Oil Surrogate (DOS) chemical kinetics mechanism for conventional diesel fuel oxidation chemistry and have been applied it in IC engine simulations. They have used C14H28 as the diesel fuel surrogate but chemical reaction pathways and number of intermediate species for C14H28 oxidation chemistry would be so high that the mechanism cannot be used in the engine 3D-CFD application. To remedy computational time issue, they introduced a global single step reaction which describes dissociation of 1.5 mol of C14H28 into 2 mol of n-heptane (C7H16) and 1 mol of toluene (C7H8):

In doing so, oxidation chemistry of the target heavy hydrocarbon fuel, in this case C14H28, was dened by two lighter hydrocarbons which rstly can represent components of the major fuel and secondly have applicable chemical kinetics mechanism in terms of number of the intermediate species and reactions. Ehleskog et al. [17] have used DOS mechanism in simulation of direct injected diesel engine using single fuel component, C14H28, and the global single step reaction in its oxidation as discussed in Eq. (1). Nonetheless, this approach still has some deciencies such as; it is not suitable for the multi-component fuel evaporation model, it considers limited number of fuel components and its difculty in determination of chemical kinetics rate coefcients of the introduced global reaction which can signicantly affect ignition delay time and combustion [17]. In this study, detailed reaction mechanism of the DOS [23] including 72 chemical species and 325 elementary reactions was incorporated into KIVA-4 to simulate multi-component diesel fuel ignition and combustion. However, the global reaction, Eq. (1), was removed from the chemical kinetics mechanism and instead multi-component evaporation and combustion were applied in simulations. Fuel components were considered to consist of n-heptane and toluene with mole basis proportions of two and one as aliphatic and aromatic hydrocarbon compounds of the diesel fuel, respectively. Initial mole fractions of liquid fuel components were introduced to the KIVA-4 code in evaporation and subsequently combustion processes. It means that in the developed model, the multi-component fuel evaporation and combustion models operate together in a way that the evaporation model performs phase change calculations and provides the combustion model with the amount of evaporated fuel for each specic fuel component. As it mentioned, n-heptane (C7H16) and toluene (C7H8) were used as the fuel components in this study. Whenever the liquid fuel is injected into the combustion chamber, these two components will represent the diesel fuel both in evaporation and combustion models. Phase change calculations of these species will be done by the multi-component evaporation model and then the multi-component fuel combustion model will perform the combustion chemistry calculations. Considering stochastic liquid fuel distribution and applied multi-component evaporation model, it is possible that in a typical computational cell n-heptane chemistry would be more dominant compared to the toluene due to differences in heat of vaporization of the applied fuel components. Nonetheless, multicomponent fuel combustion model will take into account chemistry calculations of the both fuel components in a computational cell using the dened chemical kinetics mechanism. This is because reaction pathways of both n-heptane and toluene are available in the chemical kinetics mechanism; however, oxidation and subsequently combustion strength of the each fuel components are due to their mixing with oxidant and forming of the mixture in ammability limit. Regarding to explanations above, Fig. 2 schematically illustrates comparison between developed multi-component fuel combustion model in this study and conventional single component fuel combustion model [17]. Using the DOS chemical kinetics mechanism, soot formation mechanism is dened within the reaction mechanism by a series of elementary reaction steps leading from acetylene to the rst aromatic ring, A1. These reaction steps are followed by successive stages of the HACA mechanism: H-abstraction and C2H2-addition resulting in formation of a chain of aromatic rings. Moreover, extended Zeldovich mechanism was used in NOx emission modeling [23]. 3.2.2. Mathematical framework and programming details of the multicomponent fuel combustion model Fluid ow and fuel particle simulation procedure in the KIVA code, the ALE method, decouples calculations of the diffusion and

762

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

Fig. 2. Schematic gure of applied multi-component fuel evaporation and combustion model in this study (right) compared with single component evaporation and combustion model [17] (left).

convection terms from chemical source terms. Hence, each computational cell can be treated as a homogeneously mixed reactor at each time-step. Continuity equation for species m and energy equation in terms of specic internal energy are formulated in the KIVA code as given in Eqs. (2) and (3), respectively [20].

_c _c where q m in Eq. (2) and Q in Eq. (3) are the parameters that need to be calculated by the combustion model. Mathematical descriptions of these terms are as follows:

_ _c q m W m xm _c Q
M X

4 5

   @ qm q _c _s q r qm u r qDr m m q dml @t q @ qI r quI P r u 1 A0 r : ru r J @t _c Q _s A0 q e Q

_ m Dh  x f m

m1

Combustion calculation in the default KIVA-4 code is based on a global single step reaction and cannot directly be used to include chemical kinetics reaction mechanism. To calculate the molar _ m ; the gas phase kinetics production rate of the chemical species, x

Fig. 3. Summarized ow chart of coupled KIVA4-CHEMKIN-DVODE model developed in this study.

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

763

Fig. 4. Flow chart of multi-component fuel combustion model in a particular time step developed in this study.

Table 3 Needed programming modications and parameter denition of the developed multi-component fuel combustion model in KIVA-4-CHEMKIN model compared to KIVA-3VCHEMKIN model [28,29]. Needed modication/parameter role Increase number of active chemical species to match the number of species in applied chemical kinetics mechanism Include enthalpy data of the added chemical species For the fuel specie(s) Number of other chemical species (major/ intermediate) Species placement order in chemical kinetics mechanism Data allocation schema entire the code Coupled KIVA-3V-CHEMKIN nsp was increased from default 12 to M Coupled KIVA4-CHEMKIN nsp was increased from default 12 to M

hk enthalpy data array was updated in main driver of KIVA: kiva.f First specie (nsp = 1) should be the applied fuel nsp-1 Fuel specie at the beginning of the species. Placement of the other species is optional, however it should match enthalpy array in kiva.f comkiva.i and comfuel.i external les were used to link global code parameters throughout the sub-routines of the KIVA-3Vcode

hk enthalpy data array was updated in an external le: datahk From rst specie to n (n = nspl) should be the applied fuel components nsp-nspl Fuel components (n = nspl) at the beginning of the species. Placement of the other species is optional, however it should match enthalpy array in datahk allocatedata.f and emodule.f sub-routines were used to link global code parameters throughout the sub-routines of the KIVA4 code

library of CHEMKIN II [26] was integrated into KIVA-4 code. A summarized ow chart of this integration is presented in Fig 3. In this procedure, the CHEMKIN libraries have identied introduced chemical kinetics mechanism before coupling with KIVA-4 calculations. Sample simulation starts with initialization tasks by the KIVA code such as: reading input thermodynamic, species concentration, initial and boundary conditions. KIVA continues the calculation processes of the uid ow and when it reaches to the

combustion calculations, if the temperature value for typical computational cell would be higher than 600 K as threshold value, combustion calculations will be activated for that cell. Otherwise, combustion will be bypassed by the code. Referring to Fig. 3, when the combustion is active for a computational cell an interface numerical unit will perform chemistry solutions by iterative calling of the coupled CHEMKIN-DVODE code [27] to solve the stiff system of ordinary differential equations.

764

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

Chemical kinetics mechanisms developed for IC simulations are mostly in gas phase; hence, code was developed in such a way that it would also be able to simulate premixed combustion systems such as HCCI engines. Fig. 4 is the detailed ow chart of the numerical calculations which would be done by the interface unit as a multi-component fuel combustion model. If there is an injection process in the system, rstly the spray injection, breakup and multi-component evaporation routines will be called. In a time step of a simulation, evaporation model will perform liquid fuel phase change calculations till the end of that particular time step and then the interface unit will deliver the concentrations of the gaseous fuel components to the multi-component fuel combustion model. Integrating by CHEMKIN chemistry solver, this process in KIVA-4 is not same as the earlier version of KIVA-3V. In order to this, most important modications and parameters which should be accurately introduced to the coupled code are given in Table 3. The parameter named nsp is the number of active chemical species in the KIVA code which is only 12 in the default KIVA and while integrating with the CHEMKIN code it should be increased to M number of species participated in chemical kinetics mechanism [28,29]. Moreover, there is a need to introduce enthalpy of the species into the versions of the KIVA code as it was stated in Table 3. The developed multi-component fuel combustion model in this study has a major difference with the coupled KIVA-3V-CHEMKIN in the case of number of fuel components. In KIVA4, loop counters of the chemical specie throughout the code consider nspl as the number of fuel components and nsp as the number of total species. Hence, in the Eqs. (4) and (5) domain of m can be dened as below:

The model is capable of using any number of fuel components as long as their reaction pathways would be available in the chemical kinetics mechanism. It should also be noted that programming schema of the KIVA-4 code in the case of data/parameter allocation is different compared to its previous version of KIVA-3V. KIVA-3V uses an external le to link global parameters throughout the code, whereas KIVA-4 has two added sub-routines for this purpose, see Table 3. Especially, if a new parameter is needed to be introduced in couple of subroutines, this new parameter should be included in the mentioned internal data allocation subroutines of allocatedata.f and emodule.f. Otherwise, code cannot identify an applied parameter in a typical subroutine which is also used anywhere else in the code. Useful programming information can be found in [30]. Once the combustion in a particular computational cell is activated, all processes discussed above will be done and then the interface unit will transfer the newly calculated molar production/destruction rates of the species including fuel components and other species to the KIVA code to update the combustion source terms in Eqs. (2) and (3). It should be noted that automatic time step sub-cycling of the DVODE code [27] was also activated to ensure accurate chemistry calculations when the stiffness of some of ordinary differential equations would be high. 3.3. Engine geometry and computational mesh Bowl geometry of the Toyota 2KD-FTV diesel engine is in omega shape to increase mixing of the liquid fuel and air by initiated circulation. In this study, the common rail fuel injector had 6 injector orices. To reduce the computational time, a 60 sector mesh was created based on the symmetric combustion chamber

m 1; 2; . . . ; n; n 1; n 2; . . . ; M

Fig. 5. Toyota 2KD-FTV 60 (a) medium and (b) ne sector meshes (at TDC) used in 3D-CFD simulations and calculated pressure results using mesh a and b for engine 2400 rpm and 100% load condition.

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

765

geometry of the Toyota 2KD-FTV car engine as shown in Fig. 5. The medium mesh consisted of 3540 cells at TDC with the piston crevice region included. Although grids size at wall boundaries seems to be coarse, advantage of using polar mesh over comes this issue as very ne cells (much less than 1 mm) are in liquid fuel spray path and in vicinity of injector nozzle. This ensures accurate breakup and evaporation predictions for liquid fuel droplets. Using this mesh, run times for a closed cycle simulation took about 35 h on a single processor at the High Performance Computing system (HPC) in National University of Singapore. To validate the mesh independence, a ner mesh with the same engine geometry was created as shown in Fig. 5. The cylinder pressure curves were compared at the condition of 2400 rpm and 100% load (see Fig. 5). It can be seen that with further renements on the mesh, no

signicant difference was observed on the predicted cylinder pressure. By considering the computational time, the medium mesh was used for all the simulations.

4. Results and discussion For considered experimental cases relevant numerical results are extracted to get better insight about engine operation details at engine full and mid loads and in three engine speeds. In doing so, results are illustrating engine performance characteristics such as work, accumulative heat release and indicated thermal efciency (ITE) and engine emissions characteristics for most important emissions such as NOx, Soot, CO, CO2 and UHC.

Fig. 6. Comparison of cylinder pressure and HRR results of experimental and numerical multi and single component fuel combustion models.

Fig. 7. Numerical results of performance characteristics of 100% load: (a) accumulative heat release, (b) total work and (c) numerical and experimental indicated thermal efciency.

766

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

4.1. Model validation After the model development, simulations are performed based on the input parameters of the experimental cases for engine speeds of 1200, 2400 and 3600 rpm and engine loads of 100% and 50%. In-cylinder pressure and HRR diagrams for considered cases are compared in Fig. 6. It can be seen that the simulation results of the multi-component fuel combustion model are in good agreement with the experimental measurements. In order to compare results of the developed model, simulations were also carried out using conventional single component combustion model. Using the same chemical kinetics mechanism, the global oxidative pyrolysis reaction, Eq. (1), is considered in single component fuel, C14H28, combustion model. In-cylinder pressure and HRR results show the single component combustion model predicts longer ignition delay time compared to the multi-component fuel combustion model. Comparing single and multi-component fuel combustion models, latter model shows better agreement with experimental pressure and HRR data.

By reference to Fig. 6 higher peak pressure and HRR are resulted by increasing engine load at each specic engine speed. In contrast to the higher engine speeds, at the lowest engine speed of 1200 rpm the location of the maximum HRR is delayed in the considered engine loads. This is mainly due to the timing for SOI and the amount of injected fuel in each case where injecting more fuel at higher engine speeds and loads has been resulted in formation of more combustible air fuel mixtures. This trend can be seen in the pressure diagrams that the fuel tends to burn late under the lowest engine speed of 1200 rpm. Pressure curves for this engine speed have two peaks where the pressure is dipped due to starting of expansion stroke exactly after TDC and created the rst peak. Although combustion was initiated in cranks before TDC in HRR diagrams, amount of released heat was not sufcient enough to overcome cylinder pressure reduction due to expansion. However, as combustion has got stronger and higher heat was produced, pressure curve was started to go higher and created the second peak. It should be noted that second pressure peak tends to reduce by decreasing the engine load which is mainly due to reduced fuel

Fig. 8. Mean value CAD history of (a) fuel concentration history, (b) temperature, (c) soot and (d) NOx and their temporal contours at 7, 8, 14 and 50 CAD ATDC at engine full load and 3600 rpm.

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

767

amount and weak air fuel mixtures and subsequently weaker combustion. 4.2. Performance and emission characteristics of the engine under full load Fig. 7a shows engine accumulative heat release for three engine speeds at full load condition. At higher speeds it can be seen more heat is released due to higher fuel input in that particular case. Moreover, fuel tends to burn rapidly in higher speeds. Closed cycle simulation results for the work done by the engine show that lower engine speeds are associated with reduced output work, see Fig. 7b. Nonetheless, this trend is not observed for ITE in Fig. 7c. Both simulation and experiment results indicate that a higher ITE can be observed for the engine speed of 2400 rpm compared to the other engine speeds. Using the detailed chemical kinetics mechanism incorporated with the uid dynamics code, in-cylinder thermodynamic properties and concentration change of any chemical species and their temporal and spatial variation can be tracked. Fig. 8 shows mentioned advantage of the multi-dimensional models for in-cylinder fuel concentration history, temperature and soot and NOx emissions for the engine condition of full load and 3600 rpm. Moreover, their contours are also presented in four crank angle degrees of 7, 8, 14 and 50 ATDC. Equivalence ratio contour shows that a considerable amount of fuel (lumped value including n-heptane and toluene) is trapped in piston bowl after initiation of the

injection process; where mean value of fuel concentration history shows that massive amount of fuel is burned around 10 CAD ATDC. This is why high local temperature regions can be observed in the temperature contour at 14 CAD ATDC. It also can be seen that combustion is initiated when the mean cylinder temperature reaches to 1000 K and high temperature local regions can be distinguished at the 7 CAD ATDC contour. Mean value results of soot and NOx in Fig. 8 show that soot tends to oxidize while it is encountered by high temperature, whereas NOx emission tends to increase and stabilized after 30 CAD ATDC. Soot contours illustrate fuel rich regions with moderate local temperature have the potential of producing more soot emission. It can be seen high local concentration of soot is available in piston bowl which coincides with equivalence ratio contours. Nonetheless, high local concentrations of NOx are observed where the local temperature inside the combustion chamber has higher values. Figs. 9 and 10 represent contour levels of temperature, NOx, soot, CO, CO2 and O2 for full load and three engine speeds at 20 CAD ATDC. For this specic crank angle in Fig. 10 higher NOx levels are produced by increasing the engine speed, whereas opposite trend is predicted for the soot emission. Referring to Figs. 9 and 10, it can be seen that higher local temperature expedites NOx formation and CO and Soot oxidation. This is also coincided with O2 consumption, where in high temperature local regions CO and soot are oxidized with O2 and CO2 is produced. This trend is pronounced at higher engine speed because considerable high temperature local regions are formed and are shown by the contours.

Fig. 9. Temperature (K), NOx and soot (gr) contours for engine full load and speeds of 1200, 2400 and 3600 rpm at 20 CAD ATDC.

768

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

Fig. 10. CO, CO2 and O2 (gr) contours for engine full load and speeds of 1200, 2400 and 3600 rpm at 20 CAD ATDC.

Fig. 11 illustrates mean concentration-CAD history of NOx, soot, CO, CO2 and UHC emissions compared to measured experimental data at 100% engine load and three considered engine speeds. As it can be seen acceptable agreements are achieved. Tangible production/destruction of the mentioned chemical species is observed from 10 CAD to 30 CAD ATDC; however, after 60 CAD ATDC concentrations are stabilized. By referring to Fig. 11, nal UHC emission levels for three engine speeds are almost zero both in simulation and experiments. Moreover, under full load condition NOx tends to have slightly higher levels by increasing engine speed, whereas soot shows opposite trend. It also shows that there is accumulative trend in CO2 production, whereas CO starts to oxidize around 20 CAD ATDC at 3600 and 2400 rpm. This trend is not observed at the engine speed of 1200 rpm where for this engine speed CO oxidation is initiated after 35 CAD ATDC. Mentioned trend at the lowest engine speed coincides with cylinder pressure and HRR results and mainly is because of lower in-cylinder temperature and heat release.

4.3. Performance and emission characteristics of the engine under mid load Fig. 12 shows simulation results of accumulative heat release and work done at mid load with ITE at three engine speeds. Comparing to engine full load results in Fig. 7a, it can be seen that by decreasing engine load, heat release and output work are reduced. For instance: at engine speed of 3600 rpm accumulative heat

release reaches 1020 J under 50% load comparing to 1680 J of full load operating condition. Same as the engine full load condition, speed of 1200 rpm has the longer ignition delay which is captured by accumulative heat release diagrams. The reason for this is the lean air/fuel mixtures applied for 1200 rpm at the mid engine load. Furthermore, ITE diagrams for 50% load show that the engine tends to have higher efciency at 2400 rpm. It should be noted that at higher engine speed of 3600 rpm, ITE tends to drop by 5% from full load condition to the mid engine load, compare Fig. 7c with Fig. 12c. Bar chart in the Fig. 13 illustrates the normalized emission data for 50% load condition. In order to make comparison between simulated engine emissions data at 50% load to the engine full load condition, bar chart in Fig. 13 normalized by the emission values at their corresponding engine speeds at 100% load condition. Considering this normalization, under mid load condition, most of emissions are less than 100% load except NOx at 2400 and 1200 rpm, and soot only at 1200 rpm. In addition, at higher engine speeds of 2400 and 3600 rpm considerable reduction on CO2, CO and UHS emissions are observed where the amount of mentioned emissions is halved at mid engine load. 5. Conclusions (1) Multi-dimensional CFD calculations were performed by developing a comprehensive multi-component fuel combustion model through integrating CHEMKIN II chemistry solver into the KIVA-4 code in order to study combustion and

A. Maghbouli et al. / Applied Energy 111 (2013) 758770

769

50% Load

Normalized concentration [-]

Engine speed [rpm]

Fig. 13. Normalized concentration of considered emissions for three engine speed at 50% load condition.

(2)

(3)

(4)

(5)

Fig. 11. Concentration-CAD history of considered emissions and comparison with the measured experimental emission data under full load condition.

emission characteristics of a D.I diesel engine. Application of multi-component combustion model had led to accurate simulation methodology and better validity of the results compared to the experimental data. The developed model was capable of modeling variety of operating conditions where diverse range of equivalence ratio was applied in the simulated cases. A good agreement was achieved for predicted cylinder pressure, HRR and emission results comparing to the conducted experiments. It has been observed that under low speed and mid load operating conditions, engine incurs weak combustion due to the reduction in the amount of injected fuel and reduced in-cylinder temperature. This was shown in pressure and HRR diagrams where fuel was burnt late after the injection process. Closed cycle simulation results show that if engine load and speed were reduced; output power was decreased as well. However, same trend was not reported for the ITE as the middle engine speed in considered engine loads had slightly higher efciency. Comparing to the full load operating condition, CO, CO2 and UHC emissions had lower concentrations at mid engine load. However, further reduction in the engine speed at mid engine load has resulted in production of higher NOx and soot emissions.

Fig. 12. Numerical simulation results of performance characteristics for 50% load: (a) accumulative heat release, (b) total work and (c) numerical and experimental indicated thermal efciency.

770

A. Maghbouli et al. / Applied Energy 111 (2013) 758770 [17] Ehleskog R, Golovitchev VI, Denbratt I, Andersson S, Rinaldini CA. Experimental and numerical investigation of split injections at low load in an HDDI diesel engine equipped with a piezo injector. SAE paper 2006-013433; 2006. [18] Helmantel A, Golovitchev VI. Injection strategy optimization for a light duty DI diesel engine in medium load conditions with high EGR rates. SAE paper 200901-1441; 2009. [19] Shuai S, Abani N, Yoshikawa T, Reitz RD, Park SW. Evaluation of the effects of injection timing and rate-shape on diesel low temperature combustion using advanced CFD modeling. Fuel 2009;2009(88):123544. [20] Amsden AA, Ramshaw JD, ORourke PJ, Dukowicz JK. KIVA: a computer program for two- and three-dimensional uid ows with chemical reactions and fuel spray. Technical report LA-10245-MS. NM (USA): Los Alamos National Lab.; 1985. [21] Beale JC, Reitz RD. Atomization Spray 1999;9:62350. [22] ORourke PJ, Bracco FV. Modeling drop interactions in thick sprays and a comparison with experiments. In: Stratied charge automotive engines. I. Mech E. Conference Publications 19809; 1980. p. 10116. [23] Bergman M, Golovitchev VI. Chemical mechanisms for modeling HCCI with gasoline and diesel oil surrogates. In: Proceedings of the VII. congress on engine combustion processes; 2005. p. 32334 [Munchen, 1516 March]. [24] Chen W, Shuai S, Wang J. A soot formation embedded reduced reaction mechanism for diesel surrogate fuel. Fuel 2009;2009(88):192736. [25] Ismail HM, Ng HK, Gan S, Lucchini T, Onorati A. Development of a reduced biodiesel combustion kinetics mechanism for CFD modelling of a light-duty diesel engine. Fuel 2012. <http://dx.doi.org/10.1016/j.fuel.2012.10.015>. [26] Kee RJ, Rupley FM, Miller JA. CHEMKIN-II: a FORTRAN chemical kinetics package for the analyses of gas phase chemical kinetics. Sandia Report SAND 89-8009; 1989. [27] Brown PN, Byrne GD, Hindmarsh AC. VODE: a variable coefcient ODE solver. J Sci Stat Comput 1989;10:103851. [28] Maghbouli A, Saray RK, Shafee S, Ghafouri J. Numerical study of combustion and emission characteristics of dual-fuel engines using 3D-CFD models coupled with chemical kinetics. Fuel 2012. <http://dx.doi.org/10.1016/ j.fuel.2012.10.055>. [29] Maghbouli A, Shafee S, Saray RK, Yang WM, Hosseini V, An H. A multidimensional CFD-chemical kinetics approach in detection and reduction of knocking combustion in diesel-natural gas dual-fuel engines using local heat release analysis. SAE Int. J. Engines 2013;6(2). <http://doi:10.4271/2013-010865>. [30] Dr. Randy P. Hessel, senior Scientist. ERC home page: <http:// homepages.cae.wisc.edu/~hessel/>.

References
[1] Heywood JB. Internal combustion engine fundamentals. New York: McGraw Hill Book Company; 1988. [2] Khair M, McKinnon DL. Performance evaluation of advanced emission control technologies for diesel heavy-duty engines. SAE paper 1999-01-3564; 1999. [3] Yang F, Minggao O. Experimental research on EGR in a diesel engine equipped with common-rail injection system. SAE paper 2003-01-0351; 2003. [4] Kouremenos DA, Hountalas DT. The effect of EGR on the performance and pollutant emissions of heavy duty diesel engines using constant and variable AFR. SAE paper 2001-01-0198; 2001. [5] Roy MM. Effect of exhaust gas recirculation on combustion and odorous emissions in direct injection diesel engines. SAE paper 2008-01-2482; 2008. [6] Cho K, Han M, Wagner RM, Sluder CS. Mixed-source EGR for enabling highefciency clean combustion modes in a light-duty diesel engine. SAE paper 2008-01-0645; 2008. [7] Wang Y, Li G, Zhu W, Zhou L. Study on the application of DME/diesel blends in a diesel engine. Fuel Process Technol 2008;2008(89):127280. [8] Papagiannakis RG, Hountalas DT. Experimental investigation concerning the effect of natural gas percentage on performance and emissions of a DI dual fuel diesel engine. Appl Therm Eng 2003;2003(23):35365. [9] Soylu S. Examination of combustion characteristics and phasing strategies of a natural gas HCCI engine. Energy Convers Manage 2005. http://dx.doi.org/ 10.1016/j.enconman.2004.02.013. [10] Scott B, Dennis F. Development and validation of a quasi-dimensional model for HCCI engine performance and emissions studies under turbocharged conditions. SAE paper 2002-01-1757; 2002. [11] Komninos NP, Hountalas DH, Kouremenos DA. Development of a new multizone model for the description of physical processes in HCCI engines. SAE paper 2004-01-0562; 2004. [12] Kouremenos DA, Rakopoulos CD, Hountalas DT. Multi-zone combustion modeling for the prediction of pollutants emissions and performance of DI diesel engines. SAE paper 970635; 1997. [13] Rakopoulos CD, Hountalas DT. Development and validation of a 3-D multizone combustion model for the prediction of DI diesel engines performance and pollutants emissions. SAE paper 981021; 1998. [14] Hountalas DT, Kouremenos DA, Mavropoulos GC, Binder KB, Schwarz V. Multizone combustion modeling as a tool for DI diesel engine development and Application for the effect of injection pressure. SAE paper 2004-01-0115; 2004. [15] Ramos JI. Internal combustion engine modeling, Taylor & Francis, 1989. [16] Besson M, Hilaire N, Lahjaily H, Gastaldi P, Diesel combustion study at full load using CFD and Design of Experiments. SAE paper 2003-01-1858; 2003.