2008-01-0358

Coupling of a 1-D Injection Model with a 3-D Combustion Code for Direct Injection Diesel Engine Simulations
Julien Bohbot, Christos Chryssakis, Pierre Pacaud, Adlne Benkenida
IFP Powertrain Engineering

Copyright 2008 SAE International

ABSTRACT
Modern diesel engines operate under injection pressures varying from 30 to 200 MPa and employ combinations of very early and conventional injection timings to achieve partially homogeneous mixtures. The variety of injection and cylinder pressures, as well as injector dynamics, result in different injection rates, depending on the conditions. These variations can be captured by 1-D injection models that take into account the dynamics of the injector, the cylinder and injection pressures, and the internal geometry of the nozzle. The information obtained by these models can be used to provide initial and boundary conditions for the spray modeling in a 3-D combustion code. In this paper, a methodology for coupling a 1-D injection model with a 3D combustion code for direct-injected diesel engines is presented. A single-cylinder diesel engine has been used to demonstrate the capabilities of the model under varying injection conditions. Moreover, this coupling strategy opens a new methodology for 3-D calculations that do not need to fit initial conditions but use directly a 0-D model for intake/exhaust conditions and injection conditions. Using coupling strategy makes easier to run 3-D engine simulations, reduce engineering time and allows to investigate a large range of interesting phenomena.

The development of the Common Rail (CR) fuel injection system allows more flexibility and accuracy in controlling the injection parameters and the combustion process in order to meet performance and emissions goals. The CR system applies a constant fuel pressure in the rail that is common for all cylinders. However, many researchers report a drop in the injection pressure and injection-to-injection variations when multiple injection strategies are employed [5-10]. These variations are attributed to injector dynamics, due to the needle motion and pressure waves occurring in the system. That leads to varying injection rates, which can affect the combustion and emissions formation processes, particularly NOX and soot. Typical Computational Fluid Dynamics (CFD) models for the in-cylinder processes in an engine are not able to capture these variations, since the exact injection rates are not known. In order to improve the predictability of CFD models, a coupling between the injection system and the incylinder processes has to be performed. Previous work on 1-D/3-D coupling focuses on the gas dynamics of the intake system. Sinclair et al. [11] have developed a direct 1-D/3-D coupling model to compute the gas exchange system of a turbo-charged DI-diesel engine. They have treated the intake manifold and ports as a single 3-D model while the rest of the engine was modelled by a system of 1-D components. They developed a 1-D/3-D coupling interface and have obtained encouraging results. Their work shows that 3-D modelling of a single component can be used in a full 1D system using an appropriate 1-D/3-D interface treatment. Similar coupling approaches have been followed by Borgh et al. [12], Riegler and Bargende [13] and Wehr et al. [14], who have developed 3-D CFD models for the intake manifold, coupled with 1-D engine models. In this work a novel coupling approach between a 1-D injection model and a 3-D combustion code is presented. The model is based on previous work by

INTRODUCTION
Stringent emissions regulations for diesel engines require employing innovative technologies in order to reduce in-cylinder nitrogen oxide (NOX) and particulate matter (PM) emissions. Advanced fuel injection strategies, including multiple pilot and post injections can be used to reach the limits of emissions reduction, while maintaining the advantage of high thermal efficiency of a diesel engine [1-4].

Bohbot et al.[15], who developed a coupling approach between a 1-D engine simulation software and a 3-D combustion code, for Port Fuel Injected (PFI) gasoline engines. This model is extended here by adding a direct injection model. The injection model has been developed for a typical Common Rail system used in diesel engines. Subsequently, the model is coupled with IFP's 3-D combustion code to demonstrate its behaviour in a single cylinder diesel engine, when multiple injections are used, including pilot- and post-injection events. In the following, the 1-D simulation code is briefly described, as well as IFP's 3-D combustion code. Then, the injector model development and validation are presented as well as its coupling with the 3-D combustion code. Finally, a single-cylinder diesel engine is used to demonstrate the capabilities of the model.

modeled to simulate the injection. For turbulent combustion the 3-Zone Extended Coherent Flame Model (ECFM3Z) , developed at IFP, is used [21-22]. For Diesel combustion, a full database of auto-ignition delays and reaction rates allows us to compute cold and hot flame ignition [23]. The RNG k- turbulence model [24] with either the Diwakar or Kays and Crawford wall laws are employed, depending on the fuel used (gasoline or diesel). Species (and tracers), energy and RNG k- turbulent diffusion terms are all implicitly solved by a generic diffusion routine. Parallel optimized linear algebraic libraries developed is used to reverse matrices, rendering parallelization very efficient. The SIMPLE method is applied for solving the coupled pressurevelocity system. All gradient terms used for pressure and the Reynolds stress tensor are parallely computed. Moreover, preconditioning efficiently the pressure matrix [19] drastically reduces the simulation time. The convection terms are explicitly sub-cycled. A second order upwind scheme for scalars and momentum convection is used. With the development of parallel computers the exploitation of parallelism in order to reduce the computational time has become a major goal. The data parallelism model selected, based primarily on the ratio between performance and development cost, is the OPEN-MP paradigm [25], which is easy to implement, standardized, portable, and scalable. After a profiling of the sequential version, it has been noticed that most of the CPU time is spent in the pressure solver and diffusion terms computation. Implementing OPEN-MP directives into them accounts for a parallelization of approximately 50% of the IFP-C3D code. Finally, a very good and scalable speed-up of around 3, when 4 processors are used, is reachable for the benchmark cases. Details and extensive validation on the code can be found in [18-20]. An automatic mesh generator is available in IFP-C3D to create 2-D and 3-D unstructured wedge meshes with periodic boundary conditions. The code is fully integrated in the LMS Imagine.Lab AMESim platform with a friendly graphical user interface (GUI) and automatic post-processing.

THE IFP ENGINE SIMULATION CODE

The IFP engine simulation tools are developed under the LMS Imagine.Lab AMESim platform and consists of a 1D library, IFP-ENGINE allowing the simulation of a complete virtual engine, and of a 3-dimensional combustion code IFP-C3D that can be used for incylinder spray and combustion simulations. These two parts are briefly described in this section. THE IFP-ENGINE LIBRARY The IFP-ENGINE library allows the simulation of a complete virtual engine using a characteristic time-scale of the order of the crankshaft angle. A variety of elements is available to build representative models for engine components, such as a twin scroll turbocharger, gasoline or diesel injectors, etc. Figure 1 shows these IFP-ENGINE components. Moreover, the library uses an advanced modeling approach to take accurately into account the relevant physical phenomena taking place in the engine[16]. The combustion process can be modeled with efficient phenomenological models like CFM1-D [26], but also using a mapped combustion, such as Wiebe's law for gasoline engine, and with the Barba [27] based model for Diesel engine. THE IFP-C3D COMBUSTION CODE In order to take full advantage of modern parallel superscalar machines (SMP machines), IFP has developed IFP-C3D, a solver designed for hexaedrical unstructured and parallel formalism. The IFP-C3D code [17-20] solves the unsteady equations of motion of a turbulent, chemically reactive mixture of gases, coupled to the equations for a multicomponent vaporizing fuel spray. The Navier-Stokes equations are solved using a finite volume method extended with an ALE (Arbitrary Lagrangian-Eulerian) scheme. IFP-C3D uses a time-splitting integration and the temporal integration scheme is largely implicit. Fuel evaporation, breakup, and spray/wall interactions are

1-D INJECTION MODEL

MODEL DESCRIPTION A 1-D model has been developed to simulate the injector dynamics and provide predictions for the fuel flow rate into the cylinder. The model development has been based on the operation and geometrical characteristics of a Bosch CRI2.1 used on a Diesel engine. The nozzle modeled is of the Valve Covered Orifice (VCO) type. The fuel pump and the regulator of the rail pressure are represented by very simple models assuming a constant flow rate and an ideal regulator. The electromagnetic

part has been simplified, as a first approach, by providing tables with the intensity and the voltage of the signal to the solenoid valve. The fuel return outlet has been modeled by assuming that the fuel returns to the fuel tank, where the pressure is maintained constant. The control chamber guides hydraulically the plunger and needle lift by regulating the fuel pressure inside the injector. When the solenoid valve opens, fuel leaves the control chamber from the A-hole and the pressure drops. The pressure difference on the two sides of the plungerneedle assembly moves the needle upwards, thus opening the injection orifice of the nozzle. The inlet and the outlet restrictors of the control chamber (Z-hole and A-hole, respectively) have been modeled by utilizing orifice models available in the LMS Imagine.Lab AMESim software. The part of the plunger that closes the A-hole when the needle lifts is represented with a "flapper valve" model and a piston with a crown section is utilized in order to model accurately the upper surface of the plunge and calculate the volume displacement when the plunger moves. The mechanical behavior of the plunger and the needle assembly have been modeled by assuming that the plunger, the pressure shoulder of the nozzle, and the needle, react as one piece. In order to model the motion of this assembly, the compressibility of its elements has to be considered. A simple and computationally efficient solution is to distribute the masses in a way that conserves the first oscillation mode of the system. This has been achieved by dividing the total mass in three parts, as shown in Figure 1.

injector line, and the high-pressure circuit injector.

inside the

The model initialization is based on the assumption that the fuel pressure in the high-pressure circuit is equal to the rail pressure. The initial position of the needle and the plunger depends on the fuel pressure, due to compressibility effects.

Figure 1: Mechanical model of the plunger-pressure shoulder-needle system The two conic surfaces of the needle have been modeled using two piston elements and two disks that are used to calculate the contact points of the needle inside the nozzle. The high-pressure fuel lines have been modeled using hydraulic elements in LMS-AMESim. They include the rail, the line linking the fuel pump with the rail, the rail-

Figure 2: 1-D Injector model

MODEL CALIBRATION It is necessary to calibrate the injector model to fit the different sub-model thanks to a large number of experimental points. A large part of the parameter of the sub-model are defined with geometrical data measured on the injector. For the other parameters like the diameter of the hole A, hole Z and the also for the cavitation number of the injector hole, these parameters are fitted with single and multi-injection events. A first calibration of the model has been performed by using experimental measurements for relatively long injection events (2200 s) at a rail pressure of 160MPa. The comparison between model predictions and experimental measurements for the needle lift and the injection flow rate is presented in Figures 4 and 5 respectively.

A second calibration of the model has been performed for multi-injection measurement with short injection events (260s, 330s, 430s) at 90MPa. For this multi injection events, the dynamic of the system is very important to predict the injection delay and the fuel mass for each injection. Rail pressure calibration have been optimized to predict multi-injection events and pressure oscillation in the rail and in the pipe of the injection system (see Figure 7). For multi injection events, as pilot and main spray injections, the electric command has been optimized to represent accurately the electric delay (see Figure 5). The dynamic of the needle motion is also accurately predicted (see Figure 6). The injector 1-D model has been tested on various multi-injection events and the Figure 8 plots the injection rate obtained with this model. Time injection delays and durations are well predicted. However, the pilot injection mass (short injection) given by experimental data have a strong variability, Figure 9 plots the experimental injection rate given for a pilot injection (260 s) for different multi injection events. The mass measured for this pilot have an average value equal to 3.97mg with a lower value of 3.13mg and an upper value of 4.50 mg. The parameters of the injector model have been fitted on this average value for the pilot mass. For the main injection, the injector model has been calibrated using operating condition listed in Table 1. However, the injector model do not predict a dynamic phenomenon that occurs in the main injection. This wave in the injection rate is not predicted by the injector model and more precisely in the dynamic line. The signal pressure in the pipe shows that we do not predict the high frequency of the pressure variation just before the main injection events. The line of the injection system have to be improved but our attempts to increase the accuracy of this model have failed.

Figure 3: Comparison of needle lift predictions (in meter) and experimental measurements

Figure 5: Signal Intensity of the electric port for a pilot main injection (T1=260s , T2=430 s , delay period=1777,8s / 16deg.). Figure 4: Comparison of injection rate predictions and experimental measurements

Figure 6: Needle displacement of the injector for a pilot/main injection (T1=260s , T2=430 s , delay period=1777,8s / 16deg.)

Figure 9: Discrepancies of the injection rate experimental data for the pilot injection (260s).

Point number 4 5 6 7 8 9 10 11 12 13 21

Injection 1 duration (s) 260 260 260 380 260 260 260 260 330 430 380

Dwell (deg.) 20 20 20 12 16 24 28

Injection 2 duration (s) 330 380 430 430 430 430 430

20

380

Figure 7: Signal of pressure in the pipe for a pilot main injection (T1=260s , T2=430 s , delay period=1777,8s / 16deg.).

Table 1: Experimental database used for injector calibration (1500 rpm, 90MPa)

MODEL VALIDATION The injector model has been validated by comparing the injected fuel quantity predictions with experimental measurements under various rail pressures and injection durations. The injection chamber pressure for all cases was maintained constant at 50 bar.

Figure 8: Injection rate for a pilot/main injection (T1=260s , T2=430 s , delay period=1777,8s / 16deg.)

160 140 120 Quantity [mm ]

3

1200 bar - Exp 1200 bar - Model 1600 bar - Exp. 1600 bar - Model

100 80 60 40 20 0 2200 1800 1400 1000 800 600 500

DESCRIPTION OF THE 1-D/3-D COUPLED APPROACH

The development of a direct coupling approach between two different tools requires the definition of the type of data that have to be transferred and the time scale of the synchronisation enforced by the physical system. The main objective of this development is to model with 3-D modelling only the injection and combustion processes (not the intake and exhaust phases). An important assumption of this coupling approach is the initially homogeneous distribution of species in the combustion chamber, after the closing of the intake valves. The time scale of the synchronisation is the combustion cycle duration (when both intake and exhaust valves are closed). This process was described in detail in [15] for a PFI gasoline engine. The new element here is the addition of the 1-D injection model, which provides information for the direct injection and can be used for a Direct Injection (DI) diesel engine simulation. This is achieved by sending the cylinder pressure information to the 1-D injection model, where the fuel flow rate is calculated as a function of the injection and cylinder pressure, and the signal sent to the solenoid valve. Subsequently, the calculated fuel flow rate by the 1-D injector model is returned to the 3-D combustion code to define the parameter of the 3-D fuel injection. The IFP-ENGINE library uses an assumption of three perfect gases (air, fuel, and burnt gases) whereas IFPC3D utilises a mixture containing 12 gases (fuel, O2, N2, CO2, H2O, CO, H2, NO, OH, O, H, N). To ensure the gases compatibility between the two codes, IFP-C3D computes the mass of the 3 gases needed by IFPENGINE using the fuel mass, the O2 mass, and the mass of the species produced by the combustion as described below :
3D m ENGINE = mC fuel fuel

400

300

200

Duration [s]

Figure 10: Comparison of model predictions with experimental measurements for Pinj=1200, 1600 bar for single injection events. In Figure 10, the comparison of model predictions with experimental measurements for injection pressures of 1200 and 1600 bar shows very good agreement, for a wide range of injection durations (from 200 to 2200 s). For injection durations shorter than 200 s the injected quantities are very small and cannot be measured without a significant error. A second validation has been carried out for multi injection events listed in Table 1. Mass injected and injection rate calculated by the injector model have been compared with experimental data in Figure 11. Results show a good agreement for events with a pilot injection duration equal to 260s and a main injection equal to 430s or 380s. For the injection duration equal to 330s (point 12), the injected volume quantities is not very well predicted and have a relative error equal to 20%. Because of a too small number of experiment point with 330s injection duration, we do not succeed to fit the injector model for this injection duration that can explained the lack of precision.
Injected volume
25,00 Quantity (mm3 20,00 15,00 10,00 5,00 0,00 4 5 6 7 8 9 Point 10 11 12 13 21 Expe Model

m ENGINE fresh air =

1 C 3D mo 2 0,233
i =1,11

ENGINE mburnt gases =

C 3D specie ( i )

m ENGINE fresh air

To complete the direct coupling method, we have defined a temporal staggered algorithm to manage the time lag between the two codes. As shown in Figure 12, the staggered algorithm can be split in three steps. First, the engine gas exchange system is calculated by IFP-ENGINE. When IFP-ENGINE reaches the intake valve closure (IVC) angle, the calculation is stopped and IFP-C3D starts. Then the 1-D library sends the IVC

Figure 11: Comparison of model predictions with experimental measurements for Pinj=900 bar for multi injection events

angle, the trapped mass and the mixture composition to the 3-D code.

IFP-ENGINE
Trapped mass Gases mass fraction

Intake valve closure

cycle n
3

Exhaust valve opening

cycle n+1

Heat release, Mass consumption

MPI link
LMS Imagine Lab AMESIM

IFP-C3D Time
IFP-C3D

2
IFP-ENGINE

3
IFP-ENGINE

PC Cluster running IFP-C3D in parallel mode

Figure 13: MPI link between laptop computer running LMS-AMESim and a Cluster running IFP-C3D. Figure 12: Staggered coupling algorithm The second step is the 3-D combustion calculation using the 0-D/1-D information as boundary and initial conditions. The 3-D calculation is stopped at the exhaust valve opening (EVO) crank angle. Then, IFP-C3D sends to the 1-D library the heat release histories (combustion, wall transfer) and the species mass consumption histories. Finally, IFP-ENGINE restarts the calculation from the IVC to the next engine cycle. IFP-ENGINE uses during the combustion cycle the 3-D data to compute at each time step the heat release and the mass consumption. To manage the coupled calculation, the MPI 2 (Message Passing Interface) paradigm is used to create the link between the two codes. The MPI library is used with a master/slave mode. The master LMS-AMESim/IFPENGINE creates an intra-communicator with the IFPC3D (slave) and can command and communicate with IFP-C3D at any time step. The master/slave mode allows the possibility to manage many slaves to extend the calculation for more than one cylinder but also to use 3-D modelling to other single components (such as air box, 3-D pipe, etc.). Moreover, the CPU time need by 3D calculation and 1-D simulation are largely different and generally 3-D calculations have to be performed using a parallel architecture computer with large memory. Homogeneous grid of computer can be used with MPI paradigm and LMS-AMESim calculation can be performed on a laptop and 3-D calculations on a PC Cluster (see Figure 13). This functionnality allows to reduce the calculation return time and makes the coupled calculation time return compatible with industrial requirement. VALIDATION OF THE COUPLING ALGORITHM A first validation of the coupled algorithm is done in a basic engine system. This basic system contains a single cylinder model of a direct fuel injection diesel engine. The engine model diagram is presented in Figure 14. The 1-D/3-D coupling algorithm described previously is used for the modelling of the fuel injection and combustion processes. A sector mesh is generated with the internal mesher of IFP-C3D and contains 5000 cells. This coarse grid was selected as a first step in order to reduce the computational time and demonstrate the new methodology.

Figure 14: Diagram of single cylinder model The initial mass fraction of fuel is set to null in the combustion chamber. The in-cylinder pressure is drawn in Figure 15. As shown in this figure, the engine system

is stabilising after few engine cycles when using the 1D/3-D combustion modelling that assesses the stability of the staggered coupling algorithm. As explained in the previous section, the IFP-ENGINE library uses an assumption of 3 homogeneous gases while IFP-C3D uses at least 12 gases. The gas compatibility is ensured with the assumption that IFPC3D burnt gases mass is equal to the IFP-ENGINE burnt gases mass. However, the IFP-ENGINE burnt gas mixture is obtained from a stoichiometric combustion and is quiet different from the burnt gases mixture obtained with the 3-D combustion modelling. This difference leads to a small lack of accuracy. Nevertheless, the differences between pressure obtained with IFP-C3D and the pressure obtained with IFP-ENGINE using the 3-D combustion modelling data are negligible.

combustion model is used in IFP-C3D. The in-cylinder pressure predicted by the calculation without injection is plotted in Figure 19 and is in good agreement with the experimental data. This numerical result assess that thermal effect and trapped mass in the cylinder are correctly computed with IFP-ENGINE coupled with IFPC3D.

Figure 16: Computational grid for the 72 wedge

Figure 15: Cylinder pressure evolution and stabilisation

APPLICATION ON A SINGLE-CYLINDER DIESEL ENGINE

The coupled model has been applied on a Diesel singlecylinder direct injection diesel engine with a Bosch Common Rail injection system in order to demonstrate its capabilities. The main characteristics of the singlecylinder diesel engine are given in Table 2. A 72 wedge mesh is used to describe the combustion chamber in IFP-C3D which is linked with the 1-D LMS-AMESim software that includes the fuel injector and the engine simulator (cf. Figure 17). The total CPU time for one computation on two AMD 2 GHz processors is approximately 12 hours to simulate 8 engine cycles. Mainly, the CPU time is spent by the 3-D calculations. First, the computational grid has been generated using the internal mesher of the code. This sector mesh contains 10.000 cells and is plotted in Figure 16. The mesh size is adapted during the cycle to optimize the node clustering during the motion of the geometry. For all operating points simulated, a first coupled calculation has been performed without combustion and injection to fit the wall temperature and the air loop. Secondly, the injection system is activated and the ECFM3Z[2122]

Figure 17: 1-D injector and engine simulator

Engine type Number of cylinders Bore [mm] Stroke [mm] Compression ratio [ - ] Connecting rod length [mm] Engine displacement [cm3] IVO/IVC EVO/EVC

Diesel Single cylinder Bowl IFP 1 87 92 14.7 149.9 546.9 42 BTDC/124 BTDC 146 ATDC/43 ATDC

system has converged. The in-cylinder predicted by the calculation without injection is plotted in Figure 19 and is in good agreement with experimental data. This result assess that thermal effect and mass in the cylinder are correctly computed with IFP-ENGINE coupled with IFPC3D. The next step is now to inject fuel in the system and let IFP-C3D combustion code simulate the 3-D combustion process.

Table 2: Engine set-up characteristics COUPLED 1-D INJECTION / 3-D COMBUSTION CALCULATION The operating point selected for comparison of the model results with experimental measurements of the combustion in the cylinder is given in Table 3. Engine Speed [RPM] IMEP [bar] Injection Pressure [bar] Inj. Timing [ CA ATDC] Fuel volume injected [mm3] 1500 3.44 900 Pilot: -17 Main: -5 3.9 mm3- 12,80mm3

Table 3: Engine operating conditions In the Figure 20 injection flow rate is shown, as measured experimentally for the case of a pilot injection at -17 with a main injection at -5. For these injection timing, the injection system predict with accuracy the mass of fuel injected and the injection rate. However, the model cannot predict a wave phenomena in the main injection that explain the error in the prediction of the injected mass. This wave phenomenon is observed in all injection events and we did not succeed to understand this physical process to avoid this lack of accuracy. As described before, a first calculation has been done to fit the engine system parameter. The system is stabilizing after approximately four engine cycles. The intake pressure is set to 0.1108MPa and the exhaust pressure to 0.1208MPa (experimental plenum pressure). The wall temperature are set to 450K for the piston and the cylinder wall and to 370K for the head cylinder. In Figure 18, the in-cylinder pressure and total mass histories are presented and attest that the dynamic Figure 18: In-cylinder pressure and total mass histories for coupled calculation without injection.

Figure 19: In-cylinder pressure (bar) calculated with the coupled system without fuel injection.

average mass. The Figure 26 plots the combustion heat released and we can notice the two local extremum representing the pilot injection and the main injection combustions. Fluctuations in the heat mass released are larger and attest that small fluctuations may have a larger impact on 3-D calculation prediction and in particular for pollutant predictions. Obviously, the variation of injected fuel mass may have a major effect on 3-D combustion calculation. The 1-D/3-D coupled system is able to predict the average experimental IMEP plotted on Figure 27. As seen in this figure, the dynamic system is converging after few cycles to the experimental IMEP. Figure 28 and Figure 29 present 3D view of the combustion chamber. For this multiinjection events, the spray particles impact the cylinder wall during the pilot injection. These pictures are corresponding to the last engine cycle during the pilot injection and the main injection.

Figure 20: Injection rate modeled vs. Experimental data for dwell = 12 In a second step, the full system has been simulated. For this calculation, all the initial parameters needed by IFP-C3D to compute the combustion are automatically computed and received from the injector simulator and the engine system simulator. Actually, the 1-D/3-D calculation is like a typical system simulation. Injection command is filled as the experimental test in the system and the calculation can be run. For engineers used to make 3-D calculations, this coupled way allows to avoid the difficult task to fit the initial parameters and is really a great advantage. However, the coupled calculation needs more CPU times to stabilize the dynamic system. As seen in Figure 21 and Figure 22, the system has converged after five engine cycles. The Figure 23 shows the comparison of the in-cylinder pressure predicted by the coupled simulation. The coupled simulation result is in good agreement with the experimental pressure. As seen in this graph, a small variation of the pressure can be observed and is due to the fluctuation of the incylinder mass and of the mass fraction of the burnt gases but also is due to the injected mass flow rate fluctuations. The Figure 24 shows the injected mass flow rate history using a crank angle modulo 720 abscissa and we can notice that the pilot injection and the main injection mass flow rate are influenced by the pressure in-cylinder history. The fuel mass predicted by the injector simulator for each engine cycle is varying between +/- 6% in comparison with the average value (see Figure 25). The injected volumes of the pilot and of the main injection have a cycle/cycle variation behavior. The experimental data gives for this operating condition a rate of IMEP instability equal to 1.7 % that do not fit with this large variation. In a first step, because of a lack of experimental data to validate this variation and eventually to fit the 1-D injector model, the coupled calculation used a fixed mass of fuel for each engine cycle given by the average of injected mass values. IFPC3D uses the instantaneous injection rate given by the 1-D injector model to model the fuel spray with this

Figure 21: In-cylinder pressure history

Figure 22: Burnt gases mass fraction history

Figure 23: In-cylinder pressure history vs. Experiment data for dwell = 12 (8 cycles plotted with a crank angle modulo 720 deg.)

Figure 26: Combustion heat released time history (8 cycles plotted with a crank angle modulo 720 deg.)

Figure 27: IMEP history vs. experiment Figure 24: Injected mass flow rate history (8 cycles plotted with a crank angle modulo 720 deg.)

Figure 25: Injected volume deviation history (%)

Figure 28: View liquid film thickness of the 3-D calculation during the pilot injection. It can be observed a liquid film formation during the pilot injection.

Figure 29: View liquid film thickness of the 3-D calculation during the main injection. Liquid fuel droplet are evaporating before reaching the cylinder wall.

Figure 30: In-cylinder pressure (bar) for dwell=16.

EFFECT OF INJECTION PARAMETERS ON COMBUSTION The coupled model has been tested by comparing model predictions with experimental measurements. The injection timing of the pilot injection in the single-cylinder engine has been varied, in order to study the effect on combustion. The experimental injection flow rate for all operating point are plotted in Figure 31. In Figure 32 the Indicated Mean Effective Pressure (IMEP) is shown for 5 different injection timings that leads to different injected mass (Figure 33). The main injection was held constant at 5 ATDC, while the pilot was varied from -17 to -33 ATDC. The coupled model is able to predict the IMEP variation when the pilot injection timing is varying. The injection rate for each operating condition is automatically computed by the injector model and IFPC3D computes the combustion using ENGINE results for the trapped mass and burnt gases mass fraction. Actually, the IMEP is mainly dependent of the injected mass and of the prediction of the injector model. However, spray impact have been noticed for late pilot injection timing which can explain IMEP variation. Figure 30 presents the pressure computed for a delay of 16 between pilot and main injection. The 1-D injector model have to be tested more intensively with experimental data to improve the numerical results and to understand the cycle/cycle dynamic behavior.

Figure 31: Experimental injection rate

4,5 4 3,5 3 IMEP 2,5 2 1,5 1 0,5 0 12 16 20 dwell (deg) 24 28 Experiment Model

Figure 32: Calculation vs. experimental measurements of IMEP and injected mass as a function of the dwell angle.

not sufficient to fit the model for the dynamic behaviour when taking account a counter-pressure coming from the combustion chamber.
0,7 0,6 0,5 0,4 Injected Mass (kg/h) 0,3 0,2 0,1 0 12 16 20 24 28

This coupling algorithm can be used to simulate 3-D cycle/cycle variation due to fluctuation in the injection system or in the air loop system. Moreover, 3-D calculations can compute pollutant emissions which can be used by the 1-D engine system model to predict the engine pollutant emissions. In the future, the engine system will be modified to add EGR loop and to test coupled calculation for operating conditions with EGR. This paper presents a first attempt to make a coupled calculation with a fine injection system and a basic engine system using a 3-D combustion code. This work shows that today 3-D simulations can be used in a complex engine system to simulate complex dynamic phenomena due to the dynamic interactions between all the different sub-models that can be found in an automotive engine.

Figure 33: Experimental fuel flow rate

CONCLUSIONS
This paper describes a new methodology that involved 1-D injector simulation, 0-D air loop simulation and 3-D combustion calculation coupled to simulate the full engine system with a predictive combustion modeling. This multi-scale simulation allows to compute 3-D detailed combustion using accurate and dynamic conditions given by the 0-D air loop model and 1-D injection model. This coupled algorithm can be used to make easier 3-D calculations but also to avoid the difficult task to define all initial conditions needed by the 3-D calculation. Indeed, more than one cycle is needed to converge to a stabilized solution and a coupled simulation needs at least five engine cycles. However, these kind of calculation fill already the industrial requirement of a calculation return time lower than 12 hours on two PC's processors to simulate eight engine cycles. This return calculation time may decrease with more than two processors for parallel calculation and with new processor development (high speed multi-core processor, high number of processor per node). Moreover, using heterogeneous parallel grid computer helps to manage multi-scale calculation and may also reduce the CPU cost. This coupling algorithm has been used in a Diesel direct injection single cylinder engine on a variety of injection conditions. Numerical results are in good agreement with experimental results. An injector model has been generated and has been fitted on experimental data. Simulation results have shown that the mass flow rate predicted by the 1-D model is influenced by the incylinder pressure histories. These fluctuations (+/- 6% of the injected mass) have to be investigated and more detailed experimental data are needed to validate this behavior. However, coupled numerical simulations show that dynamic behavior of the injector model is not negligible and can have a major effect on 3-D combustion predictions. Actually, the injector model is fitted using experimental data in cell and these data are

ACKNOWLEDGMENTS
We would like to acknowledge Mr. Juan Francisco Almarza for his contribution in the injection model development and Mrs Maria Alejandra Thirouard for her contribution to provide the engine experimental data. We would like also to acknowledge Mr Christophe Hanauer for experimental injection data .

REFERENCES
1. Benajes, J., Molina, S., Garcia, J.M, "Influence of pre- and post-injection on the performance and pollutant emissions in a HD diesel engine", SAE Technical Paper Series 2001-01-0526 2. Ishikawa, N., Uekusa, T., Nakada, T., Hariyoshi, R. "DI diesel emission control by optimized fuel injection", SAE Technical Paper Series 2004-010117 3. Payri, F., Benajes, J., Pastor, J.V., Molina, S. "Influence of the post-injection pattern on performance, soot and NOx emissions in a HD diesel Engine", SAE Technical Paper Series 2002-01-0502 4. Chryssakis, C.A., Hagena, J.R., Knafl, A., Hamosfakidis, V.D., Filipi, Z.S., Assanis, D.N. "Incylinder reduction of PM and NOX emissions from diesel combustion with advanced injection strategies", Int. J. Vehicle Design, Vol. 41, No. 1-4, pp.83-102, 2006 5. Ficarella, Laforgia, D., Landriscina, V., "Evaluation of Instability Phenomena in a Common Rail Injection System for High Speed Diesel Engines, SAE Technical Paper Series 1999-01-0192

6. Desantes, J.M., Arrgle J., Rodrguez, P.J., Computational Model for Simulation of Diesel Injection Systems, SAE Technical Paper Series 1999-01-0915 7. Han, J.-S., Wang, T.C., Lai, M.-C., Henein, N.A., Harrington, D.L., Pinson, J., Miles, P., "Dynamics of Multiple-Injection Fuel Sprays in a Small-bore HSDI Diesel Engine", SAE Technical Paper Series 200001-1256 8. Bianchi, G.M., Falfari, S., Pelloni, P., Kong S.-C., Reitz, R.D., Numberical Analysis of High pressure Fast-Response Common Rail Injector Dynamics, SAE Technical Paper Series 2002-01-0213 9. Bianchi, G.M., Falfari, S., Pelloni, P., Filicori, F., Milani, M., A Numerical and Experimental Study Towards Possible Improvement of Common Rail Injectors, SAE Technical Paper Series 2002-010500 10. Zhong, L., Singh, I.P., Han, J., Lai, M.-C., Henein, N.A., Bryzik, W., "Effect of Cycle-to-Cycle Variation in the Injection Pressure in a Common Rail Injection System on Engine Performance", SAE Technical Paper Series 2003-01-0699 11. Sinclair, R., Strauss, T., Schindler, P. "Code Coupling, a New Approach to Enhance CFD Analysis of Engines", SAE Technical Paper Series 2000-01-0660 12. Borghi, M., Mattarelli, E., Montorsi, L., "Integration of 3-D-CFD and Engine Cycle Simulations: Application to an Intake Plenum", SAE Technical Paper Series 2001-01-2512 13. Riegler, U.G., Bargende, M., "Direct Coupled 1-D/3D-CFD-Computation (GT-Power/Star-CD) pf the Flow in the Switch-Over Intake System of an 8Cylinder SI Engine with External Exhaust Gas Recirculation", SAE Technical Paper Series 200201-0901 14. Wehr, D., Huurdeman, B., Spennemann, A., "EGR A Challenge for Modern Plastic Intake Manifolds", SAE Technical Paper Series 2002-01-0902 15. Bohbot, J., Chryssakis, C., Marjorie, M., "Simulation of a 4-Cylinder Turbocharged Gasoline Direct Injection Engine Using a Direct Temporal Coupling Between a 1-D Simulation Software and a 3-D Combustion Code", SAE Technical Paper Series 2006-01-3263 16. Menegazzi, P., Aubret, P., Vernhes, P.-L., "Conventional and Hybrid Vehicle Emission, Fuel Economy and Performance Analysis System Simulation", FISITA 2004, 23-27 May, Barcelona, Spain 17. Zolver, M., Benkenida, A., Bohbot J., Klahr, D., Rveill, B., "CFD Tools at IFP for HCCI Engine Simulations", 14th International Multidimensional Engine Modeling User's Group Meeting at the SAE 2004, March 2004, Detroit, USA

18. Zolver, M., Klahr, D., Bohbot, J., Laget, O., Torres, A., "Reactive CFD in Engines with a New Unstructured Parallel Solver", Oil & Gas Science and Technology, vol. 58, no. &, pp. 33-46, 2003 19. Bohbot, J., Klahr, D., Zolver, M., Torres, A., "A Three Dimensional Modelling of Combustion in a Direct Injection Diesel Engine using a New Unstructured Parallel Solver", The 2003 International Conference on Computational Science and its Applications, ICCSA, May 2003, Montreal, Canada, proceedings, pp. 483-492 20. Zolver, M., Bohbot, J., Klahr, D., Torres, A., "An Unstructured Parallel Solver for Multi-Phase and Reactive Flows in Internal Combustion Engines", Parallel CFD 2003, May 2003, Moscow 21. Colin, O., Benkenida, A., Angelberger, C., "A 3-D Modeling of Mixing, Ignition and Combustion Phenomena in Highly Stratified Gasoline Engines", Oil & Gas Science and Technology, vol. 58, pp. 4762, 2003 22. Colin, O., Benkenida, A., "The 3-Zones Extended Coherent Flame Model (ECFM3Z) for Computing Premixed Diffusion Combustion", Oil & Gas Science and Technology, vol. 59, no. 6, pp. 593-609, 2004 23. Colin, O., Pires da Cruz, A., Jay, S., "Detailed chemistry based auto-ignition model including low temperature phenomena applied to 3-D engine calculations", 30th Symposium (International) on Combustion, The Combustion Institute, 2004 24. Han, Z., Reitz, R.D., "Multidimensional Modeling of Internal Combustion Engines Using RNG k- Models", Combust. Sci. Technol., 106, pp. 206-280, 1995 25. http://www.openmp.org 26. Lafossas FA, Colin .O et al., "Application of a new 1D combustion model to gasoline transient engine operation". SAE Technical Paper Series 2005-012107 27. Barba C., Burkhardt C., Boulouchos K., Bargende
M., "A phenomenological combustion model for heat

release rate prediction in high speed DI Diesel engines with common-rail injection", SAE Technical Paper, 2000-01-2933, 2000.