1. IntroductionThe requirement to meet the challenge of producing cleaner and more efficient power plants will intensify further over the next few years. This challenge requires an increased commitment to research by the transportation industry. The internal combustion engine represents one of the more challenging fluid mechanics problems to model because the flow is compressible with large density variations, relatively high Mach number, turbulent, unsteady, cyclic, and non-stationary, both spatially and temporally. Much progress has been made in CFD model development for engines in recent years.
Clean diesel engines are one of the fuel efficient and low emission engines of interest in the automotive Industry. The combustion chamber flow field and its effect on fuel spray characteristics plays an important role in improving the efficiency and reducing the pollutant emission in a direct injection diesel engine, in terms of influencing processes of breakup, evaporation mixture formation, ignition, combustion and pollutant formation. CFD modeling is a valuable tool to acquire detailed information about these important processes. In this context [2], the characteristics of ultra-high injection pressure diesel fuel sprays are simulated and validated in a quiescent constant volume chamber. A profile function is utilized in order to apply variable velocity and mass flow rate at the nozzle exit. The CFD model is also applied to an open cycle engine model to study the effects of engine flow field features such as swirl and tumble motions on the spray behavior [2].
A multi-zone direct-injection (DI) diesel combustion model has been implemented for full cycle simulation of a turbocharged diesel engine [3]. The above combustion model takes into account the following features of the spray dynamics:
• the detailed evolution process of fuel sprays;
• interaction of sprays with the in-cylinder swirl and the walls of the combustion chamber;
• the evolution of a Near-Wall Flow (NWF) formed as a result of a spray-wall impingement as a function of the impingement angle and the local swirl velocity;
• interaction of Near-Wall Flows formed by adjacent sprays;
• the effect of gas and wall temperatures on the evaporation rate in the spray and NWF zones.
A NOx calculation sub-model uses detailed chemistry analysis which considers 199 reactions of 33 species. A soot formation calculation sub-model used is the phenomenological one and takes into account the distribution of the Sauter Mean Diameter in injection process. The ignition delay sub-model implements two concepts. The first concept is based on calculations using the conventional empirical equations. In the second approach the ignition delay period is estimated using relevant data in the calculated comprehensive 4-D map of ignition delays. The model has been validated using published experimental data obtained on highand medium-speed engines. Comparison of results demonstrates a good agreement between theoretical and experimental sets of data [3].
By separating the fluid dynamic calculation from that of the chemistry, the unsteady flamelet model allows the use of comprehensive chemical mechanisms, which include several hundred reactions. This is necessary to describe the different processes that occur in a DI Diesel engine such as auto ignition, the burnout in the partially premixed phase, the transition to diffusive burning, and formation of pollutants like NOx and soot. The experimental results show good agreement for the whole combustion cycle (ignition delay, maximum pressures, torque and pollutant formation) between the two-component reference fuel and Diesel. The simulations are performed for both reference fuels and are compared to the experimental data. Pollutant formation (NOx and soot) is predicted for both reference fuels. The contributions of the different reaction paths (thermal, prompt, nitrous, and reburn) to the NO formation are shown. Finally, the importance of the mixing process for the prediction of soot emissions is discussed [4].
The KIVA code is widely used for model development in academia due to the availability of the source. However, its capability for resolving complex geometries is limited.
The KIVA engine simulation developed by Los Alamos National Laboratory was used to characterize the combustion of alternative fuels in a direct injection diesel engine. Rapeseed oil, its methyl ester and hexadecane were used in engines run at 3000 rev/min and 50% maximum torque. Approximately 40 consecutive cycles were phase averaged to derive the pressure traces for comparison to KIVA predictions. The engine parameters and the fuel properties used in the simulations are described. Simulation results were good for the methyl ester and for hexadecane which was used as a reference fuel. The model predicted lower pressures for the rape oil than those which were experimentally observed [5].
A modified CFD code based on the KIVA family of codes incorporating several strategies for reducing the computational time required for diesel engine simulations is presented. The improved code and coarse meshes are used together to simulate combustion in a heavy-duty Mitsubishi Heavy Industries diesel engine operated over a range of loads, speeds, and injection strategies. The average simulation time from IVC to EVO is reduced from around 60 hours to 1 hour through the use of 12 processors and the new strategies [6].
On the other hand, other commercial CFD codes such as STAR-CD, FIRE, VECTIS and FLUENT are frequently used by industry due to their superior mesh generation interfaces and because of their available user support. Some scientists combined STAR-CD and KIVA code for the engine simulations but they concluded that, it would be preferable to implement the advanced sub models directly into one commercial code for engine simulations [7].
The gas motion inside the engine cylinder plays a very important role in determining the thermal efficiency of an internal combustion engine. A better understanding of in cylinder gas motion will be helpful in optimizing engine design parameters. An attempt has been made to study the combustion processes in a compression ignition engine and simulation was done using computational fluid dynamic (CFD) code FLUENT, Turbulent flow modeling and combustion modeling was analyzed in formulating and developing a model for combustion process [8].
This paper describes the development and use of sub models for combustion analysis in direct injection (DI) diesel engine. In the present study the Computational Fluid dynamics (CFD) code FLUENT is used to model complex combustion phenomenon in compression ignition (CI) engine. The results obtained from modeling were compared with experimental investigation. Consequences in terms of pressure, rate of pressure rise and rate of heat release are presented. The rate of pressure rise and heat release rate were calculated from pressure based statistics. The modeling outcome is discussed in detail with combustion parameters. The results presented in this paper demonstrate that, the CFD modeling can be the reliable tool for modeling combustion of internal combustion engine [9].
4. Results and Analysis4.1. Grid AnalysisThe computational grids of the DI diesel engine are showed in Figure 1. The present study model is build using approximate 200,000 unstructured mixed cells types (hexagonal and tetrahedral) the grid is divided into two zones. The first zone is the cylinder zone and the second is the bowl zone. The first zone uses the hexagonal cells and the second zone uses tetrahedral cells as shown in Figure 1. In engine operation, valves and the piston move, so the mesh should move according to the real engine in order to simulate the charge of valve and piston position with crank angle. Piston and piston bowl movement are decided by the stroke, connecting rod and crank angle. Simulation starts at 239˚ CA and ends at 469˚ CA. This is the period from the inlet valve closing till exhaust valve opening. That means thermodynamically; the compression stroke and power stroke only accounted in this study (indicated gross work). The simulation time step is 0.5 crank angles. Each computer run spends about 12 hours on IBM compatible computer using quad duo processor with 2 GHz and 4 MB cache and 6 GB RAM. The simulation uses fixed temperature boundary conditions [21] and the initial conditions driven from the experimental work [13].
4.2. Model ValidationFigure 2 shows the comparison of the measured and predicted cylinder pressures. The ignition model scaling factor tuned for the first case (maximum load simulation) and kept constant in all cases (other load simulated). The ignition point in most cases is nearly captured. The peak pressure and the pressure gradient over the combustion period produced by CFD simulation match closely with measurements. The peak cylinder pressure is over predicated by about 6%. Figures 3-6 demonstrate the in-cylinder processes reproduced by CFD simulation using a series of distribution plots illustrating temperature, pressure, fuel mass fraction and gas velocity across the cylinder center plan respectively.
4.3. Engine PerformancesThe engine performance at different loads is shown in Figure 7. The figure shows the predicted performance from maximum load till the no load. From the pressurecrank angle diagram the peaks pressure at different loads are determined, also the start of injection, start of ignition is observed. Figure 8 shows p-v diagram which is not closed, because the simulation period did not reach to the end of expansion stroke at exhaust valve opening.
Figure 9 shows predicted rate of heat release. The curve accurately and clearly determined the different combustion zones and delay time period.
The start of injection crank angle is determined from [13] and the injection period is determined from [11,12]. The delay time period is determined to be from the start of injection till the p-θ change its slope as identified in Figure 9. This period is nearly 11 crank angles vary according to the φ (equivalence ratio) or load. Figure 10
shows the predicted heat release along with the predicted engine pressure at maximum load. From Figure 10 we can get that the total burning period is much longer than the injection period, also the magnitude of the initial peak of the burning rate depends on the ignition delay period; that means increasing delay period will increase burning rate peak [22]. Figure 11 illustrates the relative
magnitude of the gross and net heat release, heat transfer and heat of vaporization and heating up of the fuel at 1500 RPM for the base line engine. The heat release analysis method [23] is used to obtain the combustion information from the pressure data. The net heat release is the gross heat release due to combustion extracts from it, the heat transfer to the walls (omitting the crevice effect) and the effect of fuel vaporization and heat up. The energy change associated with heating up fuel vapor from injection temperature to typical compression air temperature is about 6% of the fuel heating value. The heat transfer integrated over the duration of the combustion period is about 35% of the total heat release. The small kink in the lower part of the curve shows the heat of vaporization
and heating up fuel.
Figure 12 shows the predicted temperature-crank angle diagram at different loads. The curve shows that increasing the load which corresponding to increasing the fuel mass flow rate results in increasing the combustion temperature. Figure 12 shows that the temperature of the air at the end of compression is sufficiently high for the droplets of fuel to vaporize and ignite as they enter the cylinder. Very small downward kinks found near fuel in-
