Rapid Meshing and Advanced Physical Modeling for Gasoline DI Engine Application

Rapid Meshing and Advanced Physical Modeling for Gasoline DI Engine Application R. Tatschl, H. Riediger, Ch. v. Künsberg Sarre, N. Putz and F. Kickinger AVL LIST GmbH A-8020 Graz AUSTRIA Gasoline direct injection engine development is significantly supported by advanced diagnostics and CFD simulation techniques aiming at the optimization of injection and combustion system performance. With regard to the simulation, the gasoline direct injection engine imposes severe requirements onto the performance of mesh generation tools and the applicability of mixture formation and combustion models. Rapid meshing tools and the availability of carefully validated models of mixture formation and combustion are required in order to enable the use of CFD as a reliable tool within the engine development process. The present article is aimed at demonstrating the current status of advanced rapid meshing technology and mixture formation / combustion modeling for the application to gasoline direct injection engines. INTRODUCTION Due to their promising performance with respect to fuel economy, SI engines with gasoline direct injection are subject of intense world wide research and development [1]. The complex in-cylinder processes of gasoline DI engines require the application of advanced development tools in order to meet the required goals regarding injection and combustion system performance. Besides sophisticated optical diagnostic techniques [2], comprehensive CFD models of the governing flow, mixture formation and combustion phenomena are needed in order to obtain the desired insight into the complex in-cylinder processes. In the engine development process, CFD modeling of DI gasoline systems is used to analyze the interaction of fuel injection and combustion with combustion chamber components and in-cylinder gas motion. Obviously, the aim is to minimize the variety of concepts and to identify potentially successful solutions in an early stage of system design. The success in meeting these goals is first of all based on the capability of performing a large number of parametric studies, including intake port and combustion chamber / piston bowl shape variations, within the given time-frame in the engine development process. In this context only advanced meshing tools enabling automatic generation of 3D computational grids of high geometrical complexity, including multiple moving parts, are capable of meeting the present demands with respect to short mesh generation time and ease-of-use. The second prerequisite for a successful integration of CFD in the engine development process is the adequate modeling of the individual processes governing mixture formation and combustion. The present article is aimed at demonstrating the current status in the development and application of a rapid IC engine meshing technology and state-of-theart spray and combustion models to the simulation of gasoline direct injection engines. The basic features and philosophy of a flexible, automatic meshing environment, FAME and FAME- Engine, are introduced and demonstrated on the basis of meshing the AVL Direct Gasoline Injection engine. The mixture formation and combustion modeling principles adopted in FIRE are presented together with selected results of the experimental and numerical spray validation studies. Finally, representative results of the mixture formation and combustion optimization calculations are shown for the AVL Direct Gasoline

Injection engine. Where appropriate, the simulation results are compared to experimental data. loaded from a set of engine specific start topologies, provided within FAME (Figure 1). ENGINE APPLICATION In order to demonstrate the applicability of the present methodology to the numerical optimization of gasoline DI engines, the processes of flow, mixture formation and combustion are presented for the AVL Direct Gasoline Injection engine. The engine geometry data and the operating conditions relevant to the present study are given in Table 1. AVL Direct Gasoline Injection Engine - bore x stroke 86 x 86 mm - compression ratio 11:1-4-valve cylinder head - port deactivation - injector between inlet valves Operating Conditions - engine speed 2000 1/min - IMEP 2.9 bar - piston temperature 450 K - liner temperature 400 K - cylinder head temperature 450 K Table 1 AVL Direct Gasoline Injection engine; engine geometry data and operating conditions Figure 1 Start Topology A volume mesh, referred to as the start mesh, is automatically generated, using FAME, by deleting all cells of the start topology which lie outside of the geometric domain defined by the CAD data. Between the CAD surface and the resulting trimmed volume mesh a small gap remains. This gap is automatically filled with body-fitted cell layers. Characteristic edges are detected automatically and considered during the meshing process (Figure 2). MESHING The method described here is valid for all types of diesel, gasoline (port- and direct-injection) and natural gas engines. Two grid generation tools are used for this method: FAME, a hex-based automatic meshing tool and FAME-Engine, a tool to automatically generate different geometry positions for IC-engine applications (steady state and transient). The major steps of the mesh generation process are as follows: The procedure is begun with the loading of CAD surface data and a start topology into FAME. CAD data may be transferred from Pro/Engineer, CATIA, Unigraphics, I-DEAS or other systems. The start topology is a simple volume mesh, which defines the major directions and densities of the mesh. It can be Figure 2 Start Mesh In order to properly create meshes of different crank angle positions, different volume parts of the start mesh are specified as moving, non-moving and buffer regions within FAME-Engine (Figure 3).

Figure 5 Details of the Moved Mesh Figure 3 Definition of Different Mesh Volume Conditions The mesh is automatically moved according to the valve lift diagrams and the piston kinematics. During the movement the volume mesh quality is constantly improved by use of smoothing algorithms. Also during this process, the mesh quality is continuously checked (Figure 4). Figure 4 Moving the Mesh The intention of the process is to allow the user to automatically generate the description of the complete movement on the basis of one start mesh. However, the procedure is flexible enough to enable manual interactive manipulation to improve the mesh quality at selected crank angle positions, or to use additional start geometries for the beginning of different crank angle ranges (Figure 5). There are two major methods to deal with varying engine geometries. a) Adapt the start topology and re-mesh with FAME. b) Split the part you want to exchange, mesh the new part with FAME and adapt it to the rest. The second method is used in Figure 6 to exchange an intake port. Figure 6 Exchange of Intake Port COMPUTATIONAL METHOD The numerical method applied in the present work is the CFD code FIRE. A number of publications have described the FIRE code and its application to engine related flow, mixture formation and combustion simulation [3-8]. Hence, only a brief description of the applied models relevant to the present study is given here.

Mixture Formation The commonly used high pressure swirl injectors for gasoline DI applications are designed to create a rotating hollow cone spray. The main characteristics of this type of injectors are the even fuel distribution, an early breakup of the liquid sheet into individual droplets and hence reduced penetration. Additionally, this type of injector provides the necessary variability of the spray geometry for different stratification requirements. The spray cone angle produced by the injector depends on the fuel supply pressure and on the cylinder backpressure in a way that meets the requirements, i.e. the spray cone angle is reduced under high backpressure conditions. The applicability of a CFD spray model is hence closely related to its capability to reflect the backpressure influence on the spray shape, cone angle and penetration characteristics under engine operating conditions. At AVL, experimental spray model validation data under realistic engine operation conditions are obtained in an optically accessed spray research engine. Different backpressure conditions are achieved by suitable crank-angle timing of the injection process. The relevant data for validation of the CFD method, i.e. the spray shape evolution and its dependence on the combustion chamber backpressure, are obtained by LIF spray imaging and statistical evaluation of multiple injection events. Figure 7 displays the simulated results of the spray propagation process in the spray research engine for a backpressure variation from 1 to 6 bar together with the corresponding experimental data. The images clearly show the strong dependence of the spray shape and the maximum spray diameter on backpressure due to earlier breakup of the stable liquid sheet. The earlier breakup of the sheet leads to an increased droplet / gas interaction, consequently also reducing the overall spray penetration length, as shown in Figure 8. Figure 8 Chamber backpressure influence on spray penetration length; calculation vs. measurement (rail pressure 80 bar) Combustion Combustion in gasoline DI engines at part load operation no longer proceeds via a self sustained propagating flame front like under full load operation or in conventional gasoline engines, but occurs under partially premixed and / or diffusion controlled conditions, depending on the local mixture and temperature conditions in the fuel vapor plume. Additionally, finite-rate chemical kinetic effects play an important role on the lean side of the fuel / air mixture cloud. Figure 7 Chamber backpressure influence on spray shape and cone angle; (a) CFD results, (b) statistical evaluation of fuel LIF data (rail pressure 80 bar, 3 ms after SOI) In the present case, the different combustion regimes and finite-rate chemistry effects are accounted for by a transported multi-scalar PDF method [6]. The hydrocarbon oxidation process is modeled by a global irreversible combustion reaction with an Arrhenius reaction rate expression reflecting the variation of laminar burning velocity reasonably well for a wide range of equivalence ratio, including the lean and rich flammability limits of φ l =0.5 and φ r =4.3, respectively.

Flame kernel formation is simulated through prescription of the temporal variation of the reaction progress variable and enthalpy pdf s. a) b) Engine Simulation Results Part load mixture formation in gasoline DI engines has to ensure proper supply of fuel at combustible portions to the location of the spark plug in order to enable stable initiation of a self sustaining combustion process. The temporal evolution of the mixture preparation behavior for the AVL DGI engine configuration is shown in Figure 9, together with the corresponding LIF data. The depicted iso-levels represent the local fuel vapor distribution. a) b) Figure 10 AVL Direct Gasoline Injection engine; calculated fuel vapor distribution at SA=-33 deg. CA ATDC, (a) effect of mean flow motion orientation, (b) impact of injection timing a) Figure 9 AVL Direct Gasoline Injection engine; (a) calculated fuel vapor distribution evolution, (b) statistical evaluation of fuel LIF data (section across cylinder and injector axis) It can be seen from Figure 9 that the propagation behavior of the fuel vapor cloud is predominantly determined by the induction and injection induced flow pattern and its subsequent interaction with the piston bowl. Hence, proper mixture formation is the result of a large number of system parameters, such as intake port and piston bowl design, injector characteristics and location as well as injection timing (Figures 10 and 11). b) Figure 11 AVL Direct Gasoline Injection engine; calculated fuel vapor distribution at SA=-33 deg. CA ATDC, (a) influence of injector characteristics, (b) effect of injector axis inclination A typical example of the space and time resolved combustion evolution under gasoline DI operating conditions is given in the sequence of pictures in Figure 12. Calculated results of the reaction products mass fraction distribution in a section across the cylinder and fuel injector axis are presented together with the corresponding statistically averaged fuel LIF images.

Shortly after spark break-down (-33 deg. CA ATDC) the flame kernel exhibits a rather small growth rate but significantly shows a preferential burn direction along the near stoichiometric fuel concentration iso-surface. Fuel rich zones that can mainly be found in the vicinity of the piston bowl and which are considerably outside the flammability limits remain nearly unburned during the first premixed-like combustion mode. On the fuel lean side of the fuel vapor plume, however, the reaction speed is chemically kinetics limited. Due to the ongoing intermixing of fuel vapor with the surrounding air as the combustion process proceeds, the local fuel portions at the periphery of the fuel containing zone become more and more leaned out until they finally exceed the lean flammability limits. Flame quenching in this area can finally lead to substantial formation of unburned hydrocarbons. a) b) Figure 13 AVL Direct Gasoline Injection engine; variation of injection timing Figure 12 AVL Direct Gasoline Injection engine; (a) calculated reaction products distribution evolution, (b) statistical evaluation of flame LIF data (section across cylinder and injector axis); SA 33 deg. CA ATDC The space / time resolved combustion characteristics also manifest themselves in a burn rate different to the one obtained under conventional premixed conditions. The calculated mass fraction burned data in Figures 13 and 14 indicate a rapid burning of the fuel, with an overall main phase combustion duration of approximately 15 degrees crank-angle. This is a reduction in the duration of the main combustion phase of about 50 % compared to a conventional SI engine. This behavior can be attributed on one hand to the remnant of the injection induced elevated turbulence intensity level in the spray region, but is to a larger extend due to the short flame travelling paths within the compact fuel vapor containing area. The slow burning in the late combustion phase is then mainly governed by the diffusion type burnout of the initially fuel rich zones that get subsequently mixed with air. A comparison of the calculated 2, 50 and 90% mass fractions burned with the corresponding measured data (thermodynamic analysis of the cylinder pressure trace) indicates good overall agreement for variations in injection timing and spark advance (Figures 13 and 14).

to the overall spray shape, penetration lengths and radial spray extension. The FIRE application to the full mixture formation and combustion simulation for the AVL-DGI engine clearly shows to adequately reflect the characteristic gasoline DI engine features and to provide a detailed and valuable insight into the complex processes of flow, spray propagation and combustion interaction. Locally resolved field quantities as well as global mixture formation and combustion data show good agreement with data extracted from in-cylinder pressure measurements. Parametric studies, systematically varying overall flow structure, injector characteristics and positioning as well as injection and ignition timing, have shown to coincide well with experimental observations. REFERENCES Figure 14 AVL Direct Gasoline Injection engine; variation of spark advance SUMMARY AND CONCLUSIONS The basic features of the new flexible, automatic meshing environment, FAME and FAME-Engine, have been introduced. Its application to the rapid generation of highly complex computational meshes for the AVL-DGI engine proves the superior performance when compared to conventional solutions. With FAME and FAME-Engine, meshing times can be reduced up to 70 %, assuming the availability of reasonable CAD surface data. A joint experimental / numerical approach has been presented for determination of the fuel spray initial conditions of hollow cone injectors in the FIRE simulations. Comparisons of the calculated results with experimental data show good agreement with respect [1] Wirth, M., Piock, W.F., Fraidl, G.K., Actual Trends and Future Strategies for Gasoline Direct Injection, IMechE S433/004/96 [2] Winklhofer, E., Fraidl,, G.K., Tatschl, R., Flame Visualisation in Gasoline Engines - New Tools in Engine Development, JSAE 9530850 [3] Tatschl, R., Wieser, K., Reitbauer, R., Multidimensional Simulation of Flow Evolution, Mixture Preparation and Combustion in a 4-Valve Gasoline Engine, COMODIA 94, pp. 139-149, Yokohama, 1994 [4] Cartellieri, W., Chmela, F., Kapus, P., Tatschl, R., Mechanisms Leading to Stable and Efficient Combustion in Lean Burn Gas Engines, COMODIA 94, pp. 17-24, Yokohama, 1994 [5] Tatschl, R., Fuchs, H., Brandstätter, W., Experimentally Validated Multi-dimensional Simulation of Mixture Formation and Combustion in Gasoline Engines, IMechE C499/050/96 [6] Tatschl, R., Riediger, H., PDF Modelling of Stratified Charge SI Engine Combustion, SAE 981464 [7] Tatschl, R., Pachler, K., Winklhofer, E., A Comprehensive DI Diesel Combustion Model for Multidimensional Engine Simulation, COMODIA 98, pp. 141-148, Kyoto, 1998 [8] v. Kuensberg Sarre, Ch., Tatschl, R., Spray Modelling / Atomisation Current Status of Break-Up Models, Turbulent Combustion of Gases and Liquids, - Leading Edge Technologies, ImechE Seminar, Lincoln, 1998