Piston Fuel Film Observations in an Optical Access GDI Engine by Rolf B. Karlsson B.S., Mechanical Engineering GMI Engineering & Management Institute, 1996 Submitted to the Department of Mechanical Engineering in Partial Fulfillment of the Requirements for the Degree of Master of Science in Mechanical Engineering at the Massachusetts Institute of Technology September 2000 @ 2000 Massachusetts Institute of Technology All rights reserved Signature of Author_ ]6 epartment of Mechanical Engineering August 18, 2000 Certified by Jofin B. Heywood Sun Jae Professor of Mechanical Engineering Thesis Supervisor Accepted by Ain A. Sonin Chairman, Department Committee on Graduate Students MASSACHUSETTS INSTITUTE OF TECHNOLOGY SEP 2 0 2000 LIBRARIES (This page intentionally left blank) AsKt - 4 2 Piston Fuel Film Observations in an Optical Access GDI Engine by Rolf B. Karlsson Submitted to the Department of Mechanical Engineering September 2000 in Partial Fulfillment of the Requirements for the Degree of Master of Science in Mechanical Engineering ABSTRACT Increasing pressure for reduced fuel consumption, particularly in Japan and Europe, has motivated the development of various engine concepts that improve part-load efficiency. One of these concepts is Gasoline Direct Injection, or GDI. By injecting fuel under high pressure directly into the combustion chamber later in the engine cycle and using various piston geometry and combustion chamber airflow techniques, a locally stratified, globally lean mixture can be used to obtain part load power with greater efficiency. Emissions concerns, however, particularly hydrocarbons from unburned mixture and liquid films on the piston and cylinder walls, are a major concern. The purpose of this project was to observe the interaction of a GDI fuel spray with the moving piston crown in an optical access engine, describe the interaction qualitatively, devise a means to measure the residual fuel film on the piston quantitatively, and compare the results with a Computational Fluid Dynamics (CFD) simulation. CFD was also used extensively to aid in the design of the experimental apparatus, particularly in selecting the spray and injection parameters that would allow stratified charge operation in the simple geometry available in the optical engine. The CFD simulations predicted that within the operating window for stratified charge operation, between 1% and 4% of the injected fuel would remain on the piston as a liquid film. The most influential parameter for the quantity of residual fuel film was the piston temperature. The experimental results support the CFD simulations qualitatively, but the amount of fuel film remaining on the piston is under-predicted. High-speed video footage shows a vigorous spray impingement on the piston crown resulting in vapor production. The primary GDI combustion process, a roughly premixed flame, is followed by a turbulent diffusion flame or "pool fire" in the region of the liquid film once the piston is warm enough to produce vapor at a high enough rate. Thesis Advisor: Professor John B. Heywood Title: Sun Jae Professor of Mechanical Engineering 3 (This page intentionally left blank) 4 ACKNOWLEDGMENTS I would like to thank my thesis advisor, Professor John B. Heywood, for his support and guidance over the past year. David Schmidt, Wai Cheng, and Chris O'Brien deserve credit for helping me set up an SGI workstation with KIVA-3V. Invaluable technical and administrative support has been provided by Thane DeWitt, Susan Lutin, and Karla Stryker. I would especially like to thank Benoist Thirouard for his insightful help with optical diagnostics and hightemperature adhesives. Other Sloan Lab residents who have helped to make my stay at MIT more enjoyable include Brian Hallgren, Gary Landsberg, Brigitte Castaing, Matt Rublewski, and my officemate, Ferran Ayala. On campus but not part of the Sloan Lab is Dr. James Bales of the Edgerton Laboratory, to whom I am grateful for assistance with high-speed video imaging and equipment. Thanks also go out to Brad VanDerWege, my predecessor, for his many emailed responses to my questions on the optical engine. The one person who has probably supported me the most during my time at MIT is my wife, Amira. She cared enough for me to exchange our family, friends and home in Michigan for the strange new environment of the Metro Boston area, which I don't think either of us was quite prepared for. Without her I would not have come to MIT in the first place, and it is her continued support that has enabled me to get to where I am now. Thanks also go to my family, for creating and shaping me as I am, the people who helped me get into MIT including Ruth Fukuchi, Reggie Bell, Katie Jiang, Karim Nasr, and Fred Shen, and General Motors for providing my Fellowship. The research project itself was sponsored by the MIT Sloan Automotive Laboratory Engine and Fuels Research Consortium, which includes DaimlerChrysler, Ford, General Motors, ExxonMobil, Shell UK, and Volvo Car. 5 (This page intentionally left blank) 6 TABLE OF CONTENTS A BSTRA C T ................................................................................................................................................................. 3 A C KN O W LED GM EN TS .......................................................................................................................................... 5 TA BLE O F C O N TEN TS ........................................................................................................................................... 7 N OM EN CLA TU R E .................................................................................................................................................... 9 LIST OF TA BLES .................................................................................................................................................... 10 LIST O F FIGU R ES .................................................................................................................................................. 11 INTR OD U C TION ........................................................................................................................................... 13 1 1. 1 1.2 1.3 M OTIVATION ................................................................................................................................................. 13 BACKGROUND ............................................................................................................................................... 14 O BJECTIVES ................................................................................................................................................... 16 SIM U LA TION ................................................................................................................................................. 19 2 2.1 2.2 2.3 2.4 2.5 INTRODUCTION .............................................................................................................................................. K fVA-3V M ODEL .......................................................................................................................................... SIM ULATION RESULTS ................................................................................................................................... LIQUID FUEL FILM ON THE PISTON ................................................................................................................ COM PARISON wrrH EXPERIMENTAL RESULTS ............................................................................................... 19 20 22 24 25 EX PERIM ENTA L SETU P............................................................................................................................. 33 3 3.1 3.2 3.3 3.4 INTRODUCTION .............................................................................................................................................. O PTICAL ACCESS ENGINE .............................................................................................................................. HIGH-SPEED V IDEO IM AGING ........................................................................................................................ W IDE BEAm LASER-INDUCED FLUORESCENCE ............................................................................................. 33 33 35 35 EXPER IM EN TA L R ESU LTS ....................................................................................................................... 43 4 4.1 4.2 HIGH-SPEED V IDEO RESULTS U SING ALum iNum PISTON ............................................................................. .43 H IGH-SPEED VIDEO RESULTS U SING O PTICAL ACCESS PISTON ..................................................................... 46 SUM M A RY A N D C O N CLUSIO N S .............................................................................................................. 71 5 5.1 5.2 SIMULATION WORK ....................................................................................................................................... 71 H IGH-SPEED V IDEO IM AGING ........................................................................................................................ 71 5.3 CONCLUSIONS ............................................................................................................................................... 72 5.4 FUTURE W ORK .............................................................................................................................................. 72 R EFEREN CES.......................................................................................................................................................... 75 APPENDIX A - SAM PLE KIVA INPUT DECK .................................................................................................. 79 APPENDIX B - NEW ENGINE COMPONENT DRAWINGS ............................................................................ 83 (This page intentionally left blank) 8 NOMENCLATURE ATDCI - after top dead center intake BDC BTDCF CAD CFD DISI E0I GDI HC - NOX PDPA PH PLIF PW RPM SMD SOI TDC bottom dead center before top dead center firing crank angle degrees computational fluid dynamics direct injection, spark ignition end of injection gasoline direct injection hydrocarbons oxides of nitrogen Phase-Doppler particle analysis port fuel injection - planar laser-induced fluorescence - pulse width revolutions per minute sauter mean diameter start of injection top dead center 9 LIST OF TABLES Table 2.1: Kiva-3V model boundary conditions.......................................................................27 Table 3.1: Optical access engine parameters............................................................................ 10 40 LIST OF FIGURES Figure 2.1: Com putational grid resolution. ................................................................................ 26 Figure 2.2: C om putational grid details...................................................................................... 26 Figure 2.3: Injector orientation relative to intake flow............................................................. 27 Figure 2.4: Fuel vapor distribution at ignition timing, showing injector, ignition, and fuel film locations in addition to stoichiometric boundary and centerline cross-section of mass fractio n . ................................................................................................................................. 28 Figure 2.5: 300 solid cone projected onto piston at 310' and 330' ATDCI with window area overlaid. 600 hollow cone also shown in side view for reference, with horizontal lines indicating the piston crown location marching up from 270' to 360' ATDCI. Top line above views shows droplet travel over 10' at injected velocity........................................ 28 Figure 2.6: Computed spray compared with LIF data under similar conditions.......................29 Figure 2.7: Liquid fuel film resulting from 450 injector orientation, showing film thickness of order 0.01mm with peak of 0.02mm and film spreading beyond visible window area. ....... 29 Figure 2.8a: Crank angle history of injected fuel mass from EOI to top center for lower piston temperature (350K). 9% of injected mass remains as film at 360 .................................. 30 Figure 2.8b: Crank angle history of injected fuel mass from EOI to top center for expected average operating piston temperature (400K).. 4% of injected mass remains as film at 360'. ............................................................................................................................................... 30 Figure 2.8c: Crank angle history of injected fuel mass from EOI to top center for higher piston temperature (420K). 1% of injected mass remains as film at 360 .................................. 30 Figure 2.9: Comparison of CFD spray development with high-speed video data for 300 SOI... 31 Figure 3.l a: Optical access engine diagram ............................................................................... 38 Figure 3.lb: Optical extension. ................................................................................................ 39 Figure 3.2: New Bowditch-type piston allowing optical access from bottom of combustion ch am b er.................................................................................................................................4 0 Figure 3.3: High-speed video and lighting arrangement.......................................................... 41 Figure 3.4: Wide-beam laser induced fluorescence (LIF) concept........................................... 41 Figure 3.5: Optics setup for laser beam expansion. .................................................................. 42 Figure 4.1: Pressure traces indicating cyclic variation and combustion stability.......................51 Figure 4.2a: Reference sketch showing view angle across piston crown.................................. 52 Figure 4.2b: Reference sketch showing view angle using mirror and optical access piston (note the reversal of injector and spark orientation in this view)............................................... 52 Figure 4.3a: Illustration of injected spray structure.................................................................... 53 Figure 4.3b: Illustration of spray interacting with liquid film (note: bottom view through mirror is not same scale as top iso view, and mirror view has been reversed from other data to match spray direction w ith top iso view ). ..................................................................................... 53 Figure 4.3c: Illustration of end of spray interacting with liquid film ........................................ 54 Figure 4.3d: Illustration of steady-state liquid fuel film coverage (note: bottom view through mirror is not same scale as top iso view, and mirror view has been reversed from other data to m atch spray direction with top iso view). .................................................................... 54 Figure 4.4: Cool and warm spray/piston interaction and combustion sequences (4 pages)..........55 Figure 4.5: Combustion in warm engine (82'C head) showing pool fire (3 pages).................. 59 11 Figure 4.6: Residue on piston imaged by Kodak Ektapro (high-speed video unit) in static engine. 62 ............................................................................................................................................... Figure 4.7: Residue on aluminum piston after stratified-charge operation...............................62 Figure 4.8: First spray impingement and combustion on cool piston (360K)...........................63 Figure 4.9: Film buildup after second injection on cool piston (360K), 360 . .......................... 63 Figure 4.10: Steady state spray impingement sequence and resulting fuel film coverage on cool p iston (360 K ).........................................................................................................................64 Figure 4.11: Bottom view of first spray impingement on cool piston (360K)...........................65 Figure 4.12: Bottom view at 720' ATDCI for first 5 spray impingement cycles on cool piston.. 66 Figure 4.13: Fuel film at 360' ATDCI for 4 th and 5 th impingement cycles on cool piston..... 67 Figure 4.14: First spray impingement and combustion on warm piston (400K)............68 Figure 4.15: Steady state fuel film coverage on warm piston (400K)........................................69 Figure 4.16: Residue on optical access piston after stratified charge operation........................69 12 1 INTRODUCTION 1.1 MOTIVATION While many new concepts for automotive propulsion are technically feasible today, such as electric, fuel cell, and various hybrid systems, the gasoline engine is still by far the most dominant source of power due to its combination of low cost and high power density. While the efficiency and emissions of the gasoline internal combustion engine have improved significantly since its invention, today's environmental concerns and the sheer quantity of engines in use still warrant constant efforts toward improvement. One concept that incorporates some benefits of diesel engines into the gasoline spark ignition engine is gasoline direct injection (GDI), or direct-injection spark ignition (DISI). As the point of fuel delivery has moved steadily closer to the combustion chamber, the control of fuel delivery and mixture preparation has improved. The next step from the port fuel injected (PFI) engine in showrooms today is to inject the fuel directly into the combustion chamber using high-pressure fuel injectors. This improves control of the amount of fuel delivered, charge cooling improves full load performance, and stratified charge operation at part load reduces both pumping and heat transfer losses. The stratified charge operating mode has produced the most dramatic improvements in efficiency, with claims of increasing fuel economy roughly 20% under normal operation [18, 19, 23, 24, 25]. As a result, GDI engines have been in production for several years in Japan and, more recently, Europe, where higher fuel prices contribute to a more significant portion of the operating costs of vehicle ownership. In the United States, however, tighter requirements on emissions have kept the GDI engine, which so far has exhibited higher hydrocarbon (HC) and oxides of nitrogen (NOX) emissions, out of showrooms. While lower fuel prices minimize consumer demand for the increase in efficiency, a finite global supply of petroleum and environmental concerns such as the greenhouse effect of carbon dioxide emissions do make GDI engines desirable in the longer term. It is therefore desirable to understand and improve the sources of emissions in GDI engines. Hydrocarbon emissions in GDI engines in stratified charge mode are believed to stem primarily from two sources (in homogenous operation, HC emissions can actually be reduced 13 during cold start thanks to more precise fuel metering [23]). The stratified charge preparation depends on mixing a finely atomized high-pressure spray just enough to obtain an ignitable pocket of mixture, surrounded by air or extremely lean mixture, at the ignition location. This mixing process necessarily results in richer regions within the charge pocket, and leaner regions at the periphery. Both of these regions are subject to incomplete combustion, which is the first source of hydrocarbon emissions. Incomplete combustion has been studied with conventional port fuel injection (PFI) engines. Most GDI designs, however, also target the fuel spray directly onto the piston crown (typically incorporating some type of bowl to direct charge flow), claiming it enhances mixing, vaporization, and control of charge motion. This intentional impingement of fuel onto combustion chamber surfaces is thought to result in liquid fuel films that are the second source of hydrocarbon emissions. The behavior of in-cylinder liquid fuel films and their contribution to hydrocarbon emissions has not been as thoroughly studied as incomplete combustion, although Stanglmaier and Li have shown that for HC emissions purposes, the piston is the second worst location- for liquid fuel [30, 31]. Of specific interest in this project is the formation and measurement of liquid fuel films on the piston crown resulting from late injection directly onto the piston during stratified charge operation. 1.2 BACKGROUND The concept of injecting fuel directly into the combustion chamber of a spark ignition engine only trailed the invention of the engine itself by a few decades. It is, however, only with the recent development of high-pressure fuel injectors and advanced engine control techniques that the advantages of direct injection have been realized. Pressure-swirl injectors in particular have proved ideal for use in direct injection applications, since they produce a finely atomized spray with suitable cone geometry. Techniques have also been developed to customize the cone geometry of pressure-swirl injectors for specific applications [33, 34]. Virtually every GDI engine in commercial production today uses two operating modes, and some occasional combinations of these two modes. The first mode is a simple extension of port fuel injection, and involves direct injection of enough fuel to produce a homogenous, stoichiometric mixture. Injection in this mode takes place during the intake valve open period in order to maximize mixing of the fuel with the entering air. Primary benefits over PFI include 14 greater control over the quantity of fuel delivered since there is no fuel storage in the intake port, and a charge cooling effect whereby the entering air is cooled by the evaporation of the injected fuel droplets. The charge cooling increases volumetric efficiency and allows the compression ratio to be raised as high as 12.5:1 [23], which in turn increases engine efficiency. performance increases on the order of 10% are typically quoted [17, 23, 24]. Peak The second operating mode has less in common with PFI engines, and more in common with diesel engines. Fuel injection takes place much later in the cycle, and the mixture is overall lean. Spark ignition, however, requires more mixing in order to obtain a roughly stoichiometric mixture by the spark plug, so injection is earlier than in diesel engines, though still quite late in the cycle, typically on the order of 60 degrees before top dead center firing (BTDCF). The greatest challenge in this mode has been obtaining a suitably mixed charge near the spark plug. Early designs injected the fuel quite close to the spark plug and attempted to ignite the spray as it exited the injector nozzle; this approach did not provide sufficient mixing, however, and combustion was poor and the spark plug prone to fouling [3]. The current approach is to inject the fuel spray onto some type of bowl in the piston crown, and also incorporate strong swirl or tumble airflow. The piston geometry, airflow, and spray momentum all contribute to the mixing and transport of the fuel, the end result ideally being a roughly stoichiometric pocket surrounded by air with minimal quantities of fuel. The primary benefits of this operating mode are reduced pumping and heat losses, and increased thermodynamic efficiency. Because the fuel is directly injected and the mixture is lean overall, throttling of the intake is reduced or entirely eliminated for part load operation. This greatly reduces pumping losses, which can be greater than brake power under many common operating conditions. Heat losses are reduced because an insulating region of mostly air, which has not participated in combustion, surrounds the burned charge. Finally, the thermodynamic properties of a lean mixture (the higher ratio of specific heats, y) result in greater efficiency of the conversion of fuel energy to useful work compared to a stoichiometric mixture [1]. Furthermore, various combinations of these two primary operating modes have been implemented to smooth the transition between them, reduce knock, and improve catalyst light-off [18, 24, 27]. Despite all these advantages, however, GDI engines are currently only in limited production, primarily in Japan and Europe where fuel consumption is more important and 15 emissions regulations are not as strict as in the United States. Various new catalytic converter approaches are being developed to improve the emissions [18, 24], but in stratified charge mode, where the greatest fuel consumption advantage lies, the overall lean mixture composition poses a significant challenge to catalyst operation. While all production GDI engines employ certain tactics to obtain a suitable stratified charge, such as high-pressure injection, piston crown impingement, and higher vorticity airflow, the details vary significantly from one application to the next. Mitsubishi, arguably one of the foremost proponents of GDI, has implemented an intake design that generates strong reverse tumble in a spherical piston bowl [24, 27]. This design is in relatively large production volume, but has the drawback of taller packaging requirements compared with a conventional engine, since the intake manifold extends up from the engine head rather than to the side. Other manufacturers have developed swirl systems with involute or other various piston bowl shapes [18, 23, 32]. Air-assist injectors have also been advocated by Orbital and others [14, 25], as the added mixing with air allows this type of direct injection to be retro-fitted to existing PFI engine designs. It has been hypothesized that the difference of these many approaches indicates a lack of complete understanding of the physics involved. The logic behind this argument is that if the problem were completely understood, all designs would gravitate toward the single best solution. The volume of research on GDI engines, particularly stratified charge preparation, also indicates the presence of unanswered questions. While many studies have been and are being performed on qualitative mixture preparation using planar laser-induced fluorescence (PLIF), few have focused on the interaction between the fuel spray and the piston crown. Even fewer have attempted to measure the liquid fuel film remaining on the piston. 1.3 OBJECTIVES The objectives of this project are to observe and describe qualitatively and, if possible, quantitatively, the interaction between a high-pressure direct injection fuel spray and the moving surface of the piston crown in a simplified but realistic GDI engine with optical access. Specifically, high-speed video imaging will be used to observe the fuel spray/piston interaction, and a laser-induced fluorescence technique will be designed to quantitatively measure the liquid fuel film thickness on the piston crown. Furthermore, CFD simulations of the experimental 16 setup will be run, and the results from experiment and simulation will be individually evaluated and compared to each other. The project results will increase the understanding of the fuel spray interaction with the piston crown, aid designers of GDI engines in the pursuit of optimal systems, and provide valuable feedback to the computational fluid dynamics community allowing programmers to tune their spray impingement and wall-film models. 17 (This page intentionally left blank) 18 2 SIMULATION 2.1 INTRODUCTION Computational Fluid Dynamics, or CFD, has grown from an esoteric research area into an everyday engineering tool in the last two decades. One primary enabler for this growth has been the increase of computing power and decrease of computing cost. Equally important has been the steady improvement of programming, allowing more geometrically accurate computational grids, numerous approaches to turbulence modeling, and improved solution convergence time and solution quality. Original CFD solutions typically involved the steady-state 2D flow around an airfoil cross-section, and took months of computing time on a mainframe computer. Today, complex transient 3D flows around detailed realistic geometry can be solved in a matter of days, sometimes using machines with no more power than the personal desktop computers steadily appearing in homes across the nation. Full heat transfer, moving grids, and multi-phasic flows are all routinely implemented. The value of CFD as an engineering tool is incredible. Given the geometry of a potential design, an engineer with suitable knowledge of the application (boundary conditions and material properties) can evaluate the design long before the existence of prototype hardware. Furthermore, greater insight into the peculiarities of the design's performance can be gleaned from CFD than from physical testing, since the complete flow field and various fluxes are inherent to the computed solution. However, despite these many advantages of CFD simulations, there are caveats. While any CFD software package can provide a solution, the solution will only be as good as the inputs that generated it. The program itself must be physically sound and robust, the material properties must be accurate, and the boundary conditions must be realistic and stable. Furthermore, to obtain an exact solution, direct numerical simulation (DNS) is required, which involves solving the Navier-Stokes equations with a grid fine enough to capture the smallest scale of turbulent eddies. For the typical flow around an automobile, for example, this would require a grid with many billions of cells, which cannot easily be solved even with today's formidable available computing power. For this reason, turbulence models have been developed which typically solve the Reynolds-Averaged Navier-Stokes equations, using various two- or three-equation turbulence models to account for the generation and dissipation of turbulence within the flow. While efficient and surprisingly 19 accurate, these turbulence models are not perfect and do not apply universally to every flow situation. For this reason, a sample solution is typically validated with physical data for a class of flows. Once the software and approach have been proven for one case, they can be applied with reasonable confidence to similar cases. 2.2 KIVA-3V MODEL Because CFD is a widely used tool in engine development, and no less so for GDI engine development, it was desirable to have CFD results to compare with the experimental data. Since few studies have attempted to validate the physics of spray-wall interaction under in-cylinder conditions, this comparison would provide invaluable data either validating the approach used or indicating required adjustments to the approach. Another reason for using CFD in this project was as a design tool for the experimental setup. The optical engine in question has been used at the MIT Sloan Automotive Laboratory for several decades, but it has only recently been converted to a GDI engine and then only used under homogenous operation with early injection. The engine's optical access design results in a square cross-section and flat piston and head surfaces, preventing exact implementation of any one of the current approaches to stratified mixture preparation. Furthermore, the concern existed that if one of the existing approaches was specifically adopted, the results would only be relevant for similar approaches. The variance of approaches has been discussed in section 1.2. For these reasons, a simplified approach to stratified mixture preparation was implemented, in order to make the results more generally applicable and less specific to any particular existing design. The unprecedented approach limited the usefulness of lessons learned from existing designs, however. CFD was therefore used to model the stratified mixture preparation before major engine modifications were made in order to obtain rough design specifications. At the same time, as mentioned, this type of application has not been well validated, so the engine modifications were purposely designed to allow flexibility in several parameters to fine-tune operation in stratified charge mode. Kiva-3V was selected as the CFD code, being a widely available research code with a long history in both direct injection (mostly diesel but recently gasoline as well) and port injection engine applications, and a functional built-in spray and wall-film model. The bulk fluid (gas) 20 solution is Eulerian, meaning that the grid is stationary for a given geometry and the fluid variables are solved for at each grid point. This is the approach used in virtually every CFD package currently available. The spray (liquid) solution, however, is Lagrangian, meaning that every material particle is tracked in space. This is necessary to accurately solve for the motion and interaction of the extremely small droplets produced by modem fuel injectors. In order to conserve computing resources, Kiva-3V lumps fuel droplets into parcels of droplets with identical velocity, position, and temperature, the total number of which can be specified to increase spray model accuracy or reduce solution time. In this manner, potentially millions of droplets can be modeled with only a few thousand parcels being tracked, and still provide reasonable accuracy. Typically the total number of parcels is in the range of 3000 to 5000; for this project, 8000 parcels were used to provide increased accuracy. Sub-models are included to allow the parcels to disperse, coalesce, and evaporate based on their position, velocity, size, and environment. The fuel model is single-component, which may not be entirely representative of gasoline in wall-impinging situations, but should serve the project well enough since experimental setups often use iso-octane doped with single fluorescing hydrocarbons. The wallfilm treatment includes a splash model, which takes into account the impinging droplet characteristics including Weber and Reynolds numbers. Based on this information, the droplet is either reflected (with a random adjustment to velocity to account for the momentum exchanged during the bounce) or contributes to the wall-film. Kiva-3V does not account for primary atomization, so the spray's Sauter Mean Radius (and other nozzle parameters such as cone angle) must be specified based on existing measurements. Finally, Saffman lift forces are not modeled, so the droplets' aerodynamic behavior does not vary as they approach the piston crown. The grid builtfor the Kiva-3V modeling contains a maximum of approximately 107,000 cells at bottom dead center. The collapsing and expansion of the combustion chamber (and lifting of valves) in Kiva-3V is treated by "snapping" layers of cells. The bottom layer of cells is distorted as the piston or valve moves for each time step, and once the distortion exceeds a specified fraction of the original cell height, the layer is removed or a new layer is added depending on whether compression or expansion is taking place. The grid resolution and detail are shown in Figure 2.1 and Figure 2.2. Only the intake port and valve are included in the model since combustion and exhaust are not simulated. 21 In addition to the geometry of the optical access engine, various boundary conditions must be prescribed for model simulations. The major parameters are listed in Table 2.1, and a typical full input deck is provided for reference in Appendix A. The details of the engine can be found in Chapter 3. 2.3 SIMULATION RESULTS KIVA-3V was run on a Silicon Graphics 02 workstation with an R10000 processor and 128MB of memory. Typical running time for a case resuming from 250' ATDCI, simulating injection and all fuel modeling, and completing at 360' ATDCI, was on the order of 3-4 days. Post-processing was executed on the same machine using GMV. Several scenarios were simulated to determine the operating window, if one indeed existed, for late injection, stratified charge operation of the optical access engine. As discussed in Chapter 1, typical production GDI engines employ various piston bowl geometry and charge vorticity effects to mix and transport the fuel to the point of ignition. The hypothesis for the optical engine setup was that with the injector and spark plug located on opposite sides of the combustion chamber, as opposed to a centrally located plug as in most production engines, the injected fuel would travel farther before the point of ignition and mix sufficiently by diffusion and the momentum of the spray itself, which would entrain substantial quantities of air. The initial mixture preparation strategy was based on typical production injection timing, modified to take the extra distance and spray's initial velocity into account. A few iterations were necessary to fine-tune the strategy, but the CFD model predicts that an operating window does indeed exist for stratified charge mode with the optical engine. The intake stroke generates a mild forward tumble flow, which persists despite the corners in the combustion chamber, which tend to be parasitic to both tumble and swirl. The tumble becomes more pronounced during the compression stroke, when injection takes place. The most suitable injection timing was found to be 290' after top dead center intake (ATDCI), or 700 BTDCF, which is roughly 10' earlier than typical production designs. At this point in the compression stroke, the tumble results in a mild flow across the piston crown surface, perpendicular to the injector orientation. This presumably enhances the mixing of the fuel, and also results in the center of the fuel mass distribution being slightly off-center once the fuel vapor 22 reaches the opposite side where the spark plug is located. Figure 2.3 illustrates the orientation of the intake flow relative to the injector and ignition locations, and Figure 2.4 shows (from the same isometric viewing direction) the resulting fuel vapor distribution at 350' ATDCI, roughly when ignition would take place. Significant deviation of injection timing and geometry results in under- or over-mixing of the fuel vapor, or unsuitable locations of the ignitable mixture. The fuel injector mounting angle was simulated at 450 and 30' inclined from vertical (the previous spray behavior studies utilized a vertical injector mounting, since operation was primarily homogenous). The rough orientation of the injector was based on the geometry of the engine, positions of the injector and spark plug, and estimates of the fuel vapor transport as well as the necessity to produce a liquid fuel film within the region of the piston crown observable through the optical access design. Figure 2.5 shows the projection of a 300 solid cone onto the piston crown at 310' and 330' ATDCI. Within the limitations of fuel vapor transport, this simple geometric study indicated that a narrow spray, later injection, or a lower orientation (such as 300 from vertical instead of 450) would be necessary to result in impingement within the observable area of the piston crown. Initial simulations of the late injection indicated a strong collapse of the hollow cone, primarily dependent on the ambient density (or pressure). This trend is supported by several experimental studies [3, 21, 32, 35] including VanDerWege's previous work using the optical engine, and is an improvement in spray modeling over previous versions of KIVA [21]. The cone angle collapse is a result of the spray's momentum entraining the bulk fluid; with increasing ambient density, the entraining flow has a more pronounced effect on the spray droplets. Also, the smaller droplets of hollow cone injectors are more strongly effected by ambient flows. Figure 2.6 shows a calculated spray compared to LIF data from VanDerWege under similar conditions, confirming the spray results generated with the CFD model. It is worthwhile to notice in the CFD prediction that the majority of droplet parcels closer to the axis are smaller, having partially evaporated and thus becoming more susceptible to the entrained airflow. The collapse of the spray under the higher pressures present during late injection provides the small effective cone angle required to produce impingement in the area visible through the piston window. The only potential concern was that at the 450 injector orientation, a fraction of the resulting liquid fuel film would travel as a result of momentum from the impingement 23 location to an area beyond the visible area, as shown in figure 2.7. For this reason, a 300 injector orientation was also simulated. The resulting liquid film was indeed more centrally located, but the reduction of the spray's horizontal momentum was enough to prevent effective vapor transport to the spark plug. For this reason, the 450 injector mounting was chosen. 2.4 LIQUID FUEL FILM ON THE PISTON As mentioned earlier, one of the major advantages of CFD simulations is the ability to examine the resulting solution in many different ways. For example, while the liquid droplets collapse into a narrower angle that more closely resembles a solid cone, the majority of the injected fuel remains at the outer regions of the spray in vapor form. The rapid vaporization of the fuel is one of the advantages of pressure-swirl type injectors, and is a by-product of the extremely small droplets produced. Kiva-3V provides information in the solution files listing the mass of fuel present as liquid droplets, vapor, and wall films. Plotting this data versus crank angle illustrates quite clearly that under typical operating conditions, by the end of injection the majority of the fuel has already vaporized. As the spray travels through the hot, compressed air in the combustion chamber, it continues to vaporize rapidly. Some droplets do survive long enough to impinge on the piston, however. The piston temperature is expected to be on the order of 400K during typical operation, while the Leidenfrost temperature (the point at which vapor production prevents wall wetting; above this temperature droplets hover on a cushion of their own vapor until they completely evaporate) of gasoline even at room temperature is closer to 600K [36]. Therefore, the droplets will form a liquid film on the piston. The piston temperature is, however, a major variable in determining what fraction of the liquid fuel remains as a liquid film. It should be reiterated at this point that this behavior is a result of the Kiva-3V sub-models governing droplet impingement on walls and wall-film mechanics, and not necessarily representative of the actual physics. It is instructive, nonetheless, to observe the variation in fuel mass remaining as liquid film on the piston after spray/wall interaction. Figure 2.8 shows the crank angle history of injected fuel mass from the end of injection to top center for three different piston temperatures, breaking the mass down into liquid droplets, vapor, and liquid film on the piston. The fuel mass history clearly illustrates that cooler piston temperatures result in more fuel remaining as film; as the temperature is raised from 350K to 400K and then 420K, the 24 fraction of injected fuel remaining as film is reduced from 9% to 4% and then 1%. Closer examination of the history also reveals that not only does the fuel evaporate more quickly from the hotter surfaces, but also a smaller fraction is likely to remain on the surface following the initial "splash." It should also be mentioned that combustion and exhaust were not simulated, and that the fuel film formed is representative only of a single injection onto a dry piston. In reality, combustion and exhaust will diminish the liquid film mass (liquid fuel on the piston is thought to contribute directly to HC emissions [30, 31]), and each successive cycle will add a small fraction of the injected fuel mass to the film. Potentially a steady-state equilibrium could be reached, although cooling issues and varying fuel component volatility (the heavier components would be more prone to remaining as film) place this beyond the scope of the current project. 2.5 COMPARISON WITH EXPERIMENTAL RESULTS As mentioned earlier, while one motive for using CFD was as a design tool for the experimental setup, another reason, which is perhaps more important in the long term, was to compare the CFD simulations with actual experimental data. High-speed video imaging provides an excellent qualitative fuel droplet distribution picture which can be compared almost directly with KIVA-3V results. Figure 2.9 follows the development of the spray and the resulting fuel film as predicted by CFD and observed with high-speed video on a dry 400K piston. The CFD simulation is for a start of injection at 300' ATDCI and a pulse-width of 3.3ms, or 20', while the experimental data uses a pulse-width of 4ms, or 240, in order to fire with more stability. The only other difference is a slightly lower intake pressure for the experiment; the CFD simulation used an intake pressure of 0.99bar while the experimental setup, though unthrottled, was 0.9bar. Exact matching of spray development time is difficult since the CFD data was written every 10' while the highspeed video timing results in frames every 30, so the high-speed images closest in terms of injection timing are shown. Despite this, the qualitative features of the developing spray and impingement region are in excellent agreement. Penetration rates, which can be quantified by the time for the tip of the spray to reach the piston crown (just over 10'), are also in good agreement. Only the fuel film area after spray interaction is underpredicted by CFD. 25 Figure 2.1: Computational grid resolution. Figure 2.2: Computational grid details. 26 Boundary Condition Ambient Temperature Intake Manifold Pressure Initial In-Cylinder Temperature Cylinder Wall Temperature Piston Temperature Injected Droplet Sauter Mean Radius Injected Droplet Initial Velocity Ambient Injector Cone Angle (Outer) Ambient Injector Cone Angle (Inner) Start of Injection Injection Duration Table 2.1: Kiva-3V model boundary conditions. End of injection, 310ATDC Tracers radius (walls colored by fuel mass fraction) W 0-00325 -0.00293 -0.00261 -0.00228 -0.00196 -0.00164 -0,00131 -0.000988 (8 microns) -0.000665 2 i-0.0003 Q F1.82e-05 Figure 2.3: Injector orientation relative to intake flow. 27 Default Value 293 K 99 kPa 293 K 350 K 400 K 8 pm 80 m/s 500 250 2900 ATDCI 200 CAD (3.3ms pw) 0.03g fuel injected from 290-310ATDC onto 4001 piston Fuel distribution at 350ATDC favorable for ignition Global AFR = 38 Stoichiometric boundary Fuel Mass Fraction 0.159 Rich 0.143 -0.127 Spark Ocatlon -0.111 Lean -0.0952 1 -0.0793 or location Lean -0.0476 -0.0159 Figure 2.4: Fuel vapor distribution at ignition timing, showing injector, ignition, and fuel film locations in addition to stoichiometric boundary and centerline cross-section of mass fraction. Figure 2.5: 300 solid cone projected onto piston at 3100 and 3300 ATDCI with window area overlaid. 600 hollow cone also shown in side view for reference, with horizontal lines indicating the piston crown location marching up from 270* to 360' ATDCI. Top line above views shows droplet travel over 100 at injected velocity. 28 60 (25 hollow) cone, Vinj=80 mis, halfway through 290-310 injection duration or 1.67ms ASOI (0.016 of0.03 grams) ( V 14a Tracers LIF data, VanDeawege widar cone, eariler 1nJ.) (cm) radius .0027 .00244 -0.00217 -0.0019 - -0.00163 o -0.00137 -0.0011 a*"e a -0.000831 <-81microns I'-0.000563 .000295 2.78e-0S Figure 2.6: Computed spray compared with LIF data under similar conditions. 0.03g fuel Injected from 290-310ATDC 400K piston 350ATDC (no ignition) Film Height (cm) 00 2 0. Window outline superimpcsed (0.02mm) .0018 -0.0014 .0012 .001 (0.01mm) .0008 -0.0006 -0.0004 -0.0002 M- 0 Figure 2.7: Liquid fuel film resulting from 450 injector orientation, showing film thickness of order 0.01mm with peak of 0.02mm and film spreading beyond visible window area. 29 Fuel Distribution v. Crank Angle, Tpiston=350K 3.OQE-02 2.50E-02 2.OOE-02 J c 1.50E-02 5.OOE-03 O.OOE+OO4" 0.00E+00 31ATDC 320ATDC 340ATDC 330ATDC 350ATDC 360ATDC Figure 2.8a: Crank angle history of injected fuel mass from EOI to top center for lower piston temperature (350K). 9% of injected mass remains as film at 360'. Fuel Distribution v. Crank Angle, Tpiston=400K 3.OOE-02 2.50E-02 2.OQE-02 E 1.50E-02 1.OOE-02 5.OOE-03 O.OOE+00 31 OATDC 320ATDC 340ATDC 330ATDC 350ATDC 360ATDC Figure 2.8b: Crank angle history of injected fuel mass from EOI to top center for expected average operating piston temperature (400K). 4% of injected mass remains as film at 360'. Fuel Distribution v. Crank Angle, Tpiston=420K 3.OQE-02 2.50E-02 2.OOE-02 1.50E-02 Vapo 1.OOE-02 5.OOE-03 O.OOE+00 31 OATDC 320ATDC 330ATDC 340ATDC 350ATDC 360ATDC Figure 2.8c: Crank angle history of injected fuel mass from EOI to top center for higher piston temperature (420K). 1% of injected mass remains as film at 3600. 30 Figure 2.9: Comparison of CFD spray development with high-speed video data for 300' SOI. 310 3200 30* 360 340 31 (This page intentionally left blank) 32 3 EXPERIMENTAL SETUP 3.1 INTRODUCTION The heart of the experimental portion of this project is the optical access engine, which has been modified over the years to shed light on any number of internal combustion engine phenomena. Most recently it has been operated as a homogenous charge direct injection engine with optical access on three walls. To implement stratified charge operation and a view of the combustion chamber from below, several modifications were made to the engine. High-speed video imaging was used to qualitatively study the entire spray/piston interaction. A wide-beam form of laser-induced fluorescence is also designed to measure the fuel film thickness quantitatively, although equipment limitations placed validation of this technique beyond the timeframe of this project. In the future, planar laser-induced fluorescence and phase-Doppler particle analysis (as in [3]) could also be used to further explore stratified GDI sprays interacting with the piston in this experimental setup. 3.2 OPTICAL ACCESS ENGINE The optical access engine consists of an optical extension added to a Cooperative Fuels Research (CFR) Engine, which is a single-cylinder engine with the capability of adjustable compression ratio. An illustration of the optical access engine is shown in Figure 3.1, and the engine specifications are listed in Table 3.1. The square cross-section optical extension allows complete optical access through two opposed quartz walls and limited optical access on a third wall through a quartz window. The quartz walls and windows are sealed with sheet silicone rubber gaskets attached with high-temperature RTV. A fourth mode of optical access was added for this project, by designing and constructing a Bowditch-type piston which allows optical access through the piston via a 450 mirror and a quartz window held in the piston crown with two-component high-temperature silicone RTV adhesive (Dow Coming Sylgard* 275). The two-component RTV cured properties (elastomeric) allow it to absorb differential thermal expansion of the quartz and aluminum, unlike the fairly rigid properties of high-temperature filled epoxy products. Figure 3.2 illustrates the new optical access piston assembly with a photograph of the major components, and complete drawings for the piston components are 33 included in Appendix B. A UV-coated mirror is fixed with epoxy to a supporting plate, which is held in the piston with 6 machine screws. A thermal stress analysis was conducted for the mirror and plate assembly since the mirror has poor tensile properties; the results indicated that the mirror was at risk of fracture should the operating temperatures rise from the equilibrium temperature by more than approximately 20'C, depending on the cured epoxy material properties. Cooling below equilibrium was not expected to pose a problem, since the higher coefficient of thermal expansion of the steel plate would place the mirror under compression rather than tension. Therefore the mirror and plate assembly was cured at roughly 75'C, so that typical operation would result in temperatures cooler than the assembly equilibrium point. As an additional precaution the epoxy used (Emerson & Cuming Eccobond® 45/Catalyst 15) was selected for high-grade adhesion and strength yet semi-rigid cured properties in order to both provide positional stability for the mirror and absorb some of the thermal stresses. A thermocouple was also embedded in the exposed face of the quartz window using the twocomponent high temperature RTV (which has a thermal conductivity similar to quartz) in order to obtain a piston temperature measurement. The thermocouple location is schematically indicated in Figure 3.4. Wiring for the thermocouple was embedded in the RTV holding the window, and then run through small holes drilled in the piston behind the mirror. A quickdisconnect was provided in the cavity behind the mirror in order to easily replace the segment of wire subject to reciprocating motion. Sealing of the square cross-section is accomplished with three sets of overlapping graphite bars, which provide a sliding seal without oil lubrication. Teflon bars have been tested previously but were found to leave a residue, which not only impedes visual access but fluoresces under UV laser excitation. The limitations of sealing with graphite bars and silicone gaskets prevent compression ratios significantly higher than 8:1. The engine is uncooled, and can therefore only be run for a period of several minutes before the silicone gaskets sealing the quartz walls are compromised. Engine speed is controlled with an AC electric motor with the capability to provide brake load, although the optical access engine experiences such high friction that the motor rarely needs to absorb energy. Based on the CFD simulation results, a 50' hollow-cone fuel injector with nominal flow rate of 9.2 g/s was selected (Siemens Automotive SN 33623, measured cone angle 56' and 34 measured flow rate 9.14 g/s) and mounted at 45'. Parameters were matched with one of the injectors provided to Sandia National Laboratories to facilitate future comparisons. An insert was necessary to mount the injector at an angle; a drawing is provided in Appendix B. Fuel injection pressure was 5Mpa, provided by a hydraulic accumulator (Tobul 3AT30-2-S) and highpressure nitrogen. A more thorough description of the fuel system and filling procedure can be found in [3]. 3.3 HIGH-SPEED VIDEO IMAGING High-speed video was used in order to obtain qualitative imaging of the spray-piston interaction event without the limitations of PLLF, which would restrict the view to a single crosssection as well as bringing into play the various fluorescence variables discussed in the following section. The extensive modifications to the engine allow excellent optical access for both lighting and camera viewing angles, making it a suitable candidate for high-speed video in addition to LIF. A Kodak Ektapro system was used, which contains enough memory to store 2,300 images, each containing 192x239 8-bit pixels (256 shades of gray), from its NMOS array in 2.3 seconds. This provides 1.Orms temporal resolution, which can be increased by storing images with half resolution or less. At 1000 RPM this translates to 6' of crank resolution for full size images and 30 resolution for half-height images, which were the two frame sizes used. A 90mm f/2.5 camera lens was used with the digital array. In order to obtain adequate lighting with minimal direct reflection for the extremely short exposure times, a 600W halogen light was arranged as shown in Figure 3.3. Some footage was also taken through the new piston window, using the mirror. Similar component spacing was used for this setup (the camera distance was reduced by approximately 40cm), and the camera was simply placed on the other side (exhaust) of the engine and aimed at the mirror. 3.4 WIDE BEAM LASER-INDUCED FLUORESCENCE Planar laser-induced fluorescence (PLIIF) has been used extensively in the past few years to obtain qualitative images of fuel vapor and liquid distribution in various stages of mixture preparation. With optical access, a sheet of laser light with suitable wavelength is introduced to a 35 cross-section of interest, and a gate-intensified CCD camera observes the fluorescence normal to the sheet of light. To excite electrons of the molecules typically representative of fuel, UV laser light is required, although the fluorescence is typically in the visible region. Gasoline contains heavier aromatics which fluoresce readily, but for more consistent results a non-fluorescing base such as iso-octane is typically doped with known fractions of fluorescing compounds representative of the distillation fraction of interest, such as acetone (Tb = 56'C, roughly Tio), 3pentanone (Tb = 102'C, roughly T5 0 ), or cyclohexanone (Tb = 155'C, roughly T 90). Drawbacks of this method include the need for optical access along two directions (one for the laser sheet, and one for the camera view) and high energy UV laser manipulation, but most significant is the fact that, although fluorescent yield is in general proportional to the mass of a compound present, fluorescence images are semi-quantitative at best. Much work is in progress (see, for example, [22] and [37]) generating calibration data in order to obtain more quantitatively accurate results from PLIF, but the varying intensity of excitation energy, dependence of fluorescent yields on temperature and pressure, and collision quenching by oxygen and other compounds still makes accurate calibration extremely difficult [3]. A new method that provides in-situ calibration for certain types of measurements is a wide-beam laser-induced fluorescence, which is described in this project. The new method of LIF measurement requires that the compound to be studied exist as an optically thin film on a surface with optical access. In this case, the objective is to measure the liquid fuel film on the piston, so the piston requires optical access. Excitation energy is introduced through the optical access, and the film fluoresces proportionally to its thickness [12]. The film must be optically thin, since increasing thickness of the film eventually extinguishes the excitation light and produces a fluorescence signal proportional to the area rather than the thickness of the film. Liquid phase LIF is much stronger than vapor phase, and is therefore expected to dominate [20]. The key to providing an accurate quantitative interpretation of the image, which is viewed through the same optical access as the excitation light is provided through, is in-situ calibration using grooves etched or milled into the surface of the window. The depth of the grooves must be of the same order of magnitude as the film thickness, and the width must be sufficient to provide a distinguished region in the resulting CCD image. The assumption must be made that the film will maintain a uniform surface, resulting in a band of increased 36 thickness over the groove. This will result in a stripe of brighter fluorescence, and the step increase in brightness will be proportional to the depth of the groove. Using this information, the thickness of the film can be calculated everywhere it is visible in the CCD image. Figure 3.4 shows a schematic of the wide-beam LIF concept (with a calibration groove expanded out of scale for illustration). Since the CFD results predict a film thickness on the order of 0.01mm, the grooves must be approximately this deep. While machining an exact depth is not critical, is it necessary to know what the final depth is in order to calibrate the images. Accurate resolution of the groove width with typical CCD cameras requires the grooves to be on the order of 1-2mm wide. Finally, the laser light exciting the fluorescent compounds must be expanded to the same cross-section as the window area to be studied with LIF, which is a 2" square in this case. The laser beam exits the laser as a rectangular beam 23mm high and 7mm wide. Two sets of lenses are necessary to expand this to ja minimum of 50.8mm square, as illustrated in Figure 3.5. Apertures are also included in the optic setup to contain divergent laser beams. Finally, a longpass filter, which reflects UV light between 260nm and 320nm (the laser light is 308nm), is positioned to reflect the beam toward the optical access engine. Since the resulting fluorescence is visible (above 320nm), a camera placed on the opposite side of the long-pass filter from the engine will detect the fluorescence signal through the filter. At the time of this writing, the optical access piston assembly and laser beam expansion optics setup had been completed, but a suitable gate-intensified CCD camera with controlling and timing hardware was not available to conduct a proof-of-concept run of the wide-beam LIF concept. It is anticipated that subsequent work using the same facilities will continue work on this diagnostic strategy. 37 ; S 1~ sta ndard head ass embly Co mbustion chamber $0 jare piston :0 0' 0 - E 42o wi ndow ndard piston sta ndard single inder engine embly Figure 3.la: Optical access engine diagram. 38 -- - Head Clamping Wall Weld Piston Rings Aluminum Square Piston -- Quartz Glass Walls 1 -- - - - Split Connecting Rod Weld Teflon 0-ring --- - Adaptor Base Plate n CFR Piston CFR Cylinder Figure 3.1b: Optical extensioi 39 I Bore Stroke Value 82.6mm (square) 114.3mm Displacement Valve Train Head Piston Compression Ratio Connecting rod length 0.77 liter 2 valves, push-rod driven Flat Flat 8:1 254mm Intake Valve Open Intake Valve Close 50 BTC 550 ABC Parameter 480 BBC Exhaust Valve Open 120 ATC Exhaust Valve Close parameters. Table 3.1: Optical access engine 4q zmr Aluminum piston block MIVN BIDNG REV D~liR11DN DIR. AfrDED Graphite bar seal lands Epoxy Mirror Mirror support 4- Aluminum piston block Connecting rod OPTICAL PISTON SLOAN LABORATORY R. Karlsson " ... 11/20/1999 PISTON ASSY """a'OP-000 sR.&1 OF 1 Figure 3.2: New Bowditch-type piston allowing optical access from bottom of combustion chamber. 40 4 --- - A~m - Camera - - Ka Intake side Exhaust side -20cm comb. chamber 600W lamp - Tripod Figure 3.3: High-speed video and lighting arrangement. Calibration the Combustion Chamber (cl ) Fuel Film Quart-Window Fue- Film Quartz Graphite Seal Lands From Laser and To Camera Piston (Modified) Connecting Rod Figure 3.4: Wide-beam laser induced fluorescence (LIF) concept. 41 UV UV irror 7x7mm beam mirror 1~ C.. 33mm sph. lens Aperture :mm Cyl. lens 250mm sph. lens Aperture - - 55x53mm beam Plan view of engine 7 5mm C:yl. * intake Exhausi n0 - Laser light reflected into engine Digital lens 23x7mm beam mera Visible fluorescence passed to ca mera Djecto2R7 Long-pass A"- filter Figure 3.5: Optics setup for laser beam expansion. 42 Laser Lambda Physik Compex 102 XeCI 308nm 4 EXPERIMENTAL RESULTS 4.1 HIGH-SPEED VIDEO RESULTS USING ALUMINUM PISTON During construction of the optical access piston, high-speed video imaging was conducted using the older solid aluminum square piston. While the CFD simulations indicated a combustible mixture resulting from injection starting at 2900 ATDCI using a pulse-width of 3.3ms, stratified combustion was found to be more stable using approximately 300' SOI and a 4.Oms pulse-width. This resulted in a global relative air-fuel equivalence ratio of X=2, as compared to the CFD simulations which resulted in X closer to 2.5. Based on high-speed video imaging taken during setup of the injection and spark timing, the primary reason for this was ambient density in the combustion chamber. The KIVA-3V model was initially run with an intake pressure of 0.99bar and does not model any type of leakage from the unusual sealing techniques used in the optical access engine. The intake manifold pressure, while unthrottled, was in fact monitored at 0.9bar, and some leakage past the graphite bar seals is inevitable. As a result of these two facts, the slightly earlier injection resulted in a wider cone than predicted, since the ambient density was not high enough to collapse the spray cone angle. The fuel vapor resulting from this geometry was too highly dispersed to form an ignitable mixture. By injecting 10' later, which is actually closer to the average SOI for production stratified GDI operation, the ambient density had increased enough to collapse the cone angle and provide a more localized fuel vapor distribution. Increasing the pulse-width provided a small increase in the amount of fuel, further improving the local fuel mass fraction. Despite the 10' delay in injection, spark timing was only retarded to 5' BTDCF. This may be due to a higher spray velocity than estimated for the CFD simulations, which in turn could be a result of the lower ambient density (and newer fuel injector than used for the estimate), but may also have been in part a result of the higher quantity of fuel injected. Amoco Indolene High Octane Motor Fuel (brand code 15211), a multi-component petroleum fuel formulated like gasoline but tested to meet certain property restrictions, was used for the initial high-speed video imaging to capture as accurately as possible the behavior of a real-world fuel spray. Figure 4.1 shows 50 pressure traces overlaid to illustrate cyclic variation and stability of combustion. Some non-physical noise is present in the signal from the pressure transducer, 43 which is somewhat dated. Peak pressure is observed to occur at approximately 385' ATDCI, which is 10' later than expected for optimized ignition timing, confirming that the ignition timing is significantly retarded for stratified combustion due to the geometric constraints of the optical access engine. Combustion variability is present, as expected for the unusual geometry and retarded ignition, but is not overly excessive. These pressure traces were taken sequentially in a "medium-warm" engine, with a piston temperature estimated at 380K. The engine will not fire with stability when cold (room temperature) since impingement on the cold walls and piston prevent sufficient vapor production, despite the pressure-swirl injector's highly atomized spray. Combustion stability generally improved as the engine warmed up. While stratified combustion did prove both feasible and reasonably stable, the combustion chamber geometry did not contain the fuel-air mixture in any way. The flame was also highly visible through the quartz walls and yellowish, indicating rich regions and soot production. This under-mixing may be ameliorated in production GDI designs by the higher charge vorticity. The result in the optical access engine, however, was a gradual build-up of soot on the windows (and other surfaces) in addition to the slight residue left by the graphite seal bars. Frequent disassembly and cleaning was found to be necessary to maintain suitable optical access. The collapsed cone angle is clearly visible in the footage, with the fuel droplet pattern in excellent agreement with existing LIF studies and the CFD predictions as shown in Figure 2.9. Evaporation of the spray was difficult to quantify, but it appears that evaporation did not take place as rapidly as the CFD simulations predicted. Droplets are highly visible due to light scattering, while fuel vapor appears in the footage as a slight blur. No droplets were observed to leave the piston; rather all impinging droplets contribute to and interact vigorously with a thin liquid film on the piston, which in turn produces sufficient vapor to allow stratified combustion. The liquid fuel film is observed to persist after combustion. Once the engine is warm, combustion appears more vigorous in the footage, presumably due to enhanced vapor production on the hotter piston. Pool-fires are also clearly visible once the engine is warm enough for the piston to continue vapor production, after the spray interaction, at rates comparable to its combustion. Figure 4.2a shows a reference sketch of the viewing angle used for the high-speed video imaging results discussed in this section. Tumble flow across the piston is away from the viewer, 44 or from intake toward exhaust. Figure 4.3 shows four illustrations, also for reference, pointing out key characteristics of the following images. For example, the collapsed cone structure exhibits a darker core region where the entrained air flow has created sufficient droplet density to reduce lighting, while the periphery of the cone is not as dense and therefore much lighter. The vortex structure generated by the entrained flow at the leading edge of the spray is visible from approximately 309' to 318' ATDCI, depending on the lighting (for example, in the warmed up case with the aluminum piston the increase in scattered light drowns out the vortex). Spray interaction with the liquid fuel film is typically visible as a shadowy pattern within the film in the top iso views, while the bottom view is better for viewing secondary breakup and droplet production, which is visible as increased scattered light in these views. Figure 4.4 shows two sequences of the spray/piston interaction followed by combustion during steady state operation, one for a relatively cool engine (55'C head, estimated 350K piston) and one for a warmed up engine (82'C head, estimated 400K piston). Half resolution was used for these sequences to allow a frame rate of 0.5ms, or 3CAD apart. To save space, only select frames have been included from the combustion sequence, since little of interest is visible for the cool engine combustion. Note, however, the small lighter streaks propagating away from the spark location in frames 387O-408' for the cool piston; these are believed to be soot particles generated from rich pockets or possibly residual fuel droplets. They are burned in the premixed flame as it propagates through the combustion chamber. The pool fire drowns out any such features for the warm case. Figure 4.5 shows a sequence of combustion only for the warmed up case using full resolution (1.Oms frame rate, or 6CAD apart) to more completely capture the vigorous combustion and pool fire, which appears to persist for at least half of the expansion stroke. The brighter overall image quality of the warm case may be due to increased vapor production, but it is also at least in part a result of a cleaner piston. This imaging session was the first time stratified operation was attempted in the optical access engine, and the original square piston was nearly black from deposits of prior operation. During the course of the session, the repeated impingement of Indolene near the center of the piston actually cleaned the surface, increasing the reflectivity of the piston crown and adding to the light transmitted to the camera. Figure 4.6 shows a static image taken after several runs, with the intake side quartz wall removed, using the high-speed video camera (the piston and camera angle have been lowered to provide a better 45 view of the piston crown). The impingement zone is clearly visible as a bright region on the piston, which has literally been washed by the fuel. Figure 4.7 shows a photograph of the piston removed from the engine after the entire session. An edge-filtered copy of the photograph has also been included with the deposit pattern overlaid for reference. The lighter region near the center of the piston crown is a relatively clean surface, where previously existing deposits have been rinsed away by repeated fuel spray impingement. Surrounding this region are several rings of new deposits left by the fuel film, possibly comprised of partially oxidized heavier components in the fuel, since they are less prone to evaporation and combustion. Lighter spots toward the periphery of the new deposit layers are not "clean", as the center is, but merely lighter colored deposits. It is interesting to note the remarkable similarity between the fuel film area, marked by the rings of residue, and the simple projection of a cone onto two surfaces in Figure 2.5, which was performed with a pencil and paper in a few minutes before any CFD or experimental work was done. Notice also that as the fuel film propagated away from the impingement site toward the ignition location, lateral momentum appears to have been added by mild charge tumble exactly as predicted by the CFD simulations. This is evidenced by the spreading of the deposit rings (which mark the edge of the fuel film) away from the viewer in Figures 4.6 and 4.7. 4.2 HIGH-SPEED VIDEO RESULTS USING OPTICAL ACCESS PISTON Once the optical access piston was completely assembled, high-speed video imaging was conducted using it in place of the existing aluminum piston, since the quartz window allowed more flexibility of lighting and viewing angles, and the thermocouple embedded by the window provided a temperature measurement of the piston during operation. Operation was found to be similar to that using the solid aluminum piston. Some minor differences included a much more rapid rise in operating temperatures, particularly the measured piston temperature, due to the poor thermal conductivity of the quartz window. Once a stable operating temperature was reached, continuous firing brought the piston temperature from roughly 50'C to 1500 in less than 90 seconds. For this reason run times were quite short, on the order of one minute, though with some practice capturing a particular thermal state on video did not prove too difficult. Injection timing also proved to be quite critical to stable operation, with piston temperature being an input 46 variable. Lower piston temperatures resulted in slower vapor production and therefore needed SOI advances of between one and three CAD, while with a fully warmed up piston (400K or above) SOI timing of 3000 ATDCI could be used. The most universally stable SOI timing was found to be 298 -299' ATDCI. Finally, the quartz window reflects light at low angles of incidence, so the halogen lighting used to illuminate for high-speed video had to be adjusted slightly to one side to prevent direct reflection, which would overpower any other imaging. Footage was taken at two thermal states, similar to those estimated for the aluminum piston, one for a cool "early warm-up" piston temperature of 360K and one for a warmer piston temperature of 400K, which is more typical of the expected steady-state operation piston temperature expected for production GDI engines. Again, since the quartz window is such a poor conductor, the piston temperature rose rapidly. A typical run consisted of ensuring stable operation, shutting off injection, then capturing injection again after several "dry" cycles to allow existing fuel film to evaporate. Even in rapid succession, these steps allowed the piston temperature to rise to 360K. The warmed up 400K piston was reached in less than one minute. Figure 4.8 shows spray impingement on the cool piston followed by combustion for the "first" cycle, meaning that the engine was motored for at least 10 cycles prior to this injection to allow any liquid fuel on the piston to evaporate. At 360' ATDCI all of the spray has interacted with the piston, and the mixture above the piston has been sparked just under ims earlier (3550 ATDCI) so the flame is not yet visible from this angle. Since the velocity of the piston is briefly zero at this point, the image is quite clear and shows the extent of the liquid fuel film after the first impingement. The visible fuel film area is significantly larger than predicted by the CFD simulations, as shown in Figure 2.7, although the general shape and location are in qualitative agreement. Figure 4.9 shows the same top center image after the second spray has impinged on the piston crown, showing some growth of the liquid film. From the high-speed video data, it is apparent that qualitative equilibrium is reached after approximately five cycles. Figure 4.9 shows spray impingement and the resulting fuel film on a cool piston after more than five cycles. Note the increase in lighting in this sequence compared to the "first" impingement, despite the fact that the lighting arrangement, including camera aperture setting, was not changed between these images. The reason for the increase in light is an increase in droplet breakup and vapor production. The droplets of the spray interact with the thin fuel film already existing on the 47 piston in this case, producing secondary droplets as they splash down. These secondary droplet splashes and the added vapor production resulting from them add to the scattering of light, the result being a more well lit combustion chamber. The added fuel vapor is also visible as a general blurriness of the images leading up to top center. At the top center position the extent of the fuel film on the piston is once again clearly visible, and increased significantly from the first two cycles. One revolution later, during the valve open period, the fuel film is still visible and appears the same size. This is also illustrated in Figure 4.10, with the label 7200 indicating one full revolution after top center firing. Note in this image that the lighting has decreased to the level of the first spray impingement, since there is no spray or production of secondary droplets during the valve open period. This image proves that the fuel film survives after combustion, even pool fire. After five cycles, the area covered by the liquid film does not appear to change significantly. It is impossible to determine at this point, however, whether the film thickness or composition is changing after five cycles. It is likely, for example, that for each spray impingement the lighter components of the fuel are more quickly vaporized and combusted, while the heavier components are gradually transported toward the periphery of the fuel film area, where they may be partially oxidized into the darker residue rings found on the piston after stratified operation. Figure 4.11 shows the first spray impingement on a cool piston again, but now from a bottom view using the mirror and piston window as illustrated in Figure 4.2b. The mirror is the primary camera target, and the tumble flow is toward the view across the piston crown, which translates to down in the mirror view. Note that the spray direction is reversed from the top iso views since the camera has been moved from the intake side to the exhaust side. The region of spray impact on the piston crown is clearly visible, particularly at 3360. The final three images in this sequence show a slightly darker region, which corresponds to the fuel film area. Also visible in the same images are lighter streaks, which may be breakup of the liquid film caused by turbulent shear at the surface. The location and appearance of these "white-caps" is in agreement with the direction of the tumble predicted by the CFD simulations, that is, the airflow is shearing the fuel film in the direction of the exhaust valve. It is also possible, however, that these streaks are residue build-up at the fuel film periphery lit from above. Note that Tabata et al found airflow, i.e. swirl and tumble, to be important mainly for fuel vapor transport, while only spray 48 momentum and in-cylinder temperatures had primary impacts on liquid droplet and film behavior [32]. Figure 4.12 shows the bottom views of the liquid film at 7200 ATDCI, or top center during valve open, for the first 5 cycles. The faint dark region is observed to grow during each cycle, consistent with the increase in fuel film coverage observed obliquely from above the piston. More of the lighter regions are visible during the 3 rd and 4 th valve open periods, again possibly indicating breakup on the film surface as a result of the gas exchange process. Increased secondary droplet and vapor production from the fuel film on the piston during the exhaust valve open period could have serious implications for hydrocarbon emissions mechanisms, although most current designs have the piston-bowl or other confining piston geometry located away from the exhaust valve. Finally, Figure 4.13 shows the 360' position (immediately after spray impingement) for the 4 th and 5 th cycles. The lighter "white-cap" regions are quite prominent in these images, indicating that once a liquid fuel film has built up on the piston, primary spray droplets impinging on the film cause more secondary droplets or film break-up than does the tumble flow across the film. It is possible, of course, that conventional GDI designs with higher charge vorticity generate more fuel film break-up and therefore store less fuel on the piston. Figure 4.14 shows the full spray impingement and combustion sequence for the warmed up piston at 400K. Notice in this sequence that the lighting is increased despite the lack of an existing fuel film on the piston, indicating a more vigorous droplet break-up and vapor production on the piston surface as a result of the higher piston temperature. The combustion sequence clearly shows first a roughly spherical flame propagation, consistent with premixed operation, again with luminous streaks indicating rich regions or fuel droplets producing soot which is subsequently burned in the premixed flame front. These features are well illustrated in frames 372' and 378'. Once the flame front catches up with the descending piston crown, this primary GDI combustion mechanism is followed by a transition, as seen in frame 3840, to a poolfire on the piston crown, which can be seen in frames 390' and 3960. Figure 4.15 shows the equilibrium fuel film area for the warmer piston. The fuel film is more difficult to discern in this case, mainly because the film lies almost completely on the quartz window, which reflects light in a similar manner. The closer edge of the film can be seen, however, due to refraction at the lip 49 of the pool, and the left tip of the film (closer to the spark plug) is also visible as a darker region extending from the window RTV. As with the aluminum piston, a photograph was taken of the optical access piston after stratified operation to illustrate the deposits left by the fuel film. Figure 4.16 shows this photograph with several items labeled for clarity, and also includes an edge-filtered copy of the picture with the deposit structure overlaid for reference. Note that these significant deposits are the result of relatively little operating time, less than one hour in total, since the piston window was cleaned frequently during the high-speed imaging sessions. The existence of similar deposit structures on other GDI pistons would tend to indicate similar spray impingement, fuel film buildup, and pool-fires. As an addendum to the discussion over the images taken from the high-speed video, it is instructive to describe certain qualities of the spray impingement and combustion process that are not apparent in static images. Following assembly of the optical access piston, a small lip of cured two-component RTV surrounded the quartz window. This was an expected by-product of ensuring complete filling of the gap between the window and the aluminum piston frame. The lip was very carefully removed using a razor blade, resulting in an almost perfectly smooth piston crown. No protrusions were visible on the surface, and running a finger over the RTV joint indicated a barely perceptible shift in the surface. Combustion in the high-speed video, however, repeatedly shows a small flame structure surrounding the section of the RTV joint where the spray momentum carries the fuel film across the RTV. Figure 4.14, particularly 3360 and 3420, also shows a brighter region immediately to the left of the RTV joint, resulting from a higher secondary break-up and vapor production in this area. The conclusion to be drawn from this evidence is that surface roughness, as well as existing liquid film, is very important in determining the behavior of spray impingement, as proposed by Han et al [36]. simulations did not account for surface roughness effects. 50 The CFD cIQ -t CD -t Pressure traces overlaid for cyclic variation CD -t CD -t C) CD CO C) en C) C) C) 3\ C I C) C or C C,, Tlj, -Uijilj, 7TIT7 C C,, 180 225 270 315 360 CAD ATDCI 385 405 450 495 540 Far quartz wall Figure 4.2a: Reference sketch showing view angle across piston crown. .. ......... [luminum piston (partially As Wei Figure 4.2b: Reference sketch showing view angle using mirror and optical access piston (note the reversal of injector and spark orientation in this view). 52 Spark plug n ector Far quartz wall -Spray core (dense from entrainment) Piston 315 ATDCI (2/3 through injection) Pre-existing fuel film spray periphery and vortex structure Figure 4.3a: Illustration of injected spray structure. Piston 330 ATDCI (6 degrees after EQI) Spray interaction Film with fuel film Spray Mirror looking up through piston window tlumhinum piston structure (below rrtror Figure 4.3b: Illustration of spray interacting with liquid film (note: bottom view through mirror is not same scale as top iso view, and mirror view has been reversed from other data to match spray direction with top iso view). 53 Last of injected fuel Piston Film 333 ATDCI (9 degrees after EQI) Spray interaction with fuel film (ripples) Figure 4.3c: Illustration of end of spray interacting with liquid film. Piston 720 ATDCI (top center gas exchange) Steady-state fuel film coverage area Mirror looking up through piston window Aluminum piston structure (below mirror) Connecting rod (glare from radius) Figure 4.3d: Illustration of steady-state liquid fuel film coverage (note: bottom view through mirror is not same scale as top iso view, and mirror view has been reversed from other data to match spray direction with top iso view). 54 Figure 4.4: Cool and warm spray/piston interaction and combustion sequences (4 pages). 82*C Head 55*C Head 3030 (SOI 3000) 3060 3090 3120 3150 3180 55 3210 324* (EOI) I . 3270 3300 3330 3360 3390 56 34201 3450 348*0 3510 3540 3570 3600 57 387* 3930 3990 4080 58 e 4.5: Combustion in warm engine (82'C head) showin -I1111w 3600 (50 after spark) 3660 3720 3780 3840 390U 59 396*0 4020 4200 4260 60 4380 4500 4620 4560 61 Figure 4.6: Residue on piston imaged by Kodak Ektapro (high-speed video unit) in static engine. Sootfresidue rings "Cleaner' reg ion Figure 4.7: Residue on aluminum piston after stratified-charge operation. 62 3930 (pool fire below diffusion flame) 3600 (film visible) Figure 4.8: First spray impingement and combustion on cool piston (360K) Figure 4.9: Film buildup after second injection on cool piston (360K), 3600. 63 324 330* 336" 3420 3480 3540 3600 7200 Figure 4.10: Steady state spray impingement sequence and resulting fuel film coverage on cool piston (360K). 64 3300 3360 3470 3480 3540 3600 Figure 4.11: Bottom view of first spray impingement on cool piston (360K). 65 1st qrd 5" (approx. equilibrium) I Figure 4.12: Bottom view at 7200 ATDCI for first 5 spray impingement cycles on cool piston. 66 5"' (approx. equilibrium) Figure 4.13: Fuel film at 3600 ATDCI for 4h and 5' impingement cycles on cool piston. 4th 67 -.,.,.,IIIm -. Figure 4.14: First spray impin, 4/1 if .nt and combustion on warm - iston (400K). _I iJ9U" /I 4k I L- I_______________________ 68 Figure 4.15: Steady state fuel film coverage on warm piston (400K). ReM ue nrgs Dar er a -~ r Mtt IN. PII Figure 4.16: Residue on optical access piston after stratified charge operation. 69 (This page intentionally left blank) 70 5 SUMMARY AND CONCLUSIONS 5.1 SIMULATION WORK Computational fluid dynamics simulation, using KIVA-3V, of the rather unorthodox GDI engine with optical access proved quite useful in the design of the engine for stratified operation. The operating window needed some fine-tuning for precise injection and spark timing, but given the imperfect boundary conditions initially provided to the program, the predictions were quite good. Qualitatively the behavior of the fuel spray as it impinges on the piston crown is captured well, but the amount of fuel remaining as a liquid film is under-predicted. Adjustment of the splash model may be in order, although further experimental data, possibly provided by the widebeam LIF concept discussed in section 3.4, would greatly aid the precise modifications. Further improvements to the code in the future might include modeling surface roughness effects and their interaction with potential existing liquid films. 5.2 HIGH-SPEED VIDEO IMAGING High-speed video imaging proved very useful for qualitative description of the spray/wall interaction, combustion, and liquid fuel film buildup. The spray structure is visible, despite the extremely small droplets, due to light scattering, which also reveals the location and production of fuel vapor. The optical access piston allows for two views of the impingement event, and the liquid fuel film is clearly visible. The fuel film area is seen to grown over the first few cycles, reaching a steady-state coverage area after approximately 5 cycles. After the first spray impingement, the existence of a fuel film on the piston crown promotes secondary break-up and vapor production from both the spray and the film due to interaction of the primary droplets with the liquid film (splashing). Minor surface roughness can also have a significant effect on secondary break-up and vapor production. For operation during warm up with piston temperatures on the order of 360K, the area covered by liquid fuel film is quite large, approaching half of the piston crown. For operation closer to the estimated steady-state operating point of production GDI engines, with a piston temperature of 400K, the liquid fuel film coverage diminishes due to more rapid evaporation, but the fuel film still survives combustion and is exposed to the open valves between injections. There is also evidence that the 71 gas exchange process may cause secondary break-up of the liquid fuel film, resulting in additional droplets and/or vapor being transported into the exhaust manifold and contributing directly to engine-out hydrocarbon emissions. Finally, vapor production off the piston varies significantly with temperature, as is evidenced by the combustion behavior observed. With a room temperature piston, stable operation is not possible since most of the impinging fuel forms a liquid film and insufficient vapor is produced to reliably form an ignitable mixture cloud. As the piston warms up, stable operation becomes possible with the increased vapor production, but pool fires are not immediately observed. Closer to the estimated steady-state stratified operating point for production GDI engines, however, the warmer piston produces fuel vapor from the liquid film at a rate that will feed a pool fire for approximately half of the expansion stroke. The oxygen available in cylinder may limit the duration of the pool fire. It is expected that with high enough piston temperatures, the liquid fuel film will diminish to the point of preventing pool fires, as the impinging spray is more completely vaporized. 5.3 CONCLUSIONS While GDI engines do provide substantial advantages in fuel metering control, efficiency, and performance, there is still much room for improvement. Mixture preparation during stratified charge operation in particular is not entirely understood, with fuel spray impingement on the piston crown resulting in a liquid fuel film and, under certain conditions, a turbulent diffusion flame or pool fire on the piston crown fed by the liquid fuel film. This liquid fuel film has yet to be precisely quantified, as do the mechanisms by which the fuel film evaporates during combustion and gas exchange, all of which determine engine-out hydrocarbon emissions levels. 5.4 FUTURE WORK One definite area for future work is the wide-beam laser-induced fluorescence diagnostic concept designed during this project and described in section 3.4. Several key pieces have already been constructed and assembled, including the laser beam manipulation setup and optical access piston with in-situ calibration grooves and thermocouple. Once a suitable gate-intensified camera and controlling and timing hardware are obtained, the concept can be tested. 72 If successful, quantitative fuel film thickness measurements can be conducted for a wide range of operating conditions such as piston temperature, intake temperature, fuel temperature, fuel composition, spray geometry, and piston surface roughness effects. This would provide a more concrete basis for updating various CFD sub-models. Other experimental work in the future could include planar laser-induced fluorescence, or PLIF, possibly to study the fuel vapor distribution after combustion is complete in an attempt to quantify evaporation and contribution to hydrocarbon emissions. Phase-Doppler particle analysis, or PDPA, could also be used to more accurately measure the size and velocity of the fuel droplets, both primary (exiting from the injector nozzle) and secondary (resulting from break-up of the fuel film surface due to primary droplet impingement, turbulent shear stress, or surface roughness). Finally, new computer models can be implemented in CFD codes to capture the phenomena revealed by this project, and compared with both the qualitative and semi-quantitative data now available and any further data generated by the approaches listed above. 73 (This page intentionally left blank) 74 REFERENCES 1. Heywood, J.B., Internal Combustion Engine Fundamentals, McGraw-Hill Book Co., 1988. 2. Automotive Handbook (3'd edition), Robert Bosch GmbH, 1993. 3. VanDerWege, B., "The Effects of Fuel Volatility and Operating Conditions on Sprays from Pressure-Swirl Fuel Injectors," Ph.D. Thesis, Massachusetts Institute of Technology, June 1999. 4. Meyer, R., "Liquid Fuel Transport into the Cylinder in Spark Ignition Engines," Ph.D. Thesis, Massachusetts Institute of Technology, August 1998. 5. Dawson, M., "In Cylinder Liquid Fuel Visualization during Cold Start," Master's Thesis, Massachusetts Institute of Technology, September 1998. 6. Shelby, M., "PLIF Investigation of the Fuel Distribution in Gasoline Direct Injection Fuel Sprays," Master's Thesis, Massachusetts Institute of Technology, June 1997. 7. Shin, Y., "Liquid Fuel Behavior in the Mixture Preparation Process During Starting and Warm-Up in a Spark Ignition Engine," Ph.D. Thesis, Massachusetts Institute of Technology, May 1996. 8. Amsden, A., O'Rourke, P., and Butler, T., "KIVA-II-A Computer Program for Chemically Reactive Flows with Sprays," Los Alamos National Laboratory Report LA-1 1560-MS, 1989. 9. Amsden, A., "KIVA-3: A KIVA Program with Block-Structured Mesh for Complex Geometries," Los Alamos National Laboratory Report LA-12503-MS, 1993. 10. Amsden, A., "KIVA-3V: A Block-Structured KIVA Program for Engines with Vertical or Canted Valves," Los Alamos National Laboratories Report LA-13313-MS, 1997. 11. Amsden, A., "KIVA-3V, Release 2: Improvements to KIVA-3V," Los Alamos National Laboratory Report LA-13608-MS, 1999. 12. Johnen, T., and Haug, M., "Spray Formation Observation and Fuel Film Development Measurements in the Intake of a Spark Ignition Engine," SAE Paper 950511, 1995. 13. Itoh, T., Kakuho, A., Hishinuma, H., Urushiahara, T., Takagi, Y., Horie, K., Asano, M., Ogata, E., and Yamasita, T., "Development of a New Compound Fuel and Fluorescent Tracer Combination for Use with Laser Induced Fluorescence," SAE Paper 952465, 1995. 75 14. Fraidl, G., Piock, W., and Wirth, M., "Gasoline Direct Injection: Actual Trends and Future Strategies for Injection and Combustion Systems," SAE Paper 960465, 1996. 15. Salters, D., Williams, P., Greig, A., and Brehob, D., "Fuel Spray Characterisation within an Optically Accessed Gasoline Direct Injection Engine Using a CCD Imaging System," SAE Paper 961149, 1996. 16. O'Rourke, P., and Amsden, A., "A Particle Numerical Model for Wall Film Dynamics in Port-Injected Engines," SAE Paper 961961, 1996. 17. Anderson, R., Yang, J., Brehob, D., Vallance, J., and Whiteaker, R., "Understanding the Thermodynamics of Direct Injection Spark Ignition (DISI) Combustion Systems: An Analytical and Experimental Investigation," SAE Paper 962018, 1996. 18. Harada, J., Tomita, T., Mizuno, H., Mashiki, Z., and Ito, Y., "Development of Direct Injection Gasoline Engine," SAE Paper 970540, 1997. 19. Karl, G., Kemmler, R., Bargende, M., and Abthoff, J., "Analysis of a Direct Injected Gasoline Engine," SAE Paper 970624,1996. 20. Witze, P., and Green, R., "LIF and Flame-Emission Imaging of Liquid Fuel Films and Pool Fires in an SI Engine During a Simulated Cold Start," SAE Paper 970866, 1997. 21. Fan, L., Parrish, S., Han, Z., Farrell, P., and Reitz, R., "Comparison of Computed Spray in a Direct-Injection Spark-Ignited Engine with Planar Images," SAE Paper 972883, 1997. 22. Fujikawa, T., Hattori, Y., and Akihama, K., "Quantitative 2-D Fuel Distribution Measurements in an SI Engine Using Laser-Induced Fluorescence with Suitable Combination of Fluorescence Tracer and Excitation Wavelength," SAE Paper 972944, 1997. 23. Takegi, Y., Itoh, T., Muranaka, S., Ilyama, A., Iwakiri, Y., Urushihara, T., and Naitoh, K., "Simultaneous Attainment of Low Fuel Consumption, High Output Power and Low Exhaust Emissions in Direct Injection SI Engines," SAE Paper 980149, 1998. 24. Noma, K., Iwamoto, Y., Murakami, N., lida, K., and Nakayama, 0., "Optimized Gasoline Direct Injection Engine for the European Market," SAE Paper 980150, 1998. 25. Houston, R., and Cathcart, G., "Combustion and Emissions Characteristics of Orbital's Combustion Process Applied to Multi-Cylinder Automotive Direct Injected 4-Stroke Engines," SAE Paper 980153, 1998. 76 26. Stanglmaier, R., Hall, M., and Matthews, R., "Fuel-Spray/Charge-Motion Interaction within the Cylinder of a Direct-Injected, 4-Valve, SI Engine," SAE Paper 980155, 1998. 27. Kuwahara, K., Ueda, K., and Ando, H., "Mixing Control Strategy for Engine Performance Improvement in a Gasoline Direct Injection Engine," SAE Paper 980158, 1998. 28. Preussner, C., Ddring, C., Fehler, S., and Kampmann, S., "GDI: Interaction Between Mixture Preparation, Combustion System and Injector Performance," SAE Paper 980498, 1998. 29. Ren, W. and Nally, J., Jr., "Computations of Hollow-Cone Sprays from a Pressure-Swirl Injector," SAE Paper 982610, 1998. 30. Stanglmaier, R., Li, J., and Matthew, R., "The Effect of In-Cylinder Wall Wetting Location on the HC Emissions from SI Engines," SAE Paper 1999-01-0502, 1999. 31. Li, J., Matthew, R., Stanglmaier, R., Roberts, C., and Anderson, R., "Further Experiments on the Effects of In-Cylinder Wall Wetting on HC Emissions from Direct Injection Gasoline Engines," SAE 1999-01-3661, 1999. 32. Tabata, M., Kataoka, M., Tanaka, T., and Yamakawa, M., "Measurement of Fuel Distribution in the Piston Cavity of Direct Injection SI Engine by Using LIF," SAE Paper 2000-01-0240, 2000. 33. Koike, M., Saito, A., Tomoda, T., and Yamamoto, Y., "Research and Development of a New Direct-Injection Gasoline Engine," SAE Paper 2000-01-0530, 2000. 34. Kanda, M., Baika, T., Kato, S., Iwamuro, M., Koike, M., and Saito, A., "Application of a New Combustion Concept to Direct-Injection Gasoline Engine," SAE Paper 2000-01-0531, 2000. 35. Gold, M., Li, G., Sapsford, S., and Stokes, J., "Application of Optical Techniques to the Study of Mixture Preparation in Direct Injection Gasoline Engines and Validation of a CFD Model," SAE Paper 2000-01-0538, 2000. 36. Han, Z., Xu, Z., and Trigui, N., "Spray/Wall Interaction Models for Multidimensional Engine Simulation," International Journal of Engine Research, vol. 1, no. 1, pp. 127-46, 1999. 37. Yang, J. and Melton, L., "Fluorescence-Based Method Designed for Quantitative Measurement of Fuel Film Thickness during Cold-Start of Engines," Applied Spectroscopy, vol. 54, no. 4, pp. 565-74, 2000. 77 (This page intentionally left blank) 78 APPENDIX A - SAMPLE KIVA INPUT DECK ITAPE5 MIT GDI Optical Engine irest 2 nohydro 0 1 lwall lpr 0 irez 2 ncfilm 9999 nctap8 9999 nclast 9999 ncmon 10 ncaspec 0 gmv 1.0 cafilm 10.0 cafin 360.0 angmom 0.0 pgssw 0.0 dti 1.00000e-5 dtmxca 1.0 dtmax 1.000OOe-4 tlimd 0.0 twfilm 9.99e+9 twfin 9.99e+9 fchsp 0.25 bore 11.6743 stroke 11.43 squish 1.633 1.0e+3 rpm 0.0 atdc datdct 0.0 revrep 2.0 conrod 26.67 swirl 0.0 swipro 3.11 thsect 300.0 sector 0.0 deact 0.0 1.0e-3 epsy 1.0e-3 epsv 1.0e-4 epsp 1.0e-3 epst 1.0e-3 epsk 1.0e-3 epse Critical parameter descriptions Restart segment Crank angle frequency of post-processing files Engine speed in rev/min 79 gx 0.0 0.0 gy 0.0 gz tcylwl 350.15 thead 350.15 tpistn 400.15 pardon 0.0 0.0 aO 1.0 bO artvis 0.0 ecnsrv 0.0 0.0 adia anu0 0.0 visrat-.66666667 tcut 800.0 tcute 1200.0 epschm 0.02 omgchm 1.0 turbsw 1.0 0.0 sgsl trbchem 0.0 capa 18.0 pmplict 2.0 lospeed 0.0 airmul 1.457e-5 airmu2 110.0 airlal 252.0 airla2 200.0 0.74 prl rpr 1.11 rsc 1.11 xignit 0.00 tlign -1.0 tdign -1.0 calign 999.0 cadign 9.6 xignl1 0.25 xignrl 0.75 yignfl 0.0 yigndl 0.238 zignbl 11.75 zigntl 12.50 xign12 0.0 xignr2 0.0 yignf2 0.0 Cylinder temperature Head temperature Piston temperature 80 yignd2 0.0 zignb2 0.0 zignt2 0.0 kwikeq 0 numnoz 1 numinj 1 numvel 1 tlinj -1.0 tdinj -1.0 calinj 300.0 cadinj 20.0 tspmas 0.03 tnparc 8000.0 pulse 2.0 injdist 1 kolide 1 tpi 350.0 turb 1.0 breakup 1.0 evapp 1.0 drnoz 3.1 dznoz 13.021934 dthnoz 0.0 tiltxy 180.0 tiltxz 45.0 cone 37.5 dcone 12.5 1.0 anoz 8.00e-4 smr 0.0 ampO 8000.0 nsp 3 gasoline o2 mw2 32.000 hf2 0.0 n2 mw3 28.016 htf3 0.0 stoifuel 4.0 stoio2 49.0 2 nreg "presi" 9.9000e+5 1.0353e+6 "tempi" 293.15 293.15 0.10 "tkei" 0.10 "scli" 0.0 0.0 "er" 0.0 0.0 "mfracfu" 0.0 0.0 "mfraco2" 0.2300000000 0.2300000000 Start of injection in crank angle degrees Duration of injection in crank angle degrees Total mass of injected fuel in grams Number of parcels used to model injection Square wave injection pulse Collision model on Turbulence model on Dispersion model on Evaporation model on Distance of injector nozzle from cylinder axis in cm Vertical position of injector nozzle in cm Position of injector nozzle relative to model X-axis Plan orientation of injector nozzle relative to X-axis Elevation of injector nozzle from vertically down Mean injector cone angle Differential cone angle Injected Sauter mean radius in cm Injected parcel velocity in cm/s Initial pressures in regions 1 and 2 in g/cm-s2 Initial temperatures in regions 1 and 2, in K 81 "mfracn2" 0.7700000000 0.7700000000 0 nrk 0 nre nvalves 1 vliftmin 0.05 skirtth 0.13125 tmove 293.15 vtiltxz 0.0 nlift 069 isoot 0 distamb 6.858 pamb 9.9000e+5 Ambient pressure (intake) in g/cm-s 2 tkeamb 423.0 sclamb 4.8 velin 0.0 reedin 0.0 reedout 0.0 nregino 2 nregamb 1 numpcc 2 0.0 1.0353e+6 720.0 1.0353e+6 numpex 0 82 ImIrI 0.300t 7 oli I wim I 7 0.300047 0.475 . goam // 2.250 z 0.502 0.652 0 1 S1.054 00 - C- 1.01.506 -- L 3.756 77-T z z 2.000 I - 0.600 1- 2.000 - 0,848 1.344 -1.841- -3.200 DrtL and -_Ii 1 ~ I - Drill and courrterslnk for 6-32 Mochl fl screw 6 ploces 0.250 -- top for 1/4x28 4 places M LOAN LABORATORY O PTICAL PISTON PISTON BLOCK R. Karisson I'Iril.|n"DP-001 11/20/1999 1All dimensions in inches |.* 1 OF 1 1" 0 RESIONS ZONE REY Oot" ENP1FN APF0VE DriLl and tap 6-32 x 0, 50" deep 6 pl aces A i 00 0,744 IzII 0. 248 A 45 0000 4- 1.241 ti 72484 0- 2, 0, 744 - )00 0,.250 1,241 M T OPTICAL PISTON SLOAN LABORATORY MIRROR SUPPORT R, Karl.sson OT STEEL 11/2O/1999 ,1 /20/.99 WTH A~~ E MNO.OP-002 All dimensions fn inches ISh 1 RLV np 1 DNE REY Dal CEMCIPrD4 AMROM i RO, 2 Thread 0, 5"xONF + 00 3.00 h0 - 4- * ---- 0, 25 - 1.75 - - 1,00 - MT SLOAN OPTICAL PISTON ROD LABORATORY CONNECTING R, Karlsson 11/20/1999 1 STEEL I All dimensions in inches _OP-003 Isbht 1 OF 1 4 PE ZNE 0 N$ OESCRE flN IEY Date APPRED NOTE: Material thickness to be determl ned by sprlng rate: Compressi on -Prom 0. 15' - o 0, 075' shoutcI require oapproximotel y 2 lbP, 0, 25 00 40.15 ~- 2.50 -m M - OPTICAL PSTON SEAL SPRING SLOAN LABORATORY R. MY Karlsson 11/20/1999 -1 I. M GI ND. O O P 12 SPRING ST=L P -004 All dimensions in inches Iht 1 I -DR4 ~EY OF I '-, I 4 ZME 0. 1250 -J REY DOb DEDCIPTMON WRK.D 0. 2500 4- 00 -3- 0, 28000 0,2800 -- - 3, 2500 SLDAN R. T LABORATORY Karlsson 11/20/1999 OPTICAL SEAL PISTON BAR 15 GRAPHITE I OO P-005 All dimensions in inches 1so@" 1 OF 1 2DNE IWY DEMCRI~flN D-E)5 WPWMD R0, 0625 r- 2. 200 4- 00 00 1,000 --e 2. 200 M T OPTICAL PISTON WINDOW SLOAN LABORATORY R. Karlsson 11/20/1999 QTY 1 WMAL 10 R17 SOP-006 All dimensions in inches Ia-'. 1 OF 1 I EV WK - - 250 PCNPTON AfV MI - >7 0,050 0.250 -- -- --- 10,845 ~Njfj0.750 00 0. 0,1875 -9 0.1875 A- T 1.000 i.7 0,375 - 0.29 w - -- 0.290 3/4 x 14 Stralght thread Vith 3/16' deep clearance Flat CAUgned with Injector hole) Refer to injector mounting drawing for hole dimensions ALL dimensions +/-0.00 5" M I SLOAN LABORATOR R. Karlsson 6/5/2000 INJECTOR INSERT PLUGS(30 & 45) All dism ItNJ-001 All dimensions in tnches Ist- I OF I 4- 4 3 3 4 " t- CLMP LOAD -------- n2 50s. 9 -------- 1 2 ;172 4.8 N/BE CYLINR FEVIO I "N PRESSFE A IELAE 155.E IW D-S IPV D4( 03/15"0 J GV PPV 914 07/17/9 B IDO '19443 DEA..HVNRAVING 021-EN lES OTHQWISE BEFFED PLL D ON CeNTEFLIlE TO BE WMTIN j ~j~ 3- ALL FU~EFfNK VPLLE5 AF AT n-E M]NI"k AXED~ TCLERNC 10. 6TO 1a. 3 DI M ONDIT[N D D FOR, /-AAM c c TYPICAL NOUNTING C 31. 5MAX 18. 3to. 5- (943 I I /7" ULS D11EC / B n3 AM 55 B /x I TI -- 0 7.82 t.D ED m I-i I -- I 1 DtARI D'S (022.5 4f J2. 7 w-EDD 0.8 / DWK MK Ot/15s/3 01/15/ (017.5 .25 A "O-R]NG CAV]TY mom . t0. L3 .IMA Ot/1"Sn - mum WIN (45~ 91fIFIED P.Ar WRIST FIT MET rSMInE ramBwc WRi PLATJM3 3n aLE * ICw ALL T" L-" imz IEMIV MeS ~ W ~DT EwEs 4W6 1D M1S FIN Wr iinPmF t54 IR TO ceo Bsears AMAntve Mmrlt NB, Va. (LeA) PISA (ETALY) MOUNTING A A PEMMMflI JON CAVIFY DETAIL SO&E 4 2/1 c4614 1 03 Wr MN 1 Is-Efi I m CATIA F11LE lB ZGE DuPTmH REY Dm mFdE U.351)Mia 90 745 0 25Dia 45' 0 12SDia U416D1i 0 175 RO 5 AA 21' 05 's15 1 22B 13125 45" 13125 3/1 x 24 thread 4- W=77 Section A-A MI INJECTOIR MOUNT SLOAN LABORATORY BRACKET R. K<drlsson MiNJ-D02 T All dimensions in inches ahw 1 OF 1 6/5/2000