CFD Simulations of I.C. Engines: Combustion, Internal Flows, integrated 1D-MultiD simulations T. Lucchini,, G. Montenegro, G. D Errico Dipartimento di Energetica,, Politecnico di Milano http://www.engines.polimi.it tommaso.lucchini@polimi.it
Topics Research group Research activities Thermo-fluid dynamic modelling of IC engines Example of applications/case studies: 1D-3D coupling Steady-state flow-bench simulations Mesh motion for in-cylinder simulations Diesel Combustion
I.C. Engine Group Dipartimento di Energetica Teaching and research in the fields of: Internal combustion engines; Hydraulic/thermal Machines, energetic systems; 10 People prof.. Giancarlo Ferrari (full professor) prof.. Angelo Onorati (full professor) Dr. Gianluca D Errico, Gianluca Montenegro (assistant professors) Dr. Tarcisio Cerri,, Tommaso Lucchini, Federico Piscaglia (post- doc) Msc. Daniele Ettorre,, Andrea Montorfano,, Marco Zanardi (post- msc)
Thermo-fluid dynamic modeling of IC engines Combustion Air injection system Silencers Exhaust manifold Noise Exhaust after-treatment system Intake system Turbocharger 3way CC DPF SCR DOC denox trap Integrated 1D-3D thermo-fluid dynamic modeling of S.I. engines (gasoline, natural gas, hydrogen) and C.I. engines (Diesel, HCCI).
Thermo-fluid dynamic modeling of IC engines MODELING OF IC ENGINE? Apply the state of art of numerical and physical models to study, develop and optimize new engine configurations; Improve the existing models and create new ones; In-house code development and application
Thermo-fluid dynamic modeling of IC engines NUMERICAL CODES 1) GASDYN (1D) Simulation of wave motion and chemical species transport, with reactions in the gas and solid phase along the exhaust ducts. Integrated modeling of the main after-treatment devices: 3W catalyst, DPF, DOC, SCR, denox trap, secondary air injection, etc.. 2) OpenFOAM (CFD) Application, development, customization for I.C. engine simulation;
Why OpenFOAM Open-source, freely available CFD Toolbox, licensed under the GNU General Public Licence Written in a highly efficient C++ object-oriented programming language. Easy customisation, extensions and modifications by the user. Finite volume numerics to solve systems of partial differential equations 3D unstructured mesh of polyhedral cells support. Domain decomposition parallelism is fundamental and integrated at a low level so that solvers can generally be developed without the need for any parallel-specific coding.
Why OpenFOAM OpenFOAM includes a wide range of solvers, model libraries, meshing and post-processing tools to make it attractive as research platform for fundamental studies SOLVERS Incompressible flows Compressible flows Multiphase flows DNS and LES Combustion Heat transfer Solid dynamics MODEL LIBRARIES Turbulence Large-eddy simulation Transport models Thermophysical Lagrangian particle tracking Chemical kinetics PRE-POST PROCESSING Mesh generation Mesh converters (KIVA, STAR-CD, GAMBIT, CFX) Mesh manipulation parafoam post-processor OTHER FEATURES Linear system solvers ODE system solvers Parallel computing Mesh motion Topological changes Fluid-structure interactions Numerical methods
Why OpenFOAM C++ object-oriented; oriented; New models easily developed and tested in isolation; Represents PDE systems in their natural language: ρy t tf ( ρ UY ) ( µ Y ) 0 + + = tf T tf solve ( fvm::ddt(rho, Ytf) + fvm::div(phi, Ytf) + fvm::laplacian(mut, Ytf) );
Case studies 1D-3D coupling; Steady-state, flow bench simulations; Mesh motion for in-cylinder simulations; Diesel spray combustion;
Case studies 1D-3D coupling; Steady-state, flow bench simulations; Mesh motion for in-cylinder simulations; Diesel spray combustion;
1D-3D Coupling GASDYNFoam Direct coupling of the Gasdyn and OpenFOAM code for integrated 1D-3D simulations; Fully integrated 1D/3D approach: Complete access to the source of both codes; Implementation of the same solver in the calculation tools; Flexible approach for the treatment of the domain interface; Strict coupling: 1D and MultiD codes exchange boundary conditions at each time step Commercial codes do not allow to access all the variables, solver routines...
1D-3D Coupling GOVERNING EQUATIONS Euler Equation (gas viscosity neglected) ρ r + ( ρu) = 0 t r ρu r r r + ( ρu U) = pi t ρe r 0 + = t ( ρue ) ( ) 0 pu Inviscid flow approximation widely used in 1D codes Air and exhaust gas velocities low Same 2 nd order HLLC numerical method for both domains r
1D-3D Coupling COUPLING STRATEGY Information is passed back and forth between the two codes at each timestep The Riemann problem is solved locally for each face constituting the domain interface Allows to treat flow non-uniformities coming from the 3D domain
1D-3D Coupling: application LAMBORGHINI GALLARDO V10 ENGINE Engine Displacement Compression ratio Valves V10 N.A. 4960 cm 3 10.8 4 per cyl.
1D-3D Coupling: application LAMBORGHINI V10 ENGINE: RESULTS, 6000 rpm
1D-3D Coupling: application LAMBORGHINI V10 ENGINE: RESULTS, 7000 rpm
1D-3D Coupling: application LAMBORGHINI V10 ENGINE Volumetric Efficiency Detailed validation in the SAE Paper 2007-01 01-04950495
1D-3D Coupling: application LAMBORGHINI V10 ENGINE: RESULTS, 7000 rpm Back flows from the junction to the incoming pipe with non uniform flow distribution; Strong velocity wave coming from the cylinder and passing through the junction with minor reflections;
1D-3D coupling; Steady-state, flow bench simulations; Mesh motion for in-cylinder simulations; Diesel spray combustion;
Steady-state, flow-bench simulations SEATEK PLUS CYLINDER HEAD Valve diameter = 35.5 mm
Steady-state, flow-bench simulations CARTESIAN MESH-GENERATOR ADVANTAGES Fast and automatic grid generation from STL file; Possibility to model the boundary layer; DISADVANTAGES Sharp edges can be smoothed; External surface obtained as an extrusion of the internal cell faces; High cell number required to correctly reproduce the geometry;
Steady-state, flow-bench simulations SURFACE MESH OF THE CYLINDER HEAD 900000 cells, max cell size 3 mm
Steady-state, flow-bench simulations INLET Total pressure = 0.978 bar Temperature = 303 K k = 1 m 2 /s 2 ε = 90 m 2 /s 3 OUTLET Static pressure = 0.954 bar, ouflow
Steady-state, flow-bench simulations COMPUTED FLOW FIELD Useful information provided for the ports design: flow distribution, swirl, velocity and pressure fields.
Steady-state, flow-bench simulations FLOW COEFFICIENT VALIDATION Overestimation due to the adopted turbulence model.
1D-3D coupling; Steady-state, flow bench simulations; Mesh motion for in-cylinder simulations; Diesel spray combustion;
Mesh motion for in-cylinder simulations COMPLEX GEOMETRY Unstructured grids; Moving piston and valves, ports; High mesh quality required; MESH MOTION REQUIRED Pre-processing mesh tools for mesh motion; Significant manual work required; Mesh motion is not solution-dependent;
Mesh motion for in-cylinder simulations PROPOSED APPROACHES MUMMI Multiple mesh motion and mesh-to-mesh interpolation; Reliable, widely adopted; Requires different meshes to cover the simulation; Automatic mesh motion; FAMA Fully automatic mesh adaptation; One mesh covers the whole simulation; Mesh motion combined with topological changes to keep the mesh quality high;
Mesh motion for in-cylinder simulations POLYHEDRAL VERTEX-BASED MOTION SOLVER 1) MOTION EQUATION ( γ u) = 0 2) NEW POINT POSITION x = x + u t new old Laplace equation of motion solved on a finite-element element tetrahedral decomposition of the mesh. Mesh quality controlled even in presence of extreme boundary deformations
Mesh motion for in-cylinder simulations TOPOLOGICAL CHANGES Dynamic mesh layering Attach/detach boundary Sliding interface ALGORITHM At each time step: 1) Sliding interfaces detached 2) Layers added or removed 3) Points motion 4) Sliding interfaces re-attached
Mesh motion for in-cylinder simulations MITSUBISHI-IFP IFP ENGINE (MUMMI APPROACH) Bore Stroke Con. rod length IVO IVC 81 mm 89 mm 133.5 mm 0 CA 211 CA Fully tetrahedral mesh generated with the NETGEN software from the STL file.
Mesh motion for in-cylinder simulations MITSUBISHI-IFP IFP ENGINE High mesh quality because of tetrahedral Delauneization;
Mesh motion for in-cylinder simulations The whole intake stroke is divided into a series of target meshes: es: Start End Cells Piston position at start [mm] Piston position at end [mm] Valve lift at start [mm] Valve lift at end [mm] Mesh 1 0 20 123082 89 85.4 0 1.98 Mesh 2 20 45 101216 85.4 72.2 1.98 5.15 Mesh 3 45 75 129108 72.2 48.9 5.15 7.69 Mesh 4 75 120 154970 48.9 16.6 7.69 8.48 Mesh 5 120 150 180797 16.6 4.1 8.48 6.71 Mesh 6 150 170 186766 4.1 0.5 6.71 4.78 Mesh 7 170 185 163129 0.5 0.1 4.78 2.78 Mesh 8 185 200 180350 0.1 1.8 2.78 0.9
Mesh motion for in-cylinder simulations Mesh motion from 0 to 20 CA 123000 tet cells (Mesh 1)
Mesh motion for in-cylinder simulations Mesh motion from 45 to 75 CA 129000 tet cells (Mesh 3)
Mesh motion for in-cylinder simulations Mesh motion from 150 to 170 CA 186000 tet cells (Mesh 6)
Mesh motion for in-cylinder simulations MITSUBISHI-IFP IFP ENGINE, MESH 2, INTAKE STROKE Turbulent kinetic energy Temperature Velocity field
Mesh motion for in-cylinder simulations MITSUBISHI-IFP IFP ENGINE, MESH 4, INTAKE STROKE Turbulent kinetic energy Temperature Velocity field
Mesh motion for in-cylinder simulations MITSUBISHI-IFP IFP ENGINE, MESH 4, INTAKE STROKE Turbulent kinetic energy Temperature Velocity field
Mesh motion for in-cylinder simulations MITSUBISHI-IFP IFP ENGINE, INTAKE STROKE The computed flow field reflects experimental observations Tumble generation; Turbulence intensity distribution; Simulation of a real operating condition Unsteady pressure and temperature at the inlet boundary; Direct-injection injection of the fuel; Combustion;
Mesh motion for in-cylinder simulations SCAVENGING IN A TWO-STROKE ENGINE (FAMA) COMBINED USE OF MULTIPLE TOPOLOGICAL CHANGES 1) PISTON MOTION dynamic layering deformation 2) PORTS-CYLINDER CONNECTION sliding-interface interface
Mesh motion for in-cylinder simulations SCAVENGING IN A TWO-STROKE ENGINE (FAMA) ENGINE GEOMETRY COMPUTATIONAL MESH Bore Stroke Comp. Ratio Speed Boost pressure 66.5 mm 57 mm 10.8 2500 rpm 1.05 bar PHYSICAL MODELS AND BOUNDARY CONDITIONS k-ε turbulence model; No slip at walls; Total pressure at intake; Fixed temperature at walls; Axial-symmetric No. of cells at BDC No. of cells at TDC 25000 8000
Mesh motion for in-cylinder simulations SCAVENGING IN A TWO-STROKE ENGINE (FAMA) Evolution of in-cylinder EGR and flow field Validation published in the SAE 2007-01 01-01700170 Paper
1D-3D coupling; Steady-state, flow bench simulations; Mesh motion for in-cylinder simulations; Diesel spray combustion;
Diesel spray combustion OBJECTIVES Achievement of detailed and reliable Diesel combustion models to develop a CFD tool for diagnostic and predictive purposes; HOW DIFFERENT MODELS CAN BE TESTED AND COMPARED? To be implemented into the same CFD code; The code has to be opensource to allow collaborative studies regarding both the model implementation and validation; Modified Eddy Dissipation Model (EDM+ID) compared with the Perfectly Stirred Reactor combustion model (PSR);
Diesel spray combustion SPRAY MODELING Spray is composed by a series of parcels,, evolving according to the mass, momentum and energy exchange with the continuous phase (Lagrangian( Lagrangian) Spray sub-models describe injection, atomization, primary and secondary breakup, collision, heat transfer,. OpenFOAM has a robust and efficient parallel lagrangian particle tracking algorithm; Lagrangian spray parallelized; OpenFOAM is widely used for Lagrangian spray simulation. It contains most of the available Lagrangian spray sub-models;
Diesel spray combustion MODIFIED EDDY DISSIPATION MODEL Only three chemical species (fuel, oxidant and products) Reaction rate: ( 1 ),, & ω = α & ω + αω& F F ign F mix Integral function Y I to estimate the ignition delay; Ignition delays tabulated as a function of equivalence ratio, pressure, temperature, EGR Eddy dissipation model for mixing controlled combustion: ε YO YP ρω& F, MIX = Cmagρ min YF,, β k s 1 + s Fast and reliable model;
Diesel spray combustion PERFECTLY STIRRED REACTOR MODEL Requires complex kinetics and as many transport equations as the number of the species involved; Perfect mixing assumed in each computational cell; The reaction rate for each specie (RR( i ) is computed from the stiff integration of the chemical problem in each computational cell; ISAT In-situ adaptive tabulation to reduce the computational time: RR i ( ) ( ) i Φ = RR Φ + ( Φ Φ ) i RR Φ 0 0
Diesel spray combustion VALIDATION SANDIA COMBUSTION CHAMBER DATABASE http://www.ca.sandia.gov/ecn Establish an internet library of welldocumented experiments that are appropriate for model validation; Provide a framework for collaborative comparisons of measured and computed results Case 1 2 3 4 Ambient density [kg/m 3 ] 14.8 14.8 14.8 14.8 O 2 volume fraction [%] 21 15 10 21 Ambient temperature [K] 1000 1000 1000 1300 Injected Fuel Mass [mg] 47.5 47.5 47.5 47.5
Diesel spray combustion COMPUTATIONAL GRID A quarter of the real geometry 1 mm cells close to spray axis The grid is coarsened away from it (2 mm, 4 mm) Grid refined where it is useful 90000 cells Real engine simulations require less cells Wall temperature imposed to reproduce the experimental cool- down
Diesel spray combustion EDM lift-off PSR lift-off
Diesel spray combustion T = 1000 K, O 2 = 21%
Diesel spray combustion T = 1000 K, O 2 = 15%
Diesel spray combustion T = 1000 K, O 2 = 10%
Diesel spray combustion T = 1300 K, O 2 = 10%
Diesel spray combustion Flame lift-off comparison
Diesel spray combustion The complete study will be presented at the SAE World Congress, Detroit, April 2008. It will include validation with optical images; The ignition treatment of the EDM model requires improvements (tabulated double delay, tabulated reaction rates); The PSR model correctly reproduce combustion when it is mainly controlled by the ignition delay; Future developments: subgrid-turbulence chemistry interaction (PaSR), Flamelet models (RIF, CMC.);
Conclusions OpenFOAM for I.C. Engine simulation Fundamental studies Diesel combustion Industrial toolkit 1D-3D; Mesh motion; Steady-state flow bench; Open-source promotes collaborative studies
Acknowledgments dr. Gianluca D Errico, dr. Gianluca Montenegro, ing. Daniele Ettorre, ing. Marco Zanardi dr. Hrvoje Jasak, dr. Zjelko Tukovic dr. Gianmarco Bianchi Seatek SPA, MV Agusta SPA, Lamborghini Automobili SPA IFP (Institut Francais de Petrole)
Bibliography G. D Errico,, D. Ettorre,, and T. Lucchini: "Simplified" and Detailed Chemistry Modeling of Constant-Volume Diesel Combustion Experiments" " (to be published) T. Lucchini, G. D Errico,, H. Jasak,, Z. Tukovic: : "Automatic" Mesh Motion with Topological Changes for Engine Simulation", SAE Paper 2007-01 01-01700170 G. Montenegro, A. Onorati,, F. Piscaglia,, G. D Errico: : "Integrated" 1D-MultiD Fluid Dynamic Models for the Simulation of I.C.E. Intake and Exhaust Systems", SAE Paper 2007-01 01-04950495 G. D Errico,, D. Ettorre,, T. Lucchini: "Comparison" of Combustion and Pollutant Emission Models for DI Diesel Engines", SAE Paper 2007-24 24-0045 A. Onorati,, G. Montenegro, G. D Errico:" Errico:"Prediction of the Attenuation Characteristics of I.C. Engine Silencers by 1-D 1 D and Multi-D D Simulation Models", SAE Paper 2006-01 01-15411541 G. D Errico,, T. Cerri,, T. Lucchini: "Development" and Application of S.I. Combustion Models for Emissions Prediction", SAE Paper 2006-01 01-11081108 T. Lucchini, G. D Errico,, and N. Nordin: : "CFD" Modeling of Gasoline Sprays", SAE Paper 2005-24 24-86
THANKS FOR THE ATTENTION! Dr. Tommaso Lucchini Politecnico di Milano Dipartimento di Energetica tommaso.lucchini@polimi.it www.engines.polimi.it
ρ r + ( ρu) = 0 t r ρu r r r + ( ρu U) = pi t ρe r 0 + = t r ( ρue ) ( ) 0 pu