POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Fundamental and Applied Studies for Internal Combustion Engine
Simulations by OpenFOAM
G. D'Errico, G.Montenegro, T.Lucchini, F.Piscaglia
Department of Energy, Politecnico di Milano
http://www.engines.polimi.it
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
An opensource CFD platform, OpenFOAM® :• to implement different models into the same code• to allow collaborative studies• to be applied for industrial applications
Objectives:Fundamental studies: to improve the understanding of the occurring physical and chemical process
Applied research: to provide computational tools which can help the designer in the development of future engines.
Tool:
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Fundamental studies Applied research
RANS Combustion
Spray models
Chemistry models
Mesh motion
Acoustics
Alternative fuels
DPF
1D-3D coupling
In-cylinder SI engine simulation
In-cylinder Diesel engine simulation
Whole engine system simulation
Engine component design
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Diesel Combustion modeling
Simplified chemistry (4 species)
Tabulated ignition delays
Reliable and fast
Industrial applications
TITC(Tabulated Ignition Turbulent Combustion)
Complex chemistry
ISAT algorithm
Computationally demanding
Diagnostic purposes
PSR (Perfectly Stirred Reactor)
Implementation of two different combustion models
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
SANDIA combustion chamberFundamental validation
• ¼ vessel volume• 50000 cells• Mesh size:
liquid spray region:1 mmdownstream of the liquid spray
penetration: 0.5 mmmaximum mesh size: 4 mm
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Computed vs experimental equivalence ratio distribution at auto-ignition time
x = 20 mm, t = 0.7 ms
x = 30 mm, t = 0.7 ms
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
6 cases;O2 from 10% to 21%Tamb from 750 to 1300
Liu et al. (Comb. Flame, vol 137) 44 species, 112 reactionsIt contains acetylene chemistry
The PSR model is able to reproduceboth premixed and mixing-controlledcombustion
Overprediction of the ignition delay
Poor comparison for low temperature
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Formaldehyde and PAH distributionsModerate soot conditions (T = 1000K, O2 = 21%)
• Formaldehyde (HCHO) formation close to the lift-off length. No separation between HCHO consumption and PAH formation
(SAE 2006-01-3434)
Low soot conditions (T = 900K, O2 = 21%)
• Formaldehyde (HCHO) formation upstream the lift-off length. Separation between HCHO consumption and PAH formation
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Full Diesel engine cycle simulationFull Diesel engine cycle simulation
-- 1D1D--3D Coupling with i3D Coupling with intakentake strokestroke-- oneone sectorsector compressioncompression--combustioncombustion
5Injector holes
2.6Swirl ratio
1.2 mmSquish height
-131 deg TDCIVC
10600 cm3Total displacement
14Compression Ratio
134 mmStroke
127 mmBore
6Cylinders
SEATEK 850 PlusType
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Intake stroke in a Diesel engineFourteen grids were used tocover the whole intake stroke
The grid is unstructuredcomposed by hexahedra, prisms and tetrahedra. Hexahedral cells are used in the valve region to follow the incoming flow field
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Mesh motion – Compression stroke
Dynamic mesh layering : a set of user-defined, horizontal cells removeslayers of cells according to specified maximum and minimum thickness values
Start of injection
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Mesh motion – Injection and combustion
• Mesh keeps the spray-orientation• User defined faces for dynamic mesh layering• User defined bowl points moving with piston velocity
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
In-cylinder pressure, temperature and soot
13 CAD ATDC
5 CAD ATDC
0 CAD ATDC
Temperature Soot
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
SANDIA Optical Engine: experimental apparatusTwo different bowl configurations tested
70% bowl
80% bowl
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Fuel spray evolution1 CAD 2 CAD 3 CAD
4 CAD 7 CAD 8 CAD
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Validation: non reacting case – 70% bowl – 12 ºATDC
• Influence of swirl on the fuel distribution
• Good agreement of equivalence ratio distribution on the cut planes 1 and 2
• Result on the cut plane 3 to be improved
Equivalence ratio distribution:
Cut
1C
ut 2
Cut
3
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Validation: non reacting case – 80% bowl – 12 ºATDCEquivalence ratio distribution:
Cut
1C
ut 2
Cut
3
• Influence of the bowl shape on the equivalence ratio distribution well captured
• Maximum equivalence ratio values slightly overpredicted
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Current work: flame structure validation (70% bowl)
OHHCHO + PAH
11 ATDC,Cut 1
14 ATDC,Cut 1
HCHO + PAH OHExperimental
Preliminary results
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
GDI Engine Simulation• Assessment of the Weller+EDM combustion model at engine conditions
• Mitsubishi-IFP GDI Engine test case
Complex geometry
• Simulation of the intake + compression + injection + combustion phases
0 CAIVO
133.5 mmRod length
211 CAIVC
89 mmStroke81 mmBore
• Mesh deformation close to TDC to control the grid quality close to the valves•After 90° TDC use of dynamic layering close to the piston• 9 meshes for the whole intake stroke
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Fuel injection phase – fuel vapor distribution
-50° TDC -40° TDC -22.5° TDC (spark timing)
-15° TDC -5° TDC 20° TDC
Combustion process – distribution of b
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Intake stroke investigation in spark ignition engines
• Piaggio Quasar 250
~ 11Compr. ratio
60 mmStroke72 mmBore
Sliding interface and dynamic mesh layering on canted valves
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
1D-3D couplingAt the OpenFOAM International Conference 2007 we presented a strong coupling strategy between GASDYN and OpenFOAM
The Riemann problem is solved locally for each face constituting the domain interface. It allows to treat flow non-uniformities.
This approach was further validated and integrated in GT-Power, in collaboration with GAMMA Technologies Inc.
Available applications:• Closed valve cylinder modeling (Diesel and S.I.)• Intake and exhaust systems modeling
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
POLITECNICO DI MILANO Department of Energy
• Values are passed from GT-Power to OpenFOAM at the beginning of the 3D simulation:
• Initial pressure• Initial temperature• Initial swirl and tumble coefficient• Initial turbulence level• EGR, fuel, products and oxidant mass fractions
• Other parameters, such as injection law, IVC,EVO and others, are set recurring to standard OpenFOAMdictionaries
• Istantaneous values of pressure, temperature, turbulence and mass fraction of EGR, products, oxidant and fuel are passed back to GT-Power at each time step
Cylinder coupling
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
POLITECNICO DI MILANO Department of Energy
• OpenFOAM can follow the time step given by the 1D model or it can perform subcycling according to the Courant number required by the 3D
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
POLITECNICO DI MILANO Department of Energy
• Values are passed from GT-Power to OpenFOAM at each time. 1D values can be assigned uniformly at the interface
• A coupling procedure based on the solution of the Riemann problem allows to assign non uniform fields compatible with the 3D solution
• Transport of chemical species will be implementedto track different gas compositions such as EGR inside of intake systems
Duct system coupling
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Full Scale DPF Modeling in OpenFOAM®
OUTLINE•Development of a NEW STEADY COMPRESSIBLE SOLVER FOR FLOW
THROUGH THIN POROUS LAYERS: rhoSimplePorous1DFoam
TO BEGIN WITH:•OpenFOAM ® built-in steady compressible solver rhoPorousSimpleFoam®
MOTIVATION:• Need to have a predictive tool for real world Diesel Particulate Filter design
(segmented filters, non axis-symmetric geometries) in internal combustion engine simulation;
• Channel-scale simulation cannot provide a detailed description of radialtemperature distribution;
• Simulation of thermal gradients occurring during filter-regeneration, due to non-uniform inlet flow conditions (HEAT TRANSFER PROBLEM);
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
rhoPorousSimpleFoam®:Despite the approach based on POROUS VOLUME TERM (rhoPorousSimpleFoam®) is
reliable for most applications, it may not be convenient as diesel exhaust after-treatment simulations are performed
• mesh generation (in particular the definition of porous regions) is not straight-forward• DPF channel section is discretized in a high number of cells• the resulting fine grid and the small size of porous cells (about 10−4 m) cause low
simulation time-steps HIGH COMPUTATIONAL TIME• friction between the flow and the porous wall is not accounted
Axial component of the flow INSIDE porous medium in Diesel Particulate Filters is negligible
Corning EX-80 2.66”x4.5” 100/17
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
• Each channel is discretized by a series of CV along the axial direction. Cell size over the transverse direction is taken equal to the channel size: 1D schematization of filter channels;
• Porosity has been set as a cell-face property the orientation of the porous surface with respect to the reference axes has no influence;
• Friction between fluid and porous walls has been modeled as an (axial) implicit volume sink term;
• Fully automatic setup of open/closed pattern: new utility createDpfPatch.• Straightforward mesh generation of the filter (“porous” property automatically set);• Easier definition of the whole layout;• Faster simulations, because a more efficient numerical solver may be used
New Solver rhoSimplePorous1DFoam for Porous Media
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
• The method mimics the operation of a solution procedure devised for a staggered variable arrangement, but keeping the colocated variables;
• Momentum equation is solved for the face flux φ instead of cell-centered velocity U;
• Face velocities uf are calculated from the corrected flux fields; • The cell-centered velocity is merely regarded as a secondary variable, which is used in the construction of the momentum equation;
• Pressure gradient is also calculated on cell faces.
New Solver rhoSimplePorous1DFoam for Porous Media
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
The three-dimensional mesh geometry of the monolith is generated from a 2D sketch; faceSetscustomization for DPF applications is performed by an automatic algorithm (createDpfFaceSets):
inlet and outlet ends of the monolith show a typical “chessboard”arrangement, where channels are alternatively open and closed; cell faces representing the closed-ends of filter channels are automaticallyset as ”walls”
plug ends are automatically generated by extruding inlet and outlet ends of the monolith and then adding the resulting mesh to the original one; they have non-permeable walls, that are automatically grouped and set as ”walls”
porosity is modeled as a cell-face property; porous walls dividing inlet and outlet channels of the monolith are grouped in a faceSetdefined as ”porous”
The solid region for the DPF material and the cement strips is used to model the heat exchange to the surroundings
Case Setup
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
Validation of the rhoSimple1DExpFoam solver (IBIDEN SD031 test brick)
Test case:• 34.24 x 34.24 x 142.4 mm brick;• 288 channel pairs;• Cold flow;• 198144 hexahedral cells;• 160080 porous faces.
Test case:• 34.24 x 34.24 x 142.4 mm brick;• 288 channel pairs;• Cold flow;• 198144 hexahedral cells;• 160080 porous faces.
POLITECNICO DI MILANO Department of Energy
Gianluca D’Errico Open Source CFD International Conference 2008
VELOCITY FIELD
PRESSURE FIELD
Validation of the rhoSimple1DExpFoam solver (IBIDEN SD031 test brick)