In-cylinder flows and combustion modeling: … flows and combustion modeling: application and...

In-cylinder flows and combustion modeling: application and validation to real and engine-

like configurations

T. Lucchini, G. D'Errico, A. Della Torre, T. Cerri, A. Maghbouli, E. A. Tahmasebi, L. Sforza, D. Paredi

Politecnico di Milano, Department of Energy

Second Two-Day Meeting on Internal Combustion Engine Simulations Using the OpenFOAM technology, Milan 26-27th November 2016.

www.engines.polimi.it

Topics 2

In-cylinder flows and combustion modeling using OpenFOAM® technology

CFD methodologies

Validation

Application

Next steps…

Lib-ICE

OpenFOAM


Lib-ICE 3

Internal combustion engine modeling using the OpenFOAM technology

OpenFOAM-x.x.x

Lib-ICE

Library: physical models, mesh management

Applications: solvers (cold flow, SI, Diesel, after-treatment), utilities

Mesh generation

Fuel-air mixing

Combustion

Spray modeling

Engine flows1 2 3

SI combustion

Diesel combustion

Development/validationEngine simulation workflow


Engine simulation workflow 4

Methodology

Diesel engines

SI engines

Meshmanagement

Cold flowFuel-air mixing

Combustion



Methodology

Diesel engines

SI engines

Meshmanagement


Combustion

• Automatic meshgeneration

• Mesh motion• Topological

changes

• Discretization• Turbulence

models• Mesh quality

• Lagrangianspray

• Sub-models• Nozzle flow

• Simplified or detailedkinetics

• Ignition• Pollutants



Methodology

Diesel engines

SI engines

Meshmanagement


Combustion

• Automatic meshgeneration

• Mesh motion• Topological

changes

• Discretization• Turbulence

models• Mesh quality

• Lagrangianspray

• Sub-models• Nozzle flow

• Simplified or detailedkinetics

• Ignition• Pollutants

Fully integrated approaches Open-Source code

Diesel Engines


Diesel engines 8

Methodology / model development

Automaticmesh

generation

Combustionmodels

Meshhandling

Cold flow

Nozzleflow

Spraymodels

Engine CAD data


Diesel engines 9

Mesh management

a) Main engine data b) Spray oriented block mesh

c) Combustion chamber points fit

Mesh handling

d) Spray-oriented mesh

Automatic mesh generation

Dynamic layering


Diesel engines 10

Mesh management

a) Main engine data b) Spray oriented block mesh

c) Combustion chamber points fit

Mesh handling

d) Spray-oriented mesh

Bowl #1 Bowl #2 Bowl #3

Bowl #4 Bowl #5 Bowl #6

From CAD to SIMULATION: 10 minutes


Dynamic layering


Diesel Engines: spray modeling 11

Spray A from Engine Combustion Network

10

8 m

m

2d computational mesh Spray model setup

• Injection: blob• Breakup: KHRT• Evaporation: Spalding

Test conditions

CFD setup

• Turbulence model: standard k-e with modified C1

• Fuel n-dodecane

• Nozzle diameter: 90 mm

• pinj: 50-150 MPa

• Tamb: 700-1200 K

• ramb : 7.6 – 22.8 kg/m3

VaporLiquid

Baseline (pinj = 150 MPa; Tamb = 900 K; ramb= 22.8 kg/m3)

0.00

0.04

0.08

0.12

0.16

0.0 5.0 10.0

Mix

ture

Fra

cti

on

[-]

Distance from axis [mm]

Exp

PoliMI

z = 22.5 mm

z = 45 mm

Dashed: expContinuous: calc.

0

4

8

12

16

20

24

700 800 900 1000 1100

Ave

rage

Liq

uid

Le

ngt

h [m

m]

Ambient Temperature [K]

Experimental

Computed

Parametric variations (Tamb, ramb)



FPT C11 Spray (collaboration with TU/e – DI N. Maes and Prof. B. Somers, FPT support)

TU/e optical vessel

• Liquid penetration: DBI• Vapor penetration: Schlieren

Operating conditions• Nozzle diameter: 0.205 mm• Ambient temperature: 900 K

CFD Setup

(a) (b)

• Consistent with the engine gridin size and structure

Spray

3D mesh

• Huh-Gosman atomization• Pilch-Erdman breakup

Turbulence modeling

• k-e with modified C1

Validation• Liquid: exp vs computed extinction

profiles

ANR C1 C2

Pinj [MPa] 150 80 150

ramb [kg/m3] 22.8 40 40




TU/e optical vessel



CFD Setup

(a) (b)


Spray

3D mesh


Turbulence modeling



profiles

ANR C1 C2

Pinj [MPa] 150 80 150

ramb [kg/m3] 22.8 40 40




TU/e optical vessel



CFD Setup

(a) (b)


Spray

3D mesh


Turbulence modeling



profiles

• Vapor: exp (Schlieren), calc. (mixture fraction threshold)

ANR C1 C2

Pinj [MPa] 150 80 150

ramb [kg/m3] 22.8 40 40


Diesel Engines: cold flow 15

Steady state flow bench simulations (FPT Industrial, Gilles Hardy)

Mesh generation

Cartesian, body fitted grids with boundary layer generated with cartesianMesh (cfMesh) or snappyHexMesh (OpenFOAM)

CFD Setup

• Steady-state flow• Turbulence model: standard k-e• Porous media acts as a flow straigthner in the computational

mesh: computation of the swirl torque• Verification of angular momentum conservation across the mesh

boundaries.

This work was part of Davide Paredi MSc Thesis


16

Computed flow field

Lift = 5 mm Lift = 10 mm0

0.2

0.4

0.6

0.8

0

0.2

0.4

0.6

0.8Flow coeff. Cd [-]

0

0.2

0.4

0.6

0.8

Lift = 5 mm Lift = 10 mm

Swirl number Ns [-] from Torque

Calc.

Exp.

Calc.

Exp.

Diesel Engines: cold flowSteady state flow bench simulations (FPT Industrial, Gilles Hardy)

This work was part of Davide Paredi MSc Thesis


Diesel Engines 17

Combustion modeling

Characteristic time-scale model (CTC)

Well-mixed model

mRIF modelPassenger-car Diesel engine

Medium-duty engine operating at low-load with iso-octane fuel.

• N-dodecane spray combustion at constant volume conditions• ECN Spray-A experiment• Variation of ambient temperature, density, oxygen and injection

pressure

Density Inj. Press. O2 conc. Amb. Temp

Computed

Experimental

• Kinetic mechanism: Luoet al. (111 species, 467 reactions)

Ignitiondelay


Diesel Engines 18

Combustion modeling

Characteristic time-scale model (CTC)

Well-mixed model

mRIF modelPassenger-car Diesel engine

Medium-duty engine operating at low-load with iso-octane fuel.

• N-dodecane spray combustion at constant volume conditions• ECN Spray-A experiment• Variation of ambient temperature, density, oxygen and injection

pressure

Computed

Experimental

• Kinetic mechanism: Luoet al. (111 species, 467 reactions)

Density Inj. Press. O2 conc. Amb. Temp

Flame lift-off


Diesel Engines: validation 19

Spray B in Engines (from ECN, collaboration with Dr. Eagle, Dr. Malbec, Dr. Musculus)

Optically accessible engine with three hole injector

Nozzle design very similar to spray A: possibility to perform same studies in engine and constant-volume vessel.

CFD setup and tested points

• Same CFD setup used in Spray A simulations, two different combustion models tested: mRIF and well-mixed.

Tested conditions• T@SOI: 800-1000 K• Pinj [bar]: 500 – 1500 bar• O2 [%]: 13-21 %• r @ SOI: 15.2 – 22.8 kg/m3

Fuel: n-dodecane

Validation: non reactingLiquid length

Vapor distribution

CALC.

EXP

CALC.

EXP

This work is part of Amin Maghbouli PhD project



Spray B in Engines (from ECN, collaboration with Dr. Eagle, Dr. Malbec, Dr. Musculus)

Optically accessible engine with three hole injector

Nozzle design very similar to spray A: possibility to perform same studies in engine and constant-volume vessel.

CFD setup and tested points

Automatically generated grid

• Same CFD setup used in Spray A simulations, two different combustion models tested: mRIF and well-mixed.

Tested conditions• T@SOI: 800-1000 K• Pinj [bar]: 500 – 1500 bar• O2 [%]: 13-21 %• r @ SOI: 15.2 – 22.8 kg/m3

Fuel: n-dodecane

This work is part of Amin Maghbouli PhD project

Validation: reacting



EU 6 FPT Engine MD (support from FPT, DI Gilles Hardy)

Five operating points: with different EGR levels

Grid automatically generated

75% load, 15% EGR, 2 injections


0

50

100

150

200

250

300

350

400

0

10

20

30

40

50

60

70

80

90

-50 -30 -10 10 30 50

Heat

rele

ase r

ate

[J/k

g]

Pre

ssu

re [

bar]

Crank Angle [deg]

Exp. m-RIF Well-mixed

0

50

100

150

200

250

300

350

400

0

20

40

60

80

100

120

140

-50 -30 -10 10 30 50

Heat

rele

ase r

ate

[J/k

g]

Pre

ssu

re [

bar]

Crank Angle [deg]


[rpm]

bm

ep/b

mep

ma

x

800 4000

1

Cylinder pressure validation (mRIF and well-mixed)

• Initial conditions from 1D simulations



EU 6 FPT Engine MD (support from FPT, DI Gilles Hardy)

Five operating points: with different EGR levels

Cylinder pressure validation (mRIF and well-mixed)


Full load, 0% EGR, 1 injection

0

50

100

150

200

250

300

350

400

0

10

20

30

40

50

60

70

-50 -30 -10 10 30 50

Heat

rele

ase r

ate

[J/k

g]

Pre

ssu

re [

bar]

Crank Angle [deg]


[rpm]

bm

ep/b

mep

ma

x

800 4000

0

50

100

150

200

250

300

350

400

0

20

40

60

80

100

120

140

160

-50 -30 -10 10 30 50

He

at

rele

as

e r

ate

[J

/kg

]

Pre

ss

ure

[b

ar]

Crank Angle [deg]


1

Grid automatically generated

• Initial conditions from 1D simulations



PCCI engine (support from FPT, DI Gilles Hardy)

PCCI combustion conditions

• High EGR rate (50%)• 3000 rpm

Computational mesh generated automatically with

the Polimi tool

0

20

40

60

80

100

120

140

-30 -10 10 30 50

Cyl.

Pre

ss

ure

[b

ar]

Crank Angle [deg]

Exp.

Well-mixed

0

50

100

150

200

250

300

350

-30 -10 10 30 50

He

at

Re

lea

se

Ra

te [

J/d

eg

]

Crank Angle [deg]

Exp.

Well-mixed

Combustion model: Well-mixed

Diesel fuel: n-dodecaneKinetic mechanism: Faravelli et al. (110 species, 300 reactions)



PCCI engine (support from FPT, DI Gilles Hardy)

PCCI combustion conditions

• High EGR rate (50%)• 3000 rpm

Computational mesh generated automatically with

the Polimi tool


0.2

0.4

0.6

0.8

1

1.2

1.4

0

0.2

0.4

0.6

0.8

1

1.2

OP.1 OP.2 OP.3

so

ot

(no

rmali

ze

d)

NO

x (

no

rmali

ze

d)

Calc.

Exp.


Heavy Duty engines (support from GE Global Research, Dr. Pasunurthi)


Operating conditions

• SOI variation• Soot/NOx trade-off• Dual-fuel combustion

Diesel: mRIFDual-fuel: PaSRNOx: Zeldovich, soot: Moss

Engine 1: SOI variation

-20 -10 0 10 20 30

Crank Angle [deg]

Exp.

Calc.

-20 -10 0 10 20 30

Crank Angle [deg]

Exp.

Calc.

p/p

max

p/p

max

Engine 2: pollutant prediction

OP.1 OP.3

-20 0 20 40

Pcyl/P

max

Crank Angle [deg]

SOI #1, Exp.

SOI #1, Calc.

SOI #2, Exp.

SOI #2, Calc.

SOI #3, Exp.

SOI #3, Calc.

Conventional Diesel Combustion


Diesel Engines validation 26

Heavy Duty engines (support from GE Global Research, Dr. Pasunurthi)

Setup

• Reduced mechanism for Diesel (C7H16) and natural gas•Combustion model: PaSR

-20 0 20 40

Ro

HR

/Ro

HR

ma

x

Crank Angle [deg]

Exp.

Calc.

OP 2

Validation

• Pollutants NOx : Zeldovich Soot: Moss

•Two operating points: same air/fuel ratio, different amount of injected Diesel fuel 0

0.2

0.4

0.6

0.8

1

1.2

OP 1 OP 2

NO

x/

NO

x,m

ax

Exp.

Calc.

-20 0 20 40

Ro

HR

/Ro

HR

ma

x

Crank Angle [deg]

Exp.

Calc.

OP 1

Dual-fuel combustion



Diesel Engines 27

Next steps…

…and more…Internal nozzle flow modeling (RANS, cavitation)

Spray C & Spray D

These two single holeinjectors are defined by ECNto compare the effect ofgeometry on flow behavior.

Spray Ck factor=0, sharp edge

Spray Dk factor=1.5,

rounded edge

blockMesh grid

Spray D

Spray C

Presssure across the hole axisSpray C

Spray Ccavitated

Spray DNot cavitated

ECN contributors results

Tabulated kinetics:• Homogeneous reactors• Unsteady RIF

Transported PDF combustion modeling:• Eulerian Stochastic fields with

tabulated kinetics!

10.07

10.51

11.72

10.92

9

9.5

10

10.5

11

11.5

12

Experiments POLIMI

spray C mdot (g/s)

spray D mdot (g/s)

• mRIF with multiple injections

This work is part of Ehsan Tahmasebi PhD project

SI Engines


SI Engines 29

Methodology / model development

Nozzleflow

Automaticmesh

generationSpray

models

Combustionmodels

Meshhandling

Cold flow

Engine CAD data


SI Engines: cold flow 30

Darmstadt optical engine (collaboration with Dr. B. Bohm and DI C. P. Ding)

Four-valve engine, fully optically accessible

0 10 20 30Velocity Magnitude (m/s)

PIV measurement techniques: low repetition rate planar PIV high-speed PIV (HS-PIV) stereoscopic PIV (SPIV) tomographic PIV (TPIV).



Full cycle SI: Darmstadt optical engine – automatic mesh generation

snappyHexMeshGeometry-oriented block

structured meshSTL + data

• A python program automatically recognizes the direction of ducts, cylinder and valves and generates a geometry-oriented background grid

• snappyHexMesh is then run using the geometry-oriented background mesh

• The user provides combustion chamber geometry and data (bore, stroke, valve lifts)

0

2.5

5

7.5

10

0 180 360 540 720

Lift

[mm

]

Crank Angle [deg]

Exhaust

Intake

Bore 80 mm

Stroke 60 mm

Conrod 160 mm



Full cycle SI: Darmstadt optical engine – mesh management

Full-cycle simulations:

- Multiple meshes- Mesh to mesh interpolation

strategy.

Duration of each mesh: - User defined + quality

criteria

Initial mesh at Crank angle q0

Move mesh for Dq

Mesh quality and duration

satisfied?

qcurr = q0

qcurr = qcurr + Dq

YES

NO

Move surface geometry to current crank

angle qcurr

qcurr = qend ?

End of meshing

NO

YES

Generate a new mesh with

snappyHexMesh

q0 = qcurr



Full cycle SI: Darmstadt optical engine – case setup

Models and boundary conditions

Bore 86 mm

Stroke 86 mm

Compression ratio 8.5

IVO 325 CAD

IVC 485 CAD

EVO 105 CAD

EVC 345 CAD

Speed 850 rpm

Combustion no

Engine geometry data Boundary conditions

• Unsteady boundary conditions (from exp. data) imposed at inlet and outlet ports.

CFD models

• Second-order discretization(TVD)

• Turbulence model: standard k-e



Full cycle SI: Darmstadt optical engine – case setup

Computational mesh

• 4 mm mesh size in the ducts region;• 2 mm mesh size in the cylinder and valve region;• local refinement up to 1 mm close to cylinder head,

piston and liner boundaries;• local refinement up to 0.25 mm close to the valves

boundaries.

Mesh layouts

a) Cartesian mesh (automatically generated + snappyHexMesh)

b) Flow-oriented mesh (automatically generated + Polimitool + snappyHexMesh)



Full cycle SI: Darmstadt optical engine – validation




x

y

0

-10

-20

-30

x and y velocity components along four different

measurement lines located at different distances from the

cylinder head.

Uy

[m/s

]

Flow-oriented Cartesian Experimental

Y = 0 mm Y = -20 mm Y = -30 mmY = -10 mm

450 CADmid-intake

Similar behavior

between the two

grids, flow

dominated by the

incoming air jet

Flow-orientedCartesianExperimental




x

y

0

-10

-20

-30



cylinder head.

450 CADmid-intake

Similar behavior

between the two

grids, flow

dominated by the

incoming air jet

Flow-orientedCartesianExperimental

Ux

[m/s

]


Y = 0 mm Y = -20 mm Y = -30 mmY = -10 mm




x

y

0

-10

-20

-30



cylinder head.

Flow-orientedCartesianExperimental540 CADBDC, intake

Better prediction of

the flow oriented grid:

Vortex location,

distribution of the two

main streams

Uy

[m/s

]


Y = 0 mm Y = -20 mm Y = -30 mmY = -10 mm




x

y

0

-10

-20

-30



cylinder head.

Flow-orientedCartesianExperimental540 CADBDC, intake

Better prediction of

the flow oriented grid:

Vortex location,

distribution of the two

main streams

Ux

[m/s

]


Y = 0 mm Y = -20 mm Y = -30 mmY = -10 mm




x

y

0

-10

-20

-30



cylinder head.

Flow-orientedCartesianExperimental660 CADmid-compression

Flow-oriented grid

better describe the

tumble vortex

structure

Uy

[m/s

]


Y = 0 mm Y = -20 mm Y = -30 mmY = -10 mm




x

y

0

-10

-20

-30



cylinder head.

Flow-orientedCartesianExperimental660 CADmid-compression

Flow-oriented grid

better describe the

tumble vortex

structure

Ux

[m/s

]


Y = 0 mm Y = -20 mm Y = -30 mmY = -10 mm


SI Engines: GDI fuel-air mixing 43

Stratified engine (collaboration with IM-CNR, Ing. Sementa, Ing. Montanaro)

Optically accessible GDI engine

bmep = 7.2 bar, SA = 13 BTDC; l = 1.15 (lean)

Injection pressure[bar]

SOI[° BTDC]

Chargestratification

1 100 60 High

2 100 110 Low

3 60 60 High

4 60 120 Low

Operating points


Pinj = 100 bar; SOI = 110 BTDC

Pinj = 100 bar; SOI = 110 BTDCFuel m.f.

Wall-film

Soot chemiluminescence

l distribution

0 0.12

SI Engines: GDI fuel-air mixing 44

Stratified engine

Optically accessible GDI engine: optical/computed data correlation

Possible sources of soot:

• Rich pockets• Wall-film

Correlated with optical soot chemiluminescence

Pinj = 60 bar; SOI = 70 BTDC

Pinj = 60 bar; SOI = 60 BTDCFuel m.f.

Wall-film

Soot chemiluminescence

l distribution

0 0.12


SI Engines: combustion 45

Modeling

Comprehensive combustion model•Detailed description of the flame kernel growth process

via Lagrangian approach and suitable sub-models(breakdown, electrical circuit, wrinking)• Coherent flamelet model for flame propagation in the

Eulerian phase (gas)• Strict coupling between Eulerian and Lagrangian phases

Lagrangian particles for the spark-channel

Mass and energy equations solved for

the flame kernel particles

Particles convectedby the flow,

possibility to predict restrike

Sub-models

Rp Rs

Lp LsSpark

gap

Secondary circuit energy transfer

Breakdown

ri

Ti (>10000 K) Tu (300-600 K)

hQel

cell

N

i

ii

spkV

Sf

1

Flame surface density

Possibility to predict the effects of local flow, mixture conditions, turbulence, electrical circuit properties (voltage, current). Prediction of misfire also possible.



Assessment and validation

Fully turbulent flame

(Eulerian model only)

Initial combustion stage

(Lagrangian – Eulerian coupling)

• Pressurized vessels

Applications Results

• Multi-ignition systems

• Fan-generated flow velocity and turbulence fields at the spark-gap

This work is part of Lorenzo Sforza PhD project



Experimental validation: Herweg and Maly Engine

This work is part of Lorenzo Sforza PhD project

0

0.4

0.8

1.2

1.6

2

0 0.2 0.4 0.6 0.8 1

En

fla

mm

ed

vo

lum

e [

cm

3]

Time after spark [ms]

Calc., 1000 rpm, l = 1.0

Calc., 1000 rpm, l = 1.0

Exp., 1000 rpm, l = 1.0

Exp., 1000 rpm, l = 1.0

l

l

l

l

Peripheral

Central

0

0.4

0.8

1.2

1.6

2

0 0.2 0.4 0.6 0.8 1

En

fla

mm

ed

vo

lum

e [

cm

3]

Time after spark [ms]

Calc., 1000 rpm, l = 1.0

Calc., 300 rpm, l = 1.0

Calc., 1250 rpm, l = 1.0

Exp., 1000 rpm, l = 1.0

Exp., 300 rpm, l = 1.0

Exp., 1250 rpm, l = 1.0

l

l

l

l

l

l

Flow field, flame propagation and plasma temperature details


Conclusions 48

CFD modeling of in-cylinder phenomena at Polimi with OpenFOAM

• Detailed models continuously validated and improved: Fundamental studies Applied research

• Consolidated methodologies, currently applied in the context of industrial collaborations

• …but there is still a lot to do!

Thanks for your attention!

In-cylinder flows and combustion modeling: … flows and combustion modeling: application and...

Documents

Transcript of In-cylinder flows and combustion modeling: … flows and combustion modeling: application and...