Topic 2 – Near Nozzle Flow › workshop › ECN4 › Topic2-SSom.pdf · Topic 2 – Near Nozzle...

Topic 2 – Near Nozzle Flow Presenter: Sibendu Som, Argonne National Laboratory September 5th, 2015

2

Contributors • Aachen

• Mathis Bode, Heinz Pitsch

• Argonne National Laboratory (ANL) • Qingluan Xue, Sibendu Som • Michele Battistoni from University of Perugia during sabbatical

• Bosch Research and Technology Center (Bosch) • Edward Knudsen, Eric Doran

• CMT • Pedro Marti, Raul Payri

• IFPEN • Chawki Habchi

• Sandia National Laboratory (Sandia) • Francois Doisneau, Marco Arienti, Joe Oefelein

• Experimental Data: Argonne National Laboratory • Chris Powell, Alan Kastengren

3

Modeling Objectives

• Focus on the near nozzle region (within first 10-15 mm) • Encourage high-fidelity simulations of fuel sprays to

understand the primary atomization physics • Capture the SOI and EOI physics for single and multi-hole

injectors • Capture the spray physics during the main injection process

for both single and multi-hole injectors • Study the capability of the different modeling approaches

(Lagrangian-Eulerian, Eulerian-Eulerian) and CFD frameworks (RANS, LES, DNS) for capturing the near nozzle physics

• Assess the importance of different models and modeling artifacts: Turbulence, Compressibility, primary and secondary atomization, …?

4

Conditions of Interest for Simulations • Spray A and Spray B operating conditions from Argonne:

(http://www.sandia.gov/ecn/refs.php?nam=Kastengren-2012-a#Kastengren-2012-a)

– Spray A: 675; Spray B: 201 – Priority 1: Pinj = 1500 bar; Priority 2: Pinj = 500 bar – Fuel temperature: Spray A - 343 K; Spray B – 338 K

Ambient gas temperature 303 (K)

Ambient gas pressure 2.0 (MPa) Ambient gas density 22.8 (kg/m3)

Ambient gas N2 (by volume) 100%

http://www.sandia.gov/ecn/refs.php?nam=Kastengren-2012-a%23Kastengren-2012-a

5

• Mass flow rate at the nozzle exit • Fuel spray penetration vs. time • Contour plots of projected density at 0.1 and 0.5 ms • Transverse mass distribution (projected density across the spray)

in ug/mm2 @ x = 0.1, 0.6, 2, 6, and 10 mm downstream to nozzle exit: o between 0.5 – 1.0 ms after SOI

• 2D contours of LVF at x = 0.1, 0.6, 2, 6, 10 mm at 0.5 and 1.0 ms • Mean droplet size (SMD) at x = 1, 4, 8 mm at 0.5 ms after SOI

o Mean SMD at the above axial positions vs. axial position o Distributions of SMD vs. radial position at the above axial positions

Data Needed from Spray A Simulations

Set-up conditions from Argonne

6

Outline • Spray A Comparison between all groups

• Aachen, ANL, Bosch, CMT, IFPEN, Sandia • Description of Simulation set-up from all groups • Side-by-side comparisons

• Liquid penetration, mass flow rate at nozzle exit • Liquid volume fractions, projected mass density, SMD

• Parametric variations – RANS vs. LES (ANL) – Effect of Injection pressure (Bosch) – Effect of wall temperature (CMT) – Effect of fixed vs. moving needle (CMT) – Effect of primary atomization modeling (IFPEN)

• Spray A vs. Spray B dynamics (ANL) • Concluding Remarks/Suggestions

7

Comparison between different groups

8

Institution Model approach CFD Code Compressibility Turbulence model

Aachen Two-Fluid Eulerian/Lagrangian

CIAO Low Mach LES/DNS Smagorinsky

Argonne Single-mixture Eulerian

CONVERGE Compressible liquid Compressible gas (EOS)

RANS (k-e) and LES (DS)

Bosch Two-Fluid multi-species Eulerian

Cascade Technologies

Compressible liquid & gas Peng-Robinson EOS

LES Vreman

CMT Single-mixture Eulerian

OpenFOAM (ESA)

Compressible liquid (function of P, T) and gas EOS

RANS SST k-omega

IFPEN Two-fluid model with Lagrangian-Eulerian coupling and Primary atomization

IFP-C3D Compressible liquid and gas

LES Smagorinsky

Sandia (two-phase)

Eulerian Multi-Fluid with multi-component gas and dense spray

Raptor Liquid: Singularly compressible Gas: Perfect gas EOS

LES Dynamic Smagorinsky

Model description

9

Institution Mesh resolution Needle motion Initial conditions inside injector

2-D or 3-D

Aachen Nozzle: 2.7 μm; Prim. Breakup: 1.5 μm; Lag. Spray: 0.25 mm

Fixed needle lift Fuel filled sac and orifice; 150 MPa and 343 K in Sac and orifice

3-D

Argonne 10 μm inside nozzle and first 6 mm

Transient needle motion – with wobble

Fuel filled sac and orifice; 150 MPa and 343 K otherwise

3-D Height = 200 mm D = 50 mm

Bosch Voronoi polyhedra; 10 μm inside nozzle; coarsens downstream

Fixed needle lift Liquid fuel halfway through nozzle; 156 MPa rail pressures; 900K & 300K gas

3-D 45 mm long chamber

CMT About 30 μm uniform

Fixed and moving needle lift

Fuel filled sac and orifice: 150 MPa and 343 K in sac and orifice

2-D axisymmetric Height = 12 mm Width = 6 mm

IFPEN Non-uniform hexahedra : Min. size=(5µm in the hole,2µm in the seat, 15µm near-nozzle region)

With needle motion - Without wobble

Chamber-needle-seat : (0.999, 150 Mpa, 343K) Sac and hole : (0.0001,2MPa,, 343K) Chamber : (0.0001,2MPa,303K)

3D configuration including: - control volume - Chamber (bore=6

cm, height =6cm)

Sandia

Cartesian 12.5μm Fixed needle lift 6MPa, 900K/363K 3D

Simulation set up (Spray A)

10

Institution Geometry (210675) Total number of computational cells

Computational time for first 100 microseconds (CPU hours)

Aachen Low resolution Nozzle: 300 million Primary Breakup: 900 million Lagrange Spray: 7 million

Argonne Low resolution 4.50 millions 12,288

Bosch Low resolution

3.20 million 7,000

CMT Low resolution 0.09 millions 160: fixed needle 1350: moving needle

IFPEN High-resolution (CNRS scanned)

3 millions 11,776

Sandia (two-phase)

Box 9.6x3.2x3.2mm3

50 millions 86,000

Spray A: Computational time (first 100µs)

The computational domain size, initial conditions, and needle motions are different for the institutions

11

Computational domain and mesh image

CMT

Dx = 5-10 µm

20 mm

Control volume

IFPEN

Argonne

Bosch

Aachen

Sandia

12

Mass Flow rate at nozzle exit

• In general, most groups are able to predict the mass flow rate quite well • The initial transients are not well predicted by any approach. In some cases this might be

due to the fact that fixed needle simulations are being performed • Bosch results are under predicting mass flow rate due to the equation of state that they

employed

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

0 0.25 0.5 0.75 1 1.25 1.5

Mas

s flo

w ra

te (g

/s)

Time after start of injection (ASOI) (ms)

CMT's Virtual ROI GeneratorArgonneBoschCMT-newIFPEN-ECN3Sandia (two-phase)

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

0 0.1 0.2 0.3 0.4 0.5

Mas

s flo

w ra

te (g

/s)

Time after start of injection (ms)

CMT's Virtual ROI GeneratorArgonneBoschCMT-newIFPEN-ECN3Sandia (two-phase)

13

Liquid Penetration

• Sandia simulations are over penetrating (note that the injection BC has no transient) • IFPEN: ECN3 results are shown. ECN4 results may be able to better predict liquid

penetration

• Argonne and Aachen results compare quite well with the experimental data

• Bosch results not shown, since they performed simulations with a fixed lift and the initial liquid penetration is not well predicted although vapor penetration is well captured (not shown here)

• CMT simulations considering needle transients matches best with the experimental data

0

1

2

3

4

5

6

7

8

0 5 10 15 20 25

Liqu

id p

enet

ratio

n (m

m)

Time after start of injection (μs)

ExperimentAachenANL-LESCMT-MovingNeedleIFPEN-ECN3Sandia (two-phase)

14

Mass/area (μg/mm2)

Spray A: projected mass density – 0.1ms ASOI

X-ray data

Argonne RANS

CMT RANS

IFPEN LES

All simulations predict similar mass flow rate at nozzle exit, however, quite different ensuing spray

Qualitatively dense liquid core near nozzle region is captured by all simulations

Argonne LES

RANS models tend to be overly diffusive compared to LES LES models tend to over-predict the core length and under predict spray dispersion

15

Spray A: projected mass density – 0.5ms ASOI X-ray

Mass/area (μg/mm2)

At steady-state, the observations are consistent with earlier injection time (0.1 ms)

Geometry Argonne RANS

CMT RANS

IFPEN LES

Argonne LES

Time = 0.3 ms

16

Projected mass density and fuel density - LES

IFPEN: LES

BOSCH results show liquid density on a cut-plane through jet center Argonne and IFPEN results show projected mass density, also shown in previous slides Bosch calculations show flow features like shock waves and capture flow structures Argonne and IFPEN result are line-of-sight averaged and hence do not show these features Argonne and BOSCH calculations seems to be broader than IFPEN results

Argonne: LES

Bosch: LES

0.1 ms 0.5 ms

IFPEN: LES; Time = 0.3 ms

Argonne: LES

Bosch: LES

17

0

10

20

30

40

50

60

70

80

-0.12 -0.09 -0.06 -0.03 0 0.03 0.06 0.09 0.12

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )

Transverse distance (mm)

x= 0.6mmExperimentANL: RANSANL: LESCMT: RANSIFPEN: ECN3-LES

Spray A: Radial profiles of projected density density 0.5ms ASOI

All simulations capture the shape of the radial profile, however, significant differences in the peak values and tails

The lower peak values from IFPEN (from ECN3) are possibly due to the choice of EOS

0

10

20

30

40

50

60

70

80

-0.1 -0.075 -0.05 -0.025 0 0.025 0.05 0.075 0.1

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )



0.5m

s ASO

I 0.1 mm

0.6 mm

2.0 mm

18

Spray A: Radial profiles of projected density

@ 2mm, ANL and CMT calculations can predict the projected mass density contours fairly well

@ 6 mm none of the modeling approaches can capture both the peak values and dispersion well. This may suggest that at these axial locations, transition to an Lagrangian approach may be necessary Open question: What should be the criteria for transition from Eulerian to Lagrangian?

Since projected density is already calculated by line of sight integration, perhaps multiple realizations are not necessary with LES to get smoother profiles

0.5m

s ASO

I

0

10

20

30

40

50

60

70

-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )



0

5

10

15

20

25

30

35

40

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )


x=6.0mmExperimentANL: RANSANL: LESCMT: RANSIFPEN: ECN3-LES

19

• No structures observed for 0.6 mm and 2 mm near nozzle regions • At 6 mm, turbulent and transient structures are observed • 10 mm location does not have enough resolution to show any structures

Bosch Contribution: LVF with LES 2

mm

20

LVF with LES: Bosch and ANL

Argo

nne:

LES

0.6 mm 2.0 mm

LVF

• 0.6 mm and 2 mm are axial distances from nozzle exit • Comparison between ANL and Bosch is not apples-to-apples • However, ANL LES results shows significant asymmetry • BOSCH results seem to show asymmetry and turbulent structures at further downstream

locations

BOSC

H

2 m

m

21

SMD at different locations

• Aachen provided results in terms of a droplet size frequency distribution

• CMT provided radial distribution of SMD • Sandia provided SMD as a function of axial distance • Overall, SMD is over predicted by Sandia’s approach

which is still under development • Experimental uncertainties may still persist

0

1

2

3

4

5

6

7

8

9

10

0 0.5 1 1.5 2 2.5 3

Freq

uenc

y Di

strib

utio

n (%

)

Droplet size (μm)

1 mm4 mm8 mm

Aachen

0

1

2

3

4

5

6

7

8

0 0.001 0.002 0.003 0.004 0.005

SMD

(μm

)

Transverse position (m)

1mm4mm8mm

CMT 0

1

2

3

4

5

0 2 4 6 8

SMD

(μm

)

Axial Distance (mm)

ExperimentAachenCMTSandia (two-phase)

22

Parametric Studies

23

X-ray data 15 m

icrons grid

RANS standard k-epsilon

LES Dynamics structure

Sct=0.5

Sct=0.7

LES Dynamics structure

Sct=0.7 Cs =0.05

LES Smagorinsky

Sct=0.7 Cs =1.0

LES Smagorinsky

RANS vs. LES

• ANL conducted a systematic study on the influence of turbulence model constants and Schmidt number of spray penetration and dispersion. The most promising results are shown here

• Smagorinsky model does not seem to show onset of turbulence structures

• Schmidt number seems to have a minor effect on the results

• In general, all turbulence models over predict the spray penetration

24

ANL Contribution: RANS vs. LES

• Penetration is fairly well captured by both RANS and LES models

• Liquid volume fraction results with RANS are much more diffused compared to LES, which is expected

• Higher fuel density at the spray core is observed with LES calculations

0

1

2

3

4

5

6

7

8

9

10

0 5 10 15 20 25 30

Liqu

id p

enet

ratio

n (m

m)


ExperimentRANS: k-e modelLES: DS model

LES: Dynamics structure

RANS: Standard k-ε

0.1 mm

0.6 mm

2.0 mm

25

0

10

20

30

40

50

60

70

80

-0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )


x= 2.0mmExperimentRANS: k-e modelLES: DS model

0

10

20

30

40

50

60

70

80

-0.15 -0.1 -0.05 0 0.05 0.1 0.15

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )


x= 0.6mm ExperimentRANS: k-e modelLES: DS model

0

10

20

30

40

50

60

70

80

-0.1 -0.075 -0.05 -0.025 0 0.025 0.05 0.075 0.1

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )


x= 0.1mm ExperimentRANS: K-e modelLES: DS model

ANL Contribution: RANS vs. LES

0

5

10

15

20

25

30

35

-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )


x=6.0mm ExperimentRANS: k-e modelLES: DS model

• In the near nozzle region (less than 2 mm), turbulence model effects seem to be minimal. LES with DS is perhaps performing better than RANS

• Based on previous studies at Argonne, resolution seems to influence the results more significantly rather than the choice of turbulence models

• Beyond 2 mm, LES seems to capture the dispersion well while RANS does a better job in capturing the peak values

26

560

bar

1560

bar

Bosch Contribution: Effect of Injection Pressure

• Vapor penetration curves agreed well with experiments after initial transient

• Penetration scaled as expected with rail pressure

27

CMT contribution: Effect of Wall Temperature

• All results with fixed needle lift • Wall temperature:

• Influences the mass flow rate as expected (density effect) • Has a relatively small and non-linear effect on near nozzle spray penetration

• Adiabatic wall influences spray penetration, although mass flow rate was not influenced

0

1

2

3

4

5

6

7

8

9

10

0 5 10 15 20 25 30

Liqu

id p

enet

ratio

n (m

m)


ExperimentCMT-273KCMT-303KCMT-343KCMT-343K-AdiabaticCMT-363K

1.5

1.75

2

2.25

2.5

2.75

0 0.1 0.2 0.3 0.4 0.5

Mas

s flo

w ra

te (g

/s)

Time after start of injection (ASOI) (ms)

CMT's Virtual ROI GeneratorCMT-273kCMT-303kCMT-343kCMT-343k-adiabaticCMT-363k

28

• All results with fixed needle lift • Peaks are marginally different but dispersion is quite similar for all wall temperatures • Adiabatic walls does not influence the projected density results

CMT: Effect of Wall Temperature (up to 2 mm)

0

10

20

30

40

50

60

70

-0.1 -0.075 -0.05 -0.025 0 0.025 0.05 0.075 0.1

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )


x= 2.0mmCMT-273kCMT-303kCMT-343kCMT-343k-adiabaticCMT-363k

0

0

0

0

0

0

0

0

-0.1 -0.075 -0.05 -0.025 0 0.025 0.05 0.075 0.1



0

10

20

30

40

50

60

70

-0.1 -0.075 -0.05 -0.025 0 0.025 0.05 0.075 0.1

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )



0.1 mm

0.6 mm

2.0 mm

29

273 K

303 K

323 K

343 K

363 K

Projected Mass Density (μg/mm2)

• Contour plots of projected mass density profiles looks quite similar for all temperatures

• Peaks are marginally different but dispersion is quite similar for all wall temperatures

• Adiabatic walls does not influence the projected density results

CMT: Effect of Wall Temperature

0

2

4

6

8

10

12

14

16

18

20

-1.8 -1.4 -1 -0.6 -0.2 0.2 0.6 1 1.4 1.8

Proj

ecte

d m

ass

dens

ity (u

g/m

m2 )


x=10.0mmCMT-273kCMT-303kCMT-343kCMT-343k-adiabaticCMT-363k

30

Primary atomization using Two-Surfaces density model and EE coupling with EL approach

• The liquid core length is overestimated, perhaps due to a lack of initial dispersion of droplets in the primary atomization model

• More work is needed for the primary atomization model

ECN3 Results

ECN4 Results

IFPEN results from Chawki Habchi

IFPEN: Spray A @ subcritical conditions

Iso-surface αliq = 0.05

32

Spray A vs. Spray B Dynamics

33

• Transient external spray for Spray B observed by experiments1,2,3,4

• Numerical simulation of injector to study the internal flow and near-field spray development

3

1. Jung Y., Manin J., Scott S., Pickett L.M., SAE Technical Paper No. 2015-01-0946, 2015 2. Kastengren A., Xue Q., Manin J., Habchi C., ECN3, 2014 3. Kastengren, A.L., Tilocco, F.Z., Duke, D., Powell, C.F., Moon, S., and Zhang, X., ILCASS 2012, Heidelberg, September 2-6, 2012 4. Engine Combustion Network (ECN), www.sandia.gov/ecn

Experimental Observation: Spray B dynamics

http://www.sandia.gov/ecn

34

ANL Contribution: Spray A vs. Spray B dynamics

0

0.2

0.4

0.6

0.8

1

-0.4 -0.3 -0.2 -0.1 0

Liqu

id m

ass f

ract

ion

(-)

radial distance (mm)

0.05 ms0.1 ms0.20 ms0.30 ms0.60 ms

0

0.2

0.4

0.6

0.8

1

-0.4 -0.3 -0.2 -0.1 0

Liqu

id m

ass f

ract

ion

(-)


0.05 ms0.1 ms0.20 ms0.4 ms0.5 ms0.80 ms

Hole 3

Spray B

• Spray A has negligible variation in radial distribution at this axial location. It seems to have reached a quasi-steady state

• Spray B has significant radial spreading variation at different time instants. In fact, the variation in spreading is non-monotonic

Spray A

At 20 D of downstream

Hole 3

20 D

35

0

0.2

0.4

0.6

0.8

1

-0.4 -0.3 -0.2 -0.1 0

Liqu

id m

ass f

ract

ion

(-)

radia distance (mm)

0.05 ms ASOI0.1 ms ASOI0.2 ms ASOI0.40 ms ASOI0.50 ms ASOI0.80 ms ASOI

0

0.2

0.4

0.6

0.8

1

-0.4 -0.3 -0.2 -0.1 0

Liqu

id m

ass f

ract

ion

(-)


0.05 ms0.1 ms0.20 ms0.4 ms0.5 ms

Hole 1

Hole 3

At 20 D of downstream

• Spray transient behavior observed for all holes • Hole 3 has a larger spray radial spreading and variation • Maximum variation of 60 microns for hole 3 • Overall variation in spreading from simulations smaller than that observed

in the experiments

Hole 1

Hole 2

Hole 3

20 D

ANL contribution: Spray B dynamics

36

• Turbulent kinetic energy is higher for Spray B inside the hole than Spray A • Velocity streamline shows flow turns at the inlet of the hole • Flow is more transient for Spray B vs. Spray A

Velocity at 0.1 ms TKE at 0.1 ms

Spray B

Spray A

ANL Contribution: Spray A vs. Spray B dynamics

37

Concluding Remarks/Suggestions • A well developed Eulerian-based 2D approach seems to work quite well in

predicting the spray structure in the near nozzle region. The higher fidelity approaches still need further development

• In the near nozzle region, resolution seems to influence the results more than the choice of turbulence models

• Transitioning to an Lagrangian approach somewhere between 2 mm to 6 mm is necessary. Unfortunately, we do not have experimental data in this region

• Even with the Eulerian model, upstream of 2 mm, the need for a proper atomization model (to initiate the onset of instabilities) is imperative

• Note: None of the models presented use phase interface tracking (apart from Aachen, with DNS), and hence liquid gas interface is not enforced. Most approaches instead, model small scale breakup as a turbulent diffusion process.

• Wall temperature has an appreciable influence on the spray behavior, especially on mass flow rate at nozzle exit and spray penetration. Projected density contours are minimally affected, at least with the temperature range investigated here

• Accounting for needle movement (and initial transient mass flow rate) can significantly improve the predictions, although for a single hole injector needle wobble does not have a profound influence on results

38

Future work/Suggestions • RANS can capture some of the single vs. multi-hole dynamics. LES can probably better

predict the transients observed with multi-hole injector compared to the single hole injector

• From Sandia: The uncertainty introduced by temperature via surface tension seems much more worrying. The liquid temperature gradient is very strong close to the dense region, merely stabilized by vaporization. We would like to insist on the relevance of trying to experimentally assess droplet temperature in the future.

• From IFPEN: . In future work, Chawki will use the ligament length in order to release the primary droplets farther from the liquid core, thus promote droplets dispersion. This kind of work need some special experimental data. Chawki would like to ask the experimentalists to measure ligament length/diameter in order to help for primary atomization modeling.

Slide Number 1Slide Number 2Slide Number 3Slide Number 4Slide Number 5Slide Number 6Slide Number 7Slide Number 8Slide Number 9Slide Number 10Slide Number 11Slide Number 12Slide Number 13Slide Number 14Slide Number 15Slide Number 16Slide Number 17Slide Number 18Slide Number 19Slide Number 20Slide Number 21Slide Number 22Slide Number 23Slide Number 24Slide Number 25Slide Number 26Slide Number 27Slide Number 28Slide Number 29Slide Number 30Slide Number 31Slide Number 32Slide Number 33Slide Number 34Slide Number 35Slide Number 36Slide Number 37Slide Number 38Slide Number 39Results of Spray A LES/DNS SimulationsSimulation SetupResults: Flow Profile at Nozzle ExitAcknowledgement & ReferencesSlide Number 44Slide Number 45Slide Number 46Slide Number 47Slide Number 48Spray Modeling FrameworkSpray A MeshingGoverning Equations & NumericsEquation Of State: Peng-RobinsonBoundary ConditionsLiquid PropertiesCMT: Internal Modeling CodesSlide Number 56Slide Number 57Slide Number 58Slide Number 59

Topic 2 – Near Nozzle Flow › workshop › ECN4 › Topic2-SSom.pdf · Topic 2 – Near Nozzle...

Documents

Transcript of Topic 2 – Near Nozzle Flow › workshop › ECN4 › Topic2-SSom.pdf · Topic 2 – Near Nozzle...