1. Topic 7 - Spray B in ENGINES
ECN 4
Topic 7 - Spray B in ENGINES
Louis-Marie MALBEC, IFPEN
Mark Musculus, W. Ethan Eagle, Sandia National Labs
Amin Maghbouli, Gianluca d’Errico, Tommaso Lucchini, Politecnico di Milano
Randy Hessel, Zongyu Yue, University Wisconsin-Madison

2. Objectives
Identify and understand the differences between the spray B in combustion vessels (no initial velocities, no wall interactions, constant temperature and density…) and in engines.
Spray B in Vessels (Topic 3,5)
Spray B in Engines (Topic 7)
Vessel 1
Vessel N
Engine 1 …
Unknown boundary conditions ≠
Database
Database 1
Known boundary conditions
Differences
Validation
Analyze
Analyze
Insight
Validation
Models
Models

3. Participants Engine 1 Database 1 Models Engine Vessel Engine PoliMi:
SNL: heavy duty engine
Ethan Eagle, Mark Musculus, Louis-Marie Malbec
KAIST: light duty engine
Jaeheun Kim, Choongsik Bae
Istituto Motori: light duty engine
Ezio Mancaruso
- Uncertainty on boundary conditions
- Experimental setup not suited for vapor penetration measurements
Many efforts have been done to meet the objectives of Topic 7.
Design of a LIF setup for vapor penetration (advanced)
Addition of a N2 loop to simulate EGR (work in progress)
Engine 1
Database 1
PoliMi:
Gianluca d’Errico, Tommaso Lucchini, Amin Maghbouli
University of Wisconsin:
Randy Hessel, Zongyu Yue
Models
Engine
Vessel
Engine
Carnegie Mellon:
Satbir Singh

4. Objectives
Identify and understand the differences between the spray B in combustion vessels (no initial velocities, no wall interactions, constant temperature and density…) and in engines.
Spray B in Vessels (Topic 3,5)
Spray B in Engines (Topic 7)
Vessel 1
Vessel N
Engine 1 …
Unknown boundary conditions ≠
Database
Database 1
Known boundary conditions
Sandia B211201
Differences
Sandia B211199
Validation
Analyze
Analyze
CMT B211200
Insight
Validation
Models
Models

5. Characteristics of the Cummins optical engine
Experimental setup
Characteristics of the Cummins optical engine
We use a large bore diesel engine, which limits the confinement of the sprays. Moreover, the swirl ratio is quite low. So, maybe the differences with a combustion vessel are not that important.

6. Liquid / Vapor penetration
Illumination by a pulsed red LED
Schlieren
Phantom v71
Frame rate: 25kHz
Exposure: 19us
Lens: 105mm f/2.5 2 1
Mie scattering
Color Phantom v611
Frame rate 67kHz,
Exposure: 14us
Lens =85mm f/1.4
Definition: First occurrence of normalized intensity lower than a threshold (3%) along a spray axis

7. Ignition Delay / Lift-off length
Chimiluminescence OH*
Intensified Phantom v71
Frame rate: 7.2kHz (1CAD)
Exposure 55us
Lens: 105mm UV f/4.5
Broadband Chimiluminescence
Color Phantom v611
Frame rate: 14.4kHz (0.5CAD)
Exposure: 65us
Lens: 85mm f/1.4

8. Experimental results Measures: Reference point: Parametric variations
Liquid length
Vapor penetration
Lift-off length and ignition delay
Averaged on 30 cycles
Reference point:
15%O2, 1500b, 900K, 22.8kg/m3
1200rpm, SOE = 355CAD, i.e. SOI ≈ 357.2CAD
Injector (H1:90.9µm; H2:91.7µm; H3:90.9µm)
Parametric variations
%O2: 13% and 21%
Pinj: 500b and 1000b
Ttdc: 800K and 1000K
ρtdc: 15.2kg/m3

9. Engine vs Vessel: Liquid Length
Engine SNL - B211199
Vessel SNL - B211201
Vessel CMT - B211200
Time ASOE [µs]
=> Same type of increase and then decrease of the liquid length (vapor angle variation) between SNL engine and vessel. B at CMT behaviors is slightly different.
B211200
CMT (B211200) < SNL ENG (211199) < SNL VSL (211201)
But differences come from the test rigs or from the injectors?
Malbec et al., SAE paper
=> Up to 20% variation in liquid penetration for nominally identical injectors in the same test rig!!!

10. Engine vs Vessel: Vapor penetration
CMT and Sandia Spray B vapor penetrations are similar.
Both are significantly lower than Spray A (wider initial spray angle).
2 sets of vapor penetrations from CMT and SNL, but at 22.8kg/m3, whereas in engine we have VP at 15.2kg/m3.
SNL – A201677– Vessel – 22.8kg/m3
SNL – B211201– Vessel – 22.8kg/m3
Use of the 1d spray model to generate VP at 15.2kg/m3.
CMT – B – 22.8kg/m3
1d model – 15.2kg/m3
SNL - B ENG 5.2kg/m3
1d model – 22.8kg/m3
Vapor penetrations are “similar”
But the measured rate is higher than the one obtained with 1d model

11. Expe. Setup: Vapor detection
Simultaneous IR, LIF and Schlieren images
Infra-red images
Volume LIF
SChlieren

12. Expe. Setup: Vapor detection
Simultaneous IR, LIF and Schlieren images
Very good match between 3 techniques!
Schlieren VP also seems to have a higher penetration rate.

13. Engine vs Vessel: ID and Lift-off Length
Ignition delay
Lift-off length
B211200
B211200
B211200
Combustion indicators are similar between spray B in vessels and engines

14. Engine vs Vessel: ID and Lift-off Length
Comparison with the scaling law (Pickett et al., SAE Paper )
H=C.Ta-3.74.ρa-0.85.d0.34.U0.Zst-1
This scaling law has been obtained for a free jet.
LOL of spray B in engine has a similar behavior as a free jet
Weak influence of confinement, unsteady boundary conditions, and surrounding flow

15. Expe. Setup: LOL detection
High-speed color camera
High-speed intensified camera

16. Expe. Setup: LOL detection
The exact same structures can be observed on the blue channel of the color camera and on the intensified camera.
Valid if no broadband emission from soot reflects on the background.
High speed camera + “blue” filter can be used to detect lift-off

17. Summary: Spray B in Engine vs vessels
Liquid length
LL in engine shorter than in SNL Vessel
slightly higher to CMT vessel
Same transient behavior as in SNL vessel
Vapor penetration
Similar to SNL/CMT vessels, but rate is slightly different
Ignition delay and LOL
Similar to vessels (slightly lower).
LOL: Same trends as scaling law (i.e. as a free jet).
Limitations: different injectors have been used in different facilities.
Differences between spray B in vessels and in engine
Combustion vessels are representative of engine operation!

18. Objectives
Identify and understand the differences between the spray B in combustion vessels (no initial velocities, no wall interactions, constant temperature and density…) and in engines.
Spray B in Vessels (Topic 3,5)
Spray B in Engines (Topic 7)
Vessel 1
Vessel N
Engine 1 …
Unknown boundary conditions ≠
Database
Database 1
Known boundary conditions
Differences
Validation
Analyze
Analyze
Insight
Validation
Models
Models

19. Models description BDC TDC Mesh Specification Geometry
UWM
POLIMI
KIVA-3vr2 
Open Foam
Spray breakup
KH-RT instability
KH-RT model
Droplet collision
Radius of influence (ROI) model
​No collision model applied​
Near nozzle momentum exchange
Gas-jet model
Gas-jet model​
Droplet evaporation
Discrete-multi-component model
Spalding
Turbulence
General renormalized k-ε model
Standard k-ε model
Heat Transfers
Law of the wall
Ranz-Marshall
Injection
Blob-injector model
Chemical Mecanism
Pei
Combustion Model
 Homogeneous reactor
RIF / CCM
Mesh Specification
Geometry
120 degree sector cylindrical mesh
# of cells
~ 35000
Radial
Axial
Azimuthal
Dimensions
5.4 cm
3 cm
120˚
Resolutions
2.35 mm
1.30 mm
2.2˚
BDC
Number of cells: 341562
Min volume = 1.2e-11 m3
Max volume = 1.1e-08 m3
Mesh non-orthogonality
Max: average: 24.41
TDC
Number of cells: 45916
Min volume = 1.21e-11. m3
Max volume = 9.85e-09 m3
Mesh non-orthogonality
Max: Average: 11.16

20. Operating conditions
POLIMI: Several parametric variations, only in engine
UWM: 2 parametric variations, in engine and vessel
Operating Conditions
UWM
POLIMI
Spray B
ENG, VSL
ENG
800K
1000K
13% O2
21% O2
15.2kg/m3
500bar
1000bar

21. CFD results
UWM

22. CFD results: Spray B
POLIMI
UWM

23. CFD results: Spray B
POLIMI
UWM

24. Summary: Spray B in Engine vs vessels
Liquid length
LL in engine shorter than in SNL Vessel
slightly higher to CMT vessel
Same transient behavior as in SNL vessel
Vapor penetration
Similar to SNL/CMT vessels, but rate is slightly different
Ignition delay and LOL
Similar to vessels (slightly lower).
LOL: Same trends as scaling law (i.e. as a free jet).
Differences between spray B in vessels and in engine
Combustion vessels are representative of engine operation!

25. CFD results: Liquid penetration
From UWM, liquid length in engine is shorter than in vessels, similar to what is observed in experiments (B in engine vs B in vessel).

26. CFD results: Liquid penetration
UWM
UWM
Engine CFD vs Engine EXP:
Increase in penetration from 900 K to 800 K case is captured by model, but magnitude of increase is under-predicted.
CVCC CFD vs Engine EXP:
Same comment as ‘Engine CFD vs Engine EXP’, except the magnitude of increase in going from 900 to 800 K is closer to experimental values.
CVCC CFD vs Engine CFD:
CVCC penetration is greater at times < 1.2 ms, but at times > 1.2 results are very similar for both 900 and 800 K cases.
LL shorter in engine until about 1.2ms. Hypothesis:
Swirl promotes droplet break-up and thus vaporization
At later timings (>1.2ms), swirl intensity decreases, and thus we obtain the same LL.

27. CFD results: Vapor penetration
POLIMI
UWM
SNL - B ENG 5.2kg/m3
SNL - B ENG 5.2kg/m3
Spray B in vessel
15.2kg/m3
Spray B in engine
22.8kg/m3
Vapor penetration
CFD in engine matches expe in vessel at 22.8kg/m3
CFD in engine lower than expe in engine at 15.2kg/m3
Fuel penetrates faster in vessel
Smaller bowl leads to higher pressure gradient?
Squish flow?

28. CFD results: Ignition delay
Pretty good match between expe and simu (except for LOL at 900K for UWM).

29. Objectives
Identify and understand the differences between the spray B in combustion vessels (no initial velocities, no wall interactions, constant temperature and density…) and in engines.
Spray B in Vessels (Topic 3,5)
Spray B in Engines (Topic 7)
Vessel 1
Vessel N
Engine 1 …
Unknown boundary conditions ≠
Database
Database 1
Known boundary conditions
Differences
Validation
Analyze
Analyze
Insight
Validation
Models
Models

30. CFD results: Insight in boundary conditions
The TDC temperature (core gases) in the engine is computed considering that the gases follow an isentropic compression.
The compression computed in CFD could be used to validate this hypothesis
POLIMI
UWM
Ttdc CFD
Ttdc isoS.
Spray B
922
925
800K
837
840
1000K
1067
1076
15.2kg/m3
959
965

31. LIF / Schlieren / IR Simultaneous LIF and Schlieren Schlieren
Pulsed blue LED
YAG Laser - 266nm
70mJ/pulse – Volume illumination (approx. 30x30mm)
Fuel: dodecane, no tracer
Schlieren
Phantom v71
Lens: 50mm f/11
Frame rate: 25kHz
Exposure 10µs
Dichroïc
Spatial filter (pinhole)
Band pass filter,
320nm, width 70nm
Volume LIF
Pimax (Intensified CCD)
Lens: 105mm UV f/4.5
Frame rate: 1 image / cycle
Exposure: 2µs (Gain=80)
???

32. Simultaneous LIF and Schlieren
LIF / Schlieren / IR
Simultaneous LIF and Schlieren

33. Conclusions: Wish list / Future work
Characterize the same injector in vessel and engines
This will allow a better quantification of differences between vessel and engines operations.
Define a methodology to produce CFD results
Absolute accuracy of the model is maybe not necessary, just the relative variation between engine and vessel.
Define a methodology to compare CFD and experimental results
What is to be trusted in expe/CFD results?

34. Conclusions: Benefits
Engine operation
Improvement of the control of the boundary conditions
Comparison with vessels
Use of CFD to validate the assumptions on boundary conditions
Better understanding of the phenomena affecting the jets in engines
Combustion modeling
Use of the “unique” consistent database between engine and vessels to:
Validate the impact of mesh on performances
Validate the modeling of confinement, flow/spray interactions…

35. Conclusions For this first session of Topic 7: 2 modeling groups
Accessible without too much additional work for modelers already working on spray B in vessels
1 experimental group (2 additional probably in near future)
Small bore engines are an expected next step
So……..

36. Compare AHRR reacting Engine CFD vs Engine EXP:
The model requires a little ‘boost’ of heat release at low and/or intermediate temperatures, as is seen in the bump of the red curve prior to main heat release.
CVCC CFD vs Engine (EXP and CFD):
CVCC heat release is earlier in the engine.

37. Compare Chemical HRR reacting
Same comments as AHRR slide, except this slide does not include the experimental results (ignore the green horizontal curve, it is a bug with the graphing program and it should just be a black line).