You are on page 1of 27

Combustion for Low

Emission Application of
Natural Gas

Large Eddy Simulation of MILD


Combustion Regime

Zhiyi Li, Alberto Cuoci, Amsini Sadiki and


Alessandro Parente

1
Contents

• Introduction to MILD combustion


• Combustion models
• RANS simulation
• Large Eddy Simulation
• Conclusions and Prospects

2
Contents

• Introduction to MILD combustion


• Combustion models
• RANS simulation
• Large Eddy Simulation
• Conclusions and Prospects

3
Moderate or Intense Low oxygen Dilution (MILD) combustion

2.5
HM3 HM1
2.5 2.5
2
0.1
JHC burner, 9% O2 0.1
JHC burner, 3% O2 MILD
Burner Diameter (m)

Burner Diameter (m)


400 2 400 2
0.05 800 0.05 800
1200 1200
1.5 2000 1.5 1.5
0 801200
0 0 800
1600 1200
1200 1 1600 1
1 −0.05 800 −0.05
1200
800
400 400
Isaac et al., Energy and 0.5 0.5

0.5 −0.1 Fuels 27 (2013) 2255- −0.1


0 2265 0.2 0.4 0.6 0.8 1
0
0 0.2 0.4 0.6 0.8 1
0
Burner Length (m) Burner Length (m)
0
1

⌧⌘
Da⌘ = Features of MILD combustion: Advantages of MILD combustion:
⌧c • Distributed reaction zone • High efficiency
• More uniform temperature field • Low emissions (NOx and soot)
• No visible flame • Large fuel flexibility

4
Contents

• Introduction to MILD combustion


• Combustion models
• RANS simulation
• Large Eddy Simulation
• Conclusions and Prospects

5
Turbulent combustion modelling approaches

Flamelet-like
SLFM, FPI/FGM Transported PDF
Eddy Dissipation Concept (EDC)
PDF presumed models
Partially Stirred Reactor (PaSR)
Eddy Dissipation PDF computed
Finite Rate (EDFR)
Accuracy, cost

Selected models: EDC and PaSR

• Acceptable trade-off between accuracy and cost

• Affordable treatment of finite-rate chemistry in EDC and PaSR

EDFR used for preliminary LES tests

6
EDC and PaSR combustion models

𝜏∗
Fine
structures
PSR/PFR
T ⇤ , Yi⇤ , ⇢⇤, 𝑚̇ ∗
0
Cell
T , Yi0 , ⇢0 CFD grid

Mean reaction rate:


𝛾&' 𝜌̅ (𝑌"∗ − 𝑌1) 𝜈𝜖 9/;
𝜔̇ " =
(1 − 𝛾&+ ) 𝜏∗ 𝛾& = 𝐶4 Mixing time
𝑘8 scale
(Eddy Dissipation Concept )

𝛾& - mass fraction of the fine structures 𝜏< Mixing time scale
𝑌"∗ - 𝛾
species mass fraction in the fine structures & = + chemical time
𝜏< + 𝜏>"? scale
𝑌1 - mean species mass fraction in the cell
(Partially Stirred Reactor)
𝜏∗ - residence time in the fine structures

7
Contents

• Introduction to MILD combustion


• Combustion models
• RANS simulation
• Large Eddy Simulation
• Conclusions and Prospects

8
Adelaide Jet in Hot Co-flow (JHC) burner

Flame details
Fuel: CH4/H2 (50%, 50% volume)
Co-flow: mixture of CH4/H2 /air/CO2 /H2O/CO/N2 /O2
Co-flow Temperature: 1300 K
Fuel Jet Reynolds number: 10,000
Geometry
Fuel nozzle diameter: 4.25 mm
(Dally, 2002)
Coflow diameter: 82.0 mm

9
RANS simulation case setup

centerline

240 mm
Computational details
Domain: 2D axisymmetric (254 x 1000 mm)
Computational grid: ~30,000 cells
Solvers: edcPimpleSMOKE/ pasrPimpleSMOKE
(https://github.com/acuoci/edcSMOKE)
Multi-component molecular diffusion included
120 mm
Turbulent model: RANS k-epsilon
Combustion models: EDC & PaSR

60 mm
Kinetic schemes (Nox free)
KEE: 17 species, 58 reactions 30 mm

Inlet

10
RANS simulation with EDC and PaSR models

Mean temperate profiles:

Axial 30 mm Axial 60 mm
2000
Exp
PaSR
1500 EDC
T[K]

1000
• PaSR provides more
500
satisfactory results: chemical
0
0 20 40 60 0 20 40 60
time scale explicitly
r [mm] r [mm] considered.
Axial 120 mm Centerline
2000
• Sensitivity analysis on
turbulence and combustion
1500

model parameters.
T[K]

1000

500

0
0 20 40 60 60 120 180
r [mm] Axial direction [mm]

11
Contents

• Introduction to MILD combustion


• Combustion models
• RANS simulation
• Large Eddy Simulation
• Conclusions and Prospects

12
From RANS to LES
Computational details
Domain: 3D cylinder mesh (106mm x 425mm x 2𝜋)
Computational grid: 1,715,230 cells
Inflow generation: LEMOS 2.3.x (on fuel, co-flow and
air streams)
Inlet Solver: pasrPimpleSMOKE
Turbulent model: LES oneEqEddy (One Equation
Eddy Viscosity Model)
Combustion models: EDFR & PaSR
Kinetic schemes (NOx free)
Global Mechanism: 7 species, 4 reactions (W. P.
Jones and R. P. Lindstedt)

• Cell size gradient from fuel stream to air stream from inlet to
outlet
• Mesh clustering for the interaction regions between
different streams

Mesh Computational time Boundary Condition


RANS 30,000 cells 523 CPU hours Mean
LES 1,715,230 cells 22112 CPU hours Transient inflow 13
Validation of the temperature field with the LES EDFR model
Axial 30 mm Axial 60 mm Axial 30 mm Axial 60 mm
2000
Exp
oneBC 400
1500 twoBC
threeBC

Trms[K]
T[K]

1000
200
500

0 0
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
r [mm] r [mm] r [mm] r [mm]
Axial 120 mm Centerline Axial 120 mm Centerline
2000
Exp
400 oneBC
1500 twoBC

Trms[K]
threeBC
T[K]

1000
200
500

0 0
0 20 40 60 60 120 180 0 20 40 60 60 120 180
r [mm] Axial direction [mm] r [mm] Axial direction [mm]

mean variance

oneBC: velocity inflow generation on fuel stream alone


twoBC: velocity inflow generation on both fuel and co-flow steams
threeBC: velocity inflow generation on fuel, co-flow and air streams

Inflow generation required for fuel, co-flow and air streams


14
Comparison between RANS and LES-threeBC: impact on
mixture fraction & temperature distribution
Axial 30 mm Axial 60 mm Axial 30 mm Axial 60 mm
2000
0.75 Exp Exp
RANS RANS
LES-threeBC 1500 LES-threeBC
0.5

T[K]
f[-]

1000
0.25
500

0 0
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
r [mm] r [mm] r [mm] r [mm]
Axial 120 mm Centerline Axial 120 mm Centerline
2000
0.75

1500
0.5

T[K]
f[-]

1000

0.25
500

0 0
0 20 40 60 60 120 180 0 20 40 60 60 120 180
r [mm] Axial direction [mm] r [mm] Axial direction [mm]

Mean mixture fraction (from Bilger's definition) Mean temperature

No temperature over-prediction at 120 mm axial direction with LES

15
Comparison between RANS and LES-threeBC: impact on O2
(oxidizer) distribution

Axial 30 mm Axial 60 mm Axial 30 mm Axial 60 mm


0.09 Exp
0.24 LES-threeBC

[-]
YO [-]

0.16 0.06

2rms
2

YO
0.08 0.03

0 0
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
r [mm] r [mm] r [mm] r [mm]
Axial 120 mm Centerline Axial 120 mm Centerline
Exp 0.09
0.24 RANS
LES-threeBC

[-]
YO [-]

0.16 0.06

2rms
2

YO
0.08 0.03

0 0
0 20 40 60 60 120 180 0 20 40 60 60 120 180
r [mm] Axial direction [mm] r [mm] Axial direction [mm]

mean variance

16
Comparison between RANS and LES-threeBC: mean fuel
species (CH4 & H2) profiles

Axial 30 mm Axial 60 mm Axial 30 mm Axial 60 mm


0.12
Experimental Experimental
0.75 RANS RANS
LES-threeBC LES-threeBC
0.08
YCH [-]

YH [-]
0.5
4

2
0.04
0.25

0 0
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
r [mm] r [mm] r [mm] r [mm]
Axial 120 mm Centerline Axial 120 mm Centerline
0.12
0.75
0.08
YCH [-]

YH [-]
0.5
4

2
0.04
0.25

0 0
0 20 40 60 60 120 180 0 20 40 60 60 120 180
r [mm] Axial direction [mm] r [mm] Axial direction [mm]

CH4 H2

17
Comparison between RANS and LES-threeBC: mean product
species (H2O) profiles
Axial 30 mm Axial 60 mm Axial 30 mm Axial 60 mm
Exp Experimental
0.1 RANS 0.024 LES-threeBC
LES-threeBC

YH O [-]
YH O [-]

2 rms
0.016
2

0.05
0.008

0 0
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
r [mm] r [mm] r [mm] r [mm]
Axial 120 mm Centerline Axial 120 mm Centerline

0.1 0.024

YH O [-]
YH O [-]

2 rms
0.016
2

0.05
0.008

0 0
0 20 40 60 60 120 180 0 20 40 60 60 120 180
r [mm] Axial direction [mm] r [mm] Axial direction [mm]

Mean Variance

Clear under-prediction of mean H2O beyond 120 mm axial location

18
Energy resolved with LES

Centerline
Axial 30 mm Axial 60 mm
1
1

0.8
Ratio[-]

0.98
0.6

Ratio[-]
0.4
0.96
0 20 40 60 0 20 40 60
r [mm] r [mm]
Axial 120 mm Axial 200 mm
1 0.94

resolved energy
0.8 60 120 180
Ratio[-]

Axial direction [mm]


0.6

resolved energy
Centerline
0.4
80% line
0 20 40 60 0 20 40 60
r [mm] r [mm] 𝑒𝑛𝑒𝑟𝑔𝑦 𝑟𝑒𝑠𝑜𝑙𝑣𝑒𝑑
𝑅𝑎𝑡𝑖𝑜 =
𝑆𝐺𝑆 𝑘𝑖𝑛𝑒𝑡𝑖𝑐 𝑒𝑛𝑒𝑟𝑔𝑦 + 𝑒𝑛𝑒𝑟𝑔𝑦 𝑟𝑒𝑠𝑜𝑙𝑣𝑒𝑑
Axial positions

The 30 mm and 60 mm axial locations do not meet the 80% criteria

19
Refinement of the LES mesh

Inlet
Inlet

Cells Axial length (mm) Radial length (mm) Min/Max edge length (m)
LES-coarse 1,715,230 425 106 2.934e-05 / 0.0151239
LES-fine 1,555,200 180 90 3.34864e-05 / 0.00883218

20
Energy resolved with the coarse and the fine mesh

Coarse mesh Fine mesh


Axial 30 mm Axial 60 mm
1

0.8
Ratio[-]

0.6

0.4
The fine mesh is able to solve more
than 80% of the turbulent kinetic
0 20 40 60 0 20 40 60
r [mm] r [mm]

1
Axial 120 mm Centerline
energy on sampled positions of the
0.8
mesh.
Ratio[-]

0.6
resolved energy-coarse
0.4 resolved energy-fine
21
80% line
0 20 40 60 0 50 100 150
r [mm] r [mm]
Mean temperature profiles with fine mesh

Axial 30 mm Axial 60 mm
2000
Exp
coarse-mesh
1500 fine-mesh
T[K]

1000

500

0
0 20 40 60 0 20 40 60
r [mm] r [mm]
Axial 120 mm Centerline
2000

1500
T[K]

1000

500

0
0 20 40 60 60 120 180
r [mm] Axial direction [mm]

The centerline temperature over-prediction is corrected


with the fine mesh.
22
Extension from EDFR to PaSR combustion model

Axial 30 mm Axial 60 mm
Axial 30 mm Axial 60 mm
2000 Exp
Exp 0.1
PaSR PaSR
1500 EDFR EDFR

YH O [-]
T[K]

1000

2
0.05

500

0 0
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
r [mm] r [mm] r [mm] r [mm]
Axial 120 mm Centerline Axial 120 mm Centerline
2000

0.1
1500
T[K]

YH O [-]
1000

2
0.05
500

0
0 20 40 60 60 120 180 0
0 20 40 60 60 120 180
r [mm] Axial direction [mm]
r [mm] Axial direction [mm]

Preliminary results: PaSR model should be revised in the LES framework

Under-prediction of H2O profiles: effect of kinetic mechanisms to be investigated

23
Contents

• Introduction to MILD combustion


• Combustion models
• RANS simulation
• Large Eddy Simulation
• Conclusions and Prospects

24
Conclusions
• In RANS, PaSR model provides more accurate predictions than
EDC, since the chemical time scale is explicitly taken into
account.
• Turbulence inflow generation is necessary for the co-flow and
air streams in LES, to correctly reproduce RMS values.
• LES-EDFR model provides satisfactory predictions of
temperature and species mass fraction.
• Under-predictions of H2O mass fraction profiles on the
centerline probably come from the kinetic mechanism.

Prospects
• Extend LES model to include finite-rate chemistry.
• Couple the simulation with tabulation and mechanism reduction
methods to improve computational efficiency.

25
Aknowledgement

This project has received funding from the European Union’s


Horizon 2020 research and innovation program under the Marie
Sklodowska-Curie grant agreement No 643134.

26
Thank you!

27

You might also like