Low-Dimensional Manifolds in Direct Numerical Simulations ...purl.tue.nl/696285172893447.pdf · Direct Numerical Simulations of Autoigniting Mixing Layers Jeroen van Oijen, Ugur Göktolga,

Low-Dimensional Manifolds in Direct Numerical Simulations of Autoigniting Mixing Layers Jeroen van Oijen, Ugur Göktolga, Philip de Goey 4th IWMRRF San Francisco, June 2013

MILD combustion

•  Promising combustion concept •  High efficiency •  Low emissions •  Homogeneous temperature •  Low noise

•  Preheated and diluted reactants •  Exhaust gas recirculation •  High initial T: autoignition •  Low flame T: low NOx emissions

/ Mechanical Engineering PAGE 1 6/21/13

MILD

Conventional

Courtesy IFRF

Objective

Develop a numerical model for MILD combustion •  Flamelet based chemical reduction (FGM) •  Large eddy simulation (LES)

Multi-scale approach

•  Detailed simulations of micro-scale reaction structures (10-4 m) •  Reveal fundamental processes •  Develop reduced chemistry model (FGM)

•  Simulation of lab-scale flames (10-2 m) •  Develop turbulent combustion model (LES/FGM)

•  Application in industrial-scale burner systems (100 m)

/ Mechanical Engineering PAGE 2 6/21/13

Jet-in-Hot-Coflow (JHC) burners

•  Berkeley, Cabra et al. •  Adelaide, Dally et al. •  Delft, Oldenhof et al. •  …

/ Mechanical Engineering PAGE 3 6/21/13

DNS of turbulent mixing layers

•  Investigate reaction structures

•  Compressible equations •  Detailed chemistry: DRM19 •  Constant non-unity Lewis numbers

•  Compact finite difference •  6th order diffusion terms •  5th order convective terms

•  Low-storage 3rd order RK •  Time step equals 10 ns

/ Mechanical Engineering PAGE 4 6/21/13

Fuel

Hot

cof

low

Numerical implementation

•  2D domain 20x20 mm •  521x521 grid points, h=38µm

•  Validated with 1041x1041 •  Conditions similar to experiments of

Dally et al. 2002 •  Ufu - Uox = 67 m/s

Re = 3850 •  Three cases (HM1/2/3)

/ Mechanical Engineering PAGE 5 6/21/13

CH4-H2 1:1 T = 305 K

O2, N2, CO2, H2O T = 1300 K

Per

iodi

c b.

c.

Outflow b.c.

2 m

m

20 m

m

Ufu

Uox

J.A. van Oijen (2013) Proc. Combust. Inst. 33:1163-1171

DNS results 2D mixing layer

/ Mechanical Engineering PAGE 6 6/21/13

H2 H ΔT = T – Tmix

HM1: YO2 = 0.03 Ka = τig / τη= O(10)

Green lines: Zst = 0.0067

Ignition time

•  Maximum YH and temperature rise ΔT as a function of time •  Comparison with laminar mixing layer and 0D homogeneous ignition

/ Mechanical Engineering PAGE 7 6/21/13

Scalar dissipation rate

/ Mechanical Engineering PAGE 8 6/21/13

Grey dots: Scatter Symbols: Cond. average log χ Bold line: Laminar

t = 0.2 ms

Scatter plots in mixture fraction space

/ Mechanical Engineering PAGE 9 6/21/13

t = 0.2 ms t = 0.4 ms

Scatter plot conditioned at Z = Zst

/ Mechanical Engineering PAGE 10 6/21/13

Progress variable Y is a normalized linear combination of reaction products

Igniting laminar mixing layer with FGM

•  Assume a 2D manifold •  Create 2D manifold from 1D igniting mixing layers •  Solve transport equations for controlling variables

/ Mechanical Engineering PAGE 11 6/21/13

|∇Z | � |∇C|

C =Y − Yu(Z )

Yb(Z )− Yu(Z )m

m0= 1− Ka

m = ρsL

m

m0= 1− Ka +∆h

∂ lnm0

∂h+�

j

∆Zj∂ lnm0

∂Zj

m

m0= 1−MKa

∆Zj = cjKa ∆h = chKa

φ = 0.7− 12∆φ cos(2απx/Lx)

ω̇Y = �m0�Σ

�m0�1 =�

m0(Z )Pβ(Z ; �Z ,�Z ��2) dZ

�m0�2 =�

m0(Z )Pδ(Z ; �Z ) dZ = m0(�Z )

�m0�3 = m0(φ = 0.7)

�m0�Σ from DNS

�m0�Σ modeled

Y = f (Z ,Y)

Y (x , t) �→ Y (Z ,Y)

1

|∇Z | � |∇C|

C =Y − Yu(Z )

Yb(Z )− Yu(Z )m

m0= 1− Ka

m = ρsL

m

m0= 1− Ka +∆h

∂ lnm0

∂h+�

j

∆Zj∂ lnm0

∂Zj

m

m0= 1−MKa

∆Zj = cjKa ∆h = chKa

φ = 0.7− 12∆φ cos(2απx/Lx)

ω̇Y = �m0�Σ

�m0�1 =�

m0(Z )Pβ(Z ; �Z ,�Z ��2) dZ

�m0�2 =�

m0(Z )Pδ(Z ; �Z ) dZ = m0(�Z )

�m0�3 = m0(φ = 0.7)

�m0�Σ from DNS

�m0�Σ modeled

Y = f (Z ,Y)

Y (x , t) �→ Y (Z ,Y)

1

|∇Z | � |∇C|

C =Y − Yu(Z )

Yb(Z )− Yu(Z )m

m0= 1− Ka

m = ρsL

m

m0= 1− Ka +∆h

∂ lnm0

∂h+�

j

∆Zj∂ lnm0

∂Zj

m

m0= 1−MKa

∆Zj = cjKa ∆h = chKa

φ = 0.7− 12∆φ cos(2απx/Lx)

ω̇Y = �m0�Σ

�m0�1 =�

m0(Z )Pβ(Z ; �Z ,�Z ��2) dZ

�m0�2 =�

m0(Z )Pδ(Z ; �Z ) dZ = m0(�Z )

�m0�3 = m0(φ = 0.7)

�m0�Σ from DNS

�m0�Σ modeled

Y = f (Z ,Y)

Y (x , t) �→ Y (Z ,Y)

1

FGM validation: scalar dissipation rate

•  Temperature rise in 1D igniting counterflow flames (Cabra exp.)

•  FGM created with strain rate a = 100 s-1

•  Comparison FGM (dashed) vs detailed chemistry (solid)

/ Mechanical Engineering PAGE 12 6/21/13

LES of JHC: H2 addition

/ Mechanical Engineering PAGE 13 6/21/13

CH4 25% H2

Artega Mendez & Roekaerts TU Delft, The Netherlands

25% H2 CH4 10% H2 5% H2

E. Abtahizadeh

Conclusions

•  MILD combustion occurs in thin reaction diffusion layers •  Due to the very low values of Zst, the reaction layers lie at the edge

of the turbulent mixing layer •  Ignition chemistry appears to proceed along a lower-dimensional

manifold •  A 2D FGM can accurately predict the ignition delay time

•  Counterflow flames: various strain rates •  JHC flames: capture effect of H2 addition

/ Mechanical Engineering PAGE 14 6/21/13

Outlook

•  More realistic DNS •  Include variations in coflow composition and temperature •  Include turbulence in the coflow •  3D turbulence

•  Validating FGM in DNS •  Quantitative comparison of LES/FGM results with experiments •  Investigate fuel effects •  NOx predictions

/ Mechanical Engineering PAGE 15 6/21/13

Thank you!

•  My coworkers

•  Dutch Technology Foundation

/ Mechanical Engineering PAGE 16 6/21/13