1. University of Wisconsin -- Engine Research Center slide 1 Flamelet Modeling for the Diffusion Combustion in OpenFOAM ME 769 Final Project Presentation Guangfei Zhu & Sihan Jin 5/7/2015

2. University of Wisconsin -- Engine Research Center slide 2 Overview Project Introduction Simulation Setup 1. Governing equations 2. Beta-PDF 3. Creating Chi table 4. Turbulence model 5. Model Verification and Short Summary 6. Mesh generation for Large Re simulation Simulation Results Conclusion and Summary Reference and Future Work

3. University of Wisconsin -- Engine Research Center slide 3 Project Introduction Flamelet combustion model has some advantages over other types of combustion model when simulating turbulent flame. OpenFOAM has an extensive range of features to solve anything from complex fluid flows involving chemical reactions, turbulence and heat transfer as well. Flamelet model has been built in OpenFoam by other people for turbulent flame simulation. When Reynolds Number is relatively low, great performance could be achieved by this model from others’ paper. However, large Reynolds Number flame simulation has not been done by other people so far. Our objective is to simulate Sandia Flame D under large Reynolds Number condition with Flamelet model.

4. University of Wisconsin -- Engine Research Center slide 4 Simulation Setup

5. University of Wisconsin -- Engine Research Center slide 5 Simulation Steps Fig.1 Simulation step flowchart

6. University of Wisconsin -- Engine Research Center slide 6 Simulation Setup 1. Governing Equations

7. University of Wisconsin -- Engine Research Center slide 7 Simulation Setup 1. Governing Equations

8. University of Wisconsin -- Engine Research Center slide 8 Simulation Setup 1. Governing Equations

9. University of Wisconsin -- Engine Research Center slide 9 Simulation Setup 2. Beta-PDF

10. University of Wisconsin -- Engine Research Center slide 10 Simulation Setup CFD Simulation Inputs The burner dimensions: Main jet inner diameter, d = 7.2 mm; Pilot annulus inner diameter = 7.7 mm (wall thickness = 0.25 mm) Pilot annulus outer diameter = 18.2 mm Burner outer wall diameter = 18.9 mm (wall thickness = 0.35 mm) 3. Creating Chi table The bulk flow and scalar boundary conditions: Coflow velocity (Ucfl) = 0.9 m/s (+/- 0.05 m/s) @ 291 K, 1 atm Main jet composition = 25% CH4, 75% dry air by volume Pilot flow composition = 0.734 N 2, 0.056 O 2, 0.092 H 2 O, 0.110 CO 2, 0.0022 OH by mass Main jet kinematic viscosity = 1.58e-05 m2/s (from ChemKin) Main jet velocity @ 294 K, 1 atm = 49.6 m/s Pilot flow velocity @ 294 K, 1 atm = 11.4 m/s

11. University of Wisconsin -- Engine Research Center slide 11 Simulation Setup 3. Creating Chi table Cantera Inputs:

12. University of Wisconsin -- Engine Research Center slide 12 Simulation Setup 3. Creating Chi table

13. University of Wisconsin -- Engine Research Center slide 13 Simulation Setup 4. Turbulence Model In our case: For RANS simulation, we use k-epsilon model. For LES, we use Smagorinsky’s model which is an adaptation of Prandtl’s mixing length theory to subgrid-scale modeling.

14. University of Wisconsin -- Engine Research Center slide 14 Simulation Setup 5. Model Verification In the small Re simulation, the initial condition is main jet velocity = 10 m/s and pilot jet velocity = 3.5 m/s. This corresponds to the Re = 6329 which is also turbulence flow. Time = 0.01 s Time = 0.02 s Time = 0.03 s (a) my case(b) Prof. Hagen’s case Fig.2 Comparison of temperature distribution between our case and Prof. Hagen’s case

15. University of Wisconsin -- Engine Research Center slide 15 Simulation Setup 5. Model Verification

16. University of Wisconsin -- Engine Research Center slide 16 Simulation Setup Summary 1. During the research, we need to check every term to find the difference between our case and reference case in order to lay a foundation for the future study; 2. We should try to discard the unreasonable data and create or choose more reasonable data.

17. University of Wisconsin -- Engine Research Center slide 17 Simulation Setup 6. Mesh Generation for Large Re Simulation Fig.5 Sector geometry

18. University of Wisconsin -- Engine Research Center slide 18 Simulation Setup Nodes along radius: 10, 10, 32, respectively. Nodes along axis: 100. Total number of cells is 15256. Fine mesh Nodes along radius: 40, 40, 60, respectively. Fig.6 mesh structure 6. Mesh Generation for Large Re Simulation

19. University of Wisconsin -- Engine Research Center slide 19 Simulation Results

20. University of Wisconsin -- Engine Research Center slide 20 Simulation Results Prof. Hagen’s case Our case LES T

21. University of Wisconsin -- Engine Research Center slide 21 Simulation Results Fig.7 Temperature change with time for Sandia D flame of LES (our case) Time = 0.01 s Time = 0.02 s Time = 0.03 s Time = 0.04 s Time = 0.05 s Fig.8 Temperature change with time for Sandia D flame of LES (Prof. Hagen’s case)

22. University of Wisconsin -- Engine Research Center slide 22 Simulation Results Prof. Hagen’s case Our case LES chi

23. University of Wisconsin -- Engine Research Center slide 23 Simulation Results Fig.9 Chi change with time for Dandia D Flame of LES(our case) Time = 0.01 s Time = 0.02 s Time = 0.03 s Time = 0.04 s Time = 0.05 s Fig.10 Chi change with time for Dandia D Flame of LES (Prof. Hagen’s case)

24. University of Wisconsin -- Engine Research Center slide 24 Simulation Results Prof. Hagen’s case Our case RANS T

25. University of Wisconsin -- Engine Research Center slide 25 Simulation Results Prof. Hagen’s case Our case RANS chi

26. University of Wisconsin -- Engine Research Center slide 26 Simulation Results Fig.11 Temperature and species comparison between two cases (5e-3 from inlet)

27. University of Wisconsin -- Engine Research Center slide 27 Simulation Results Fig.12 Temperature and species comparison between two cases (0.288 from inlet)

28. University of Wisconsin -- Engine Research Center slide 28 Conclusion 1. Sharpe change of variables will cause problem in Flamelet modelling because it affects the interpolation during the simulation; 2. Mesh size is very important for large eddy simulation because bad mesh will cause strong gradient for some terms such as scalar dissipation rate; 3. Near the nozzle, the region around the reaction zone could be regarded as being laminar which is dominated by the molecular diffusion. 4. High scalar dissipation rate appears in layer like structures and direct inward.

29. University of Wisconsin -- Engine Research Center slide 29 Future Work and Reference Reference: [1].http://www.sandia.gov/TNF/DataArch/FlameD/SandiaPilotDoc21.pdf.http://www.sandia.gov/TNF/DataArch/FlameD/SandiaPilotDoc21.pdf [2]. Barlow, R. S., and Frank, J. H., Proceeding of the Combustion Institute 27:1087 (1998). [3]. D. A. Lysenko, I. S. Ertesvag and K. E. Rian, Flow Turbulence Combust 93:665-687 (2014). [4]. Pitsch, H. and Steiner, H., Phys. Fluids 12(10), 2541-2554 (2000). [5]. Pistch, H. Combust. Flame 123(3), 358-374 (2000). [6]. Pitsch H. and Helfried Steiner. Proceeding of the Combustion Institute. 28:41-49 (2000). Future work: 1. Change the initial conditions manually to minimize the difference; 2. Change the turbulence model to dynamics structure model (Prof. Rutland’s turbulence model); 3. Modify the model to calculate the scalar dissipation rate according to the thesis of Yuxin Zhang (one of previous master student of our ERC).

30. University of Wisconsin -- Engine Research Center slide 30 Thank you for your attention! Questions?