1. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 1 Measurements and simulations of mixing and autoignition on an n-heptane plume in a turbulent flow of heated air C.N. Markides, G. De Paola, E. Mastorakos(em257@eng.cam.ac.uk)em257@eng.cam.ac.uk http://www.eng.cam.ac.uk/~em257/

2. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 2 Introduction Structure of presentation: Experimental –Apparatus –Bulk observations Simulations –The CFD –The CMC model Results –Ignition lengths –Explanation of trends –Implications Conclusions & suggestions for the future

3. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 3 Why is autoignition important? LPP gas turbines: Premixing for low NOx, but danger of autoignition! Diesel & HCCI engines: Fast mixing for low emissions, but need to predict autoignition!

4. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 4 Experiments 1. Apparatus Air in, hot grid Fuel in, cold Atmospheric pressure Air T up to 1100K Bulk velocities up to 30m/s Fuels: H 2 /N 2, C 2 H 2 /N 2 C 7 H 16 /N 2 Techniques: Hot wire for initial conditions PLIF of acetone for  2D image of OH* with ICCD Turbulence intensity boosted by grids. “Diffusion from point source”. (Markides & Mastorakos, 2005, Proc. Comb. Inst. 30)

5. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 5 Experiments 2. Visualization Ignition spot appears and then disappears. Location of ignition spot is random. Fuel Hot air OH chemiluminescence (0.2 ms exp.): Individual spots, not connected flame Ignition spot development at 20kHz: nothing, spot, spherical flame, nothing (consistent with DNS!) C2H2 ignition, natural light (1/125s exp.)

6. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 6 Experiments 2. Visualization Qualitative regimes of operation (for all Ujet/Uair tested between 1 and 5): T U Random Spots Flashback No Ignition Lifted Flame Individual short-lived autoignition kernels Continuous flame sheet ? Stabilisation in mixture “almost ready to ignite”? This regime more likely at high Ujet/Uair. Similar to “Cabra” burner Quick propagation back to nozzle Autoignition not happening due to high strain?

7. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 7 Experiments 2. Visualization Localised autoignition Statistically-steady If ignition happens close, then it happens often They always come in bursts

8. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 8 Experiments 3. Mixing Mean and variance of mixture fraction as expected Two-component scalar dissipation measured at Kolmogorov resolution satisfies global conservation (Bilger, 2004) Data used for validating CFD & CMC model (Markides & Mastorakos, to appear in Chem Eng Sci)   

9. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 9 Calculations 1. CFD – for mixing, neglecting reactions STAR-CD k-  model Very good resolution close to nozzle needed Use experimental initial conditions Use experimental C d in model for

10. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 10 Calculations 2. CMC model Conditional Moment Closure equations: Conditional convection Conditional turbulent flux Diffusion in  -space & chemistry, closed at 1 st order

11. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 11 Calculations 2. Formulation of the CMC model for plume Averaged across plume:

12. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 12 Calculations 3. Code, chemistry, validation 31-scalar reduced heptane chemistry (Bikas, PhD Thesis, Aachen) Ignition times of homogeneous mixtures OK Ignition times of spray with CMC OK (Wright, De Paola, Boulouchos, Mastorakos, Comb. Flame, to appear)

13. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 13 Results 1. Mixing: Good agreement <><> <><>  variance

14. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 14 Results 2. Autoignition lengths: reasonably good agreement Physics: As U increases, ignition length L increases, but also L/U increases. Hence, not simply chemistry-controlled! Trend captured by model

15. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 15 Results 2. Conditional statistics Ignition at the most-reactive mixture fraction, not at stoichiometry. As L increases, P(  )   (  -  well-mixed ). Ignition time becomes long as P(  MR )  0. Low T: long L High T: short L  MR

16. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 16 Results 3. Discussion Flamelet or CMC: ignition time increases as N increases In our flow: N increases with U Hence: Ignition time in our flow increases as U increases Also: N<N critical hence 2 nd -order CMC not needed

17. 10001886:2001-03-30 Engineering DepartmentUniversity of Cambridge 17 Conclusions A novel autoignition rig is operational and has produced results for various fuels Intense turbulence can delay autoignition due to increasing scalar dissipation rate CMC model can capture all experimental trends Crucial aspect: modelling of scalar dissipation Future:Transport equation for 2D-CMC to capture spatial diffusion / flashback conditions LES, PDF calculations