CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Unsteady Flamelet NOx Post Processing (http://www.cfd-online.com/Forums/fluent/107653-unsteady-flamelet-nox-post-processing.html)

clarkie_49 October 3, 2012 06:05

Unsteady Flamelet NOx Post Processing
 
Hi All,

I have a working natural gas combustor model for my undergraduate thesis which uses the Zimont BVM model coupled with laminar flamelets.

After running the transient unsteady flamelet update, Fluent creates a temperature field defined as "mean temperature" under the "unsteady flamelet" section. I would like to use this temperature field with the in built NOx post processor instead of the steady flamelet "static temperature". Could anyone advise on how this can be done?

Thanks in advance.

clarkie_49 October 4, 2012 05:35

More info re: Unsteady NOx Post Processing (UDF required?)
 
Must this be solved with a UDF? After searching and searching it would seem so, but i have no idea on how to create a UDF for this purpose (or any other).

I have looked at the case "data" file and it looks as though the variables are placed under the following headlines:

Static temperature (from steady flamelet):

(0 "SV_T, domain 1, cell zone 40, 2240 cells:")
(300 (3 40 1 0 1 3584 5823)
( 6.47804810e+02
6.47804810e+02
6.47804810e+02 ..............................

Mean temperature (from unsteady flamelet):

(0 "ufla-t, domain 1, cell zone 40, 2240 cells:")
(314 (1 40 1 0 1 3584 5823)
( 6.47804810e+02
6.47804810e+02
6.47804810e+02 ..............................

It would seem as though all that is needed is to set:

sv_t = ufla-t

for each node.

But there are 300,000 cells and manually editing these text files is next to impossible! Is there an "easy" way to do this?

cguy October 4, 2012 21:13

I would say it is easier (and more accurate since turb-chem interactions are accounted for) to include NOx in your chemical mechanism that you use to solver the unsteady flamelet...

clarkie_49 October 4, 2012 23:19

Quote:

Originally Posted by cguy (Post 384992)
I would say it is easier (and more accurate since turb-chem interactions are accounted for) to include NOx in your chemical mechanism that you use to solver the unsteady flamelet...

Hi Graham, thanks for your response. The NOx predicted by the unsteady flamelet model is approximately 10x greater than the NOx post processing and that produced by normal gas turbine combustors. In the following document the same situation is found and this 3 step process which i am trying to use (steady-unsteady-nox) is suggested, but not described in detail:

http://books.google.com.au/books/abo...kC&redir_esc=y

It is my understanding that the unsteady flamelet model can do a good job of predicting medium-slow forming species such as CO, but cannot account (enough) for the slow reaction rate of NOx. Below are my findings:

Steady flamelet NOx: 1400ppm
Unsteady flamelet NOx 240ppm
NOx post processor with N2O pathway and reburn (Steady flamelet mixtures and temperature used): 15ppm

However the difference in the steady and unsteady temperature field is quite significant, with the maximum temperature reducing from 2380k to 2240k for the steady and unsteady cases.

This is why i would like to post process the NOx using the mean temperature field from the unsteady solution (ufla-t).

Any help would be greatly appreciated.

cguy October 5, 2012 06:34

Brendan,
You really don't want to write a UDF - it would be hundreds or thousands of lines. However, the UDF should ultimately work the same as including NOx in the unsteady flamelet mechanism. There are two possible sources of discrepancy:
1) The mechanism is different. You could type the Fluent post-proc NOx mechanism into your Chemkin mechanism and import that. I would guess that this is not causing major differences
2) More likely is the choice of turb-chem interaction. With NOx post-processing, you can choose "none" or "mixture-fraction" (the other options should not be used). I suspect that you are using "none", which ignores fluctuations and hence the predicted NOx is lower. You could get Fluent to approximate this by setting the mixture-fraction variance to zero as follows. Type this exactly in the tui:
(rpsetvar 'cgvar 0)
solve your steady flamelet
plot contours of mixture-fraction variance - they should be zero
re-run the unsteady flamelet with NOx in the mechanism

clarkie_49 October 5, 2012 07:43

1 Attachment(s)
Thanks again Graham, i should have included more information.

It is not the NOx post processing set-up that i am having issues with as i have always had that set to "mixture fraction". Given the equivalence ratios that i am using (phi < 0.6) for premixed mixture and the fact that the primary injector is also modelled as largely premixed, i am expecting these low numbers of NOx emissions; in fact it has taken me months of geometry changes to get it to this state.

I am fairly happy with the results, but it would make more sense to post process using the unsteady temperature field. I know it can be done, as the author of the paper i linked used this same method as they found the same issues with the steady/unsteady flamelet GRI3.0 NOx prediction. Attached is a snapshot of the relevant pages of the paper which can be found at http://search.proquest.com.ezproxy.l...89944/abstract if you have access.

A few more details in regards to my model:

2D lean-premixed axisymetric combsutor
Chemical Mechanism: GRI 3.0 (also tried GRI2.11 with very similar results)
Turbulence Model: k-omega SST (RANS)
Combustor avg. pressure: 12 bar

Andrea1984 May 12, 2014 05:03

Hi Brendan,

I am facing exactly the same issue so I was wondering if you managed to use the NOx pollutant model with the unsteady flamelet temperature field (i.e. using the Mean temperature field instead of Static temperature to evaluate NOx formation)

Cheers

Andrea


All times are GMT -4. The time now is 04:16.