Does anyone had experiences into simulation of laminar combustion and checked via experiment the adiabatic temperature predicted by CFX ? I found that CFX calculates a temperature from 100 to 150 K degrees ( according to different excess air simulations ) higher than the adiabatic one (obtained by fine chemkin calculations) . Does anyone has corrective cp factors or any suggestion for a more correct temperature prediction ? thanks a lot Marc

Hi Marc,

the "global" adiabatic flame temperature (based on inlet temp, mixture, enthalpy of formation, et) is just a re-statement of the first-law of thermodynamics. This always applies globaly but would only apply to a local flame temp limit if there were a 1-D plug flow reactor, with adiabatic walls and complete combustion.

In CFD the adiabatic flame temp (as a physically plausable limit) would only then apply on a cell by cell basis given the inlet mixture, enthalpy of formation, Cp, etc. in EACH cell. So it is basically a senseless metric for physical temp limit in CFD....although an incredibly pervasive belief exists that it should be an upper bound!

A simple thought experiment will suffice to dispell this:

1. Consider a combustor with an outer and inner tube. Both contain the same combustable mixure, premixed, non-premixed it makes no difference.

2. The pervasive myth is that no-where in the device should the temp exceed the adibatic flame temp based on inlet conditions.

3. If I use the outer tube as an early flame that pre-heats the mixture in the inner tube. In the inner-tube I inhibit combustion along the length some ways so that the mixture has heated up to Tnew.

4. Downstream I then combust the inner tube mixture...maybe an ideal 1-D adiabatic plug-flow reactor. This clearly can reach a higher temp than the adiabatic flame temp based on the inlet conditions.

Ok...well what happens in reality....well in most cases the above does not happen (ie burning held off locally) but it could as a poorly mixed turbulent flame...fuel and air could be pre-heated as it passes through the flame and lead to some high temp spots.

Basically there is some other reason your predictions are wrong. In general CFD with simple models over-predicts temp compared to experiment...that is another chapter...........Bak_Flow

hi bak

Yes, I agree that there is a belief about adiabatic temperature as upper bound, I have that belief too !

A realistic simulation of a methane combustion should have NO region or cell with a temperature higher than the adiabatic temperature. YES, I agree with your experiment, but you "cheated" pre-heating the inner tube mixture. In that case I agree that the temperature can be higher than the adiabatic one. Let's talk about methane combustion with standard inlet condition ( e.g. 288 K for the inlet mixture ).

CFD worldwide calculates a higher adiabatic temperature that the one measured or calculated with different fine chemical simulator ( e.g. chemkin ) .

The reason can be : 1) the reduced kinetic mechanism in a CFD program usually has from 2 to 5 reaction , and do not properly compensate for the thousands of chemical reaction in real combustion . The energy release of the Harrenius formulas of these few reaction are not so correct/usefull . 2) sometimes a specific heat capacity "correction" should compensate for the energy over prediction , but I still have not found usefull parameters ! 3) ????? some suggestion ?????

I would be really happy to have simulations with temperature NOT exceeding the adiabatic one.

Any suggestion from anyone ?

Hi Max,

ok rather than argue about a belief lets really really understand this.

1. Ok if I cheated....why. How do you do your adiabatic flame calculation.....based on inlet conditions right....so did I?

2. If you took all the rignt numbers for inlet conditions, equivalnce ratio, Cp, enthalpy of formation, etc. By applying the first law of thermodynamics what can we determine? Adiabatic flame temp....right! What are all the underlying assumptions in that number we get?

3. What is the difference between what I did and 2.

4. Can a real flame approximate the difference? How and why?

Hint think global vs local! Regards,

Bak_Flow

The calculation of the adiabatic temperature is related to kinetic gas theory. Most of the chemical simulators can handle hundreds of chemical reactions and so reproduce temperature profile, species profile, and more, along a single direction ( they are 1-D in general ) . See www.acad.carleton.edu/curricular/CHEM/courses/tferrett/Chem343_W05/PSet3bkey.pdf

for reference of all the mistakes possibli made during flame temperature calculation.

My first question still reside : what can I do to have good adiabatic flame temperature still using reduced CFD combustion mechanism ?

Hi Marc,

ok well now we are on the same page. So you have a very detaied mechanism from Chemkin. There are some methods to attack the problem in a similar way with CFD but we simply cannot solve the hundreds or thousands of species transport equations. So people come up with reduced mechanisms which capture the "essence" of what you need to know. Also many turbulent flames can be solved by doing the 1-D detailed chemistry which generates a flamelet library with all of the "flame state" known as a function of mixture fraction. Since the flame is turbulent we then have to account for the turbulent fluctuations by integrating this over some "presumed" probability distribution function or PDF. Usually a beta or double delta PDF is used. This then gives us a "chemistry" table which can be imbedded in the solution given that we solve transport for the mixture fraction and its properties (usually variance or some higher order terms). So we get all the details you want kind-of stuck into the flame at the right spots.

There is a great book by Warnatz that covers the whole range of models.

Good Luck........Bak_Flow

My combustion problem is LAMINAR ! All said and done for turbolent combustion is good.

My problem is about adiabatic temperature in laminar methane-air premixed combustion .

Still the problem reside : CFX predicts a maximum temperature 100-150 K higher ( changing according to air-methane ratio changes) than the one calculated with different chemical programs.

Ok I guess you said laminar earlier but I missed that! Actually I have only done a bit with truely laminar flames...they are hard! Without the dominance of turbluent transport you have to get ALL of the molecular transport and reaction mechanisms right.

CFX-5 can only apply bulk diffusivites so the transport will be hard to get correct. Also the Soret and Dufour effects may be significant. Then there is reaction mechanisms. My best advice is to start to do some reduced reaction mechanisms in Chemkin....say 10-20 reactions on 10 species and see what that gives you on the 1-D code. Then add these reaction mechanisms to CFX. Compare the CFX results in a 1-D equivalent geometry to the reduced mechanism. This may take some iterations to get close for Engineering purposes.

Best.........Bak

Marc,

I agree with Bak_Flow. Laminar flames are hard. To get correct answers you need to have the right transport properties implemented for viscosity, thermal conductivity and species diffusivity. CFX-5 does not have any of this right.

Also, you are not considering the effects of numerical approximations to the advection terms in the mass fraction equations. These approximations can cause boundedness and monotonicity issues which may cause temperature to overshoot.

Try CFX-5.8, it is much better for the latter issue. Transport properties are still a problem though, you will have to do that yourself by specifying the mixture thermal conductivty and viscosity and the individual species diffusivities (probably from kinetic theory models).

Neale.

Thanks. I have already tryed using individual species diffusivities . To use right specied diffusivities is important to describe flame shape, but not so much important for temperature ( changes in temperature +/- 10 K ) . I'll try thermal conductivity function of T as soon as possible . Does anyone has already written ccl subroutines for K = f(T) for CO2, H2O, O2 and N2 ???

Hello Marc, did you compare only the temperatures or also the composition? What differences do you see in the compositions between chemkin and CFX?

Best regards M stein

I have some strange interesting behaviour . I'd like to discuss in a private mail ! write me if interested marc

[Jozef]

Hi , I ma trying to find out the thermal diffusiviy of a mixture of air and diesel. Do you know any calculating procedure by thermal conductivity, density a and cp using?? thank you for answering me

[Milan2013]

Originally Posted by marc
;71333

hi bak

Yes, I agree that there is a belief about adiabatic temperature as upper bound, I have that belief too !

A realistic simulation of a methane combustion should have NO region or cell with a temperature higher than the adiabatic temperature. YES, I agree with your experiment, but you "cheated" pre-heating the inner tube mixture. In that case I agree that the temperature can be higher than the adiabatic one. Let's talk about methane combustion with standard inlet condition ( e.g. 288 K for the inlet mixture ).

CFD worldwide calculates a higher adiabatic temperature that the one measured or calculated with different fine chemical simulator ( e.g. chemkin ) .

The reason can be : 1) the reduced kinetic mechanism in a CFD program usually has from 2 to 5 reaction , and do not properly compensate for the thousands of chemical reaction in real combustion . The energy release of the Harrenius formulas of these few reaction are not so correct/usefull . 2) sometimes a specific heat capacity "correction" should compensate for the energy over prediction , but I still have not found usefull parameters ! 3) ????? some suggestion ?????

I would be really happy to have simulations with temperature NOT exceeding the adiabatic one.

Any suggestion from anyone ?

3. Plot the specific heat vs temperature, and you will see they are rising up. If a CFD solver is setup to work with constant values (at T0=298K) it will overpredict the maximum temperature, compared to using a propper mean value for the temperature range T0-Tmax (Tmax being the average maximum) or using cp=cp(T) localy.