[Les]

Hi All,

I am currently involved with CFD combustion modelling simulations and wanted to know what is the acceptable accuracies for the simulations results. The predicted results are being compared with measured industrial data. Are there guidelines for acceptable combustion modelling results, for example the percentage difference model & measured data:
if within 10 % very good result
10 - <20 % good
20 - <30% acceptable results
>30 % unacceptable results

Any assistance will be much appreciated.

Kind Regards,
Les

[RobertB]

What are you modelling? What are your key performance parameters?

From the IGT premixed world,

NOx. Depends on the exact configuration, for near fully premixed should be able to get trends. When you have piloted systems it is harder as it is difficult to get a mech. that does both the rich and lean chemistry.

CO, hard

Temperature. For backside cooled components reasonable results can be obtained. For film cooled components it is more difficult.

Exit Profile, reasonable estimates can be obtained but the wall gradients are difficult (but then they are also difficult to measure!).

In general you need to consider what you are trying to do with the information. If for example a simulation shows an obvious hot spot and the physical reasoning behind it is correct then it does not have to be 100% accurate to be useful.

Expecting to set up a simulation where you press the button and obtain a perfect answer is optimistic. The same could also be said for the experimental world but this truth seems to be less well recognized in that arena (having operated in both). The goal of engineering is to gain credible data to understand the design and move it in the correct direction, analysis can certainly do that.

[Les]

Dear Robert,

We are modelling pulverised fuel fired boilers which are used for electrical power generation. It is a 600 MWe unit combusting high ash coal (>30 % wt). The performance parmeters we monitor at Furnace Exit is flue gas temperature, gas species concentrations for oxygen, carbon monoxide, sulphur dioxide, oxides of nitrogen and percentage combustible residue in flue dust. Essentially we obtained profiles for the parameters above using a water cooled probe during full-scale tests and are now validating the CFD combustion model results using the Furnace Exit results. Other detailed measurements were obtained during the full-scale tests e.g fuel, air, and steam flows and flue gas temp's at different particle residence times. Before we even setup the simulation a heat & mass balance for the unit was developed to check the inputs we intended to use in the CFD simulation.

Once the model is validated we intend to determne the furnace performance under various operating conditions e.g fuel/air maldistribution, low excess air level, changes to coal feed. The results will not be used for design changes but rather to inform the station personnel of the possible outcome of running at the simulated conditions e.g. temperature excusions at furnace exit or lower combustion efficiency etc.

These are the average percentage errors I obtained during the validation exercise (model & full-scale results) at furnace exit level:
flue gas temperature <10 %
oxygen 26 %
carbon monoxide 38 %
combustible matter in flue dust 50 %

I think the first two are good and the latter two are acceptable. Thanks for your comments.

Regards,
Les

[RobertB]

If you are that far out on O2 concentration I am surprised you are that close on Temperature. Are your errors on Trise and delta O2 or on the absolute values?

Or do I just not understand the chemistry?

[Les]

Dear Robert,

The percentage errors are calculated on absolute values.
E.g. For O2 comparison
% Error = (O2 model - O2 measured)/ O2 model
% Error for O2 profile 11.81 % to 53 %. I averaged the errors for the profile to get 26 %.

If you like, I can send you the profiles?

Regards,
Les

[RobertB]

Would it make sense to use

(O2air-O2model)/(O2air-O2measurement) or vice versa

For example on an IGT the exhaust )2 concentration is ~12-14% depending on firing class (for more modern machines)

If my calculation gave ~17%O2 then by your system the error is only ~25% when in fact you had only consumed half the oxygen you should have.

That was why I was surprised you could have significantly better temperature agreement than O2 as the O2 is consumed 'generating' the temperature rise.

Since I have never done anything with a boiler I think I can add little value looking at the profiles.

[Les]

Thanks Robert,

That was the only percentage error equation I found in literature for theoretical & experimental value comparisons. Please provide me references for the formula used by you, would like find out more.

Kind Regards,
Les

[RobertB]

I don't have a reference, however...

Combustion is an oxidation process, you can only use up a certain fraction of the oxygen based on the amount of fuel you input. In your case I would guess around 4-5% unless you have a lot of EGR.

Therefore approximately 75-80% of the oxygen present can never take place in the combustion and in that sense is equivalent to the nitrogen and minor species.

If you measure 17% O2 and calculate 16% for example, you have reacted 25% more O2 than the experiment. It makes no sense to say this is an error of 1/16 or 6.25% as the remainder of the oxygen was never going to be involved anyway.

My feeling is that it is defined that way to make the error levels look better in a comparison of analytical with experimental results.

I still do not understand your results, what is your average exit O2 level? I don not understand how you can have large O2 errors but small Temp errors, are you using a similar definition or a Trise definition (which would seem more reasonable).