I am an academic and I have been working on combustion modelling and CFD projects for a number of years. Some of the comments below are a little honest and therefore I wish to remain anonymous in order to preserve future funding for my department.

Generally, my colleagues from outside the field are cynical about the usefulness of adopting such complex simulations as invariably we need to "optimise" critical model constants â€" which can vary considerably depending on operating conditions and application.

Of course, ideally we would never need to "optimise" (although we would be out of the job!!!) and currently such an approach is necessary in the long term to help develop more reliable predictive tools.

At present, in my eyes, it appears to be a "chicken and the egg" situation as rates of combustion can not be predicted reliably without first optimising with experimental data recorded measured at almost identical conditions! This means the model is totally not predictive.

I understand that a model must be developed initially using some experimental data however is anybody actually adopting totally predictive combustion models for conceptual work? If so, for which applications and adopting which combustion model? Lastly of course, how predictive and how successful are your methods?

Id like an honest debate.

Hi, even me too working in accademics, developing and testing combustion models, i am really not understanding where all this models is are using to give productive results, offcorse the people who r working in the industry shoule give their view..

yea man, don't be surprised when you get no answer to last two questions. 1) everyone protects his know-how (the less he has the better the protection) 2) no one will say "I've made xyz model and found it totally faulty, it really can not predict anything because the chosen approach is not appropriate enough". such a statement would lead him to being asked "how long time and how much money have you spent on this totally useless work?". consequently, in academic field the result would be funds reduction, in commercial field the job loss. as well, none of the companies will ever accept that their solution is invalid, or will accept that the solution of a competitor is better. the better established company the greater lies. simply, every product must be sold, thus must be "...the leading product in its field.." just note the news, many terms have disappeared, many have been changed for "more acceptable" (interruption instead of abort, friendly firing instead of killing ours, aggressor have been foiled instead of killed..........)

(I am also academic, very closely cooperating with industry, at the moment sitting in one such company working for them "on something")

I guess, the best use of "useless models" is they give chance to improve and make something useful. For those who think and belive that the current things are useless, they have all the freedom to improve and make everybodies life better. This is very much true if we see the past, but to improve or to come to a place where we can say, yes, i have some thing which is very useful. You most probably have to start from something useless.

I guess, this is the best use of current combustion models if they are of no current use.

that's what are we (I presume all) doing: improving useless to usable, bad to fair, fair to good ...

So dear Andy, as You can see - modern combustion models - we shall say the next - "STILL IN DEVELOPMENT STAGE"

and moreover - combustion chamber - it's not an airfoil or wing of Airbus

I am sure that I'll perform reliable simulating of combustion chamber, but not before 2015

But academy work should be continued, because it's need to answer to following questions:

- how to make work LES near very comlex walls

- how to resolve all cooling holes and dillution holes

- how to resolve very coplex wall configuration

- how to model liquid fuel phase dispersion, evaporation and etc.

- reaction models are still rather poor

and in general - it's simply not enough memory and processor's time to solve such task

I'm not sure of what you mean by combustion models, but I started using CFD for modeling combustion in industrial boilers about 20 years ago. These boilers are relatively large - about 30 feet by 30 feet in cross-section and 100 feet high - and are designed to burn a variety of fuels like waste wood, Municipal Solid Waste, spent pulping liquors, etc. It is difficult to make measurements inside the combustion chamber, or even see into it.

The boilers face many problems including meeting environmental regulations, availability, capacity limitions, and corrosion. One might argue - well, we can't even model a candle flame so why try to model a whole boiler?

I think the basic answer lies in whether we are better off with doing the CFD modeling or in not doing it. After modeling about 50 industrial boilers, my conclusion is that we are better off doing the modeling.

I am generally dealing with combustion of particles - drying, volatiles release, and char combustion, along with gas phase reactions of combustibles. In the gas phase, we use about 15 species and track a dozen or so reactions which can be controlled by stoichiometry, kinetics, or turbulence. Yes, it is frustrating at times that even simple oxidation of CO cannot always be well represented with kinetic models of say Howard or Dryer & Glassmann, and that little fundamental research seems to be going on to fill in the gaps. However, with some trial and error using available boiler data, our CFD models of the boilers work reasonably well in predicting CO and other emissions.

I guess a lot of the benefits of doing CFD modeling of combustion in boilers is to get a better feel for how some of the more basic elements work - how do you get the fuel and oxygen in contact in such a large box, and what happens when you have cold spots in the furnace?

Part of the nature of studying something is that as you look at it in more detail, the more questions it will likely generate. I guess advances in combustion modeling involve the use of more computational cells, more intermediate species, more complicated turbulence models, and better temporal resolution.

If you start to lose perspective due to the complexity, people like myself modeling combustion in industrial situations can always use help.

George Box once said:

"All models are wrong, but some are useful"

When asking if "any" combustion model is useful, you are best served by stating for what purpose you are using your model for. This might stimulate a more productive debate.

Thank you. Some very interesting views - I find it particularly interesting that there are a few other academics also asking themselves what they are trying to achieve. I also agree that we can deliver better combustion models more rapidly, by being honest with one another about their problems and shortcomings.

Hopefully, an accurate predictive combustion model will be developed well before we move to alternative technologies!

My own view is that predictive combustion modelling is one of the most difficult problems about - this also makes it extremely interesting. Of course, it is better to try and fail, than to never try at all!

Accurate predictive combustion is a long way off and is necessarily what we should be working towards. On the way, we will learn and presently this forms the "use" of predictive combustion, it is a tool to better understand a complex process.

What about those of you in industry - what sorts of applications do you model? What combustion models do you adopt?

I thought your question was about the usefulness and effectiveness of CFD applied to combustion sytems?

Of course combustion modelling is still being developed and will continue to be until the effect of the modelled terms is trivially small on every parameter of interest. In my judgement this has no chance of occuring by 2015. However, this does not mean that CFD has not been useful and effective in the industrial design and development of combustion systems for the past 20 years (in the high tech end).

The answer to the question in the title of this thread is: yes in the hands of someone who understands the assumptions within the simulation and, most of the time, no in the hands of someone who simply runs the package and believes everything in the output. (I have observed what tends to happen with the latter type of engineer is that they oscillate between believing everything and believing nothing a few times and then give up on CFD after they have acted on something wrong in the simulation once too often.)

In regard to the last point, I agree that only a fully trained engineer with significant experience of CFD and combustion modelling is better than an inexperienced engineer unaware of what they are doing and that all results should definately be taken with a "pinch of salt".

Are we daring to suggest, that the blame lies with the operators themselves? You may well be right, "a bad workman blames his tools"!

No I would not say the blame lies with the users. If using CFD as a black box is not providing useful and reliable information then stopping using it is a perfectly reasonable course of action. Others reasonable courses of action would be to get educated or employ directly or indirectly educated users.

If blame is to be attributed, in my experience, it almost always lies with management deciding to use a tool without getting sufficiently educated in its use.

Nice debate going on "Combustion models are useful" -yes they are useful. Many theories and models both based on physics and empiricism are developed and they are developing. One thing we need know most of the industrial complicated combustion already simulated and gave satisfactory results. Of course most of them are confidential projects. So, we can't find much literature on them. So, if a new model comes and if we test with some experimental data to validate and then to implement for industrial application, for sure it will need at least more than five years. More ever if we develop a combustion model with some constants and variables which may not easily available, then it may be junk for many. So, after some 10 years we may have little domain to do CFD combustion simulation. Combustion is definitely a very good science a lot to explore, but I do believe that from my experience models aged 20 years are also giving results which industry is needed. So, younger and more intelligent (models which address many factors which influence the combustion) are not really doing better than the older when we use them for multiple application. So, I believe physics driven models can be applicable to large domain with. Industry doesn't look 100% accuracy. Industry use CFD as predictive tool. But it never rely solely on CFD for design or modeling.

We had a vast literature on COMBUSTION. So, this literature extending day by day but I feel what ever is happening not really leading to develop a tool (or model) which can predict as it is in reality. So, instead of developing new models at many corners of the world independently, it is always recommended to share your ideas, collaborate, and develop a single model. Every body is correct but criticism is always needed to say that your model is wrong and why it is wrong, and then only a better model is development is possible. If we see some literature in CFD, they change this and that constant and publish some new paper. That may be useful only for publishing. When we change some constants it will definitely give some new prediction. So, this type of work may not healthy to develop new model.

Best thing is (I may be too much greedy regarding this) some revolutionary thinking for new physics for combustion is needed. It is not easy but I believe some solution must be there to define combustion in a simpler way and most efficient way. Hope that thing will be come very soon, and we . And will give all solutions and we will be take 100% hold on all types of combustion.

We can do our combustion problems in 2 ways. 1. U can use same existing models …for that u have to do as follows….as you tune constants in k-e model , similarly u can tune to your model constants to validate model and u can use same parameters if combustion parameters or same. For this you need at least some experimental data. That can be any data like temperature at some location.

If don't have any data whether to rely on the predictions, always there is other solution. This depends on individual's should develop your own method to validate your model, whether it is predicting reasonably good or not.

So, I conclude that combustion models are good and useful. How extent means that depends for what you are looking in your solution. So, u can use existing models without much worry.

Have fun Cheers!

Dear Combustion!

As You have a lot of literature about combustion models, may be it will be possible and not difficult for You to write a topic about combustion modelling into the CFD Wiki?

We will be very grateful to You fou Your contribution

Actually michail, there are lot of people who can contribute, but they do not do so. some of them are too busy, some of them are too lazy, some of them think can not contribute without even trying to do so. i really appreciate the effort you have put on wiki and i really hope some more people come forward and make some contribution, no matter how small it is. it is for our own good.

Thank You for the compliment ;-)

I am plain Russian Air Force's interceptor's technic, and I am trying to learn CFD and to prove to myself that I am capable to do something on the international level.

Also I am very glad that I am working in CFD-Wiki editor's team.

It's as already noticed Jonas Larsson - the best way to learn CFD and in the same time fun. As soon as I have some spare time now, I shall try to contribute to CFD-Wiki as much as my abilities allow me.

Very interesting discussion! Thank you Anonymous for starting it. Next year I will celebrate my 10th anniversary of working in the field of combustion modeling. Okay, I would not suggest this way for a young engineer who wants to start a rapid and successful career, however the time was never boring.

So far, the introduction, but now to very important point: How useful are the turbulence models? They are very nicely tuned for some wings. For backsteps it starts to get difficult and for jets they are coming to their limits. The final break down comes with swirl, which is very important for many combustion applications. K-epsilon fails, reynolds stress is to instable for practical use and LES has other short comings. So, how useful are the turbulence models?

Yes, I've used the quote the odd time myself in CFD presentations. But, the "all" part of the statement has bothered me. If I have two apples and go to store and buy four more and then want to know how many apples I have, I represent this with a simple model: 2 + 4 = ?. I can't imagine a case where this simple model is wrong (e.g. when is 2 + 4 not equal to 6). So if this is a model that is never wrong, doesn't it invalidate the statement of Box?

"2 + 4 = ?. I can't imagine a case where this simple model is wrong (e.g. when is 2 + 4 not equal to 6)"

Well, just to spice it up... there is an implicit assumption in your assessment that 2+4=6. I can argue that 2+4=sqrt(20) - if we recognize that these numbers represent vectors in euclidean space. I'm sure there are other equally valid answers )

Adrin Gharakhani

Hi everybody,

I think it is the same like with the combustion models. They are limited by many factors, giving better or worse results in different cases, but still giving better answer to the questions than our intuition gives. Or to use add like slogan "it gives your intuition wings".

The best asnwer would be to look around the industry labs. Still growing number of them are using CFD to improve their products. None of them would not pay expensive programs and CFD guys without any profit of it.

Still CFD will hardly give you the answer how the world outside your office works,(yet sometimes it does).

As someone already said above - you must allways know the limits of the models.

Anyway I'd like to see this question in wiki for particular models. Otherwise it does not have any sence, anyway. ).

matej