Dear colleagues

I am studying CFD and still wondering - how useful and effective CFD is?

The masterpieces of Modern Aerospace Technology - SR-71 , XB-70 , Su-100 , B-1B , Tu-160 , F-15 , Su-27 , GE90 , F100-PW-229 , Al-31 were developed without CFD.

For example designing of turbojets turbine blade cooling system or developing of turbojets combustion chambers - all are made using experimental methods...

I am not experienced in designing bureaus work, but it seems to me that CFD for designers - just little fun.

I would like to hear any opinions. We can begin from our respectful Jonas Larsson, who works at Volvo Aero Corporation.

Hi,

You may find this paper "Future Trends for Computation Fluid Dynamics in the Process Industry" interesting.

CFD isn't a revolution in terms of what products you can develop - you can always use experiments or elaborate hand calculations instead. It might be a bit more expensive to do tests, it might take longer time and might limit the number of designs you can try, but the physics and the psysical limiations on a design isn't different just because you now have a tool called CFD. Also the Egyptians buildt impressive things although the lacked most of the tools and machines we have today.

However, in todays very competitive world CFD can give you an edge. With CFD you can optimize your design quicker, test more concepts and reduce the risk of unexpected problems later on when you finally do test your product. A company which refuses to use a new tool like CFD and instead relies only on expensive and time-consuming testing will not stay in business for long.

In the aeropace and turbomachinery fields CFD now plays a key role in the design process of many components. I know this from first hand experience - the development of the wings and the engines to coming new aircrafts like the Airbus A380 Superjumbo and the Boeing 787 Dreamliner has relied heavily on CFD. The examples you mention are areas where CFD is still not reliable enough to stand on its own and needs to be complemented with tests. There are other areas where CFD is more reliable and is used without any testing.

Yes, it's true, but my dream is - to perform complete modelling of turbojet's combustion chamber. I've seen some of such works, but they are rather primitive and can't be used in the designer's work process, just for academical use. Confining reacting swirling flow with recirculation zone in complex geometry. I've made some tries, but unsuccessfuly, because it's impossible with modern computer facilities...

I suppose it's the most coplex task in CFD

As well as the oft-stated remark that CFD can reduce experimentation time/cost, there's the other consideration that CFD models allow engineers to model and visualize things that simply cannot be measured. (How do you measure instantaneous heat transfer coefficients?)

In the automotive industry, CFD plays a huge part in many areas - well beyond the obvious aerodynamics field: combustion chamber, manifolds, exhaust systems, intake systems, coolant system, underhood and many more. We even use 1D CFD for performance and intake/exhaust noise prediction.

> For example designing of turbojets turbine blade cooling system or developing of turbojets combustion chambers - all are made using experimental methods...

This statement is incorrect if it implies numerical simulation is not being used in the design. It would be unwise not to perform experimental tests of complex designs but numerical simulation is increasingly being used to reduce the amount of experimental testing required.

Have you talked to anyone from, say, GE, P&W or R-R about the use of CFD in the design of combustion chambers? It started being used seriously to help solve real engineering problems in the 80s.

3D RANS CFD simulations of combustion system from compressor exit to turbine entry (i.e. including the diffuser and both sides of the chamber wall) were being performed in the mid 80s when computer memory was a serious limitation. A similar simulation could easily be performed on a current PC although, of course, today one would almost certainly choose to perform a larger and more reliable LES simulation on a parallel computer.

I disagree with you about combustion chambers.

Integral charachteristcs, such as

- hydrodynamical losses

- emission of NOx, CO, smoke emission

- outlet temperature

- outlet temperature distribution

- dillution air distribution

- cooling and burn-through problems

- high altitude stall charachteristics

can be estimated only during the experiment

Dear Andy, have you ever seen the modern turbojet's combustion chamber?

I had draughts of one of modern combustion chambers (unfortunately I can't say whose), and tried to build a grid for it. It's absolutely impossible because of its complexity. Besides we need to perform a complete calculation from diffuser till outlet of combustion chamber, because otherwise amounts of the dillution air will be left unknown...

Also I talked to desiners, and they couldn't say anything about real use of CFD in their work.

Of course, you can find such tries almost everywhere (I mean combustion chambers computations), but they are completely unuseful.

> Integral charachteristcs, such as:

> - hydrodynamical losses

> - emission of NOx, CO, smoke emission

> - outlet temperature

> - outlet temperature distribution

> - dillution air distribution

> - cooling and burn-through problems

3D CFD simulations were performed by combustion engineers to study these effects for current and prototype combustion chambers during the 1980s when I worked in the combustion research department of one of the main aero gas turbine suppliers.

- high altitude stall charachteristics

This was not studied directly with CFD at the time but probably is today. I do recall CFD studies being performed to help locate igniters.

> Dear Andy, have you ever seen the modern turbojet's combustion chamber?

Yes.

> I had draughts of one of modern combustion chambers (unfortunately I can't say whose), and tried to build a grid for it.

Sending real CAD files to grid generation companies was a common trick. It was surprising how poor many of the mainstream grid generators are at working with real CAD files full designer junk as well as the relevant 3D geometries.

> It's absolutely impossible because of its complexity.

Nonsense. It requires one to implement rational interfaces between CAD systems and numerical solvers to handle the filtering, conversion and clean up but this is simply part of modern engineering. It is a good test. If it takes more than a day to generate a reasonable initial grid then I would strongly suggest you need to change/fix your software system.

> Besides we need to perform a complete calculation from diffuser till outlet of combustion chamber, because otherwise amounts of the dillution air will be left unknown...

As referred to earlier, I first performed such a 3D CFD RANS simulation in 1986 in gas turbine manufacturer's research department and not academia. 20 years on it is fair bet that CFD capabilities have improved somewhat!

> Also I talked to desiners, and they couldn't say anything about real use of CFD in their work.

I am not sure I understand this point. Are you saying they would not talk to you about CFD? or that they do not use CFD (i.e. someone else performs the simulations)?

> Of course, you can find such tries almost everywhere (I mean combustion chambers computations), but they are completely unuseful.

Why?

About grid generatrion.

It took two months (!!!!) to describe half of the flame tube section. Fragments of the holes for dillution and cooling air left unresolved with very bad grid.

Do you want to say that 10x10 nodes is sufficient to describe dillution air hole? Moreover, we should consider at least 3 combustors (swirlers) segment (of annular chamber which is widespread now) of the flame tube.

How many holes for dillution air? (2 chains at least)

How many small holes for cooling air? (in the front section and along the wall of the flame tube)

How to resolve very complex configuration of the flame tube walls? Of course, there were times when very simple configuration was used, but it's 2005 now.

And You would like to assure me, that such results can be relible?

Besides everywhere is written, that k-e model gives bad results for such types of flow and we should use Reynolds Stress Models or LES, but their use is very expensive in computational cost.

How LES will work near the very complex wall, where in any case, won't be fine grid resolution???

How about reaction and liquid fuel fase and evaporation models??? Fuel dispersion models???

I had a lot of troubles because I had to perform such modelling. Therefore the task was simplified, but I my try was unsuccesful anyway.

About talks with desiners.

They have such funny story. Once upon a time a Government commision arrived in the Designer Bureau. Looking through the facilities they noticed "porn-star" sitting without a work.

"What is she doing here?" - "It's our direct connection with a God".

"How it can to be?" - "When we don't know, where to make a hole, we give her a draft and she point it"



-------------------------------------------------

They described their work. New combustion chamber is delivered at the test-facilities and the test begins. After hours of work they obsergving it and look for "burn-through". Then there they made a hole for cooling air.

Of course they know about CFD, and some of them even perform MSc and DSc theses, but they don't use CFD. They even have not such computational facilities.

> I had a lot of troubles because I had to perform such modelling. Therefore the task was simplified, but I my try was unsuccesful anyway.

Why have you concluded that simulating a combustion system is impossible rather than concluding that simulating a combustion system is difficult and you were inexperienced, possibly inadequately supervised and using inappropriate software tools for the task? Surely your literature review and/or a brief perusal of the web sites of active university and government combustion research groups indicated that others were not having the same level of difficulties.

> They described their work. New combustion chamber is delivered at the test-facilities and the test begins. After hours of work they obsergving it and look for "burn-through". Then there they made a hole for cooling air.

Is this a story? The main aero gas turbine manufacturers would normally use thermal paint to identify hot spots. Failing a combustor on a rig test would be rare and failing a combustor in an engine test is going to be very expensive.

> Of course they know about CFD, and some of them even perform MSc and DSc theses, but they don't use CFD. They even have not such computational facilities.

I do not recognise your designers but I can assure you that those at GE, P&W and R-R have access to very large computers and a wide range of numerical software to support their research and development work. Certainly some people in the group focus on numerical simulation, others experimental work and others more general issues.

Here I agree completely with You.

"Why have you concluded that simulating a combustion system is impossible rather than concluding that simulating a combustion system is difficult and you were inexperienced, possibly inadequately supervised and using inappropriate software tools for the task?"

I agree that I am inexperienced, I agree that I was pretty bad inadequately supervised (I even had a conflict with my supervisor because of that), therefore I am studying further and wish to perform sumulating I talked about.

About literature review. I spent an year in the library. But made a conclusion that adequate sumulating will be available just in some 2015 year, when computational facilities allow this.

I am not an expert in combustion, but I do pick up some logical key points of your conversion. For example, I agree with andy on a number of things: 1) the fact that you have been unable to do something, doesn't prove that task impossible for everyone else, 2) the fact that you are not aware of something, doesn't mean it does not exist. CFD is used for some of the combustion problems that were mentioned, and andy's comments should make perfectly clear that that's a matter of fact. Another point worth repeating from Jonas' comment: CFD may never be able to completely replace experiments, and the results may be questionable, but that doesn't make it "completely useless". There is no design tool that completely stands on its own, so why should we expect that from CFD? It's used as a tool in coordination with other tools and methods. It's not perfect, but it would be unwise to try and completely do without it.