CFD consulting work pricing
 

My company (in the U.S.) is considering hiring a CFD consulting company to do a CFD project. But we have no idea how much it will cost. The project has very complicated geometry and is a compressible flow/heat transfer problem. We have CAD geometry file and we need a CFD simulation to get the solid surface temperature.

My partner says it should cost around $25k. But I think it gotta cost us at least 40k. I surely agree with my partner that we should keep shopping, but meanwhile we want to know some ideas from some experts.

Anyway, anyone can give me a rough estimate what is the price for a typical CFD consulting work?

Thanks a lot for your input.

Assuming there is no experimental component then you are paying for consultants time, plus computer time, plus overhead. If you have a complete set of CAD files in a format usable by the consulting company and a good idea of operating conditions then the preparation time should be modest. But the amount of to and froing between client and customer gathering and checking data is usually something of unknown.

You have not said what you want delivered in terms of numbers of cases, reporting, simulation data, training, building relationships, etc... which will have a bearing on the price and the type of consultant employed. I suspect as someone new to the process you will have to talk to a consultant or two to get a feel for this sort of thing.

I can confirm that $25k to $40k looks a reasonable figure for a set of simulations but I lack sufficient information to say with confidence. Instead of asking how much will it cost for a piece of work you have not adequately defined (and are unlikely to adequately define in posts like these) it may be wiser to turn it around and ask prospective CFD consultants how much will I get for a budget of $25k or $40k.

I think you can get this work done in India at much cheaper rates.
License and hardware costs will be same but the man hour charges will be much lesser. There are many companies in India doing decent CFD consultancy work.

It may make an interesting thread to find out people's experiences. I am aware of the stereotypical view of "sweat shops" using stolen software and inexperienced workers with no technical supervision. Perhaps it is now an unreasonable one?

Not all companies are "Cheap chop shops" in India.
Following are 2 companies doing very nice development and consultancy works with their own CFD solvers entirely developed inhouse in India -

http://www.zeusnumerix.com/
http://www.sandi.co.in/v2/home/

Note- I dont have anything to do with either of these companies.

Here is my 2 cents from my own personal experience in the U.S.

The overall overhead rate is about 2.9 for a company. So if someone earns $50 dollars an hour, their rate is $145 an hour. That is just short of $1200 a man day. The overhead rate for a one man shop will be less. About 2.

If you don't have confidence in a company, or individual, you should throw in a simple test case that is representative of some of the physics you are interested in. You should be aware of what to expect from a CFD analysis of that problem. If the company or individual does not meet your expectations, pass them up or find out why.

CFD is tricky and it is very easy to get misleading answers. It would be a shame if the decisions you make are based on misleading answers. Experience counts a lot.

Determine how accurate you would like the margin of error to be understood. The more you would like to nail down the uncertainty, the more effort you'll require. This can be a VERY man hour consuming task. Especially if you are trying to back up the results with data and trying to understand the data comparisons. This is an absolute pain. Be cautious of data comparisons that look good or a company that does not show poor results. Poor results sometimes tells more of a story than good results.

Whatever code is used, make sure it is compatible with your needs. For example, if there is some sort of certification process involved, you should generate the results with a code the certification body feels comfortable with.

Make sure you get the input and output files and that they are compatible with mainstream CFD codes in case you would like to change vendors.

No one likes this subject, but think about what happens if there is a failure down the road. If there is, you'll need to retrace your steps. This may not be so easy if an in house code is being used.

It's important to get face time with the CFDer(s). You want to be able to sit down with the vendor in a working meeting atmosphere and and spend a few hours mining the results right there at the computer. Ppt presentations, email, and data file transfers is just not enough.

Again, just my opinion.

Would you mind expanding on this point because I would have thought it was the other way round. However, I can see the knowledge of the user and the kind of internal and commercial software may have a bearing.

The issue is about documentation (i.e. versioning) and knowing the snapshot of the code that was used for the analysis. Sometimes, when a third party is using their own in house software things get a little sloppy and you need to trust what they are saying. There is the possibility that, down the road, others, including their own people, will not be able to recreate the results. For example, they may use a feature which is activated by a hard coded switch, i.e. the code is a branch off the main trunk. Once they use that code for your analysis, that branch must be permanently archived and appropriately documented. Or, they may use a newly implemented feature (i.e. alpha or pre-alpha release of the main trunk), such as turbulence model, etc.

Of course, if a special in-house version of the code is used, others may not be able to recreate the results since they do not have that version.

I can see your point although using development code for production work would not have been allowed in my old company.

I was looking at it more from the point of valid code but inappropriate modelling assumptions. The coding details of things like wall functions, numerical fixes, etc... are usually not made available by the suppliers of commercial code and yet may have a starring role when things do not behave as expected. It is straightforward to sort this out this for code written internally, harder work but doable for unfamiliar open source code and usually pretty difficult for commercial code that has to be examined as a black box if the commercial company does not collaborate.

Very true.

I have done simulations for the industry some years ago, and one in this year after a break. My two cents:
- The most important point is that you get reliable results. An experienced engineer will do an large amount of effort to secure the validity of the results
- It may be that your CAD file are (nearly) useless. The more details you put in a single simulation the more complicated it gets. Complicated in the sense that the simulation gets less valid. I have good experiences with simulation different parts of the problem with different geometries. I call this region of interest method. It is much easier to prove the validity for simpler cases.
- If you find someone you have confidence in you should allow him (her) to use his own simulation software.
- Normally I made a price for the project. I expect from the client however, to pay something more if the problem gets more complicated than we thought first. It may surprise you, but I had never problems in this point.

I have an example. My simulations (that were simulations of the electrical field with non linear boundary conditions) costs some thousand dollars for a not so complicated problem (much less then 5000). Even simulation for convective mass/heat transfer are not much more expansive.
I had once a customer which did not follow my recommendations and dictated what to simulate. "Simulate this geometry with this b.c." and so on. Normally I make a non dimensional analysis which may be used for a whole class of problems. In this case, I was some kind of simulation slave. It costs the company around 20000 $, because some simulation rounds were necessary. It was wrong to dispense my experience: Much more costs and teh result is only valid for a special case.

As said already it is a good idea to check the possible engineer with a smaller problem, for which you have some idea of a correct solution.