I work for a company operating in the oil and gas industry. Currently we are preparing to buy a CFD license. Requirements for the software are:

- Meshing of complex geometries possible - Multiphase models - Easy gridgeneration - Efficient computing

There are several codes on the market. Can anyone give his/her opinion on a good code for the process industry ? Thanks in advance.

John,

I am an application engineer with CFD Research Corporation. We are using our CFD-ACE+ code for a variety of problems including: * Pump analysis (some with cavitation) * Valve analysis * Analysis of flows with chemical reactions * Flows involving either mixing or separation of phases

Geometry can either be set up in our pre-processor, or geometry can be read in from a DXF or IGES file. For complex geometries I generally use IGES files generated by a CAD package (I use Pro/E). Once the geometry has been imported, grid generation does not take very long.

In terms of efficency, I think CFD-ACE+ is comprable to other codes on the market. I have solved to convergence a 150K node flow/ heat transfer/ chemical reaction problem in ~6 hours on a P3-700.

You can get more information from our web site <www.cfdrc.com> or send me any specific questions that you have.

Regards, Alton

(1). The issue has been discussed several times here. Do a search of old messages here to find some answers. (2). What medicine to take? It is safer to see a doctor first. So, in this case, hire someone with CFD background to look at the company's need first and provide the plane to use CFD. (3). The fundamental problem in CFD is: there are some codes, functions, modules, availables to do a lot of different things. As a result, for the complex geometry problem, you will be using a geometry and mesh generation code. For multi-phase problem, it is really a research problem. And in most cases, people are still using test data and correlations for the pressure loss and heat transfer calculations. (even the test data is hard to find) And for the post-processing, there are also some good codes out there. (the integrated post-processor sometimes is limited in capabilities) (4). Even if you have all of the CFD codes on the market available to you, there is no guarantee that your problem will be solved. (5). My suggestion is: hire an experienced CFD engineer to take a look at the company's need first. Then create a research project to find out whether the problem can be solved by using CFD approach.

I think John's made some good points. Also I think its fair to say that the 'initial' results of the CFD project are very important, in the sense that if its successful it will be relatively easier to persuade more investment etc in it later . . . conversely if the results are poor (for whatever reason) convincing management to continue with it will become difficult. Hence setting small achievable goals and generating realistic expectations in those who will decide future direction/projects is quite important.

Greg

My suggestion would be to give same problem to all the companies who's licence you want are considering buy and make it clear that you will buy the software only if you find the results satisfactory. compare threir results and decide on which licence to buy. If a company had already done this work it would not take them long solve the problem ..., but if it is something new for them, they will not be able to give you a definite time frame and will probably charge you with based on actual man hours required instead of agreeing for a definite lump sum amount. Atleast this has been our experience in past while deciding which code to buy for our necessity.

The licence charge of the comercial CFD codes is quite hefty around 20,000 USD per annum.., so risk involved is quite high

And dont always believe in result they have obtained, becuse my experience is, that if some one is not so good in CFD, these people are quite smart to convince them.

I agree with John's view that first you have to look for a good CFD engineer who can interpret the result and additionally i think a good technical expert in your field to judge the quality of result. Some one who has a mix of both would be the best in my view.

I hope i had made the point clear.

Regads

Apurva

John,

The requirements you have set up are quite general and will probably be satisfied by most general purpose CFD codes available on the market today. Which code is best for you will depend on the specifics of the problem you intend to solve. If you don't know what applications you will be working with, it will be utmost advantageous to try to establish this, at least to some degree, before you purchase an expensive software license.

There are many ways to do this. One is to have someone analyse the possibilities and potentials of using CFD in your company. This can be done by a consultant firm, as an academic work, or by employing/hiring someone to do it as an internal project with clear focus and confinement. This will generate some costs before the advantages show up, which is often hard to motivate if your company (or rather your management) is inexperienced with CFD. Nevertheless, it will be worthwhile, allowing a much more focused allocation of efforts and resources once you go to the step of purchasing a license.

You can also decide yourself upon some applications that appear interesting, purchase a license and have someone in the company trained in the code and begin simulating. This will generate some results as the experience is acquired, and is thus more appealing to the management. However, this is also the hard way to do it. Unless your company has quite a lot of luck, there will be setbacks due to a combination of ill-chosen applications, useless results, and unrealistic expectations. Sadly, the poor sod of a CFD engineer, who is most likely to be on his own since the company wants to keep down the costs in this phase of CFD introduction (and forever after that), will have to face the disappointment of the management as well as his own, and most of his efforts will be as much about justifying his existence in the company as about developing its CFD use. In my belief this approach is also the least efficient and will therefore eventually be the most expensive one.

My advice is therefore to confine the area of applications as much as possible before you try to generate any CFD results at all. Knowing the focus of your activities, you should have an external partner, such as a consultant or an academic environment, analyse all or the most important applications to assess if the ratio between efforts and use is acceptable for the company to bring the activities in-house. In some cases the continued use of consultants is going to be more cost-efficient. All these activities should be carried out on a project basis with confined goals and a confined time schedule.

If you decide to still bring CFD in-house, get in touch with a selection of different software vendors (for instance after a quick survey of their products on the web), present your applications and establish the ability of the different codes to deal with exactly the kind of problems you need to analyse. Many vendors will perform free demo calculations on your cases to get you as their customer. Select a code which suits your specific needs.

Once you start your in-house work, keep the focus on the identified core applications, at least to start with, and concentrate on generating some experience and best-practice approaches whithin those areas. Oh, and one more thing: it is an absolute requirement for efficient utilisation of CFD to direct resources in accordance with the level of activity you choose. There are limits to how much one person can do, a fact that is poorly recognised in many areas these days. Moreover, CFD is a full-time job for an engineer and combines rather poorly with having the same person working with a multitude of tasks. If you haven't got resources or work for at least one full time CFD'er, this could be one reason for using consultants in stead. If you did a good job in the previous parts, this should come automatically, though.

Good luck,

Lars Ola