Recommendation for CFD software
 

Greetings,

I am part of a small and growing company which specializes in energy systems for a variety of end users; primarily those which have different types of biomass wastes.

We have funds set aside for CFD software which we hope to acquire soon. At this point we considering only Ansys (CFX and or Fluent) and CD-Adapco (Star-CCM+) based on several factors. As we begin conversations with these companies I want to come up with some good questions to ask which will help me see through their smooth salesman ship. Ultimately I need to find a product that is best for our company and its growth. We we will use CFD for custom high temperature heat exchanger development, gas mixing, gas cooling, system output, duct sizing, etc.

Currently we do not have any engineers who have extensive CFD experience. I have some Star-CCM+ exposure but none for an Ansys package, and it was used in an academic setting.

Here a few general questions that have come to mind:
-Quality and dedication of support (subjective, I know...), but what can provide and when the help stops
-Meshing of ducts and tubes. For the most part all of out surfaces will be cylindrical in nature with transitions, joints, refractory brick, curves, etc., which I know meshers can break down. Our CAD department is small (uses AutoCAD Inventor) and cannot easily make continuous surfaces on all of our parts. The meshing process is very important to us...
-Geometry manipulation (adding or removing features)
-Post processing features

If you have any other thoughts or recommendations please let me know. I want to be as objective with the sales people as possible. I appreciate any feed back.

James

hi James
Instead of doing the hard work, let them do it for you. Put together a reasonably complex problem that is typical of the type of work that you will do and ask both CDadapco and Ansys to run you through the whole set up/run/postprocess. That should make it clear to you what you should purchase.

All the best
ACmate

James,

Just to add to ACmate comments: if you choose to get Ansys/CD-adapco to run through a case for you, I would suggest that you define the parameter(s) that the CFD model should predict and assess a range for the predicted parameters (good, acceptable and not acceptable). It usually better to have an idea before hand rather than once you get the results.

Sincerely,
Julien

Personally, I don't have experience with either code. But here is my two cents.

You are planning to use CFD for heat exchanger design, gas mixing, etc. You may be pushing the limits of CFD. The answers you get may be more qualitative than quantitative. So whichever you go with, the technical support should have an understanding of the physics and how to model it and also how to create a simple model of your complex system to at least get resonable answers for your design trade offs. Also, the sales person should be straight forward with you in regards to what your CFD expectations should be.

It sounds like your company has limited experience with CFD. I would strongly suggest that you hire a consultant to help you out with your decision.

Good idea Julien (and ACmate). There are a couple of cases we could easily have them run through and I am familiar with the operating conditions as well what we would want to see and know. Thank you!

Martin, I agree with your stance on technical support; its a good point. I am surprised however by your comment that heat exchanger design may be pushing the limits of CFD. Why do you say this? I have a few ideas of why this could be, but I am interested in what you are thinking. Thank you.

Wouldnt say that heat exchanger analysis is pushing CFD by any means really is an odd comment as CFD is widely used in this area.

But I would get both codes to do a test case and present to you then get a evaluation license of both codes and see how easy it is to run your case yourself and reproduce the results you have been provided. Then you can really assess all aspects from accuracy to support/training and ease of use and go from there.

I would also look at the workflows of each program and see which one is likely to save you guys the most time in the long run too.

James, What do you mean by work flow of each program? I am guessing the drop down trees (or ribbons, or similar) in which you set up conditions, meshing, visuals, etc.

Also, we currently have Ansys for FEA of mechanical assemblies. But we will have different people running each of these types of simulation (the FEA and CFD users will be different). So, it won't necessarily save time having the same menu options (if that is what you are implying).

First I'll ask you a question and then go on with my comment. What is your expectation in regards to CFD? A little more detail than "design a heat exchanger". Are you looking to optimize the shape of the heat exchanger? Are you looking for the amount of heat exchanged by the heat exchanger? Are you looking for the amount of pressure drop between the ends of the heat exchanger? Are you looking for the pressure exerted on the heat exchanger (i.e. you don't want it to crack or blow up) or the temperature of the heat exchanger (i.e. melt)

OK, on to my comments.

How good CFD is depends on what you are using CFD for. Are you using CFD for a qualitative or quantitative analysis of the physics? It also depends on the fidelity of the data you are looking for. Is it overall or is it local information.

I assume for a heat exchanger, and you mentioned mixing gases, you'll have issues with turbulence models (what's your Reynolds number? LOL, how does one even define the Reynolds number for the internal flow of a complex heat exchanger?), compressibility and buoyancy (depending on your application), unsteady flow (how much cpu power do you have?), and mixing gas properties. All this leads to uncertainty of what your temperature profile at the boundary is. Of course the wall temperature is also affected by the thermal properties of the heat exchanger and the temperature itself affects the fluid flow. It's a coupled problem. And, in the end, I gather that a heat exchanger is all about pumping out heat. Seems like the uncertainty levels on that metric, if that is your metric, may be large.

Personally, I'm an external aerodynamicst. But from what I understand, CFD codes can easily give different velocities in a pipe with multiple U shaped bends. I know that was true at one time. Things may have changed. That was one of the motivations for the curvature correction in the Spalart Allmaras turbulence model. Unfortunately that correction, as far as I understand, has adverse consequences in other areas. I usually don't use the corrected form for external aero. My mind is blanking about the ins and outs of the SST model in regards to pipes, but I'm sure that is in papers.

OK, that was fully turbulent models. What about transitional turbulent models? Geez, I sure don't know. Does anyone know?

Please take what I have said into account when you talk with the tech or sales support. If I am wrong, I'm wrong. So be it. No skin off your back. However, it seems so easy for some to say "Wouldnt say that heat exchanger analysis is pushing CFD by any means really is an odd comment as CFD is widely used in this area." How would jbritton know if he doesn't know your problem? And such statements are absolutely 100% false. That's like saying aircraft analysis is not pushing CFD. And I know that is false. CFD has its issues. If those issues don't impact your problem, great. If they do, beware.

Umm, is that the only reason you find my comment odd?

Personally, I didn't think that internal fluid dynamics was so much easier than external aerodynamics. In fact, I thought it was more difficult. Sure, there are a lot of wonderful success stories with CFD and external aerodynamics. However, there seems to be a lot of bad results, at least for external aero, that never make it out in papers, except for hints. I'm personally aware of many.

Martin Hegedus
www.hegedusaero.com

OK, another comment.

I assume you are interested in capturing the velocity and thermal boundary layer. Are you planning to go with a structured or unstructured solver? For external aero, fully unstructured solvers can be more "finicky" to use to capture drag than structured solvers. Thus I assume they are more finicky in regards to thermo applications.

For example, take a look at the summary presentation of DPW-4
http://aaac.larc.nasa.gov/tsab/cfdla...sentations.htm

http://aaac.larc.nasa.gov/tsab/cfdla...2009-07-20.pdf

There is much more scatter with the unstructured solvers than structured. Of course how you plan to grid your problem will affect your gridding tool choices. Also, I'm not sure if the issues of structured vs. unstructured carry over from external aero to internal. But I assume they do.

Thanks for the thought Martin. In terms of heat exchanger design (which I know is broad) is to improve our existing exchanger through the tube sizing and orientation and passes. While I agree that the end "heat transfer rate" may be questionable, we are hoping it will allow us to compare one design case to another. These exchangers are expensive and difficult to modify in real life. So, we want to use CFD to look at how certain parameters effect fluid flow, pressure drop, etc. Our current prototypes were designed using textbook formula for Nusselt numbers for internal and external flow. Bottom line CFD has to get us closer than these equations with are riddled with assumptions.

I understand that CFD has is faults, and I do not plan on taking the end result as the holy grail. But I do understand how it can help us optimize sizing, orientation, etc.

Again, thank you for your help!