CFD software 2

[samv]

I am very excited at the prospect of becoming a "CFD specialist" for the company I work for. We currently design on more of an experience and trial and error basis. We use Pro/E. We are a manufacturer of ventilation products.

I am interested in simulating:
-fans and their effects up/downstream (fluid and acoustic),
-heat exchangers (plate heat exchangers with surface geometries)

I am currently looking at Fluent. My current understanding is this is perhaps the best and the most expensive??

I'd like to hear all of your comments on any CFD software regarding:

capabilities
user friendliness
technical support
...and cost
and any other topic you feel is important...

thanks!
samv

 

[Tunalover]

samv-

FLUENT has the best products but, as the old adage goes, "you get what you pay for." I understand that they offer a CFD package tailored specifically to HVAC applications. I've worked with FLUENT on and off for ten years and I recommend them for the following reasons:

A. Their solution kernel is the most robust in the industry. This produces the shortest convergence times.
B. Their service and technical support is second to none.
C. Everyone you'll ever talk to for customer service or technical assistance is a PhD in Mechanical Engineering.
D. They are nice, "down to earth" people.

Tunalover

 

[samv]

Thanks Tunalover - glad to hear your positive comments on fluent...

I've spent some time this morning looking at cfd-online site and have concluded that the three main commercial codes are Fluent, CFX, Star-CD. Any comments on any of these three?

Also - having limited experience in fluid dynamics (I have a bachelor in Mech. Eng, with basic fluid courses) is using these codes going to be difficult? I do expect to have to do some training and a learning curve - but it seems like a lot of the CFD users have Masters and PhD's??

thanks in advance for your comments!

 

[amorrison]

Look at

http://www.cfd-online.com

 

[samv]

I have been there...

anybody here have comments?

 

[Tunalover]

samv-
In my experience there is a hefty learning curve and, unless you're a genius with a photographic memory, don't expect to be good at it if you use it only occassionally. The best experience in heat transfer analysis with CFD is to analyze, test, analyze, test, and so on until you have demonstrated to your management that you and the tool are competent. As you become more proficient and knowledgeable, the test element becomes less important, your experience kicks in, and your company begins to increasingly rely on simulation for product development. It's tough to completely get rid of testing because the test is what instills the greatest confidence in a new design. At least, the CFD analysis will greatly reduce the development time by eliminating most of the traditional iterations to get to the final design and test. I know this sounds pretty vague, but that's how I see it.

Regards,
H. Bruce Jackson
aka Tunalover

 

[hendrixf]

We use Fluent to bolster our research work with process heat exchangers (shell/tube, fired heaters, air coolers, economizers, plate, etc.). The learning curve is pretty steep and you'll probably want to dedicate yourself for quite a while to become effective. This is the case with CFD in general. I previously worked for a finite element analysis company that developed mechanical design software and computational mechanical design is not as tricky as computational fluid dynamics.

There are several nuances that will take some experience before grasping. We have the benefit of several years of experimental research that we use to bring the models closer to reality. For example, we'll use our film coefficient models instead of the built-in Fluent correlations so that the simulations match our experiments.

Validation can often be a bit harder than model construction. We use our heat exchanger design software whenever possible to validate the CFD results.

Hope this helps. Good luck.

Fred Hendrix, Software Developer
HTRI

 

[PHYFE]

I am a user of both CFX & FLUENT...

At this stage I prefer FLUENT over CFX - although I will say the CFX developement team is very responsive to improvement requests.

I find the UI for FLUENT to be less confusing... I find Fluent's pre-processing and solving steps to be better... However, CFX-Post is better (easier) then Fluent's post-processor... But the money in CFD is on the front end - mesh quality and boundary conditions.

In general both solvers are equally as accurate - your mesh and boundary conditions will dictate your accuracy more then the solver...

Good Luck,
-Duncan

remember... CFD=Colorful fluid dynamics!

 

[ko99]

samv,

I use Flotherm by Flomerics

http://www.flomerics.com

It's a CFD program but not a general purpose program like Fluent. The user interface and features are optimized for modeling electronics cooling.

Fluent has a competing product in this market called Ice Pak. I believe Flotherm leads the market and Ice Pak is next. They are both excellent. I am a consultant specializing in electronics cooling and prefer Flotherm. Their user interface, built-in features, and product support are excellent.

Flomerics has another product called FloVent. I know very little about it but it may be more appropriate for your applications.

Good Luck,

ko

 

[samv]

Thanks everyone for your comments, & keep'em coming!

Anybody have any insight on flovent from flomerics, or starcd?

samv

 

[Tunalover]

samv-
Here's the soapbox again! One profound lesson on computational engineering was (fortunately) suffered by a coworker of mine years ago to my benefit. We were doing FEAs of hermetic "can" assemblies for use in towed array sonar with a brand new copy of ANSYS and a 3rd party pre/post processor. My gung-ho coworker (and not the sharpest knife in the drawer) was enthralled with the new tools and arranged an open house presentation to include top management.

His results plots were beautiful but had one small problem. He had hosed the boundary conditions and the results, although visually-stunning, were just plain wrong. As he went forward to show off the results, a sharp EE (go figure) in the back of the room interrupted him and asked "shouldn't the maximum stresses be over there instead?" He was right. The presenter's face turned a deeper shade of scarlet than on the fringe plots he was showing.

People tend to be so impressed with the quality of the post-processing that they sometimes blindly believe the results. Always think about the problem, use hand calculations for rough estimates, etc. before sitting down at the tube. The tool can be challenging to master and can detract you from the engineering with embarrassing or expensive consequences.

Tunalover

 

[mameli]

Here is a comparison I made between ICEPAK and Flotherm just for electronic cooling applications. Obviously the final answer is very much depend to your specific application.
-------------------------------------


ICEPAK Graphical user Interface (GUI) seems to be the Achilles heel. Setting up a model in ICEPAK considerably takes longer than Flotherm. Working with mouse and drag and dropping the objects (like Flotherm) is not possible or it is too hard to be worth to be used regularly). Working with coordinates of the points (as in ICEPAK) definitely is very advantageous and prevents the misalignments that occur in other way of drag and drop but in fact in 9 out of 10 cases, thermal guys do not mind about those very tiny misalignments. By the way this way of model setting up is also very much easy to use in Flotherm if user wishes.

As the model is being made up, ICEPAK does not generate a real perspective figure of the model as Flotherm does and this in addition to the before mentioned subject, make it much more difficult to work with objects that have too many parts and details.

Since ICEPAK is able to work with non-Cartesian meshes, Meshing is a very critical and time consuming process in Icepak. Meshing is naturally almost the most important part of a CFD analysis but Cartesian system of Flotherm along with its powerful automatic Mesher, lessens a lot of meshing hassle.

In Icepak it seems that convergence is more likely to achieve than in Flotherm and in shorter time as well and it probably means that Icepak has a more powerful solver.

In post processing, apparently Icepak gives user more options and flexibility than Flotherm although lack of perspectives, still remains a big problem.

In terms of support, Flotherm looks better at least in Europe. Icepak has only a single person to cover northern Europe for support (not sale) while Flotherm office in UK is a very active and busy office.

The ability of making automatic reports in Icepak sounds great but the format of prepared possible reports, do not look that good and useful.

For working on liquid cooling systems and liquid cold plates, Icepak seems to be much easier and beneficial to implement than Flotherm. Making liquid cold plates and pipe flow of liquids in Flotherm is not easy at all.

Capability of importing solid models into the Icepak does not sound to me as a merit at all. It is a very time consuming process and especially if the object is a bit more complicated than a single heat sink and cuboid, definitely it is much better to start from scratch.

Generally speaking…

• For liquid cooling systems and cold plates, Icepak is fantastic and much easier to implement than Flotherm.
• Almost in every case, there is a way to simulate round shapes as cuboids without causing any considerable errors occur in results. If there is such a way and it is not too hard and time consuming to figure it out, it is easier to work in Flotherm.
• If you are a professional Icepak user, you are a professional Mesher and have a good sense of CFD. (and goal keeping)
• Icepak is more useful in component level analysis than system level.
• Definitely it is worth to learn Icepak and benefit from its capabilities along with Flotherm. Maybe doing component level analysis in Icepak and System level analysis in Flotherm if it is worth to make two models because Icepak and Flotherm are not compatible and importable to each other.
• If a case is supposed to be solved in Icepak, I guess the best way is to import its solid model into Icepak, render its major parts to Icepak objects and doing the rest of it using drawing tools of Icepak itself.
• If a good CFD package means an optimum collection of GUI, Mesher, post processor and solver, it means Flotherm


M. Ameli

http://www.thermacore-europe.com

 

[ko99]

Thankyou for sharing your evaluation. Very helpful. I currently use Flotherm but also evaluate alternatives from time to time (including Ice Pak). A couple of questions & comments:

1. I don’t quite understand why you place so much importance on the perspective view.

2. You mentioned Ice Pak is better at component-level analysis. What do you think of Flotherm's component tool (called Flopack)?

3. You found Icepak convergence better and faster than Flotherm. I've had the opposite experience and assumed it was due to Flotherm's simpler orthographic grid, or perhaps my experience with Flotherm (Flotherm has many tricks for improving convergence, I assume Ice Pak does too).

4. Agree Flotherm support is excellent (i’m in USA west coast)

5. Agree that liquid cooling is difficult in Flotherm.

6. Agree that gridding/meshing is very easy in Flotherm

7. Agree that it's usually faster to create geometry from scratch, but there are some cases where I find importing a solid model is helpful. How does Flotherm MCAD interface compare to Ice Pak?

8. Agree that the automatic reports (Flotherm has them too) are of little value.

9. I find Flotherm’s optimization tool is very useful. Do you know if Ice Pak has anything like this yet?

10. Did you have a chance to compare radiation and/or transient modeling?

Thanks again,


ko (

http://www.ecooling.biz)

 

[mameli]

Thank you for very useful comments.

1- I find it extremly hard making models with holes and openings and too many components like a PCI unit, without checking the perspective in Flomotion continiously. BTW spending sometime on rotating and looking at perspective, normally gives me some idea about what to do next.

2,7- I have never used Flopack and Mcad.Cad importing is a standard part of ICEPAK not with a seperate license like MCAD.

3- Using all the convergence tricks, I rarely get a full convergence in complicated models (and rarely need to get it)in Flotherm. I have just started working with ICEPAK but it has all the convergence tricks and my first few models converge quite fast.

9- ICEPAK does not have optimization although I guess they have added it in its newest version.

Optimization in Flotherm is a bit of confusion. If your model is not fully converged but you accept the results and try to do an optimization, you notice how strange it behaves. Sometime everything goes well, all the scenarios run and you get the optimum results and sometime you define the optimization and leave it over night to be solved and in the morning see it has stopped after the first scenario due to unconvergence.

10- Apparently calculating the shape factors for radiation takes longer in ICEPAK than Flotherm.

For Transition, ICEPAK gives you the power to define which set of results to be saved and you can see how a parameter changes during transition in post processor while in Flotherm you got to run and save it manually for every step and obviously you can not animate it. (Flotherm has added this in its newest version)

M. Ameli

http://www.thermacore-europe.com

 

[Maxham1]

I've worked with both Icepak and Flotherm and find both of them to be excellent tools. I think Flotherm and Icepak are designed for electronic packaging, whereas both companies offer another tool for larger scale modeling as in HVAC. The Flomerics tool for HVAC is called Flovent and I forget the name of the Fluent HVAC tool.

I think the big difference between the electronic packaging and HVAC CFD tools is the manner in which turbulence is modeled. You may be able to get away with the electronics-packaging tools (Flotherm or Icepak) by tweaking the turbulence model, but I've never tried nor had the need to.

I think anyone considering modeling for CFD should get the CAD interface tool. An earlier version of Flotherm that I started with about 6 years ago did not have the MCAD interface and I found that you can draw almost anything on your own, but you'll spend a heck of a long time doing it.

If I had to list any significant differences between Flotherm and Icepak I would say that Icepak can handle non-cartesian geometry better than Flotherm. But Flotherm gives the user better control over the meshing. For us these two points are not show stoppers, especially when you consider the excellent support staff at both Flomerics and Fluent. They really know CFD and are very friendly and professional. They can help you through almost any problem you'll have with convergence, modeling, sanity checks of results, etc.

 

[Bgoldstein]

You might also want to look at "easier" to use CFD packages, like CFDesign from Blue Ridge Numerics. No, it doesn't have all of the capabilities of a package like Fluent or Star-CD. However, it is MUCH easier to use, and cheaper. It has a great interface to Pro-E and SolidWorks, and has great technical support. I've used Flotherm and Icepak before, and I prefer CFDesign to these 2. For what you are doing, it may be all that you need.

With that said, if you are going to become a full time (or close to that) expert user of CFD software, then you'd probably be happier with something like Fluent, where you'd have the most control.

 

[Tunalover]

Maxham1-
As an aside, you said that both companies offer tools to handle "larger scale" problems as in HVAC. Are you referring to the fact that dimensions are larger in an HVAC problem? If so, note that a CFD code really doesn't care what system of units it uses. ICEPAK and FLOTHERM could be used on HVAC problems. The difference are that the products designed for HVAC problems have HVAC-specific menu picks, library elements, and GUIs.




Tunalover

 

[stanier]

Has anyone used Algor's suite of programs that cover flid flow, heat transfer and mechanical event simulation? How does the product/price compare with the others being mentioned?

 

[Maxham1]

Tunalover-

I see your point about the HVAC products having HVAC-menu picks. I didn't realize that. I do realize a CFD tool doesn't care what units you use.

I was under the (wrong?) impression that the critical difference between electronic-packaging CFD tools like Flotherm/Icepak and HVAC CFD tools like Flovent is the way turbulence is modeled. I've forgotten most of what I learned in graduate CFD courses where we actually wrote Fortran programs to model the Navier-Stokes equations, but I remember the challenges with large-scale turbulence and how it's different than small-scale turbulence found in electronic packaging. The large-scale turbulence issue is the reason we're not 100% happy with the weather reports. Predicting the weather - now that's a very-large scale HVAC problem

Anyway, it's probably best to check with the suppliers themselves to confirm what I've said above, but I thought the difference between electronic-packaging and HVAC was more than menu-picks and library elements.

 

[Tunalover]

Maxham1-
All the codes have robust turbulence solvers (because the solution kernels are the same regardless of the application; for example, ICEPAK uses the UNS/FLUENT kernel) but I don't know if the HVAC code will do laminar flow problems.



Tunalover