I am interested in getting some user feedback on the FLOTRAN CFD software from ANSYS. Opinions, experiences, etc. would be greatly appreciated.

Thanks.

Mark Snyder

I'd like to start out by stating that I used to work for a company that was an ANSYS distributor, and I've used FLOWTRAN. I'd also like to think that my opinion is mostly unbiased.

FLOWTRAN is ok. It isn't great (the tools I have now are), and it isn't terrible (I have used a code that is).

FLOWTRAN is good for doing 2D work. The key to a good CFD solution is a good grid. ANSYS is a finite element package, and the things that make a good FEA mesh and a good CFD grid are different. A good FEA mesh is very uniform with aspect ratios cloae to 1. A good CFD grid has clustering in the boundary layer and aspect ratios are often not 1. ANSYS doesn not have good tools for creating CFD grids. You can, with a bit of effort, make a decent 2D grid. For complex, 3D geometries, you'll find grid generation very difficult at best.

The solver is not bad, it has features that you'll find in most CFD codes. It will handle compressible or incompressible flow, laminar or turbulent flow, and multiple species. It doesn't have some of the physics found in some higher-end codes like cavitation models and the ability to handle moving parts.

Post processing is not bad, but nothing spectacular.

In the end, it depends on what you intend to do with it.

If you have any questions, or want more information, feel free to e-mail me.

Alton

For more accurate feedback from users of ANSYS/FLOTRAN subscribe to xansys@onelist.com.

Flotran is great depending on your applications and the release of ANSYS you have access to. Recent version 5.6 has new physics enhancements which might be helpful. You have the added advantage of getting the multiphysics capability of ANSYS.

Hi Mark; I've been using Ansys/Flotran in my PhD research and I am almost done. It is very resonable if you get to use it in the proper way and adjusting the convergence and stability parameters. My problem was a fluid-structure interaction. I advise you if you need to use it to use frequently the on-line help. Further there is a person whom you can e-mail in case of help: rich.lange@ansys.com. Good Luck. Ramadan Texas A&M University

Alton --

Thank you for your comments and advice. Having worked for a number of years in the solid mechanics area, I really need to get myself up to speed on the types of analyses that you can and cannot perform with current CFD tools.

Could you recommend any good books (the CFD equivalents of the solid mechanics-oriented FEA texts by authors like Zienkewicz, Bathe, Hughes, etc.)

One understanding that I have about FLOTRAN is that the flow field must be either all laminar or all turbulent. Is this really true? Is it limitation with other codes as well? If so, how do you tackle problems where you suspect that both regimes will be present.

Mark Snyder

Achuth --

Thanks for your advice. I've tried for two days to connect to "xansys@onelist.com", but haven't had any luck. I'll keep trying, and hope that the site hasn't gone away permanently.

Could you elaborate on why you think FLOTRAN is a great code?

Thanks.

Mark Snyder

Mark,

With the right tools, you can perform many, many different types of analyses. My "thing" is fluid structure interaction (FSI), which combines CFD analysis with a good old-fashioned structural (and maybe thermal) analysis. There are people here who are analyzing chemical vapor deposition for semiconductors, and others working on pumps and other things with moving parts. The possibilities are almost limitless.

A somewhat classic text is "Computational Fluid Dynamics, the Basics with Applications" by John D. Anderson, Jr. (ISBN: 0-07-001685-2) published by McGraw-Hill. Like most good engineering books, it is full of equations that will make your hair hurt. What I think you'll find really useful is the first chapter which is titled "CFD: What is it?". The chapter contains a good description of what CFD is, and how it works. It also has several charts that will give you some insight into the kinds of decisions you will need to make when setting up and running your analysis.

Ah, turbulence, one of my favorite things. Most CFD codes handle turbulence via a "device" called a turbulence model. In order to directly analyze the phenomena that cause turbulence, you would need to resolve the flow field to a minute level of detail. There are codes that can do this (a technique called Direct Numerical Simulation [DNS]), but they require a grid that is so fine it isn't practical for "real world" geometries. Instead a turbulence model is used. The turbulence model uses 1 or 2 (or occationally more) differential equations to relate fundamental quantities that describe the level of turbulence to other properties of the flow field (like distance from the wall, local velocity, and fluid properties). These models contain several constants that are set based on experimental measurement. That's why you'll sometimes see turbulence models being described as good for some particular type of flow, that models parameters were selected based on that flow situation.

So what does that mean for you? If you know that the flow you are analyzing is laminar (molases flowing at low velocity), you can turn off the turbulence model. This speeds up your computation, because the solver does not have to solve the turbulence equations. If you think the flow is turbulent, or if you aren't sure, you should turn the calculation of the turbulence equations on. This should not introduce any "error" in regions where the flow is laminar, but will give a much more accurate answer in turbulent flow regions. You should also select a turbulence model appropriate for you flow situation (most codes have a generic k-epsilon model that is reasonably accurate for most situations). The one thing you should be aware of is that using a turbulence model adds some requirements for your grid. The first point off the wall should be located close enough to the wall for the turbulence model to give accurate results. That distance is going to vary depending on the turbulence model and other bits of "magic" the vendor has built into the code.

Alton

A very nice overview of the subject... let me add one thing though. The vast majority of all turbulence models used are not very good at handling laminar flow. Most often you will get turbulent boundary layers etc. from the beginning. Capturing the laminar parts and transition demands that you use a very well tuned model and have very good numerics. With the type of usage we are discussing here I think that it is safe to say that all codes are incapable of handling mixed laminar/turbulent flows and transition.