Important programs and programming languages for engineers...

Baseliner said:

Today at 10:02 pm edited 10:06PM

I took an FEA class senior year and you really begin to appreciate the power of ANSYS after doing just 2D simulations by hand.

ANSYS is IMHO ridiculously powerful and beautiful at the same time. It is like playing the guitar - how good you become is not a function of whether you take a class on it (I find companies selling 1-day courses on a product like CFX pretty ridiculous and disrespectful to the software), but rather how much interest and time you are willing to spend on it.

What is your guys’ takes on ANSYS simulations?

ANSYS is an great tool. We have guys here at work that use it for thermal stress, fracture mechanics, and plastic deformation. It’s amazing what it can do, compared to the FEA class I took 25 years ago (we actually wrote code that analyzed 2d plates with about 50 nodes)

@baseliner

Yes,and the development at ANSYS is expanding amazingly too. Someone once told me that they reinvest 10% of profits to R&D. Sure is a lot.

ANSYS has a blog where you can go over the new stuff coming up. HFSS is a product that “wow-ed” me.

Obviously it is only as good as the person using it. I kinda want to see their “back-end” code just to see it. Obviously I wonder when I run some simulations taking up 20+ hours, what exactly is it trying in the process to solve Navier Stokes?

Here’s an idea of the theory behind the NS numerical solution.
https://math.berkeley.edu/~chorin/chorin68.pdf

@baseliner If you search on “Ansys Theory”, you will find the pdf file for their theory manual. You will not find the code, but you will see how Ansys treats fluid flow.

How ANSYS “solves” the Navier-Stokes equations depends on whether you are using Fluent or CFX. I believe both of them are finite volume codes and they generally solve a set of equations that model the N-S equations rather than the equations themselves (e.g. the Reynolds-averaged Navier-Stokes equations). The set of equations are then “closed” using any number of different turbulence models (e.g. k-ε). I believe CFX and Fluent handle these differently (i.e. use different models and different sets of equations).

Of course, if you think these simulations take a long one, try doing large-eddy simulations (LES), and if you think that takes a long time, then try direct numerical simulation (DNS).

WARNING: non-engineers follow this thread at your own peril! :))

I am an actuary but I studied a bunch of physics and mathematics so I can follow along! :slight_smile:

Fluent and CFX were developed independently (ANSYS bought Fluent) and have some differences. Like bonehead said, both are control-volume based and use a pressure-based solution technique (for wide range applications - they are in it to make money after all). They differ mainly in the way they integrate the fluid flow equations and solution techniques.The CFX solver uses finite elements (cell vertex numerics) to break the domain into pieces. Fluent solver uses finite volumes (cell centered numerics). Think back to the theory and remember differentiating nodes and elements. CFX focuses on one approach to solve the governing equations of motion (coupled algebraic multigrid - how you probably learned it using hand), while Fluent leaves it up to the user to choose from several solution approaches.

From the usage standpoint, the way I understand it, if you were working at a refinery, food processing plant or something of that nature, Fluent would be the way to go. CFX is more for “mechanical engineering” type stuff like turbines, pumps, heat exchangers etc.

But over time both of them are converging. I believe combustion was a major reason why ANSYS bought Fluent, but in my ANSYS 15.0 I can simulate combustion on the CFX product.

Oh I know 20 hrs isnt that long. Some convergences can take as long as 6 weeks (my professor used to run these)!

What I meant I want to see their back end code with reference to my 3d model. As in, watch the colors change on my own model, with my boundary conditions - as it happens. That would be pretty cool to see. Not just some negatively sloping colored curves…