[federernadal]

Someone was talking to me about how "insanely powerful" ANSYS is, and in 10 years it is going to be pretty much synonymous with the word "engineering." What would you guys (a lot of whom are seasoned experts in this field) say ab out that?

While in my senior FEA class I too was really impressed with ANSYS' power in getting correct answers (for tests that could be checked easily against by hand), I was wondering how correct the simulations would be for tests that solely rely on the ANSYS :

How do we know ANSYS's modeling is correct in areas that cannot (or have not) been tested in real life (say for example deepwater tests in the arctic), or supersonic speeds through short diameter pipes, etc)? Just curious as to whether to take answers with a grain (or jar haha) of salt or not...

FWIW, I am thoroughly enjoying the universe of ANSYS and am really impressed with it so far!

[LuckyTran]

On the FEA side, numerical tools are extremely powerful. I would not say "in ten years" I would say "ten years ago". FEA has been used in engineering for decades already and is a very mature field. There are lots of different types of engineering, but yes I would agree that in most part FEA is synonymous with engineering.

Actually it is even worse, sometimes ANSYS is synonymous with FEA! My professors don't even know that ANSYS is a company or that the actual name of the FEA package is called Mechanical. Instead of saying let's do some FEA or structural analysis, they say "let's do some ANSYS."

What you need to be careful with when using these physics based software is to be able to tell what are the underlying physics and what are the models. The physics won't change (hopefully) no matter where you go. Force balance on Earth will probably be the same as a Force balance on Mars. This will be true until one day we all decide that Newton's laws no longer hold. But models/correlations used by the software for predicting a certain outcome obviously have some some limitations to them. The correlations are only good over their intended range for specific processes and even then they're not 100% accurate within their own intended range.
One example of what can be improved is failure modes analysis.

On the CFD side, the situation is very similar in terms of success. Except that turbulence models and such should be taken with entire buckets of salt. But not everyone is accepting of this fact.

One of the issues that will probably explode soon is that a new generation of engineers will enter the workforce that are all trained in numerical techniques that are not aware of its drawbacks and a very different engineering culture is going to arise out of that. A good example is laminar & turbulent flows. In the old days, flow over a sphere or flat plate was called sub-critical versus super-critical when laminar to turbulence transition occurred somewhere. In super-critical regime both laminar flow & turbulent flow occurred. However, because of CFD and it's limitations of being able to only model laminar-only versus turbulent-only flows many in today's generation feels that flows are either strictly laminar or strictly turbulent. Why? Because there is no option to do laminar & turbulent flow at the same time in the code!

It is indeed impressive watching someone else (the computer) do your work for you. You are basically managing a minion army. But the qualities of a good manager are not necessarily the same qualities of a good minion. The code can so truly impressive things but that won't mean anything if the user is not compatible.