CFD Research: Mainly using commercial codes or developing new codes?
 

For graduate school research, do most graduate students doing CFD research use commercial codes or develop their own codes?

its definitely your decision..
you have to take your own decision..
in the Msc level may be you can want to learn the advance analytical fluid dynamics and then numerical methods.. later than you can learn commercial cfd packages and cfd application engineering..

after these steps may be you can want to develop some new coding techniques..

this is the my own way;)

I am a graduate student and I have developed my own code, but I took a lot of time to learn it. However, most of my friends are using software packages.

The process of developing a code could be really stressful and time consuming. Moreover the software packages and related tutorials are readily available. Therefore, I think, most of the people go for it.

And now that I think of it, if I had worked with software packages first and got to know how cfd is used, how the results are verified and validated, then I might have developed my code much faster.

But then, I think I would have to keep my focus and constantly remind myself that I am using this package in order to understand different aspects of cfd otherwise I would loose my focus from coding.

Writing code is pure overhead and in itself contributes nothing to CFD research. In my experience the best quality research performed by graduate level students on CFD methods involves taking an existing CFD research code being used by others in the research group and modifying it to include what the student is studying.

That's what I did on CAFFA code. This helped a lot in my understanding of numerical methods, the concepts behind them and their range of applicability. I wouldn't have learnt or understood them if was working with a package.

mmmhhh.... I think you can not really contribute in modifying succesfully an existing code if you have not previously learned to build your own-made CFD code... at least a simple cartesian structured code is a must for anyone involved in CFD researches!

It's done step by step. I had the chance to work with distinguished professors and they guided me in the best possible way. We began by discretising simple equations, then I wrote a few codes in fortran and C++ for 1D problems and then I worked on Mr. Peric's CAFFA code.
I feel that this procedure helped me to understand CFD deeper than working on Fluent for example from the beginning.

The new researcher needs to be able to program but that should have been taught at undergraduate level if not before for students suitable for CFD research. For such a person to then spend time writing training programs of whatever kind rather than progressing their research study is unlikely to be wise. It is both typical and more efficient to crack on with understanding the software and tools that will be used to progress the research study and talking to fellow CFD researchers about why this and that is done. Everything to do with writing code is overhead and should be minimised if CFD research is the objective.

Of course there may be other objectives such as what the researcher would rather do such as developing practical skills and knowledge instead of getting on with their research studies. Also writing code can have a more important role outside established CFD groups or in newly formed CFD groups. But new researchers are normally pretty poor at writing useful CFD code from scratch because they lack experience and practical knowledge.

You will definitely write some training codes if you take graduate courses. The sooner you take the course, the better.

Then you use flexible software packages where you can modify stuff. No need to write a complex multiphysics code from scratch: there are teams of mathematicians doing it at Fluent, CFX, Comsol, etc. You think you're better/faster than a team of experienced code developers? IMHO, no way.

My 2 cents.