[zdunol]

Hello good people,

I saw cool SpaceX presentation about parallel computing which turned out to be about combustion CFD modelling. I guess I can't post links here so just type SpaceX gpu parallel computing in yt, the video name is SpaceX Merlin (&Raptor) Engine R&D, GPU-Powered.
So, guys there created their won solver which is pretty cool if you ask me, and I got inspired by this video and started to dig into basics of CFD. I have found cool cfd course in python (12 steps to Navier Stokes with Python) and I feel like i learned a lot.

So I would like to ask you if you could recommend me any course like that, where I need to write everything on my own in Python, or Matlab. Or any textbook which is about writing CFD codes. I'm going through Anderson's CFD The Basics with Applications but it is lenghty and I would really like to do more coding.

Thanks!

[kaya]

I was in your shoes 2 years ago. Unfortunately I can't answer your question but what I can tell you is; do verify every step you take. Have some analytical functions aside where you know derivatives or nonlinear operations and check convergence before you add an extra line at each stage. It may sound a bit costly in time but believe me, it actually costs less then you would spend in finding bugs at the end when you put the time stepping for-loop.

Good luck.

[mprinkey]

I can't recommend Prof Murthy's course notes enough:

https://engineering.purdue.edu/ME608/webpage/main.pdf

That book covers start to finish how to write unstructured finite volume CFD codes in all of the gory detail. She worked at Fluent when what is still the core of Ansys Fluent was being designed and coded, so her perspective is instructional but fused with a ton of practical experience. The version I linked to is from 2002. There are likely more up-to-date references available, but I have yet to find a better place to start new would-be CFD devs.

[zdunol]

Thank you guys a lot

[zdunol]

so I have found something which looks to be cool. I don't like posting links because they go invalid, so it's a wiki book called Parallel Spectral Numerical Methods

[mprinkey]

Originally Posted by zdunol

so I have found something which looks to be cool. I don't like posting links because they go invalid, so it's a wiki book called Parallel Spectral Numerical Methods

As a rule, Spectral Methods are not great for any flow field that has shocks or discontinous interfaces (like flame fronts). Gibbs phenomena is a very real problem and attempting to use modes/basis functions that cross discontinuities just breaks down. If you are interested in Spectral-type formulations, look at Discontinuous Galerkin methods. These share a common ancestry with spectral methods and with finite volume methods. In some ways, they leverage the best aspects of both. BUT neither DG or spectral methods is where you should start to learn about CFD, IMO. Learn the basics of the Finite Volume Method first.