Pinned Tweet
Learn #FEM for #CFD through hands-on coding in Octave/MATLAB.
I have uploaded two lessons on solving the diffusion equation and advection-diffusion in 1D using the FEM. Will be adding for 2D advection-diffusion, Stokes and Navier-Stokes soon.
Playlist at
youtube.com/playlist?list=PL…
Chennakesava Kadapa retweeted
For typical industry problems in aerospace, automotive and biomedical industries, CFD solvers are relatively easy to develop and parallelise when compared to solid mechanics. Contacts is one of the most difficult problem to implement and efficiently parallelise for large scale problems. On top of that, plasticity is a beast to deal with. Without plasticity and contacts, solid mechanics solvers are useless at solving practically relevant problems.
If you are interested in developing parallelised solvers for CFD, solid mechanics, as well as multiphysics, I offer my expertise and consultancy services at a fair price. DM me.
Just read a paper. China has developed CFD codes to be able to run production scale simulations on a single GPU. For comparison a typical ANSYS or Dassault Automotive CFD simulation would take 8-16 GPUs for the same run time and slightly higher accuracy. And the author of the paper and the PI is this person.
China is paying well to bring researchers from abroad to innovate and research in STEM for low cost CAE products while maintaining reasonable accuracy.
This is the difference. China pays well for global STEM talent to leverage it to solve its national problems(GPU shortage, costs). India too has a cost problem. Indian OEMs can’t budget for 150-200 GPU clusters exclusively for CAE workloads. But India wants to solve and innovate for low cost by paying low cost for talent itself. In the end India optimises for cost and loses on performance and time.
I guess I am going to start rejecting papers for journals that invite me to review without mentioning the time frame but give me two weeks once I accept to review.
I am going to mention it as the reason in the comments to authors so they won't submit to that journal again.
Oooh, flutter!
What can a glowing wing tell us about the future of flight? ✈️
At NASA Langley, researchers used unsteady pressure sensitive paint for the first time on a large-scale, freely moving model in Langley's Transonic Dynamics Tunnel, turning changes in airflow into data researchers can see.
Learn more: go.nasa.gov/4BiRkuz
Video credit: NASA
The authors: We choose this benchmark to validate our method.
The benchmark: the properties of fluid and solid were inspired by aliens whose blood and heart valves are about 7.3 times heavier than mercury.
Me: just because you can pass any values to the solver, you should not ignore practical aspects.
Chennakesava Kadapa retweeted
The goal of an engineering college is to graduate engineers.
Producing scientists, while nice, should not the goal of an undergraduate engineering program.
This is important since having clarity in framing makes all the difference in deciding the curricula for such programs.
Man, how do people get funding for doing the same space-time formulations for wind turbines for almost a decade???
What a weird argument!
Any advanced textbook in my field knows more than me. I am sure it's the same for anyone.
What does it mean?
The worst thing about reduced order modelling is the amount of crap that exists as literature but solves no practically useful problems!
One ends up wasting so much time reading all of that crap! Add AI-exacerbated ROMs to that now!!
I have been reading papers on ROMs for FSI problems. Those based on POD/DMD consider weak nonlinearities and simple systems for which the proposed data-driven methods are an overkill. Actually, the wake oscillator based models are still better for many rigid-body FSI problems. They are simple and come with known expressions for the governing equations, useful for stability analyses. The recent ML-based ones are totally useless! Yet, the same authors keep publishing almost the same bullshit! It's so infuriating!!
🤦
Why do physicists have such affinity for AI slop!!