Realtime CFD ?

[freebeard]

Open VDP
It seems like this is limited to animation.
They're not measuring anything.
Their algorithms have been cribbed from actual kinematic phenomena recorded and stored in front a green or blue screen.
'Air' cannot be 'observed.'

OpenVDB
It's a physics engine used in animation
They're not measuring they're generating data.
It's computational. no green screen.
"air' can be described mathematically.

[aerohead]

Originally Posted by freebeard

OpenVDB
It's a physics engine used in animation
They're not measuring they're generating data.
It's computational. no green screen.
"air' can be described mathematically.

here's some context:
* digital video interactive ( DVI ) undergirds today's viewer-controlled video.
* it was invented in 1988 by the David Sarnoff Research Center, Princeton, New Jersey, by Director, Arthur Kaiman and his team.
* it was the first time that motion images were created by a computer, using images originating from either CGI or a camera.
* digital processing and storage was the breakthrough.
* 'real looking' film was converted to 180,000 bytes/frame.
* then digitally compressed for storage on a laser-CD-ROM memory disc, by a compression algorithm, requiring a mainframe computer.
* decompression was handled by two, decompression algorithm, custom integrated circuits inside a personal computer, with joystick control, handling video and audio, memory block and logic block.
* computer software allowed the user to change colors, reposition objects, scan 360-degrees, or point to an object and get more information.
* The Radio Corporation of America ( RCA ) was involved.
* General Electric Corp.
* Microsoft.
* VLSI Technology Inc., San Jose, California.
* GE would donate Sarnoff Labs to SRI International.
* GE, retaining the rights to DVI ( DVI Technology Venture ).
* October 14, 1988, Intel Corp. acquired DVI Technology Venture.
--------------------------------------------------------------------------------------
* the computer can reshuffle digitally coded bytes quickly manipulate what the viewer sees by simple commands.
* a panorama view can be created from any existing fish-eye photograph in memory.
* textures can be draped over any CGI object.
* flight simulators function this way.
* military training materials.
* business training materials.
* music videos.
* all images are generated from the CD-ROM.
* internally, a minicomputer is linked to a graphics workstation, and block-level architecture.
* as of 1989, a videodisc data rate was 4 Mb/s.
* Application software is outsourced to qualified vendors.
------------------------------------------------------------------------------------
* the 'dynamics' of physical phenomena appearing within the imagery, to the best of my knowledge, are simply 'artifacts' derived from scans of actual kinematics and dynamics which are stored in retrievable form within a lookup menu.
* the viewer experiences the 'look' of a dynamical systems which have been scanned and pre-recorded, not 'calculated' in the scientific sense of the word.
( George Lucas actually scanned film footage of real aerial dog fights in order to capture the dynamics later used in model, motion-capture-created dogfights appearing in Star Wars).
* While extremely sophisticated, even in 3D, the 'product' is very different from CFD, and I'm still unsure about the 'grid.'
* as of 1989, NASA's Numerical Aerodynamic Simulation Systems Division @ Ames Research Center reported that, 'the effect of turbulence, (is) something that no computer can yet simulate accurately.'
they went on to say that, to solve 'turbulence' would require a computer which was 1,000,000 times more powerful than the Cray Y-MP/832, which, according to Moore's Law, wouldn't be around until 2019.
One of my tasks today, is to see if I can find that Cray's capability in 1989.

[freebeard]

I shall adress two of your points

* Application software is outsourced to qualified vendors.

Qualified vendors see the value in Free and Open Source software and the GPL. Evil vendors will attempt to 'embrace, extend and extinguish' as Microsoft is currently doing.

https://en.wikipedia.org › wiki › GNU_General_Public_License
GNU General Public License - Wikipedia
The GNU General Public License (GNU GPL or simply GPL) is a series of widely used free software licenses that guarantee end users the four freedoms to run, study, share, and modify the software. The license was the first copyleft for general use and was originally written by the founder of the Free Software Foundation (FSF), Richard Stallman, for the GNU Project.

I've met and talked with Richard Stallman error: it was John Gilmore.

* the 'dynamics' of physical phenomena appearing within the imagery, to the best of my knowledge, are simply 'artifacts' derived from scans of actual kinematics and dynamics which are stored in retrievable form within a lookup menu.

Adversarial Generative AI will generate speculative solutions en mass and pit them against a fitness alogrithm to find an optimal result. Kind of like a thousand monkeys with typewriters vs a literary critic.

[aerohead]

I used 'qualified' within the context of an outside provider which has the technical ability to write all the code based on applied mathematics.
SONY Corp. would be an example:
They created code which interrogated two, pre-broadcast, adjacent lines which would emerge, placed in between the real signal material, across a video screen, interpolating a 'new', artificial broadcast line from their algorithm, creating a high-resolution image from low-res..
Not every entity has chops like that.

[aerohead]

There is no need of AI with respect to 3D, viscous-flow, bluff-bodies, in ground proximity, at super-critical Reynolds number.
The 'word salad' I mentioned was specifically created to 'do' sophisticated fluid mechanics.

[freebeard]

They created code which [interrogated two, pre-broadcast, adjacent lines] which would emerge, placed in between the real signal material, across a video screen, [interpolating a 'new', artificial broadcast line] from their algorithm, creating a high-resolution image from low-res..

Reviewing https://en.wikipedia.org/wiki/Digita...ve#Compression

The original video compression scheme, called Presentation Level Video, was asymmetric in that a Digital VAX-11/750 minicomputer was used to compress the video in non-real time to 30 frames per second with a resolution of 320x240. Encoding was performed by Intel at its facilities or at licensed encoding facilities set up by Intel.[4] Video compression involved coding both still frames and motion-compensated residuals using Vector Quantization in dimensions 1, 2, and 4. The resulting file (in the .AVS format) was displayed in realtime on an IBM PC-AT (i286) with the add-in boards providing decompression and display functions at NTSC (30 frame/s) resolutions.

So they capture a still frame and then extract the differences in subsequent frames to upscale from 320x240 to NTSC (640x480).

Currently upscaling is available as a free or paid service online: duckduckgo.com/?q=AI+upscaling&t=raspberrypi&ia=web

The new hotness is outpainting:

https://www.dpreview.com › news › 8850471712 › openai-adds-outpainting-feature-dall-e-allowing-users-to-make-ai-images-of-any-size
OpenAI adds 'Outpainting' feature to its AI system, DALL-E, allowing ...
2 Sept 2022Outpainting can extend original images, create larger images in different aspect ratios, and much more. Plus, the system considers existing shadows, reflections and textures, to help make newly-generated content appear natural. Outpainting is now available to all DALL-E users on a desktop. If you'd like to join the public beta, join the waitlist.

[aerohead]

Originally Posted by freebeard

Reviewing https://en.wikipedia.org/wiki/Digita...ve#Compression


So they capture a still frame and then extract the differences in subsequent frames to upscale from 320x240 to NTSC (640x480).

Currently upscaling is available as a free or paid service online: duckduckgo.com/?q=AI+upscaling&t=raspberrypi&ia=web

The new hotness is outpainting:

The outcome of their synthesis was a doubling of the image resolution, which may have required a higher resolution screen to begin with, in order to be able to add the intervening lines.

[Cd]

Several months ago, I came across a video on some beta CFD software. I think it Autodesk that was testing it.
What excited me, was that it has "predictive CFD " perhaps using A.I.
You would input a model, and the system would make suggestions on problem areas and solutions.
You could also test out certain additions, such as fairings and covers.
They came in a library much like the primatives used in 3D programs.
Apparently, the testing didn't go anywhere, and that's the last I have read of it.
I bookmarked it and forgot about it ( lost somewhere now )
However, I see AirShaper is offering something slightly similar.
It runs the test and can optimize your creation .
https://youtu.be/cZAhPQFINZ8
I'd love to see some of us do some testing on our cars in CFD.
I think we take a lot of crap because eventhough our aeromods can be proven to work by simply showing an increase in MPG, common people are more impressed to see some pretty CFD pictures alongside some cD data that shows a change.

[aerohead]

Originally Posted by Cd

Several months ago, I came across a video on some beta CFD software. I think it Autodesk that was testing it.
What excited me, was that it has "predictive CFD " perhaps using A.I.
You would input a model, and the system would make suggestions on problem areas and solutions.
You could also test out certain additions, such as fairings and covers.
They came in a library much like the primatives used in 3D programs.
Apparently, the testing didn't go anywhere, and that's the last I have read of it.
I bookmarked it and forgot about it ( lost somewhere now )
However, I see AirShaper is offering something slightly similar.
It runs the test and can optimize your creation .
https://youtu.be/cZAhPQFINZ8
I'd love to see some of us do some testing on our cars in CFD.
I think we take a lot of crap because eventhough our aeromods can be proven to work by simply showing an increase in MPG, common people are more impressed to see some pretty CFD pictures alongside some cD data that shows a change.

It sounds like a virtual extension of 'cyber-physical systems ( CPS )', where non-stochastic search algorithms interact with virtual mechatronic structural model CFD solutions.
The algorithm searches along the domain of the model, morphing it's shape, then analyzing it's dynamics in an iterative process, compiling all the data, which can be automatically sorted for a 'best' outcome.
As long as 'Paris Dressmakers' dictate automotive shapes, there'll be a place for polishing turds, which CPS could certainly be a part of.
If, however, low-drag ever becomes a priority, then there's little to nothing new to be learned. Virtually all conceivable aerodynamic architectures had been investigated within the 80-years following the Wright brothers' flight.

[freebeard]

It sounds like a virtual extension of 'cyber-physical systems ( CPS )', where non-stochastic search algorithms interact with virtual mechatronic structural model CFD solutions.

AKA Adversarial Genrative AI.

I have yet to write Python code, but I'm convinced a solution to volk-CFD is in the future of Blender. Curently, the Geometry Node Editor is a graphical code generator that recently added animation nodes. A strategy of shrinkwrapping a curve over a mesh manifold, then converting it to a mesh and solidifying it might be a start.

OTOH, GTP Chat could make a good frontend.

GPT Chat's AI is writing Blender Python now! and you're gonna love it!

Not ready for it's closeup yet, but Blender has the Asset Library to populate with vehicles and mods, and OpenVDB to accommodate the far field that aerohead's favorite equations require. It could come down to writing a verbose description.

all conceivable aerodynamic architectures

Mine would involve a vehicle that morphs from a Volkhart-Sagitta V2 above the axle line to Aston Martin Valkyrie below.