Great work! It's really nice to see complex experimental topics covered in such a manner. Your graphics rival (and are honestly better than) a lot of the big million+ subscriber science channels so that is super impressive. As a young experimental aerodynamicist who hasn't done my masters or phd yet your videos are great for giving me a better understanding of techniques I've encountered at work but don't fully grasp the math behind.
Thanks for the compliment! It was a lot of effort to put it together so I am happy it's appreciated!! As I get time I'll be covering more of those cool techniques!
I hope you enjoyed this video! Feel free to comment if you have any questions or if you find some mistake! Also, if you're not yet familiar with POD, please check my previous video on the technique! th-cam.com/video/axfUYYNd-4Y/w-d-xo.html The Pokemon game scenes come from a fully-playable Pokemon mod you can check out in my Github page github.com/3dfernando/SPOD-Pokebattle I'll keep a list of observations that were brought up by my colleagues and or commenters: - 9:24 You should multiply the coefficients by exp(i*theta). That pi shouldn't be there! - 15:55 The peak at f=22.2kHz shows backwards wave propagation. This is because of Aliasing! The mode represents the 4th harmonic of the impingement tone which is at 24.7kHz, but the high-speed camera captured the phenomenon at 46.9 kfps, so its Nyquist frequency is 23.45kHz. The peak will still appear in the data, but it will fold back into the resolvable spectrum. This is not a limitation of only SPOD but more generally a limitation of the Fourier transform.
Fernando great Job! i have tried hard to develop an a MATLAB algorithm for POD analysis of turbulent flow relevant to my research work, i could not please any help from you would be appreciated.
@@ademolajayeola9846 Ademola, I hope this was useful! Performing regular POD in Matlab is fairly straightforward as you only need to use the svd() function. For a SPOD implementation, my favorite is the one by Schmidt and Towne (2019); link in the video description for a Matlab Exchange file.
Um salve do Brasil! Excelente vídeo! Realmente faz falta um conteúdo desse nível no youtube, normalmente eu só encontro coisas bem elementares sobre análise de fourier em geral
Thank you for sharing, it helps me a lot! But I have not fully mastered his mathematical theory. for the filtered correlation matrix S, if the filtered length Nf=0, does it mean that the number of blocks =1, which is POD? If Nf is equal to data length, number of blocks is equal to data length, is this the discrete Fourier transform?
I was wondering if you know if there's any sort of tutorial on how the SPOD can be applied to the data acquired from a PIV system? i have 1000 files in ".vec" format from TSI PIV system and I can't figure out how can I convert it into something that can be read by the SPOD algorithm
I'm a big fan of the streaming SPOD code implementation in Matlab, by Prof. Oliver Schmidt. It's pretty easy to follow and modify to your input data set You can see it here: www.mathworks.com/matlabcentral/fileexchange/69963-streaming-spectral-proper-orthogonal-decomposition
Hey! Loved your video. Huge fan of Steve Brunton as well. Was just curious if you could link some good, cheap float sensors. I'd really like to attempt the experimental setup you show in your video! Thanks!
Hey Noah, I'll take your comment as a compliment to my Blender abilities 😊 That scene is CGI, I generated the data for the heights of the wave using the dispersive wave equation and used it as a displacement map on a plane in the 3d animation software Blender. Pretty cool, huh?
@@fzigunov that is pretty awesome! Would you recommend a similar route for myself? It'd be cool to not have to pay for sensors and create a whole physical setup etc.
Thanks for the dense effort put into making this video. Is there a way to apply the method to unevenly distubuted sensors such as rainfall statiıns in a region
Absolutely. If you look at 10:52, once you flatten the sensors dimension, the technique is agnostic of how the sensors were distributed. So if you have a point cloud of sensors, your SPOD modes should capture spatially sparse samples of the spatial waves in the weather.
This is like opening a flood gate to me. The ideas behind these techniques are extraordinary, and I believe you did amazing when setting up examples for us to understand and visualize. The quality here goes beyond the roof! Did you use blender for the animations? Did you transfer the color data for the cubes from your code to the graphics editor? Or did you use simulations built into (what I assume is ) blender to extract data or a mix of both?
Hi, Mich, thanks for the feedback! I'm happy this video was helpful! I used Blender and Matlab for most animations. I simulated wave propagation with a wave equation solver I wrote in Madlab, then the data was read through a script in Blender Python scripting to color the cubes. I thought it would be cool to use Blender to run the collision simulations to add some fun special effects 😊 The waves on the water at the beginning are a displacement map applied in Blender, from a video exported from the Matlab simulation, so that was significantly more straightforward to do.
Yeah, the mathematical machinery behind both is the same. I believe there's a minor difference in interpretation (i e., POD includes how to organize your data and what you do with the information you get from SVD; whereas SVD is just the barebones mathematical operation).
Awesome video. Did you use the 3blue1brown animation framework to achieve the visuals? Hope to see more in the fluid dynamics field. Its quite rare to see animated examples of journal topics which can get quite abstract or convoluted. Visualisation always helps drive everything home and form some kind of intuition.
Thanks!! I've used Manim (the python library 3b1b uses in his videos) for some of the visualizations, but also Matlab and Blender. Blender is really cool for nice 3d visuals and it's free software!
So all waves in the universe behave the same. Light and darkness, sound and stillness all the same just have different weight. That section where sound expand but I also contracts to something the steady motion it was in... Nothing does coming from something nothing is something.
Great video! I have suscribed ro your channel. I will be pursuing a PhD in fluid dynamics, and it would be great if I could send you a private message. Let me know if that is possible! ;)
![[Old] Renormalization: Why Bigger is Simpler](http://i.ytimg.com/vi/yfdboKFnjac/mqdefault.jpg)
As a PhD in Fluid Mech in my first year, your channel is an epic find!
The production value in this video is off the charts!
Thanks, Anibal! I thought the silly animations would make the math a little more bearable =)
production value is insane
Great work! It's really nice to see complex experimental topics covered in such a manner. Your graphics rival (and are honestly better than) a lot of the big million+ subscriber science channels so that is super impressive. As a young experimental aerodynamicist who hasn't done my masters or phd yet your videos are great for giving me a better understanding of techniques I've encountered at work but don't fully grasp the math behind.
Thanks for the compliment! It was a lot of effort to put it together so I am happy it's appreciated!! As I get time I'll be covering more of those cool techniques!
I hope you enjoyed this video! Feel free to comment if you have any questions or if you find some mistake!
Also, if you're not yet familiar with POD, please check my previous video on the technique! th-cam.com/video/axfUYYNd-4Y/w-d-xo.html
The Pokemon game scenes come from a fully-playable Pokemon mod you can check out in my Github page github.com/3dfernando/SPOD-Pokebattle
I'll keep a list of observations that were brought up by my colleagues and or commenters:
- 9:24 You should multiply the coefficients by exp(i*theta). That pi shouldn't be there!
- 15:55 The peak at f=22.2kHz shows backwards wave propagation. This is because of Aliasing! The mode represents the 4th harmonic of the impingement tone which is at 24.7kHz, but the high-speed camera captured the phenomenon at 46.9 kfps, so its Nyquist frequency is 23.45kHz. The peak will still appear in the data, but it will fold back into the resolvable spectrum. This is not a limitation of only SPOD but more generally a limitation of the Fourier transform.
Fernando great Job! i have tried hard to develop an a MATLAB algorithm for POD analysis of turbulent flow relevant to my research work, i could not please any help from you would be appreciated.
@@ademolajayeola9846 Ademola, I hope this was useful! Performing regular POD in Matlab is fairly straightforward as you only need to use the svd() function.
For a SPOD implementation, my favorite is the one by Schmidt and Towne (2019); link in the video description for a Matlab Exchange file.
@@fzigunov okay, thanks
This is an amazing video! Great job breaking the concept down
PeriTune no. 204, Piano Sad2 (Strings)
This video is gold.
Amazing video!
Um salve do Brasil! Excelente vídeo! Realmente faz falta um conteúdo desse nível no youtube, normalmente eu só encontro coisas bem elementares sobre análise de fourier em geral
Valeu Bruno! Fico muito feliz em poder contribuir! Cada comentário me motiva a continuar produzindo quando me sobra um tempo!!!
THIS IS AMAZING PLEASE KEEP DOING THIS
Very nice!
Great video. Do you need time resolved PIV measurements to conduct spectral POD? Thanks
Hi, Honcho, you're correct, you need time-resolved data (PIV, PSP, Shadowgraph, etc.) to perform SPOD
At the end you said "I know this is not much", but I thought all the visualizations you included were a lot more impressive than "not much"!
Outstanding!!!
Thank you for sharing, it helps me a lot! But I have not fully mastered his mathematical theory. for the filtered correlation matrix S, if the filtered length Nf=0, does it mean that the number of blocks =1, which is POD? If Nf is equal to data length, number of blocks is equal to data length, is this the discrete Fourier transform?
I really liked the video.
BTW what is the music that you used?
Hey, in 12:00, since you're dealing with complex numbers, wouldn't the V^T be V* instead (meaning the complex conjugate transpose)?
Hi, Gustavo, you're correct. I'm sorry the typo made it to the final cut.
I was wondering if you know if there's any sort of tutorial on how the SPOD can be applied to the data acquired from a PIV system? i have 1000 files in ".vec" format from TSI PIV system and I can't figure out how can I convert it into something that can be read by the SPOD algorithm
I'm a big fan of the streaming SPOD code implementation in Matlab, by Prof. Oliver Schmidt. It's pretty easy to follow and modify to your input data set You can see it here: www.mathworks.com/matlabcentral/fileexchange/69963-streaming-spectral-proper-orthogonal-decomposition
Hey! Loved your video. Huge fan of Steve Brunton as well. Was just curious if you could link some good, cheap float sensors. I'd really like to attempt the experimental setup you show in your video! Thanks!
Hey Noah, I'll take your comment as a compliment to my Blender abilities 😊
That scene is CGI, I generated the data for the heights of the wave using the dispersive wave equation and used it as a displacement map on a plane in the 3d animation software Blender. Pretty cool, huh?
@@fzigunov that is pretty awesome! Would you recommend a similar route for myself? It'd be cool to not have to pay for sensors and create a whole physical setup etc.
Thanks for the dense effort put into making this video. Is there a way to apply the method to unevenly distubuted sensors such as rainfall statiıns in a region
Absolutely. If you look at 10:52, once you flatten the sensors dimension, the technique is agnostic of how the sensors were distributed. So if you have a point cloud of sensors, your SPOD modes should capture spatially sparse samples of the spatial waves in the weather.
This is like opening a flood gate to me. The ideas behind these techniques are extraordinary, and I believe you did amazing when setting up examples for us to understand and visualize.
The quality here goes beyond the roof!
Did you use blender for the animations?
Did you transfer the color data for the cubes from your code to the graphics editor?
Or did you use simulations built into (what I assume is ) blender to extract data or a mix of both?
Hi, Mich, thanks for the feedback! I'm happy this video was helpful! I used Blender and Matlab for most animations. I simulated wave propagation with a wave equation solver I wrote in Madlab, then the data was read through a script in Blender Python scripting to color the cubes. I thought it would be cool to use Blender to run the collision simulations to add some fun special effects 😊
The waves on the water at the beginning are a displacement map applied in Blender, from a video exported from the Matlab simulation, so that was significantly more straightforward to do.
Is POD the same as SVD (singular value decomposition)?
Yeah, the mathematical machinery behind both is the same. I believe there's a minor difference in interpretation (i e., POD includes how to organize your data and what you do with the information you get from SVD; whereas SVD is just the barebones mathematical operation).
The video is very helpful. The reason I subscribed, though, is because the HP and max HP increased when brain leveled up. 🤣
I thought some people would subscribe because the brain is level 42 - the answer to the ultimate question of life, the universe, and everything 🤣
Thank u man!!
Awesome video. Did you use the 3blue1brown animation framework to achieve the visuals? Hope to see more in the fluid dynamics field. Its quite rare to see animated examples of journal topics which can get quite abstract or convoluted. Visualisation always helps drive everything home and form some kind of intuition.
Thanks!! I've used Manim (the python library 3b1b uses in his videos) for some of the visualizations, but also Matlab and Blender. Blender is really cool for nice 3d visuals and it's free software!
Sir I need your help in processing some of my images
Hi, Bikash, am I mistaken or do you work at FSU? If so, I can definitely provide help, just drop by my office (AME 229).
So all waves in the universe behave the same. Light and darkness, sound and stillness all the same just have different weight. That section where sound expand but I also contracts to something the steady motion it was in... Nothing does coming from something nothing is something.
i wanna be as cool as you someday
Great video! I have suscribed ro your channel. I will be pursuing a PhD in fluid dynamics, and it would be great if I could send you a private message. Let me know if that is possible! ;)
I'm happy you found it useful! If you want to DM me, I'm on Instagram and LinkedIn with the same name. Good luck with your PhD!