Aerodynamic Shape Optimization - The Adjoint CFD Method

Поделиться
HTML-код
  • Опубликовано: 28 дек 2024

Комментарии • 39

  • @eduardodavalos791
    @eduardodavalos791 3 года назад +7

    Excellent video Wouter! Thanks for your hard work explaining in an easy way the fascinating world of aerodynamics!

    • @AirShaper
      @AirShaper  3 года назад

      Thank you very much Eduard! I you liked this one, perhaps our video on Machine Learning for Aerodynamics might also interest you?

    • @eduardodavalos791
      @eduardodavalos791 3 года назад +1

      @@AirShaper I will take a look at it!

    • @AirShaper
      @AirShaper  3 года назад

      @@eduardodavalos791 Great, looking forward to your thoughts on it!

  • @benykla
    @benykla 3 года назад +6

    Always clear & inspiring. Thanks Wouter!

    • @AirShaper
      @AirShaper  3 года назад +2

      Thanks Halfina! I'm glad it was clear :) I have to admit this one really wasn't easy (still today it's a very complex topic for me as well :) )

  • @grekiki
    @grekiki 2 года назад +2

    That's awesome, didn't expect to see "back propagation" in CFD :)

    • @AirShaper
      @AirShaper  2 года назад

      Thanks Gregor! If you want to learn more, we did a paper together with the developers of this code:
      airshaper.com/research/from-scanned-cad-to-an-optimized-car

  • @tchezao0248
    @tchezao0248 3 года назад +2

    Awesome Video, my friend. Didn't know about this method and now I am even more fascinated.

    • @AirShaper
      @AirShaper  3 года назад

      Thanks a lot!! It would work on classic Porsches as well :)

  • @marcnobel3938
    @marcnobel3938 3 года назад +3

    Great! I will try that out when my next project is finished.

    • @AirShaper
      @AirShaper  3 года назад +1

      What's the next project about?

    • @marcnobel3938
      @marcnobel3938 3 года назад +1

      @@AirShaper A whole body kit which should be a TÜV/Streetlegal version of a real Rallye Version. Just turned down enough to make it road useable. This includes a massive (in relation to the car) rear spoiler.

    • @AirShaper
      @AirShaper  3 года назад +1

      @@marcnobel3938 Ok that's right at the heart of what we like & do :) Looking forward to get in touch as soon as you get to the aero part!

    • @marcnobel3938
      @marcnobel3938 3 года назад

      @@AirShaper I guess in roughly 3-4 weeks I have modified all scans to the needed standards. Some parts are already in Geomagic to get them ready for SW. We are very excited to try out your analysis.

    • @AirShaper
      @AirShaper  3 года назад +1

      @@marcnobel3938 Hi Marc, ah that's pretty soon then, nice! Keep in mind that in theory, there is no need to make the models watertight for AirShaper. But of course, gaps & holes that are large enough to impact aerodynamics should be closed. Feel free to reach out at wouter@airshaper.com to continue the discussion!

  • @tomaspedro1999
    @tomaspedro1999 3 года назад +2

    Very good insight!

  • @daryoushmehrtash7601
    @daryoushmehrtash7601 3 года назад +1

    thanks, good explanation.

    • @AirShaper
      @AirShaper  3 года назад

      Thank you, you're welcome

  • @danielniels22
    @danielniels22 3 года назад

    HAHAHA HOW LUCKY I AM 🤣 what a coincidence that today i have just had an assignment about optimization and you upload this haha thank youuuu 👍👍👍👍👍

    • @AirShaper
      @AirShaper  3 года назад

      Good timing then 😎

  • @351kostakis
    @351kostakis 2 года назад +1

    Fantastic

  • @lontongstroong
    @lontongstroong 6 месяцев назад

    Can the algorithm do adjoint optimization on *transient* simulation?

  • @bitsurfer0101
    @bitsurfer0101 Год назад

    Can manufacturers apply design boxes in order to make a bicycle frame stiffer in one area versus another?

    • @AirShaper
      @AirShaper  Год назад +1

      This method is to optimize the aerodynamics, not the mechanical properties. For that, similar methods exist, like OptiStruct or Inspire.

  • @ejegbavworuemu2129
    @ejegbavworuemu2129 3 года назад

    Wow I actually considered optimising shapes following a similar technique. I didn't know it was already implemented. Any ideas on how I can apply and say read more?

    • @AirShaper
      @AirShaper  3 года назад +1

      I would recommend exploring the adjoint optimization libraries of SU2 and OpenFOAM. Not easy, but very powerful.

    • @ejegbavworuemu2129
      @ejegbavworuemu2129 3 года назад +1

      OK then thank you

  • @ZachBZera
    @ZachBZera 3 года назад

    How do I learn how to do this? Because it sure as shit wasn't taught to me at one of the top mechanical engineering schools in the world lol

    • @AirShaper
      @AirShaper  3 года назад +2

      One way is to dig into OpenFOAM (www.openfoam.com) and first learn how to run CFD simulations and then how to run adjoint shape optimizations. Definitely doable (even I can do it, which says a lot).
      The other option is to run such simulations via our platform, where we have automated literally everything, from 3D model to finished 3D morphed model. Just give me a shout at wouter@airshaper.com!

  • @watermenlon3617
    @watermenlon3617 2 года назад +1

    can this also be used for reducing noise when an object is subjected to airflow? trying to do shape optimization where the goal is noise reduction. Is this possible to work with?

    • @AirShaper
      @AirShaper  2 года назад

      Yes, shape optimization can also be used to reduce noise! Any specific topic you're working on?

    • @watermenlon3617
      @watermenlon3617 Год назад +1

      @@AirShaper currently working on an undergrad thesis for ducted turbofan blade noise reduction by adding trailing edge serrations. we'll be trying to optimize not the blades but the serrations attached on the turbofan blades for noise reduction

    • @AirShaper
      @AirShaper  Год назад

      @@watermenlon3617 sounds really cool!! The NTUA does have adjoint code including MRF for rotation, but they haven't made it public yet :)