Page 1 of 2 12 LastLast
Results 1 to 10 of 14

Thread: CFD (Computational Fluid Dynamics) / Flow Simulation Questions

  1. #1

    Join Date
    Jan 2014
    Posts
    159

    CFD (Computational Fluid Dynamics) / Flow Simulation Questions

    Please post your CFD/Flow Sim Questions on this thread. Be sure to include screen shots if you have them. Also, we encourage you to try doing a web search for your issue to see what is available before posting here. If you find a solution and think it is a common problem, please post your findings out here to help others.

  2. #2

    Join Date
    Apr 2020
    Posts
    11
    Thank you!!

    Regards,
    Joe T.

  3. #3

    Join Date
    Mar 2018
    Posts
    6
    I would like to use the SW CFD capability to experiment with some design changes. A key to that is having an adequate video card. What cards have proven adequate for this type of work?

    Alan
    Winder, GA

  4. #4
    SOLIDWORKS Support Volunteer Jeffrey Meyer's Avatar
    Join Date
    Nov 2011
    Location
    Israel
    Posts
    209
    Hi Alan,

    Like most CAD software SW demands a combination of ALL computer resources - RAM, Disc, Clock speed, etc. including hardware graphics, and you can't have too much of any of them.
    In that respect CFD is no different.
    I do most of my work (including some CFD stuff) on a good laptop equipped with an NVIDIA Quadro RTX 3000 graphics card. So far I haven't noticed that the graphics card is the bottleneck.
    I have no experience with other graphics cards, so I can't compare. I would suggest you look at the SW web site to see what graphics cards have been tested and choose according to your budget, not your requirement to work with CFD.
    BTW - The Quaro RTX was a very significant part of the cost of my laptop.

    Jeffrey

  5. #5
    Just going throw out there that a Quadro isn't necessary for Solidworks. Not for CAD, for rendering, for running CFD or any other simulation. In school/FSAE we did everything on OLD gaming hardware such as i7-2700 and GTX 660. Just last month I ran some CFD on my personal machine which is an AMD Ryzen machine and a GTX1060. Previous to that I ran Solidworks on an AMD FX-6300. All the listed hardware has worked for me even when opening assemblies with several hundred parts. Admittedly all these setups had at least 16Gb of ram and the old i7 chugged but everything worked just fine.

    All you are missing out on is 'certified' hardware and some performance. If your system is older or not as powerful just be aware that it may be slower.
    Last edited by Michael S; 02-12-2021 at 12:16 AM.

  6. #6
    It looks like the EAA version does not allow CFD on an airframe or for example new cowling. What can you do / not do with the version available through eaa with CFD?

    Thanks.

  7. #7
    SOLIDWORKS Support Volunteer Jeffrey Meyer's Avatar
    Join Date
    Nov 2011
    Location
    Israel
    Posts
    209
    Are you trying to run a CFD on a whole airframe? If so, stop trying, because without knowing what compute resources you have, I can safely assume you haven't got enough.

    Also, CFD on a cowling is a highly non-trivial analysis. A cowling is probably right behind a propeller with a spinner resulting in highly non-trivial turbulent inflow to the cowling. What geometry do you want to put behind the the cowling to make the analysis significantly meaningful or accurate?

    Turbulent flow is by definition time dependant thereby requiring orders of magnitude more compute resources and/or inaccurate guesswork shortcuts.

    Last but not least, maybe I'm mistaken, but as far as I know with regard to CFD, the "EAA version" of SW is no different from the regular educational edition that in turn is no different from the commercial edition.

    Bottom line: Could you give us some indication of what you mean by "... EAA version does not allow CFD ..."?

  8. #8
    I do not have the skills to model a propeller by myself (I am totally new to SolidWorks), but there does seem to be a model of a CH 750 available. So this would be dependent on a propeller manufacturer sharing their model as well, which now that I think about it seems very unlikely. I assume I could find a representative propeller, but I am not sure if the analysis would then be useful in that case. It doesn't seem like the exact propeller shape would be critical to get at least some useful data. I am looking more for "better or worse" rather than explicit accurate values, but maybe there's no way to halfway do this and get any useful data.

    I do see the Premium version has the features I'm looking at using, which is great. When you say haven't got enough, what hardware would be required to do these analysis in a reasonable period of time (say overnight) assuming the other hurdles could be achieved? A 5990X/3090 is relatively affordable in the context of an airplane. My flight sim is a 5900X / 2080 Ti but sounds like that is not going to cut it? Would that be sufficient to model things like designing fiberglass fairing on a wingtip for example (clean air ahead of it).

    Thank you for your help!
    Last edited by JohnMyers; 08-08-2021 at 05:55 PM.

  9. #9
    SOLIDWORKS Support Volunteer Jeffrey Meyer's Avatar
    Join Date
    Nov 2011
    Location
    Israel
    Posts
    209
    Hi John,
    I like using SW so please don't get me wrong if this post takes the wind out of your sails and enthusiasm, but I feel the necessity to explain the order of magnitude of the problem you're trying to solve.

    1. A propeller is a rotating wing with varying airfoil profiles, angles of attack, and airspeed. It is not difficult to model the geometry in SW - I can gladly help you with that. Your much bigger problem is modeling the moving turbulent air around the propeller. For accurate CFD you need many millions of small air particles (mesh elements in CFD) particularly in the close vicinity of the wing leading edge.

    2. If I could describe CFD in a single sentence, it is would be a digital algorithm for the inversion of a massive double precision floating point matrix.

    3. Turbulent flow is by definition time dependent, meaning that at any given instant those tiny particles of air could be moving in a completely different direction from the previous instant. Accurate CFD would require solving the above matrix inversion repeatedly at very small time increments.

    4. You can simplify and speed up the process by making some drastic assumptions, like assuming the flow is fully laminar (no turbulence), or assuming relatively large CFD elements to lower the size of the matrix, etc. These assumptions significantly reduce the accuracy and validity of the CFD analysis.

    5. The above applies to a static wing. Now add the fact that you have a rotating wing with another rotating wing following closely in its wake. And then in the turbulent wake of this combination we add a static cowling containing cooling ports and geometrically complicated engine parts.

    6. Recently I was faced with a much simpler problem that I was unable to solve: I have a cylindrical chimney throwing out polluting smoke next to an apartment building with an open window. Problem: With the wind blowing at a known speed and direction, how much pollution goes into the window?

    7. In the light of the above I think we can understand that even the most powerful pc available is like a drop in a bucket and all the pc performance enhancements will make it like two drops in a bucket.

    8. But I have a positive suggestion for getting the feel of what CFD can do for you (I'd appreciate feedback on this): Get some wind-tunnel test data/polars on a known airfoil and run a CFD analysis on a thin 2D slice of this airfoil. See if you get some (or any!) similarity between the analysis and the test data - like best L/D and at what AOT. Try different mesh sizes and localized mesh refinements and measure the solve time. How much memory did it use?

    Hope this helps.

  10. #10
    Quote Originally Posted by Jeffrey Meyer View Post
    Hi John,
    I like using SW so please don't get me wrong if this post takes the wind out of your sails and enthusiasm, but I feel the necessity to explain the order of magnitude of the problem you're trying to solve.

    1. A propeller is a rotating wing with varying airfoil profiles, angles of attack, and airspeed. It is not difficult to model the geometry in SW - I can gladly help you with that. Your much bigger problem is modeling the moving turbulent air around the propeller. For accurate CFD you need many millions of small air particles (mesh elements in CFD) particularly in the close vicinity of the wing leading edge.

    2. If I could describe CFD in a single sentence, it is would be a digital algorithm for the inversion of a massive double precision floating point matrix.

    3. Turbulent flow is by definition time dependent, meaning that at any given instant those tiny particles of air could be moving in a completely different direction from the previous instant. Accurate CFD would require solving the above matrix inversion repeatedly at very small time increments.

    4. You can simplify and speed up the process by making some drastic assumptions, like assuming the flow is fully laminar (no turbulence), or assuming relatively large CFD elements to lower the size of the matrix, etc. These assumptions significantly reduce the accuracy and validity of the CFD analysis.

    5. The above applies to a static wing. Now add the fact that you have a rotating wing with another rotating wing following closely in its wake. And then in the turbulent wake of this combination we add a static cowling containing cooling ports and geometrically complicated engine parts.

    6. Recently I was faced with a much simpler problem that I was unable to solve: I have a cylindrical chimney throwing out polluting smoke next to an apartment building with an open window. Problem: With the wind blowing at a known speed and direction, how much pollution goes into the window?

    7. In the light of the above I think we can understand that even the most powerful pc available is like a drop in a bucket and all the pc performance enhancements will make it like two drops in a bucket.

    8. But I have a positive suggestion for getting the feel of what CFD can do for you (I'd appreciate feedback on this): Get some wind-tunnel test data/polars on a known airfoil and run a CFD analysis on a thin 2D slice of this airfoil. See if you get some (or any!) similarity between the analysis and the test data - like best L/D and at what AOT. Try different mesh sizes and localized mesh refinements and measure the solve time. How much memory did it use?

    Hope this helps.
    Thanks for this info, I am very interested in learning SolidWorks but it is intimidating especially the CFD stuff. FEA and CFD has always been super interesting to me as a layman.

    To ask the question backwards, what is possible with CFD in SolidWorks as it relates to small experimental aircraft? My assumption was, given the relative performance of a $10,000 workstation compared to one 15 years ago, that you could use it to assist with a cowl. Most cowls seem to be designed with the "looks about right" method, so the bar seems low. If a propeller is too complex to simulate, how about just clean airflow? I would think that for a basic shape, something that works well with clean airflow would have a better shot at working behind a propeller than one that does not.

    Some other problems I would love to use CFD for:

    - determine if fiberglass fairings are worth the effort to construct in several areas. For example a wingtip.
    - cowl design, or some simplified version of this
    - heat flow in engine compartment after shutdown
    - ducts and airflow for cabin heater
    - ducts and airflow for air conditioner (I'm trying to put one in a Zenith, but that is another topic!)
    - determine drag with/without wheel pants a different speeds

    From your last point, it sounds like even assuming no prop, flow over the entire airframe would be impractical. What would the limit be, could a wheel with/without fairing be possible? I've seen Boeing landing gear cfd videos and things from years ago that show some complex interactions with the gear doors, etc. I would think a modern $10,000 PC would be similar in performance to a private supercomputer that Boeing would have access to from 20 years ago. Thanks for entertaining my questions!

    Mark Patey said he used SolidWorks CFD extensively in designing his wing, with similar hardware. Maybe he was just using sections?

    Separately, would it be possible to use FEA to figure out of an engine mount and firewall attach points are strong enough for certain engines? I don't have a model of the engine so maybe just an object of similar mass and density would suffice.

    First task for me is just to figure out how to use SolidWorks to start with, but it's very interesting to know what is/isn't possible after I learn it!
    Last edited by JohnMyers; 08-09-2021 at 12:44 PM.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •