Fluid Simulations in the Kitchen Sink

In an age of ultra-powerful GPUs and cheap processors, computational techniques which were once only available to those with a government-sized R&D budgets are now available to the everyday hacker. An example of industry buzzword turned desktop software is the field of “computational fluid dynamics”, which put simply allow modeling how gasses or liquids will behave when moving through a cavity under specific conditions. Extensive utilization of these fluid simulations are often cited as one of breakthrough techniques which allowed SpaceX to develop their engine technology so rapidly when compared to Apollo and Shuttle era methods.

But just because anyone with a decent computer has access to the technology used for developing rocket engines doesn’t mean they have to use it. What if you prefer to do things the old-fashioned way? Or what if, let’s me honest, you just can’t figure out how to use software like Autodesk CFD and OpenFOAM? That’s exactly where [Desi Quintans] found himself when developing GUST, his cooling duct for i3-type 3D printers.

[Desi] tried to get the big name fluid simulation projects working with his prototype designs for an improved cooling duct, but had no end of trouble. Either the learning curve was too steep, or the simulation wasn’t accurate enough to give him any useful data. But remembering that air is itself a fluid, [Desi] took his simulation from the computer to the sink in order to better visualize what his cooling duct was doing to the airflow.

[Desi] printed up a box with a hole in the bottom that would connect up to his nozzles under test. As the volume of water in the box would be a constant between tests, he reasoned that this would allow him to evaluate the different nozzles at the same pressure. Sure enough, he found that the original nozzle design he was using caused chaotic water flow, which backed up what he was seeing in his experiments when mounted onto the printer.

After several iterations he was able to tame the flow of water by using internal baffles and fins, which when tested in water created something of a laminar flow effect. When he tried this version on the printer, he saw a clear improvement in part cooling, verifying that the behavior of the air and water was close enough for his purposes.

We’ve seen other projects that successfully used fluid simulations in their design before, but the quick and dirty test procedure [Desi] came up with certainly has its charms.

18 thoughts on “Fluid Simulations in the Kitchen Sink

  1. >”he found that the original nozzle design he was using caused chaotic water flow, which backed up what he was seeing in his experiments when mounted onto the printer.”

    Unfortunately, the results are still inconclusive because nozzles behave differently when the fluid changes. The change from laminar to chaotic turbulent flow depends on the Reynolds number of the fluid, which is dependent on the density, and dynamic/kinematic viscosity of the fluid, and it has an annoying hysteresis where turbulence starts at a higher flow velocity than it stops, so not seeing turbulence doesn’t mean it can’t occur due to some other disturbance that sets it off.

    So you can’t just put the same nozzle under water and expect to see any meaningful results. You need to do a bunch of math to translate the design from air to water and keep it analogous, if possible at all because of the differences in viscosity and density parameters.

    Hence why fluid simulation software.

    1. The thing is, when you change the fluid density by a factor of 1000, that alone can induce the chaotic flow, and the improvements that you may observe by changing the design might have a completely different reason.

    2. I’m 100% with you that in this day and age this is awfully primitive (your iPhone has more than enough computational power to generate useful simulation data from any given model). On the other hand, I distinctly remember seeing a variation of this method (with more controlled flow obviously) and some sort of coloring agent being used in a documentary (I want to say by the Soviets during the space-race but I could be wrong, so, ANSYS wasn’t really an option then). Fluids and compressable flow were my two weakest points (and since then I’ve only used diffeqs in control systems so I’m hardly qualified to assess this project)… but just devil’s advocating for this dude — there *might* be some scientific credence to this at least at the 0th order and/or being able to reduce the algorithimic complexity by giving a 2 parameter heuristic model (some rhos here, some etas there?) to design against.

  2. This article should be deleted. It is misleading for several reasons, not the least of which is the incorrect claim that laminar flow is more desirable for heat removal (it isn’t).

  3. I think it’s a silly custom to call air a fluid, or should I say to not have a separate term for liquids.
    There is just a density where you get surface attraction and that changes behavior significantly.
    Ask dax about it, he seems to know the issue.

  4. Air is compressible, water is not, even that is enough to mess up the results, espesially since it’s using an air pump through a nozzle that is restrictive, there will be compression occurring.

  5. It probably works well because it compresses the air a little bit. Also it allows the air to escape very fast. But….it might have negative effects. It might blow the molten plastic in one direction. Also it might be loud. All the negativity is silly. Just print one out and test it, thats the whole point of a 3d printer! Prototype people!

  6. Sigh.

    This may have produced some sort of perception of a positive result, but there are more than a few problems with the method, and the conclusion that laminar flow in heat transfer is preferable is demonstrably wrong-headed.

    If you’re going to use an incompressible fluid to model a compressible one, you have to be very careful about interpreting the results. This is done, by the way – for example,snow accumulation around buildings has been modeled using sand settling in water in a scale-model city.

    You have to account for the differences and generate a scaling method (usually involving dimensional analysis/the dreaded Buckingham’s Pi Theorem) though Re is a good start…which was ignored in the test.


    You also have to quantify the results pretty carefully – I don’t see much of that here.

  7. Let’s thank Desi Quintans for sharing his work. I ran an Autodesk CFD simulation and resulting image is below. (This is my first time running a fluids simulation, so it might not be perfect…) It looks to me like the air on the far side is continuing to flow outward rather than being directed inward. If true, that portion of the airflow is “wasted”.

  8. Crazy negativity on this one, I don’t get it…

    Nobody ever claimed the method Desi used was necessarily accurate, or even a good idea. The post even starts out saying that he had to do this because he couldn’t figure out the “right” way to do it. Even if the principle is flawed, he still got the result he was hoping for.

    That sounds like the definition of a hack to me. Finding an alternate solution to a problem that works, even if it isn’t ideal Probably 50% of the projects on this site would fit that description.

    I’ll agree with one of the previous comments that smoke would have been a better idea, but I can see how there might have been logistical issues with setting that up quickly/easily.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s