I want to be clear that I am very bad at building things, and am humbled by the process of setting up a real experiment. A whole lot goes wrong with doing the things in practice. But I am still happy enough with how things are going to share this attempted 3-divider setup:
Phases of Experimentation
There's a lot to think about for putting together the water-medium bucket-scale simulation of flow dividers. I usually don't really think about those things until it goes wrong when I try and actually spin it.
For the first overall geometry, I tried to do a floater-sinker combination on an open cylinder. So nothing holding the cylinder in place, just some things attached that keeps it vertical. This always went wrong because it's fundamentally not stable - it was interesting to measure significant decrease in drum speed, reflecting increased drag from the wobbling state. But even before I got far into this observation, I observed that my first choice for floater material dissolved in water. That went so badly it was almost funny.
Moving on, clearly it was time to experiment with some "spacer" approaches. I tried to Jerry-rig this with a few household items, such as toothpicks. These items were not up to the task. The one exception of note was ping-pong balls. I could believe those would work to hold a divider in place and not add too much extra friction. However, those things are maybe 1/100th the density of water, and I had more practical problem that the divider didn't extend out of the water, vertically, far enough. So with enough speed these just flew out. But a taller divider might still work. Partially sinking the balls was a cute idea, but if it works, it requires less-sticky materials than what I have.
Still not realizing exactly the extent that stickiness (general friction coefficient) is playing, I tried some cardboard spacers. In retrospect, why did I bother? It gave me some good numbers on what happens when something substantially increases the friction the motor is seeing.
This all led to the current approach, which is to cut a circle for the top and bottom of the cylinder that has a hole in the center just large enough to let the metal shaft through. This seemed to address the wobbling issues I was seeing. But the first time, I incorrectly made the thing too short.
In the meantime, the drum itself had a tremendous amount of wobble because I was just bad at drilling the holes. After getting some help with re-doing that, the reference speeds showed dramatic improvements and the fluid movements were a lot more gentle.
Finally, to what was hopefully my last critical oversight, after building the 3 color drums, I included a spacer on the bottom for the drums to "sit" on. I did not realize that they didn't need to sit at all. After a certain speed, there we get a significant lifting force. If you look at what I just wrote on the negative pressure designs, this is already foretold. Fluid forces will push the ends in if the dividers are open to the fluid. Well the connection I'm using is definitely not water-tight so we can be pretty sure this is happening. Because only 1 of the 2 cylinder's end are under water, this... pushes it up. There was nothing to stop its movement up, so the "hat" I created for them would all start flying off and everything went off the rails.
Pretty Sure Working just a Little Bit Now
To set expectations, I have ran a few references with water and no water in the bucket. This gives a sense of the fluid-dynamic specific contribution to the torque. The speed moving the drum (due to the bearings and stuff) was quite significantly less than the motor free spinning the shaft.
For adding a single divider, I have a best-case scenario of adding about 4 RPM speed. This is if the divider itself doesn't add additional friction (which it will) or generally cause chaos as all the other experiments have.
Running all 3 dividers nested in each other (before the hats flew off), it was hard to tell what was going on but speed measurements seemed to be at around parity with the reference. This was encouraging. I did another go with a single divider, and again, before the lifting problem it seemed to be touching parity or maybe sometimes just a little better. It was unclear when the lifting problem was becoming terminal so it's hard to say what range the effect was expected to be measurable.
So onto this configuration, where I might have corrected enough things for it to not be terrible.
Just running with the middle divider present (blue). Down at the bottom, there is a spacer both under and above the bottom circle of the drum. I obtained some numbers, and they're not much:
But we are now hitting around 1 of the 4 RPM. I'm using an automotive tachometer, and because it's doing a timing measurement, I actually believe it is accurate to the 0.1 RPM. The larger problem is the physical speed stability, and it tends to get worse before the dividers break apart. Otherwise it is >1 RPM roughly, in good conditions. So this still might be better to call it a statistical suggestion of a speedup.
I need to go higher voltage & speed, but still have a problem with the hat flying off... right around the 10 V mark, which is also the speed the expected 4 RPM speedup would apply to.
This is still extremely underwhelming, but these experiments have seen a gradual improvement from the dividers making the speed drop, to reaching parity with the reference, to now, maybe just dipping into the improvement territory. Still lots of build quality issues to address. Once I get a much more satisfying speedup, and generally ability to spin the thing faster, that will open up a number of theoretical questions to testing.
No comments:
Post a Comment