CFD simulations of air flow

In Notes on CandyFab and heater design (1.0), Windell lays out the goals of the CandyFab project and what he sees as a (maybe "the") design limitation standing in the way, namely, the design of the heater. He put out a call for ideas. Ultimately, I'd like to see it be able to melt a single grain at a time. Also it seems there is interest in melting more than sugar. So any heater design should be scalable and not be simply focused on sugar. I posted that I thought a second flow of air, not as hot, not as fast, around the first one might solve several problems.

It may preheat the area near the melt so the melt energy required of the central flow is not as great.
It may "focus" the flow and tighten the melt area and increase resolution.
It may act as a buffer to reduce the velocity of the spreading air and reduce "blow away."
Windell added that it may anneal the media and prevent thermal stress and cracking.
A later thought was, to speed things up, the outer flow temperature could be raised to increase the size of the melt and "paint" large areas inside the part.

And there may be more. Of these ideas, the biggest question is, would adding more air to the equation reduce "blow away?" Everything else is gravy. The short answer is that it looks like it can and there looks like there are other ways, too. I don't have a CandyFab machine and I don't have the type of hot air gun it uses. But I wanted to test this idea and others. So I ran some simple Computational Fluid Dynamics (CFD) simulations.

I used the aquarium pump information of 1 L/min. and the fact that Windell said it was throttled by about half. So 0.5 L/min. equals 0.509 in^3/sec. Windell stated that they had to use a 1/8" dia. nozzle to keep from blowing sugar everywhere. That's an area of 0.0123 in^2. This gives a velocity of 41.4 inches per sec. (2.35 MPH). This is an ideal number, not that it's the goal but that it doesn't account for the expansion of the hot air or the viscous drag induced by the tubing, etc. I assume that a 1/16" dia. nozzle is closer to ideal so I use that in nearly all the simulations, but since the estimated 1/8" nozzle velocity seemed to work okay for them so far, this is the nozzle velocity used throughout. These simulations were run to satisfy vague "what ifs," so there is no added heat and the media surface is assumed to be smooth and non-porous.
	We begin with the original setup (illustrated to the left): 1/8" nozzle, 41 inches per second, and 1/2" from the media. The images from the simulations shown below are available in much higher resolution; click through to get to medium resolution. (From there, full resolution is available if you click "all sizes," above the picture.)

So what do we see? These are obviously two-dimensional side views of the air exiting the nozzle and blowing into the media. These are not thermodynamic images. Heat is not involved here. These give us velocity and direction vectors at different mesh points. And they show us pretty much what we already know, that the air strikes the media and forms a stagnation area of low velocity (in blue), high pressure air directly beneath the nozzle. This is what deforms the puddle and wedges the surrounding air away. The nozzle flow (in orange) draws air from around the nozzle and ultimately creates a region of flow (in red) that is actually faster than the nozzle exit velocity.

The same setup, but only 1/4" (6.4 mm) from the media.

Again, the same setup, but 1" (25.4 mm) from the media.
	Next, consider the following setup: We have a 1/16" diameter inner nozzle flowing at 41 inches per second and a 1" outer nozzle flowing at 4.1 inches per second (1/10th the velocity), both at 1" from the media. The outer velocity and diameter were chosen arbitrarily. There is a dramatic change in shape of the airflow:

	So if the outer nozzle flow buffers the inner flow from drawing in so much air, what if the inner nozzle flow is just shielded physically by a plate? This is 1/16" dia. nozzle, 41 inches per second flow, 1 inch from the media with a 1 inch plate around the nozzle tip. Notice the air flow around the edges of the plate toward the center. (The close up looks good but the big picture is messy.)

	So, let's go crazy and make the plate as big as the envelope. 1/16" dia. nozzle, 41 inches per second, 1/2" from the media with a 4 inch plate around the nozzle.

This creates a ceiling to the media's floor. The big picture looks strange but the reality is that both the size and color of the arrows indicate velocity. No outside air is drawn in. The close up shows that the vortex acts like the outer nozzle flow from before. This could be a more clean (cleaner) solution to the blow away problem. A plate around the nozzle tip will shield the hot nozzle from gathering and then re-depositing the black sugar drops. It could be removable for cleaning. The downside is that one loses separate control of the outer nozzle air for speed building or tuning. And unless the plate is made of pyrex, the melt can't be seen, and face it, that's the real fun in this whole endeavor. Enough with simulation-- what about when air really meets sugar? I wanted to see what velocity sugar was stable at. So I used a three speed table fan and a tray of sugar. I turned the fan on and brought the tray closer until it was right up against the cage of the fan, and the sugar didn't move. This surprised me, because I half expected to be cleaning sugar from all over the place. I bump the fan up through all three speeds and the sugar stayed. I moved the tray around and started to get some movement near the lip due to vortices. I turned off the fan and blew the sugar around myself and it behaved normally. So the question was how fast was the air from the fan going? I shot some video of the fan as I released tiny cotton balls (the fluffy tips from some cotton swabs) in front of the fan. By putting a ruler in the video for scale I could now figure out roughly the velocity. Although it feels faster, the fan at its highest setting puts out about a bit over 55 inches per second (3.125 MPH). I don't have an actual speed limit for when the sugar moves but I have something in the ballpark of when it doesn't. That's useful when looking at the simulations. Comments and other ideas are welcome.

CFD simulations of air flow | 4 comments | Create New Account

The following comments are owned by whomever posted them. This site is not responsible for what they say.

CFD simulations of air flow

Authored by: Alan S. Blue on Friday, August 24 2007 @ 04:46 PM PDT

Several comments.

1) Excellent :D
2) Did you also do heat transfer? The 'double nozzle' looks excellent from a 'wind' perspective, but does the hot part of the air reach the media?
3) Can a cooler piece be closer to the media without getting splattered and gummed up? The 'current heater' picture has _red_ sections. A cool washer-like disc (with a big annulus, as big as the blue section if you like) avoids the _worst_ aspects quite easily. A trickier job with the washer could make it such that none of the red, orange or yellow could reach the media. If, and only if, a cool disk in that spot --very close to the media -- won't get gummed up.
4) I can imagine combining the disk _and_ the return nozzle too. If the gumming can be avoided. But restricting the air-flow pattern itself should dramatically reduce the amount of flying hot sugar. (The mental picture I have here is with the outer-tube-with-inset-restricting-ring displaced down to nearly-touching. I don't see this working if it can't get quite close without gumming.)

Again, this is quite excellent, any pointers on the tools you've used here?

[ Reply to This | # ]

Authored by: Windell on Friday, August 24 2007 @ 04:56 PM PDT

The idea of using a secondary outer nozzle is very interesting from a thermal perspective alone-- if we use warm (if not hot) air in the outer region.

It's come up in conversation here a few times that preheating the sugar would make everything easier-- if the sugar is already warm, than the center hot spot has less work to do to melt the sugar. Also, keeping a larger area warm may help to prevent cracking, since thermal stresses will be spread out over a larger area, not concentrated at one hot spot. For working with *glass* beads, it may be an absolute necessity on that basis alone. =)

So, we could imagine a heater geometry that consisted of a hot center tube (1/16" ID), with its heating element and insulation, and we fit that into a piece of thin-wall steel tubing-- maybe 5/8" OD. Outside of that, we flow air downward to (1) cool the outside of the steel tube, and (2) to produce the warm shielding air flow. It might work well.

CFD simulations of air flow - Authored by: Anonymous on Saturday, August 25 2007 @ 07:42 PM PDT

Authored by: Anonymous on Thursday, June 26 2008 @ 06:55 PM PDT

Pictures are not available any more :-/

What's New

Stories

Comments last 7 days

Links last 2 weeks

Topics

User Functions

Older Stories

Tuesday 26-May

Monday 05-May

What's Related

Story Options

Poll

Who's Online