Thursday, July 29, 2010


Continuous belt production

As opposed to continuous-belt production...

I think that (as with so many things) the original idea for this came from Ed Sells. And lots of people have subsequently had ideas of reprapping on a continuous conveyor belt running over a flat surface. See here and here and here.

The big advantage would be that you could print for as long as the plastic filament lasted, continually throwing reprapped parts off the end into a bucket. They would be split from the belt as it ran over the winding roller.

There are two problems that have to be overcome:
  1. Keeping the belt flat against the tendency of parts to curl away from it, and
  2. Driving it accurately without slip.
The first problem is significant; as Nophead pointed out in the forum discussion, you need a tension of 100 kN to keep the belt flat to 0.1mm. Vacuum doesn't help, as it would create too much friction preventing the belt from moving.

But the second problem is quite easy to solve (usual apologies if someone has though of this before) - you do continuous production without a continuous belt:

By winding and unwinding between two rollers you can cover the entire length of the bed, and pull the part off by running it over the end roller. You can then rewind to print the next part. A simple flag on the belt passing through an opto-switch would allow the system to be zeroed.

Driving the idler roller might require some thought. As the belt winds and unwinds, the roller diameters change, and so does the required velocity ratio between them. The stepper can compensate for this effect by changing the number of steps-per-millimeter it uses depending where it is on the belt to get the build right. But the idler has to do the inverse.

The first thing to say is that - with a thin Kapton film belt - this probably wouldn't matter and you could just run the two rollers with a timing belt between them relying on the elasticity of the system to take up the slack.

But if it did become a problem there are at least three solutions:

  1. Drive both rollers with steppers. I don't like this as they're bound to end up fighting each other.
  2. Put a wind-up spring on the idler roller to take up the slack. I'm not too fond of this either. It would probably load the stepper too much, and the forces required from the spring to achieve sudden movements would change with the mass of the object being reprapped.
  3. Keep a tension on the film by driving the idler clockwise (in the diagram) with a DC motor on a controlled constant current source. Constant current should give constant torque. By changing the current you could compensate for increased mass needing to be moved.
But there's still the problem of keeping it flat...


can't we just use a continuous track [ ] and mount the heated bed in it split among the track modules (maybe even only powering the ones in use for eficiency's sake)?
that should at least allow for an easy to move with constant force horizontal surface, not to mention it would provide "infinite" feed with minimal mucking around
This comment has been removed by the author.
I couldn't think of a solution to the first problem, which is why I abandoned work on it. Vacuum doesn't work because atmospheric pressure is too low, not because it stops the belt from sliding, you could simply turn it off during moves.

I had a simple solution to keeping it under tension though. The belt only has to move the length of the table. If you put pulleys on the ends of the rollers say 1/3 the diameter and link them with a cord wound the opposite way to the belt, then it would only move 1/3 of the distance. You could then have a spring in the middle tensioning the cord.

Or you could do what they did on old fashioned wireless sets. You have the spring pulling on the cord from inside the pulley.
Picking up where aeth left off, why not use a continuous rigid track, made up of plastic pieces that can themselves be printed with RepRap?

* No flatness problem, because the track is rigid.
* The roller diameters are constant because the track is continuous (no winding and unwinding).
* No slippage, because you can incorporate gear teeth into the track pieces.
* No need to rewind after every n pieces.

Only problem is how to deal with the cracks between the track pieces. Maybe glue on a thin film on top of the rigid track?
sorry if i didn't explain myself well enough, i did mean rigid continuous tracks

i'll take the opportunity to make clear what i was thinking as best as i can, i was thinking of lozenge-like modules, each equiped with a heated bed top that rested one atop the previous module, maybe a rough ascii art sketch will help

(sorry, no pre tag allowed)
top view:


side view:
____ ____ ____
/| .-. .-. /|
| | | o | | o | | |
|/ __'-'____ _'-' |/

the modules could be further encouraged to provide a horizontal surface by being tensioned or resting upon a fixed track as they slide like a train (and this same track could also be used to provide them with power to heat the top, though some thought would need to be focused on avoiding thermal gradients in the track/print bed before the new print had started, at first a pause to let the new modules heat up to par should be enough to avoid most of the diference)

i hope this helps someone

I think, just like a continuous belt isn't needed for continuous production, neither is any kind of belt at all. What we really need is a way to automatically peel parts from the print surface. I suppose that a belt offers one potential way to do that, but could it also be done by having some kind of automated scraper for a rigid print bed?
Also - sorry for the double post - if scraping the bed turns out to be impractical, what about using a kapton belt with thin steel strips going widthwise under it, which can be clamped by permanent magnets?

It might be a bit pricey, but probably not much more expensive than a vacuum...
Has anyone considered keeping the belt system seperate from the x-axis?

That is to mount the heated bed and belt system onto the exisiting x-axis.

Advantages are that it does not compromise the accuracy of existing x-axis drive. The belt is only driven to drop the part(s) off at the end of a build. The belt does not move during a build, so accurate positioning is not required.

Disadvantages are that an extra drive is required to move the belt, but this would be very simple. Also, the height under the extruder may have to be increased to accomodate the belt drive system.
I think jbayless is on the right track here, the actual problem is how to remove the built pieces from the bed efficiently - a continuous belt would do that nicely, but is very hard to implement in practice, so it might make sense to think outside that box.

Maybe simply mount a barrier just beyond the printing area in such a way that the bed can slide under it, pushing the pieces loose from the bed and allowing them to drop down into the space under the bed (which would be left open to allow a bin for completed parts to be placed under the printer).

This would require a longer movement path for the bed to one side, but doesn't require any other moving parts.

It also assumes that pieces can be loosened from the bed by pushing at them from the side. Some suitable stiff foam material could be used on the barrier to protect fragile parts and provide a larger contact area when pushing irregular parts.

This might work best for individually printed parts - a bed full of parts would provide more resistance and the parts could be crushed against each other if fragile, but on the other hand with a continuous printing system it would anyway be better to print one part at a time to minimize losses if a print fails half-way.
I agree that a belt isn't the problem, automatic separation is.

How about a flat bed consisting of a thin steel plate tempered to make it a spring? You then have a highly geared-down DC motor that bends the bed into an arc (within its elastic limit) to release the parts.
Yes that is what I did manually here:

Two problems are:
I can't find a source of the metal and I don't know exactly what it is.

Small parts don't release when you bend the ends because the part braces the middle causing a flat spot in the arc.

My idea to solve that was to run a ripple down the plate using offset ball bearing rollers at the edges.

In the end I run my Mendel 24/7 by taking advantage of the fact my wife works part time and arranging my build trays to match.
Why can't it be like a treadmill? Have your kapton belt slide over a metal plate. You don't need as much force to keep it flat.
Yes you do, look at the discussion linked to above, where I estimated 100KN.
To me it seems like a fairly simple solution would deliver most of the value: just have the table able to move one build-platform's width back and forth.

Build on the left half of the table, then move to the other end for the next print and alternate.

Add a sensor or switch to avoid moving back if the parts have not been cleared. Then you have the entire time of a print to clear the parts on the other side.

Though this is not 100% automated, practically close enough. If you are printing for a long time, your build platforms are probably going to be chock-full: so the builds will take a long time. So, you have plenty of time to remove the parts from the other side while the next print is running.
That would have to be two tables, rather than one long table. I wouldn't want to be prying parts off of a print bed that's still being used to print stuff on the other side of it. It would certainly mess up the print in progress.
really you only need one motor as your tension is only strongly applied in one direction, either in removing the part, or in tensioning the belt.

imagine if you take your two rollers, one is attached to a dc motor capable of applying the 100kN force (at least.) this one will pull the belt to remove the part.

the other roller will simply be spring loaded to return the belt to the start position, doesn't need to be anywhere near 100kN force though.

Use a solinoid to lock that roller when you are to begin printing, apply your 100kN with your DC motor to acheive flatness.

complete your print, release both your 100kN tension and your solinoid lock. run the dc motor to pull the part off of the build area.

allow the spring to return the belt, there shouldn't be much force against this besides the dc motor, and if that is too much just run it lightly in reverse.

once the belt is back into position, relock the first roller and apply your build tension.

technically this will degrade one area of your belt every time, but if you employ an encoder you can scale this to longer belts if necessary.
>the other roller will simply be spring loaded to return the belt to the start position, doesn't need to be anywhere near 100kN force though.

Yes it does. The massive tension is needed because to hold a membrane flat by pulling its edges requires an infinite force. To hold it nearly flat requires a near infinite force. Please see the forum discussion Adrian linked to.
How about instead of bending just an arc in a thin flexible plate to cause parts to pop off, instead push a roller bearing back and forth underneath it causing a bending wave to go through the platform? I'm thinking of something like a massage chair. See:

Then once the parts have been sufficiently massaged and they are nice and relaxed and their muscles are soothed, tilt one end of the platform about 45 degrees so that the parts slide off the side and down into a box under or beside the printer.
@nophead Is that what you meant when you said offset ball bearing rollers at the edges?

the return spring doesn't need to apply that force because it is only active when you are returning the belt into position.

you only need the belt to be flat when you are building on it. in that case you would have your DC motor side pulling against your "idler" which is in fact stopped by a solenoid or latch. the spring does nothing during building/tensioning.
I think any worthwhile solution has to be fully automatic. That would allow machines to run unattended for long periods making many complete sets of parts. A solution that requires even minimal attention at the end of each build would not be much of an advance on what we have now, which only requires four clips to be undone at the end of a build.

Tempering thin steel plate to make it springy is not difficult - all it needs is a heated sand bed and a quench bath. I don't know how flat it would stay, though...
The belt could be made from printed links reminiscent of a wristwatch chain, held down by rails at the edges and driven via cogs directly below the build area. The actual printing surface might need to be painted onto these links.
If you are doing a segmented belt, then you have to deal with the "seams" between the segments. This is because the whole point of a belt is to bend it to get the part off. If you have to deal with seams, why bother with a belt?

Use narrow sections, say 2mm wide. Have every other one raise up 1/2mm, breaking the adhesion to 1/2 of the secitons, and then the alternate ones raise up, breaking the other half of the sections off. And then they lay flat again while a simple (and weak) sweeper pushes the completed parts off one end.

I think that'd be simpler than a segmented belt. Since there's no getting around the seams...

The funny part is that this design could be made seamless by putting a flexible surface on top... Which interestingly would make it quite like nophead's proposed rolling bearing thing, but all at once. This design would require much higher gearing and that means stronger materials, so I think nophead probably has the right design, with a single moving ripple.
Hello Everyone! I haven't posted here before, but I have been following Reprap development - very interesting.

I like the conveyor belt idea, and I have a possible solution that might work well.

In astrophotography, astronomers would mount film to a grid-grooved flat (or sometimes curved) metal plate using a vacuum.

The grooves allowed air to continue to flow, and still hold the film in place without pulling the film too tight.

The beautiful thing about this system is the conveyor belt can still move over the grooved metal plate using a gear-reduced stepper or servo drive assembly.

You could also switch to pumping air into the plate to float the belt slightly to aid in conveyor movement.

The metal plate can still be heated.

The grooves should help distribute the low pressure across a wider area of the belt without deforming it.

By having the plate extend beyond the sides of the belt, air can still flow through to the pump.

Making such a plate would be easy-breezy - anyone could make one with a rotary tool mounted to a jig, or even with a Dremel attached to a reprap.
The thing about a continuous belt, or a tread, is that arbitrarily long parts can be built, if the layers are put down on the diagonal, at a gradual-enough angle to not interfere with the print head.

If the actual build surface is applied as a paint, the gaps do not need to be dealt with explicitly, so long as they are narrow enough to be spanned by raft material.

I agree with the rolling wave of loosening-from-the-bed; it's how all low-energy fractures proceed.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]