Saturday, November 21, 2009


Moving towards production

The Java driver software now handles multiple part printing reasonably well, though wrinkles remain. This tray of Mendel parts was printed on my old Darwin in six hours (my new Mendel is still, as web pages say, Under Construction...). Erik and Nophead's experiment of reversing the extrude motor at the end of each write has led to minimal PLA string.

Ages ago Zach defined an XML multi-part file format for RepRap (and Fab@Home) that should be pretty easy to implement that would allow one to store a whole tray of parts like this in a single file. (The G-Code file generated to build all of them is also a single file of course.) I have a few immediate things to do on the software/firmware front (like having an M code that sets the temperature and waits till the extruder gets there, as opposed to returning immediately - one needs both), then I'll implement that. That in turn should mean that Mendel is represented by between four (Mendel) and six (Darwin) files that one just has to load and print...

Labels: , ,

I think easier build management is really becoming important at this stage.

I really like how the java host is progressing, but in a way it's one app trying to do everything. It's both its strong side (the ONLY integrated toolchain for RepRap) and its weak side (no possibilities of using other, best-of-breed tools).

Wouldn't it be better if you had a queue manager and can just drop a folder of STLs into a queue directory? A functional version could be made relatively easily, the required footprint of each of the objects can be used to automatically allocate a part of the print platform. One of these footprint layouts could be put in a file specifying all the relative orientations and positions. The RR-Fab@Home file format can then be written (a zip of STLs with an XML file). Keep in mind that STL2 is being drafted right now (input is asked!), so these issues might be addressed by it.

About build management: this could be extended with an genetic algorithm that will optimise the use of space (even by stacking objects with fairly symmetrical top-bottom faces!)

The queue manager would allow you to prioritize the prints and perhaps even tell you when you've removed objects during a build in another corner so that it can start reusing this area again. That way the machine can continue producing round-robin style, the platform effectively is your 'buffer'.

I actually started writing something in PHP. It's a standalone webserver that you can access from any PC on the LAN (or your Android/iPhone on the road if you configure your firewall for it). It is supposed to control the machine, but also allow you to configure build jobs and do distributed skeinforging with the rest of the network and possibly other users of the same system (like distcc). The queue manager would also be responsible for sending IM/text messages when something finishes or human intervention is required or can be monitored from any internet enabled device.
Using RR-Fab@Home just for saving the time for multiple objects in a build might not be a good idea. The footprint of machines' build base can be very different (e.g. Makerbot, but even between Darwin's) and people might want to fit a nut and bolt before printing more of a certain object. IMHO, it's power lies in multi-material objects and also storing some more production related data that is the same across most machines.

Sorry that you had to read all that ;)
P.s. STL2 input can be given here:
I wonder how important multi-part builds are in the long term. I'd think that eventually the reprap will be able to clear the printed part off of the build area itself. If it can do that, then multi-part prints are all liability (more parts ruined if something bad happens) and no gain.

So short term, yah, it's important. But I wouldn't waste much time on getting too fancy.
John Gilmore I disagree. Multi part production has its place for cleaning parts off.

If you are using materials that take so long to set fully like some forms of epoxy. Waiting on the print dry would be critical. So a RapRap with a simple to remove tray would be an advantage.

Of course a full set on a tray has an advantage for packing into boxes and the like for shipping.

Basically simplest way for massive production using Raprep style printers is have a system to insert new trays into RapRap automatically and remove completed trays and take them to storage for cooling/setting. Then some machine/humans to remove the parts from tray. Ie 3 step process will suit more materials and allow more production in the same amount of time. Printing setting and removing can all be performed parallel to each other.

Past a particular scale molds and pour filling become beneficial due to the speed factor. Ie its faster to pour. To do this interconnects between multiable objects can save on the numbers of pours that have to be performed.

One of the biggest mistakes you can to is attempt to make one machine do every job if you are after fast production. Now I am not saying that the machine that removes the items from tray and prints the item to tray could not be twins just tooled differently. This is still two machines.

Multi material printing may make it simple to print two halves of a mold on a tray. So far a machine that is open spec for mold filling and extraction either.

Stacking to print I don't agree with the risk of defective parts will go up massively.

Also stability of print ie the most functional parts produced. Option for each item in a multi print to be completed before starting the next object. So you don't end up with a whole stack of half done object due to head stopping putting out. Yes I know this means leaving space so the print head can get down without hitting objects already produced.

The width and the way the current meddle and Fab@Home print head moves does not reach max effectiveness for this. Lower head to work not lift/lower rack that guides print head left to right. Rasing and lowing the print platform in Fab@home and darwin is also not perfectly suitable. Same issue large area that has to be avoided from hitting work when lowered.

The question comes down if head lowering system would be stable. It might be a physical limit we cannot over come. I know the top frame of mendel.

Once you get into multi part production things become a whole lot more complex.

I would not mind seeing CNC functionality for raprep making it a more all round tool.

So far I have not built a raprep each time I have seen an annoying design flaw. I have to say you guys are heading in the right direction. Mendel is a lot closer to a 3 axis milling system.

Something to be aware of some 3 axis has them layed out like this.

Rack and pinion on the tray. Allowing the tray to be expanded in 1 direction. The over left to right at fixed height and at tool lowering system from that.

Now the Rack and pinion means trays can be pushed out of machine when they are filled with the items they need printing and a feed location can push a new try with nothing on it into the machine. Simple light sensors to detect edge of tray prevents it going to far.

Basic frame of Mendel quite suitable to alter into lowering head machine with eject able trays.

Now of course making either work stable is going to be a lot of messing around.
To expand on the CNC similarities and the rack driven table oiaohm described could easily be expanded to be either very long or changeable. Production CNC machine shops have used changeable "pallets" for a long time to swap finished parts for fresh raw material. I don't see why we reprap couldn't swap a table of finished parts (perhaps still "wet" or cooling) for a naked table to start the next batch.
I missed something else important with the changeable head support planed latter on in the raprep project one of the heads could be a laser cutter. Using a laser cutter also need the the means to lay out the parts effectively on the surface.

One of the old rapstraps had its parts laser cut. I do wonder if raprep will ever get to the point were it can be a mixture of laser cut parts and extruded parts in 1 unit all make able on one machine. Particularly thinking pre build scrap exists to convert into a laser cutter if you have the head. (lot of scanners end up dead all over the place)

So good layout code could suit two jobs not one. laser cutting and tray layout of mult parts .

There is no point having a self replicating machine if its material picky and picky about its parent.

Change tray to match different head and materials would be good.

3 machines in one on the table should be possible. Basic CNC, extruder and laser cutter. CNC is particularly handy if in future Raprep wants to make it own boards. Ie cut the boards is an option. Laser cutter is handy from the point of view of what materials you can get from a local hardware store simply.

Space effective layout also applies if you need a pack of different PCB boards out of one board.

Ie CNC work Extude work and laser work all need effective layout of parts to save time and costs. All can benefit from removable trays to stack up complete work as well.
One thing that would be useful is to provide several sizes in layout so that the the correct size for your machine could be selected to maximize part generation.
Ages ago Ed had the idea of a machine with a continuous belt as a build base. The parts would separate as the belt went round its cylindrical drive at the end.

MakerBot's CupCake has a magnetic build tray that can be swapped in an instant. Short-term, I think that's the way to go.
I like the idea of a replaceable platform or even a rolling platform; even though i still have to build my first RepRap/RepPrep.

I am currently working on a set of modular programs that slices STL and copyouts the generated G-code. If combined with a part sheduler, the whole build train would be there.

But wouldn't it be better to also focus on a faster printing machine? I.e. a faster moving machine? What is holding up the speed? The speed of the extruder? The speed of the X/Y/Z axes? the solidification of the materials used?
Adrian Bowyer Sorry to have to rain on a belt drive. Its not used that often in molding for the reason of flex of belt that can lead to warped parts produced. Set of solid plates may hard. But solid plates give you a solid surface for the materials to set on when removed. Its also possible to laser scan a solid plate after use to detect warping. Since its outside the machine is a fairly simple process to write off warped trays and put new ones in. Compared to the nightmare effort of belts getting them tight and so on.

Belt drive system its still a limited area thing. When the belt is full of items cooling you have to wait. Plates the area can be expanded by using more plates. Its more limited space compatible. Yes I know that seams stupid but it is. You can have racks to place parts on when cooling with plates. Think oven trays.

magnetic build tray not the wisest ideas. magnetic trays can slip slightly out of position reason why they are not used normally in even small scale CNC. Ie you either don't have a removable tray or use mechanical locking to table of some form is required. Either rack and pinion with tracks or some form of latachable support plate. Main advantage of rack and pinion is the means to have many different size trays for different jobs. Small job small tray. Large job large tray. Its a queing thing as well many trays using a rack and pinion can be lined up before the machine to complete. And feed and removal machines can swap the plates over. With robots the completed pushed out trays can be taken to storage for cooling and the like. Yes this is large scale.

Trays get called pallets in CNC because some cases you don't bother remove them from tray before shipping to customer. The tray protects the item until its at the customer then tray is returned.

Ie solved the having time to remove parts problem. Out source it.

Its key not to use magnetic stuff because this can limit materials raprep can consider using. There are some iron based pigments and the like you don't want mixing with magnets.


The speed of the extruder?/solidification of the materials used? Both related. Its more speed material can be melted and cooled at. There is a limit how fast the melted material will flow and hot burns or sets solid again ie jamming print head. Then how long before is set enough to print on top of. Put layers on to fast it will bite the big one.

Hotter the material the long it will take to cool to go solid before you can print the next layer. So there will always be a speed limit here.

The speed of the X/Y/Z axes? Really if this is a issue is simple to fix. Faster step pulses.

There is a big reason why CNC shops have the means to mold there produced parts. Pour beats CNC most days for numbers of items created. Same will apply to raprep.
Post a Comment

<< Home

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

Subscribe to
Posts [Atom]