Help designing a self guided automatic 3D printer

AI Thread Summary
The discussion focuses on designing a self-guided automatic 3D printer, emphasizing the need for a topographic model of the printed material to prevent collisions with the extruder. Participants suggest using a raster pattern for printing layers and highlight the importance of an algorithm to predict potential collisions. There is uncertainty about the capabilities of the current microcontroller to map the model during printing, prompting inquiries for suggestions. A proposed solution involves processing G-code on a separate computer to simulate the printing process and identify collisions before actual printing. Overall, the conversation stresses the necessity of collision detection and modeling for successful 3D printing operations.
Boltzman Oscillation
Messages
233
Reaction score
26
TL;DR Summary
I need to design a 3D printer where the extruder should not be allowed to touch any material that has already been printed or the printing stage. The extruder does not move, the stage does in XYZ. Every solution I can come up with does not stop the extruder from touching printed material. Right now the best solution I have would be to use a laser but that laser can easily be blocked off by printed material. Anyone have any insight as to how this is usually done?
All in the summary, can't copy paste on phone :(
 
Engineering news on Phys.org
Why can you not rasterize the volume, then print the bottom layer on the stage first by scanning in a raster pattern. Lower the stage one step, print the next layer, repeat until done.
 
Baluncore said:
Why can you not rasterize the volume, then print the bottom layer on the stage first by scanning in a raster pattern. Lower the stage one step, print the next layer, repeat until done.
Ah, well I should have probably been more clear with my situation. The final product will be an AI controlled printer but for now it is manually controlled. My job is to make sure the user cannot push the stage unto the extruder or into any printing that has already been printed. There is no input model of a volume, for now.
 
Boltzman Oscillation said:
There is no input model of a volume, for now.
Then you must build one as an object is printed.

You must have a model of the material on the stage if you are to keep it away from the extruder. That will require an array(x,y) containing a topographic map of z over the work stage. You will need an algorithm with the ability to predict collision with the physical profile of the extruder.
 
If you want to prevent a destructive crash that might damage the machine, but you are happy to dent or melt part of the object being printed, then;
Hang the ejector in a trivet mount, with it's mass and possibly a light spring from above to hold it on the three locating seats. Any lateral or vertical contact will lift the ejector from at least one point on the trivet, which can be detected, to alarm the controller.
https://en.wikipedia.org/wiki/Tribrach_(instrument)
 
Baluncore said:
Then you must build one as an object is printed.

You must have a model of the material on the stage if you are to keep it away from the extruder. That will require an array(x,y) containing a topographic map of z over the work stage. You will need an algorithm with the ability to predict collision with the physical profile of the extruder.

I see what you mean, currently the printer works using an msp microcontroller so i don't know if it will be able to map out the model as it prints but I have no experience on this topic so i really don't know if it is capable. Any suggestions?
 
Baluncore said:
If you want to prevent a destructive crash that might damage the machine, but you are happy to dent or melt part of the object being printed, then;
Hang the ejector in a trivet mount, with it's mass and possibly a light spring from above to hold it on the three locating seats. Any lateral or vertical contact will lift the ejector from at least one point on the trivet, which can be detected, to alarm the controller.
https://en.wikipedia.org/wiki/Tribrach_(instrument)
That sounds like a simple way of getting this done. Ill definitely relay the idea to my team and hopefully we can see some design come out from it.
 
Boltzman Oscillation said:
I see what you mean, currently the printer works using an msp microcontroller so i don't know if it will be able to map out the model as it prints but I have no experience on this topic so i really don't know if it is capable. Any suggestions?
I would guess the microcontroller on the printer receives G-code (or a similar primitive instruction code) that the microcomputer then uses to direct the positioning motors and extruder.
What is that control format?

Before the generated G-code control file is used, it should be possible to process the G-code file on another computer with a simulator that builds a virtual model and searches for any virtual collisions. At the same time the software can predict the minimum run time and material consumption for the real process. Such verification software might already be available in the public domain. It is orders of magnitude easier to write a brute force simulator than a rasterizer.

Control codes are so simple that for a 3D x-y-z printer, an array(x,y) = z, initially with z at stage height could be constructed, then updated as z is printed. Once you have defined the shape of the extruder head, a check can be made at each “stepper step” to see if any point on the head has a coordinate that collides with the stage, the object under construction, or the axis limits.

Unusual coordinates systems such as a polar axis would need a trivial change to the simulator mapping strategy.
 

Similar threads

Replies
1
Views
1K
Replies
10
Views
3K
Replies
20
Views
6K
Replies
1
Views
1K
Replies
8
Views
5K
Replies
23
Views
7K
Back
Top