A Cable Robot 3D-Printer
Вставка
- Опубліковано 3 січ 2024
- The video presents a cable driven 3D-Printer, based on the parallel kinematic delta plattform. The machine uses dyneema fishing line as cables and a custom kinematic model running on the 'smoothieboard' printer board.
Carbon fiber preloading rods are used to keep the cables under tension, providing the necessary stiffness and allowing for rapid accelartion.
The parallel kinematic design has potential benefits for large workspaces.
![Building a Small and Quiet Air Compressor [4K]](http://i.ytimg.com/vi/aUEZkS5FLxU/mqdefault.jpg)
UPDATE: Printable STL Models, Step files and the modified Smoothieware firmware is now on github: github.com/0x23/CableRobot3DPrinter. The host/calibration software code will follow...
The project (software, firmware and 3D models) will be put on github, as soon as I find some time - so if you are interested stay around.
This is very cool and novel! I had the thought just before you said so that "This might be decent for scaling!" There are so many offsets and radii involved in the cable drives that I'm impressed you got calibration down to 0.2mm inaccuracies ... those could easily be due to component tolerances or frame being slightly out of true, although printing the second build with the first build when the first wasn't proven accurate could also have something to do with it :P All in all, super-impressive. Thanks for sharing!
If you look around on the internet there was a cable delta years ago that used a single rod for even more weight savings. Wanting to say it was around 2013-2014.
I designed an over constrained 5-axis, 6-motor one like that more recently, based on a sort of octahedron tensegrity structure, but I haven't found any other such mechanisms.
Wow, this is wild! Love this concept and your execution and ground up work on everything is incredibly impressive!
Excellent work, and very inspirational. Thank you for sharing it ! It has got me thinking of ways to adapt this to a table-top CNC router (for woodwoorking). I live in a very warm and humid area which is not kind to linear bearings of any sort. Your design eliminates them entirely.
Scaling up is definitely the way to go.
The end effector on the cables at the FAST Radio Telescope in Pingtang, China weighs 30 tons, so you still have a fair way to go yet.
Very nice engineering! Would be nice to have a comparsion to a regular delta. Probably the cable-version is much more cost efficient at huge scale...
An awesome project! Subscribed!
Excellent work!
if you haven't heard of it you'll probably love the hangprinter
Snap! Nice work, wow!
Use Dyneema cord and an Gyn tackle or Threefold purchase to limit your drawbacks
Its already Dyneema cord and gyn-tackle would add alot of complexity. Thanks for the advice though.
if you need insparation the positron v3 uses cables
Now that is something you don't see everyday !
Thank you for sharing! Question, what program did you use to model the kinematics?
The kinematic model is coded in plain C++. The code is then used in the printer firmware as well as in the Printer host software, which I also wrote in c++.
You can simplify the kinematics and maybe get rid of those last inaccuracies by modifying the winch in the following way:
Instead of winding the cable on a cylinder, use a screw. Like a trapezoidal thread screw. The cable rests and winds in the thread. The screw rests in bearings and is guided by a nut ( or, use ballscrews to get both ). Therefore, the winch axis slides axially on each turn by screwing into the nut, keeping the cable exit stationary, instead of moving about slightly ( like is shown in your kinematic model )
Its a really good approach to reduce cable wear and winch size for larger machines if implemented well. I believe the sources for inaccuracies lie somewere else though. Its a really good idea but makes the mechanics more complicated - but I would love to see someone try it.
@@DiffractionLimited you used a braided cable, so it must compress a bit when going over the pulleys. Do you think that contributes to the inaccuracies? Or would it be something unaccounted for in the geometry ?
I also have the impression the preload rod might be wobbling sideways as it is supported only on one end rigidly and on the other only by the cable ( is that right? Can't tell for sure from the video)
@@bschwand Yes the preloading rods are wobbeling around alot - BUT they are not really part of the kinematic chain when you think about.
@@DiffractionLimited if I understand the geometry correctly, I think they are (at least dynamically) because their sideways wobbling will exert great leverage and ever so slightly make the final print platform position vary.
To make an analogy, my Linear delta printer frame is not part of the kinematic chain. However, pushing a little on the frame corner caused the printhead to noticeably move. After rigidifying the frame, my printer is a lot more accurate. Plausible ?
@@bschwand The balljoint at the endeffector will not transfer any torque to the plattform - so no leverage is exerted. But you are right it is and accellerating mass and thus will dynamically apply some (hopefully small) amount of force to the balljoint when wobbeling. A reason, why I made them out of carbon fiber to be as lightweight as possible.
That is awesome! Would love to see a speed benchy printed on it.
If you were to build one today, how much would it cost?
Not quite sure and it depends on the machine size but you can try to estimate it youself. The main components are:
- 3x Nema17 0.9deg Stepper Motor
- 1x Smoothieboard
- 1x Powersupply 24V
- Hardware (Screws, Nuts etc)
- 3xCFK Rods
- Frame and Baseplate
- Fishing Line: Spiderwire Stealth Code Red 0.4mm
- Extruder and Hotend
- 3D-Printed Parts
@@DiffractionLimited Wow, you got that much accuracy with fishing line?! That stuff is designed for strength, and for most materials the easiest way to do that is to let it plastically deform (lengthen) when under load. The winch setup and speed means you're probably putting significant dynamic load on the line causing not just backlash, but the line to act like a spring.
The dynamic accuracy is probably worse, dependeing on the accelleration. The system has no backlash since it is preloaded but you are correct the cables act like a spring - hence the lowish stiffness. Steelcables might do better but require larger bending radii.
@@arthurmoore9488 Dyneema has very low stretch and high strength. Not obvious that steel would do better. And steel would be heavier and harder to bend. To be tested, I guess
Woah nice. Can you share the kinematic visualization code? That looks sick. And calibration does it send to the printer from that window widget?
Max accel you tested?
If I remember correctly it was about 10000mm/s^2 in the clip at the beginning.
Looks great! Will you publish the design anywhere?
If people are interested in this, I can put the 3D-Models and the source code for the kinematic model on github. I probably will not have the time to make it a nice self-contained project with documentation and build instructions though.
@@DiffractionLimited I'd be very interested, if you're willing to do so! Would you mind including a bill of materials as well? Thanks so much!
@@DiffractionLimited Please do! I'd love to take a look & modify / make my own design of this printer :))
@@DiffractionLimited I'd really appreciate taking a look at your design as well. It seems perfectly suited for a room scale pick and place task
@@tinkrmind keep in mind that the tension rods are an essential component and they would have to extend outside the room.
Wow I cant eaven get the calibration of my normal delta printer corect. The delta radius is always of casuing not all parts of the print to stick to the bed and it sometimes just speeds up for no reasen and generaly does wierd stuff. And you wrote a custom calibration software and kinematics wow
Very awesome 👌
Nice to see you here. Its funny because the successor to this printer is actually a modified vzbot, which works very well for me - thanks for the great design !
what's the advantage compared to a rotary delta robot? like in pick-n-place machines?
Direct-drive without the expensive gearboxes and associated backlash required for a well working rotary delta robot. You find some rotary deltas using RC-servos on youtube. You will notice that they are usually quite wobbely and not so accurate.
Awesome. Im not highly experienced with cable driven motion. Why choose a polymer cable over a stainless steel one here? I guess creep isn't a big deal since they're preloaded with a separate spring that can account for creep right? Cheers
Very good question. Polymer cables are acturally easier to get and work with. Steel cables would be even better but require very thin strands for the bending radii used in this construction. In prinziple the design could be implemented with steel cables, especially for larger machines.
@@DiffractionLimitedDo you think the stiffness issue that reduced print quality came mostly from the forces caused by the accelerations of the print head, or forces caused by the drag of the nozzle against the previous layer? Like, if you slowed the printer way down do you think that the low stiffness would be any issue?
This seems like a interesting starting point to use belts just like how you're using cables. It seems like belts are just many cables in parallel over molded with polymer to give higher stiffness. It seems like you can use a belt and have the same unique and simpler kinematics with the same stiffness as a standard delta printer. What do you think?
@@neverwipe Actually the print quality is fine, expecially with HW version 2. Belts would be great but the cable has to bend in two directions at the deflection assembly. Belts only bend in one plane, unless you find some clever way to twist them. However all strands within the belt would have to be loaded equally, otherwise you loose the advantege.
@DiffractionLimited Interesting you say print quality is fine. You mentioned stiffness is an issue and I assumed that it would show up as low print quality. What is the issue related to low stiffness?
If stiffness isn't a problem then the belt idea wouldn't be as interesting. Here a potential way to do it while keeping the belt straight:
You can mount the motor and pully with a tension ball joint to the frame. The rod would be mounted to the this assembly with a linear bearing, since the other degrees of freedom are allowed by the ball joint. Because the mass of the drive train is right next to the ball joint, the inertia of the drive train shouldn't be an issue. If it is an issue, you might be able to use a flexible shaft to mount the motor to the frame and have the pully/linear bearing assembly on the ball joint.
@@neverwipe Since the endeffector is lightweight, the lowish stiffness is not a problem for normal sprint speeds (60mm/s) - since 'low stiffness' is quite relative its best look at the slow motion footage and assess the situation for yourself. Also today we have input shaping, which would improve the situation too.
Good idea for the belts, but it would add a lot of complexity, which i always try to avoid if possible.
I'm so early this is still 360p. I'll come back a little later. Just FYI, the speed of the audio on this seems to vary a lot. It's very strange.
I am using text to speach, so sometimes it sounds just funny. Also adjusted some audio clip speeds to match the video content better.
@@DiffractionLimited I see? May I ask why text to speech? I kinda understand as having a voice print out in the wild can be very problematic these days.
Ingenius
Wow!!!
Make a pick and place from this.
Snap! Nice work, wow!