Introduction
Hello all, we are looking to gauge interest and get feedback on an AI integration for Klipper.
Current Features
Visual Failure monitoring: uses frames from the webcam/IP camera/RTSP camera and detects spaghetti failure and notifies the users OR pauses/takes action on behalf of the printer.
Anomaly Detection: using live data from the printer, monitors and tracks the current health of the printer and it’s subcomponents. Used for scheduling preventative maintenance and detecting+diagnoising issues with the printer.
Comments
If you are interested in a feature/plugin/integration that provides you access with these AI features, please like or comment on this thread.
If you have requests for features that aren’t listed and fall under the scope of AI + Data, please leave a comment or reach out via direct message/email.
This is exactly one of the things that we’d like to be able to detect/prevent/fix. Ideally we want to detect these types of issues with kinematic feedback variables from the steppers and from the heating units. For example, if we have some closed loop feedback, accelerometer feedback, and temperature feedback, we would be able to identify those issues on the corners.
From your original post, it sounded like you were looking to use AI to monitor the print operation as well as printer health and stop the print/generate an alert when there was a problem. I’m not sure what all these “feedbacks” have to do with AI and how they are going to do correct a problem or do any kind of better job of preventing them - I’m nervous that adding more hardware to the system will make it more expensive, complex and difficult to work with.
Personally, I think just having the ability to detect print errors along with stopping the print and reporting them ASAP (rather than a few hours down the road) would be a major win.
The Anomaly Detector for monitoring the health is a rather broad model that has multiple end uses that could address the issues you showed above. It’ll be interesting to see if we can get it to do so with a high accuracy…
We already have the capability to detect and report the spaghetti type errors, we just have to implement a frontend component for Klipper to be able to use it.
AI, great idea.
Scan the spool with the camera and have it select the right settings for that spool automatically
Do a full mapping (or 5mm x 5mm and interpolate) of the surface of the bed when cold and when hot, and calculate the differences with temperatures to be used online with the Z axis while printing.
A “teach in” where a predefined object should be printed, compare to the actual result and make/suggest changes to the profile. Object should amplify possible issues.
Maybe repeat to “learn” enhancements.
For different plastic types.
All this will run on a PC, right?
At least until printer is “tuned” by above.
That profile can then hopefully be used in conjunction with the only life factor, the temperature, by the 3D printer’s CPU to adjust.
My Ender doesn’t print yet, (see my cry for help) so I don’t know how errors look like.
Thanks
I believe this feature already exists on some printers with their proprietary spools of filament that have NFC tags. Most reasonable way would be to build a database of filament’s with unique ids and to have the user query that id for all info regarding the filament.
Im not sire of this thread has moved with Obico but i would like to see this implemented direct on the pi/linux/ubuntu based devices without the need for cloud connection.
For basoc failure analisis (spageti detection) it seams simple enough bit i would like to see a deeper dove into fault analysis using cameras redily available. Using an Ir enabled stereo camera for pi with the combination of the rendered Gcode. A trained algorythim can monitor layer by layer looking at the extrusion in relation to the toolpath specified in the gcode looking for faults.
In the gcode you have the ability to pass all parameters such as speed, flow, material, expected temp etc.
With it onboard the device running klipper it automaticaly has direct access to the machine firmware where it sees all curent printer parameters. With this in combination we could use a simple Qr sticker on the build plate to identify orientation distance and angle for telling the model the current camera orientation. Or via a single layer stl that prints a calibration model. This again will simplify the calibration without increasing costs.
In the future it could be adapted for multi camera capabilities to take advantage of multiple frame mounted cameras and an optional nozzle mounted camera like comonly used on many vorons.
We are seeing many consumer printers installing lidar and cameras at the nozzle already.
