Best/easiest "Blob of Doom" detector?
Reading about another poor soul's experience with the Blob of Doom ( https://shop.prusa3d.com/forum/general-discussion-announcements-and-releases-f61/my-mk3-s-extruder-was-engulfed-in-a-solid-blob-of--t23095.html ) got me to thinking we should be able to fairly easily make a Blob of Doom detector that can pause printing if a blob of doom is suspected. Maybe an electric eye that goes from one side to the other, where the Blob of Doom interrupts the beam? Or something easier? What would be the easiest way to instrument for it?
After my own Blob of Doom experience, I've never felt completely comfortable leaving the printer to print by itself (though I sometimes do it anyway).
Re: Best/easiest "Blob of Doom" detector?
Maybe the easiest way would be to monitor RPM on the cooling blower. If the RPM drops below its typical values, then the assumption might be that a blob of doom is blocking the blower's vent. All this could maybe be done in the firmware.
Re: Best/easiest "Blob of Doom" detector?
Coming from a background of HVAC tech (heat pumps, electric heating etc)...
Imagine the plastic blob of doom is forming - it covers the heater block. This would have the effect of thermally insulating the heater cartridge. The software would find it is having to pulse the heater cartridge less often in the "PWM loop" / "PID control" of the heater to maintain the setpoint temperature. If we detect a % slow down change of the PWM ratio once the head heat is stable, and there is no % change in ambient room temperature (we have this sensor as a sanity check), AND we know the fan speed % has not changed within x mins, we can then assume the blob is forming.
I would imagine it would be possible to implement this in software.
Issues would be
- Calibration of this detection. We might need to run a PID tuning in the normal state, then a seperate "Blob PID" again with a heatproof insulator covering the heater block to simulate the effect of the blob.
- Users of E3D socks might have to do their own calibration, given how well the sock already insulates the heater block.
- Our little 8bit Atmel CPU might struggle to keep up with this task as well as all the other tasks it performs!
Another possible way would be using OctoPrint and some sort of image recognition software on the timelapse photography option. If the heat bed is moved to a set position and the head moved away to take a still image for each layer printed, and the software performs an "image regognition" of the print bed and sees a rapid change, then this could halt the print and turn off the heater block (detatched print from bed senario). Obviously there would be a requirement to constantly light the print bed area consistently to prevent false positives.
My low tech solution so far is to use an E3D sock, watch the first layer sticks like the proverbial to the bed and check on the printer often.
Re: Best/easiest "Blob of Doom" detector?
Reading about another poor soul's experience with the Blob of Doom ( https://shop.prusa3d.com/forum/general-discussion-announcements-and-releases-f61/my-mk3-s-extruder-was-engulfed-in-a-solid-blob-of--t23095.html ) got me to thinking we should be able to fairly easily make a Blob of Doom detector that can pause printing if a blob of doom is suspected. Maybe an electric eye that goes from one side to the other, where the Blob of Doom interrupts the beam?
I have been thinking the same thing, and I am going to experiment with a through-beam laser sensor. I "just" need to get more skilled in handling a CAD program, so I can redesign the X-axis plastic parts to incorporate the laser emitter and receiver.
Re: Best/easiest "Blob of Doom" detector?
Coming from a background of HVAC tech (heat pumps, electric heating etc)...
Imagine the plastic blob of doom is forming - it covers the heater block. This would have the effect of thermally insulating the heater cartridge. The software would find it is having to pulse the heater cartridge less often in the "PWM loop" / "PID control" of the heater to maintain the setpoint temperature. If we detect a % slow down change of the PWM ratio once the head heat is stable, and there is no % change in ambient room temperature (we have this sensor as a sanity check), AND we know the fan speed % has not changed within x mins, we can then assume the blob is forming.
I would imagine it would be possible to implement this in software.
Issues would be
- Calibration of this detection. We might need to run a PID tuning in the normal state, then a seperate "Blob PID" again with a heatproof insulator covering the heater block to simulate the effect of the blob.
- Users of E3D socks might have to do their own calibration, given how well the sock already insulates the heater block.
- Our little 8bit Atmel CPU might struggle to keep up with this task as well as all the other tasks it performs!
Another possible way would be using OctoPrint and some sort of image recognition software on the timelapse photography option. If the heat bed is moved to a set position and the head moved away to take a still image for each layer printed, and the software performs an "image regognition" of the print bed and sees a rapid change, then this could halt the print and turn off the heater block (detatched print from bed senario). Obviously there would be a requirement to constantly light the print bed area consistently to prevent false positives.
My low tech solution so far is to use an E3D sock, watch the first layer sticks like the proverbial to the bed and check on the printer often.
when its detectable that way via temperature, its already way to late, and you already have a sufficient blob of doom to cripple your printer.
op is looking for a way to predict a forming blob event, and trigger a detection to cancel or pause the print before the blob even happens. optical detection with light barriers around the nozzle base sounds really like the most plausible solution. its rather complex though and you might need grid detection to get any form of sample size to judge if its just a string of filament or a real blob blocking the light, and the question if the firmware has still enough capacity to hold such detection code, or if you run it via external hardware which can send a e-stop event to the printer.
i can see it being part of the print cooling nozzle
Re: Best/easiest "Blob of Doom" detector?
I wondered how that would happen, then yesterday I got home from work and BAMM.. Blob of Doom... note..don't try to trick your mk3 into thinking it has a mmu... for some crazy reason it goes nuts... a shame too. I would sit there and do the MMUs job until I get mine..... but after that.. my printer has never been the same
I have a Prusa,therefore I research.
Re: Best/easiest "Blob of Doom" detector?
I think the best way would be a camera with software that knows what your print should look like and pauses the print if its wrong. Kinda of facial record for prints.
Re: Best/easiest "Blob of Doom" detector?
I think the simplest way would be to mount a laser parallel to the x-axis and looking through the gap between the heater block and the print, just under the heater/thermistor wires. It would be mounted onto, and just below the Z-axis/X-axis couplings, and If it's blocked for more than say 15 seconds, the printer would pause and ask for help.
Re: Best/easiest "Blob of Doom" detector?
While a laser beam sounds ideal, maybe even just sending a pulsed signal from an LED to a photodetector (as many electric eyes do) would be enough. The pulse would tend to mitigate the effects of ambient light. Either way, translucent plastic may be challenging.
RE: Best/easiest "Blob of Doom" detector?
Hmm, it seems to me that the 'Blob of Doom' (BOD) forms when the print comes off the bed and sticks to the hot-end assembly. So, detection should focus on that event (print coming off the bed). Some sort of laser beam across the surface of the print bed that gets blocked by the print itself?
Has anyone experienced a BOD where the print DIDN'T come off the bed?
Frank
RE: Best/easiest "Blob of Doom" detector?
Blobs can form higher up on parts - supports fail and begin collection on the nozzle - and usually end in a crash - but not always. I've usually caught the blobs very early, before they become BoD, so far I've been lucky. Detecting this type of fail would be tricky. No laser method I can think of would be smart enough to know something has gone wrong: it'd be like detecting a single tree in a forest is missing; perhaps even as hard as finding a tree under the visible canopy has fallen over.
A vision based system looking at the four sides of the heater block would be the way I'd investigate. Any visual creep up the block would be detectable and a fair sign something is wrong.
RE: Best/easiest "Blob of Doom" detector?
Detecting the blob after it formed would be like an accident detector that goes off after all the cars have stopped bouncing around.
RE: Best/easiest "Blob of Doom" detector?
Detecting the blob after it formed would be like an accident detector that goes off after all the cars have stopped bouncing around.
jency53 is a spambot, no point replying to him. I know I've reported that account before...
RE: Best/easiest "Blob of Doom" detector?
jency53 is a spambot, no point replying to him. I know I've reported that account before...
I've noticed he posts a lot of duplicates, but I wonder if he's hitting "reply" or "submit" repeatedly without waiting for a response. The actual responses are coherent enough, just lots of repetition.
RE: Best/easiest "Blob of Doom" detector?
jency53 is a spambot, no point replying to him. I know I've reported that account before...
I've noticed he posts a lot of duplicates, but I wonder if he's hitting "reply" or "submit" repeatedly without waiting for a response. The actual responses are coherent enough, just lots of repetition.
No, it's stuff other people have previously posted in the thread. (Look at the third post down). It then inserts links/spam into the quotes when it quotes itself.
RE: Best/easiest "Blob of Doom" detector?
Back to topic. I think the only good working way is visual. From the gcode it's possible to render the expected result for each layer position. Then it's just a matter of camera calibration to be able to compare actual result with the projected one. This way you can always check for issues like blob of doom or detached object.
Often linked posts:
Going small with MMU2
Real Multi Material
My prints on Instagram
RE: Best/easiest "Blob of Doom" detector?
The Spaghetti Detector plugin for Octoprint seems to be a good starting point. It's looking for strings, but I wonder if it could be trained on blobs around the extruder just as easily.
On a rudimentary level, would a simple door sensor along the X axis work? Mount a reflector on the hotend on one side (right to avoid wires presumably). Shine a light on it along with a light sensor from that side along the X axis. If the light reflection is disrupted, it's not anything normal so send an alert.
Or a simple switch underneath the fan. Nothing should be pressing from below under normal circumstances.
I've only had one partial BoD, so can't say I've got enough experience with 'em to think of anything that would work in all circumstances.