I haven’t really started on it yet, but I wanted to run the idea past people to see if anything like it had been attempted before or for obvious problems.
A workpiece mount failing, whether clamps or tape, can shake a machine apart fast if the piece gets picked up by the spindle.
The idea is to use a relay and accelerometer (like an MPU-6050) through an Arduino to check for excessive movement over a certain period (say 3 seconds) and to kill the power to the motors and router.
Use case is for when you run back into the house “for 5 minutes” to use the bathroom or when you’re just across the garage with your hearing protection on and it takes you a minute to notice that the workpiece came off the table.
Has anyone tried this before? Other than the obvious complacency (and that it won’t necessarily account for the collet getting buried and starting a fire that way), can anyone think of why this wouldn’t work or would otherwise be a terrible idea?
Well I think it is a bad idea to leave the machine unattended for any amount of time, and building something to allow/encourage the user to do so is an even badder idea.
Shaking of the machine isn’t the only thing that can happen when the machine is left unattended. Fires can (and have) started. Bits can break. Steps can skip, and if caught early enough the part may be salvageable, but if left to wreak havoc, then not so much.
Personally I’m not a fan of the idea of building complacency into a system that has the potential to burn your house down or injure somebody.
I can certainly get a baseline vibration when it’s running well. From my observations, setup failures are violent when they happen.
This isn’t to prevent ruined work as much as to prevent a wrecked machine or worse. There’s a slight oscillation and vibration when the machine is running, but nothing like when a piece gets caught on the spindle and starts spinning off axis.
Even things like steppers losing connection on a single conductor create a jackhammering effect that I’ve already had break printed parts. I would want to cut off power immediately in that case. Soomething like a sustained movement in the X and Y over a certain time would trigger it. I wasn’t really planning on filtering anything other than that.
Do you think there will be enough electrical interference that it wouldn’t work? I could also make a simple mechanical accelerometer with ball bearings held in place with springs and count how often they lose contact in a given period to determine if the router is shaking badly.
I got the idea from a video someone posted on here showing the workpiece getting picked up and shaking the machine apart in 6 minutes or so (might have even been your video I can’t remember).
Accelerometers aren’t as intuitive as you might think. The router vibrating can have a pretty high acceleration even though the magnitude of the oscillations are pretty small. A big jump may not have a high acceleration, despit having large movements.
It seems possible, but you would need some good data analysis tools and skills.
The acceleration will be big jerks, for a few milliseconds with (relatively long) pauses in between. But you can do something like threshold measurements against some limit and then add something up when it is tripping, and subtract something when it is not. When the total goes over some larger value, it can trip. Essentially low pass filtering it, but not requiring a continuous 3s of high acc.
Interesting. I agree with all the sentiments about not leaving it unattended, but I think this might be some very early and interesting work. Maybe accelerometer is part of a solution, along with stall detection on the motors, and potentially things like video monitoring, heat, or other sensors. Someday I wonder if it would be possible to build a machine that stops the spindle and machine movement far faster than any human can react, even if they’re standing there watching it.
I know it’s not something that is likely in the short term, but the work would need to start somewhere. There’s also likely to be a lot of dead ends along the way.
But I do agree that we can’t leave the things unattended. I know what happens when you give a robot a machine that can kill you… I’ve seen the movies about it.
Unlike preceding comments I do think detecting failures and taking automated actions can be benefical
It can never be a substitute to prevention, but in case everything else fails, I’d be happy to have it…
Saying such measures would encourage user to let the machine unattended sounds an awful lot like arguments you hear about some car safety features…
Having an airbag never encouraged anyone to run into a tree
Having auto emergency brake never encouraged anyone to drive 80mph near schools…
Those who do ignore basic security/safety rules will still ignore it whatever…
Detecting vibrations seems very hard though.
On a simpler scale, I’d begin with detecting skipped steps and pausing on such event
Coupling the power/contrôler with a heat/smoke detector above the machine could also be a great and easy place to start
Thank you all for your comments. You’ve given me a fair amount to think about. And this is going to be significantly more complicated than I initially thought. But I still think I’ll pursue it when I have the time. I like the idea of stall and missing steps detection, and a relatively simple fire suppression system is also something I’ve been thinking about how to implement. I’ll keep you updated when I get to actually tinkering. Thanks.
The features that actually make it to market do have a lot of “safety” design considerations though. It is worse to have a safety system that doesn’t work than not have a safety system. I have worked with safety engineers trying to move ground robotics into production and it is really hard. Early driver assist features encouraged drivers to ignore the road, but weren’t smart enough to actually improve performance by enough to cover that loss. There are studies that show alerting inattentive drivers to recover from a robotic failure would put them at a huge disadvantage and it causes more accidents than it solves. So a lot of those early systems didn’t make it to market.
I don’t think we have to worry about it at this stage. There are enough technical challenges and a low enough cost to get started that we can shelve those concerns and just do it for fun.
I’m reminded that the introduction of anaesthesia during surgery initially increased mortality - when the patients weren’t screaming in pain the surgeons were free to take their time which led to more infections.
As with anything else, safety features are good - but should be treated as the last resort and are never a replacement for safe practices. If you’re nipping out of the room for 5 minutes there’s a HOLD feature, that’s far safer and it’s already there.