2 Attachment(s)
Re: Jerk Control in machines
Hi Jack - Here's a couple of papers. Shapers are not a compete solution. They "solve" the mechanical vibration side of things in a simple way, they do not improve jerk, they are like a noise cancelling circuit. You have to identify the potential vibrations of the machine and address those with the cancelling circuit. It does not change the trajectory to improve jerk.
Once machines have accelerometers on the spindles (such as printers do) we can feed back the head accels and use jerk algorithms hand in hand with the shaper to solve it even better. Sophisticated machines have been doing this for some time. Peter
https://www.mdpi.com/1424-8220/22/6/2186
Re: Jerk Control in machines
Quote:
their trajectory planner (it's open source, you can check it out) has no realtime requirements ... trajectory can be computed entirely offline
a cnc system that does not adapt to inertia caused by part's mass change, can be considered to be "offline", or runing in only one dynamic mode; have the algorithms for such, and you can estimate toolpaths at a pc, with pretty acurrate time, without miscelanous like tool changes, magazines, etc
Quote:
I fear they are not up to the task for a machine tool where cutting loads are constantly changing
basically, that's what a diff control is :)
Quote:
Once machines have accelerometers on the spindles ... Sophisticated machines have been doing this for some time
another trick is the diff map, for controlers that are to operate at minimal cutting loads, like a high speed finishing, or a cam lathe, so to identify hot position, where motion is a bit out of control
and in last 2 months, i did such applications :
... lathe sinusoisal acceleration rpm, for heavy parts, as basic acceleration is way to agressive
... harmonic control for live tools, like to destabilize the rpm, instead of using an endmill with variable flutes / kindly :)
Re: Jerk Control in machines
Hi All - I have been researching on and off this subject for a while. Been 7 months since I've spoken about jerk here. I have started a collaborative design with a company for a large router (3mx1.5m) and they want to use servo motors and run fast. So I dug out my notes and did a bit more digging on controllers. I found an old note about Dynomotion and started reading their stuff. I also sent a note to Duet3D and started reading their stuff. I found Duets docs and forum was like walking through mud. However Dynomotion was clear and happy. Plus Tom from Dyno answered an email even though its Easter... I like Kflop and its now the front runner for the AA machine. Good diagnostic software and jerk control to boot for steppers and servos. Shall know more after reading more. Its also open architecture so maybe able to write an input shaping or active control for it... Have to brush up on C. Peter
https://www.dynomotion.com
Re: Jerk Control in machines
Hi All - I didn't find this at the beginning :( would have saved a lot of time ... I'd be using Kflop by now. So will get one on order and test it very soon. Peter
https://youtu.be/LdkyiH-v_9A
Re: Jerk Control in machines
hello :) recently, i talked a lot with shineworld, from rosetta cnc, on different topics; that guy covers a wide range of things
their controller being shared by different cnc machines configurations, he can simply switch defaults, post-analyze and adapt, then send to the real machine; implement custom functions if needed; very good development, on all sides, and he knows his stuff
i was after some abstractizations, related to cad-cam-cnc combo, and he managed to point me towards a good direction
i wonder where is mactec ? all the best peteeng :)
Re: Jerk Control in machines
Thanks DK - I think I have resolved the jerk issue. Some time ago I looked at Rosetta stuff and yes its impressive but outside the hobby scope/cost area. I'm sure MC is lurking but there's not much meaty stuff on the forum at the moment for him. I'm having fun with structural design and AI, adaptive and generative optimisation design for my two new machines. Looking fwd to using Dynomotion on the upcoming machines... I'll have to brush up on my 35 year old C programming.... Peter