High Speed Machining (HSM) for 3d printing

[SkyKorp 4]

Back in my days of CNC milling all the rave was about High Speed Machining (HSM). It meant that a typical machine tool could now go twice as fast with the same part quality. It was huge for large scale manufacturing.

It worked like a sausage grinder for GCODE and transformed them using new acceleration and deceleration and feed rates where needed. It was huge for small business owners and it took the industry by storm.

It's old tech today, and standard practice now.

What if a UM2 could go 3X faster?

I watch my UM2 and wonder as I play with the speed. I can see that there is a limit, and you can't push it too far without consequences like overruns, jerks, or processing and hiccups. This is what HSM solves. There are many ways to do this, and I could not find anyone playing in this technology for 3d printing. I feel that it would dramatically increase the speed and smoothness of the print and machine operation. A game changer.

I have some ignorance as I have yet to study the controls for accell/decell and other limitations. I am looking to learn and have good discussion.

Edited September 20, 2015 by Guest

[NonSequor 1]

If I understand what you're saying, the deal is that you when you change speed you want to retune other parameters as well to get results comparable to the speed your more familiar with.

Basically, all of your settings inputs form a configuration space. The problem is that this space is very large in terms of number of dimensions, and it's hard to tell how the parameters interact. If we want to move into the higher speed portion of the configuration space, what other things do we need to change? It seems like changes in settings are largely propagating by trial and error relative to each hardware design. To work out how to push into the higher speed territory, I suspect you need a more organized exploration of the configuration space.

I'm guessing that for CNC, the manufacturers did exhaustive testing relative to a small number of reference runs on mature machine lines. Meanwhile, most hobbyist 3d printer manufacturers are less than 5 years old and are in their second or third iteration. I would expect most of the manufacturer driven R&D effort to be dedicated to hardware design for now.

But maybe there's something you could do to push into the higher speed configurations using crowd-sourcing rather than paying people to print the same 3 models every day for a year with different settings.

So let's say you rate your print quality on a scale from 0 to 1. For each point in the configuration space, we'll say that configuration has an average print quality somewhere in that scale. For the moment, ignore the fact that quality may be both subjective and multidimensional. We have to run roughshod over that to make the problem manageable.

Let's say that we have Cura collect a rating from users for the quality of their print job and then submit that rating to an online database, along with the settings used and maybe a hash of the source STL file. Using the database, we want to look for a couple of different things:

(1) Are there points in the configuration space far from the default settings (particularly in speed related settings) where there are a large number of user reports of high quality prints?

(2) For settings very close to the defaults, what directions can you move where the change in quality is 0 or positive?

Over time, you want to come up with ways to change the defaults to reflect good reported outcomes (which also pushes more people to look at the region surrounding the new defaults) and also to try to come up with a lower dimension subspace of the configuration space which gives good results (i.e. when you ramp up the speed, other settings are changed appropriately).

[SkyKorp 4]

No not really what I am talking about. You post process the Cura generated GCode file. Run it through a parser that can read ahead, this breaks the motion up with slower feedrate around a corner. You can set the incremental space to insert a slower speedrate. Instead of letting the machine spend time reading ahead with buffers, this is cheap but very good solution to older cheaper printers.