[Jammin]

Hi all,

I'm an engineering graduate, new here, and I'm interested in parallel kinematic machining. On paper at least, it seems to be an elegant way of improving dynamic performance of machine tools as well as improving accuracy and stiffness.

I know there have been a few conversations in the past around here about hexapod machining in 6 degrees of freedom (6DOF), and I'm aware that some companies have even made some (with varying degrees of success).

Does anyone here know anything about simpler 3DOF non-cartesian mechanisms? I'm thinking of something more like a constrained 3-axis delta robot with dual arms which keep the tool pointing downwards. This would be driven by 3 actuators which would make control much simpler and the mechanism cheaper.

One day I'd love to design and sell one of these, so I'd be really interested to get your thoughts on which markets this kind of thing would suit. For example, if I made a lightweight desktop system that could compete with a Roland on accuracy but at 5x the speed, who would want it?

I hope this can be an interesting conversation and I look forward to hearing your thoughts.

[LongRat]

This is a very nice way of making a machine tool. A constrained mechanism to give 3 translational axes of motion from 3 actuators is exactly what I work on for my day job - see http://www.renishaw.com/en/incise-sc...oftware--10249. This is a machine for milling dental crowns etc and has a small volume but is extremely accurate, not going to cut any tough metals though. The actuators are friction roller drives, so backlash is non existent even after the drives wear.
To get motion accurately in X,Y,Z directions is more difficult than in a cartesian machine (even after you've applied the kinematic transforms) because tiny length differences in the fixed-length struts cause 'dishing' in the motion and divergence from linearity. To correct for this we do 3D volumetric error mapping, i.e. we use the machine to probe the locations of a large array of accurately mapped balls (previously measured on a CMM). Depending on the accuracy you are looking for, you may not have to do this. Our machine is tiny (roughly spherical working envelope about 90mm diameter) and without error mapping can be out by 0.3mm or more. With error mapping it is accurate to sub 10 microns.

[Jamison]

Nice renishaw machine LongRat.

I've been working on deltas, tripods, hexapods for the past couple years for Delta Tau. I have a couple decent vids on youtube
[nomedia="http://www.youtube.com/watch?v=lkbFCVcRP88"]YouTube- Fast 140ppm mid sized delta robot[/nomedia]
[nomedia="http://www.youtube.com/watch?v=Gv5B63HeF1E"]YouTube- Mini Delta Robot with background noise to verify actual speed[/nomedia]
[nomedia="http://www.youtube.com/watch?v=bQtT4EDyQVU"]YouTube- Delta Tau's Brick controlled Hexapod[/nomedia]

These mechanisms are optimal for loading that results in axial stresses... tension/compression. They are essentially a moving triangulated truss... strong as hell if loading stays tension/compression. Keep these things axially loaded and they are fast, light, rigid as hell, cheap and easy.

Tool clearance is the fly in this soup. Point loading a delta geometry mechanism tooltip (that is cantilevered 2"-5" from the end effector plane for clearance) transmits big torsion to the shoulder cranks. Depending on your preferred geometry, the benefits of a PKM mechanism begin to go away in a hurry. If you compensate with mass, it might as well be a knee mill...

I stopped compiling a list of industrial PKM machine tools a few years ago... so many have tried, only a couple Tricepts and Toyoda mills remain.

[LongRat]

Ah, those are your videos. Really impressive, seen them before.
I completely understand your concerns about the directionality in stiffness with these designs. In our niche market machine, we are using long, thin tools and soft materials so the cutting forces are very low and it doesn't cause a problem. If you want to compete with Roland you'll need to cut aluminium, perhaps with tools up to 6mm diameter or so. I'm no expert on Roland's machines though.

[Jamison]

@LongRat, I'm embarrassed, my post appears directed towards you... but was for Jammin's benefit... I suspect you could teach me a thing or three about PKM machine tools. The Renishaw machine is very nice... the mechanism moves the part or the tooling? Controls code must be amazing.

Now if I could get my PKM machine tool collection (pdf file) to attach... it will demonstrate just how many have gone before and failed. PKM machine tools are truly the domain of the masochists.

http://www.ecuriechotis.com/pkmgallery.pdf

[LongRat]

The mechanism moves the material while the tool is fixed.
Before it became a commercial product, the machine at one stage had two 200,000 RPM spindles and a third running at 65k, each pointing in a different orientation. 4 material index positions and a built-in scanning probe. That generation of the machine really was complex. Now we have only 1 spindle and a touch trigger probe rather than a scanning probe.
There is room for PKM machine tools in the market, just maybe not every sector of it. Our machine is commercially successful because it is cheap and extremely small, which is what the dental market demands.
Nice PDF by the way. I've never heard of most of those machines. I guess that would be because even all together, they would have such little presence in the field.

[Jamison]

That video of the Renishaw is so cool. Great little machine, and a demonstrator of PKM benefits applied. I'm jealous. I'll admit it. Marketable applications for PKM machine tools are so rare. Bravo!

[Jammin]

Good work, LongRat! And thanks for the PDF Jamison. This is really interesting and you guys really know your stuff.

So why do you think so many PKM machine tools have tried and failed? Is it because of the compromised performance, or do you think it's a perception thing? I'd love to hear your thoughts!

[LongRat]

I'm really not sure of the answer to that. I am not an expert on the industrial machine tool market, but maybe marketing has a lot to do with it. I would think if some of the major players never developed a PK machine they would strongly discourage potential customers from investing in one.

[Jamison]

Controls for PKM's have only come around lately. Kinematics for deltas are now common, but forward/inverse kinematics for many are very complex and require serious computing power. Investment in the controls side is still a heavy one.
Then there is the compensation for non perfect mechanics... on a typical mill, it is a constant (or linear) compensation on the X axis for compensating X error, Y axis for Y, and Z axis for Z. PKM's will require that the kinematics be involved in the error compensation... where linear or constant is not the norm. The kinematics can be adjusted to match the real link lengths or axis locations, or brute force3D error mapping can be used to interpolate/extrapolate position.
Lastly, there is the support side of the machine equation. If you have ever had a machine go down... you know the story. Not many techs would be comfortable troubleshooting a machine where motors can't be jogged (rigidly coupled PKM's like the hexapod) or where unintuitive algorithms add complexity to input->output troubleshooting.
So in a practical sense, all levels of PKM machine developing and marketing require a significant infrastructure... we aren't there yet.

[Jamison]

1

[Jamison]

Good observation LongRat... I have a year old Fanuc VP interview where he rails against the "Spider robots" PKM's.
IMTS last week brought a zoo of new Fanuc PKM's... lol.

[dcartier]

Does anyone here know anything about simpler 3DOF non-cartesian mechanisms? I'm thinking of something more like a constrained 3-axis delta robot with dual arms which keep the tool pointing downwards. This would be driven by 3 actuators which would make control much simpler and the mechanism cheaper.

There is a PKM design that meets most of your criteria. It offer 3 DOF, though it also has Cartesian control, which was not on your requirements, but it does make it much easier to use with all types of software. It also very simple and cheap components as it requires only 1 DOF per joint.

The name of the PKM is the Orthoglide and was developed in a university in France and Montreal Canada. You can see it here Orthoglide: A 3-Axis Parallel Machine Tool for High-Speed Machining.

I had planned on building one of these at some point. Time will tell if I ever get around to it.

Dennis

[Jamison]

There is another version of this at UC Riverside... interesting how the 3axis PKM (Delta) can have so many different types of geometry... Festo R&D has an optimized geometry for a 3DOF PKM as well.
Unfortunately, the challenges remain the same. All three actuators are involved kinematically in jogging in any one axis. Torques from a tool tip load propagate through the system proportional to the distance the tool tip is away from the common end effector plane. Calibration will continue to be a challenge, not as bad as a rotary actuated delta, but still will have several variables.
this would be a fun project Dennis, keep us posted.

Jamison

[Jamison]

Just compared the Orthoglide to the UC Riverside version called the UCR in the PKM gallery pdf found in this thread.

The Orthoglide axis are simply coupled, but the UCR is completely uncoupled. So the UCR is in fact a cartesian PKM. Simple calibration, simple offsets, simple jogging... and simple CAM posts!
Interesting. There is some monster bending forces on those cantilevered arms.

[samco]

Emc2 (free and open souce) has support for kinematics.. How about this style?
YouTube - hg5bsd's Channel

sam

[Jamison]

Great PKM work samco. Linear deltas (tripods) are a great trade off of speed and footprint for rigidity and simplicity.
It isn't often one guy has that much software and mechanical ability.
How do you calibrate? I see your Z is pretty good! Do you error map or calculate actual link lengths and axis locations with tool tip data points?

Jamison

[samco]

As much as I would like to take credit for it - I did not make it. (I am just a EMC2 lover - yes I said it) Maybe you could ask though youtube..

sam

Someday when I have some time - I want to build a hexapod

Great PKM work samco. Linear deltas (tripods) are a great trade off of speed and footprint for rigidity and simplicity.
It isn't often one guy has that much software and mechanical ability.
How do you calibrate? I see your Z is pretty good! Do you error map or calculate actual link lengths and axis locations with tool tip data points?

Jamison

[dcartier]

There is another version of this at UC Riverside... interesting how the 3axis PKM (Delta) can have so many different types of geometry... Festo R&D has an optimized geometry for a 3DOF PKM as well.

The version at UC Riverside looks like an implementation of the Tripteron. You can see this and a related PKM, the Quadrupteron at Robotics Laboratory - Tripteron and Quadrupteron.

My concern with these was that they would require angular contact bearings to handle the side loads perpendicular to the axis.The Orthoglide removes the need for these types of bearings by only presenting parallel loads along the axis.

Dennis

[Jamison]

Good point.
One less joint/link is always a good thing mechanically, but adds a level of complexity in the controls... which is a good thing.
I always prefer controls complexity over mechanical complexity. Another couple lines of code seems to be cheaper and easier to manage than range of motion, deflection, and backlash problems... code revisions tend to take less time and resources than mechanical revisions as well.

If I were to choose an architecture for a mill, I would choose the linear delta similar to the DS Sprint Z3, from DS Technologie, Germany.