Use robot for plasma cutting instead of gantry machine ?

Never occured to me to use a robot to plasma cut until today when I saw a video of such online. Seeing as older robots can often be purchased way cheaper than typical gantry type CNC plasma cutters...just wondering... I guess the main advantage of the gantry type would be more size capacity (unless the robot was a really big one) and possibly easier programming since you are only concerned with 2 1/2 axis really. Thoughts ?

A gantry CNC plasma machine includes a bunch of stuff you would have to add to the cost of the robot- machine torch and interface to turn it on and off, voltage sensing for auto torch height, water table, and sometimes fume extraction.

Voltage sensing torch height feedback is pretty much mandatory for production, as the metal being cut moves up and down as its being cut, so a preprogrammed height aint gonna be enough.

So what the robot would be replacing would be about 1/4 to 1/3 the cost of the overall setup, I would guess. You would save on the actual rack and pinon system, and the motors and motor controls, but you would have to reinvent the feedback system, torch on and off, air on and off, and interface electronics, all of which are off the shelf, already troubleshot systems in a Gantry table.

So, I guess if you can get a complete Robot for ten grand or less, and then are an electronics whizz, you could build your own table, design your own interface, and write your own software, and save some money- but my guess is that the subset of potential plasma table buyers who can do all this, know how, and have the time, is pretty small.

Another one of those "theoretically possible", but usually more trouble than its worth type things, which works great for garage geniuses with no overhead and no actual work to do daily.
Me, I need a machine that a 22 year old employee can run, that I can order parts over an 800 number, and will run 359 days a year.
Which my gantry table has, since 1992.

Ries said:

So, I guess if you can get a complete Robot for ten grand or less, and then are an electronics whizz, you could build your own table, design your own interface, and write your own software, and save some money- but my guess is that the subset of potential plasma table buyers who can do all this, know how, and have the time, is pretty small.

Click to expand...

I was hoping that the same specialized electonics, interface and software for robot mig welding would work equally well, or nearly so...for plasma cutting. And seeing as robots set up for welding sell pretty much as cheap as ones set up for manipulation, I was hoping one could just buy a welding robot and replace the weld power supply and mig gun with plasma torch/plasma supply. But the actual cable connections could be tricky unless a robot was designed with potential double duty from the get go....guess I'll contact Miller to see what they have to say about it.

There are many applications for plasma on robots......and plasma systems can be rather simple to install on an industrial robot...most of which have I/O cpability that works well with the plasma start and arc transferred signals that are necessary.

Of course torch height control can get a little complicated.....the best robot/plasma integrators monitor plasma arc voltage and use that to adjust the torch to work distance through the 6 axis of motion.

Another difficulty with robots is with offline programming software.....which is hard to use, or non existant. Robots usually have to be "taught" a cut path.....and work very well in an environment where you always cut the same parts. It is not quite as simple to dowload autocad files into a robot controller....as it is to import these types of files into a gantry style cutting machine.

Jim Colt Hypertherm

Jim colt nailed it, besides the torch height control. The real difficulty would be in accurate and repeatable programming of the weld robot to cut complex 2d shapes with any real tolerance. Software is generally vendor specific and I've never heard of a welding robot that could automatically convert 2d cad files into 3d controller information. That being said, there is a lot of work industrial work with automated cutting operations, including lasers and abrasive water jet cutters, I don't see why it would not be possible to convert such a system to plasma arc cutting. But the general application is for automotive or occasionally piping.

Gantry systems are designed specifically for accurate and simple 2-3 axis cutting operations. Using a 6 axis robot seems like the long way around the block to produce a 2d cut. Not to mention that any system will generally have dedicated CAM software that will calculate and control torch height, travel speed, etc based on the cutting profile and extrapolate the information from CAD files.


I don't doubt it's possible to plasma cut with a 6 axis welding robot but it wouldn't be more cost effective in the long run against traditional gantry style beds.

In my years with Hypertherm I have worked with hundreds of Robot/Plasma applications....only 1 was for a flat plate bevelling application.....all of the rest were for 3 dimensional jobs such as trimming excess material from forming operations (car door skins, motorcycle tank halves, etc) as well as tube and pipe coping and trimming....and casting flash trimming.

As I said earlier, robots are very good for repetitive jobs as the cut path teaching process is rather time consuming. If someone eventually develops a good offline software that can easily develop cut paths for a multiaxis robot....for both 2d and 3d cutting applications then we will likely see much more use.

One of my favorite applications with plasma and robots involved 9 robots. 1 large robot was used to lift and position a 40' piece of pipe (for an irrigation system)....and the other 8 robots each had a plasma torch and a mig gun, the plasma would cut a hole and the mig would weld a bung in the hole (the robots also had a custom gripper for the bung)....then the positioning robot would move the pipe to a hot dip galvanized tank. I worked o this project about 15 years ago with a major robot manufacturer.

Jim Colt



Jim

I was doing some work out at Oregon Steel Mills and they had a plasma robot from Hypertherm Automation. Really neat machine. It was used to cut out sections out of the pipe they made (Spiral formed and welded seams) for metallurgical testing. You would set a section of the pipe in front of it and it would cut out various sections. I want to say it was hooked to a 400 amp plasma cutter I think a HT4400.

Keep in mind that I'm incredibly new to WAF - but wouldn't the real issue with the robot be converting the language to run in multiple axes? Thus, not having the software to do it for you...


It seems like any of the plasma control software has some form of arc length compensation, and that's nothing more than an offset adjustment in what's typically Z (by definition, it has to be Z). Well, you'd write your program to add that compensation to whatever axis was functioning as Z and then include the traditional offset compensation input. This gets tricky when Z is actually traveling about an axis in a circular ineterpolation.

Writing the code, you'd just keep in mind what direction the torch was aimed when inputting said compensation command, and you're no longer bound by the 2 1/2 limitations.

Arc on/off, air on/off - simple M code assignments; you're dealing with a function that only needs a do or don't input on the line proceeding the motion to follow. Communicating this with the plasma power source should be trivial. Here again, it's done in all the current gantry software.

Now keeping your head around programming 6 axes about a 3D object could be enough to drive you insane. The flip side is that it's been done for you with the software that runs gantry style 5 axis systems. I don't think it would be as hard as some here might suspect to replace the gantry with the arm, so long as the arm operates in the same planar space as we're accustomed to programming from.

I'm curious to hear what Miller has to say about this.

Jim, I was talking to a head engineer over at wolf robotics. They gave me the name of the software -which I now forgot- That they use for all offline programming, including remote programming. The modern CAM software works in conjunction with a CAD model of the part to perform reach studies as well as program work paths without the traditional "teach" method. The end user calibrates the software against the real robotic cell and then they can perform all the "Teach" operations in software.

It was an interesting system that was used by several large manufacturers including Cat and John Deere for robotic Laser cutting and laser-hybrid welding.

*edit*
Found a link http://www.wolfrobotics.com/products/options_toolbox.htm
robot studio is the name of the software. Still though I don't think it's practical to use a six axis robot for 2d shape cutting operations due to the calibration and programming difficulties

I totally agree with you that it's silly to use a robot in the place of a gantry when all you did was cut plate, but if you were able to achieve the travel rates and maintain the tolerances on the work and regarding the torch height control, it seems like you wouldn't have a reason not to once initial cost was resolved.

What milacron is talking about makes complete sense to me. Take a robot out of a welding cell, and plunk a plasma torch in it's hand. Now you can cut spirals out of pipe for the same money as a gantry would cost new.

So this begs the question, How tight of tolerances can a robotic arm hold while working in 2D?

Tolerances of course depend on the capability of the robot mechanics and the robot controller. A robot that can weld well cannot necessarily cut as well. Plasma needs constant speed and very fluid motion....no vibration or speed changes. Also many robot controllers do not have the ability for circular interpolation.....so holes and radii are a bunch of short straight lines. This is from my past experiences with older robot technology...and I know the newer equipment has much better motion capability.

The one application that I worked with for flat plate cutting actually is at a Caterpillar location....and it may well be using the Wolf software. It is plasma beveling (weld prep bevels) flat plate. The robot is "wall" mounted (90 degrees to the horizon) on a moving gantry to allow for more travel. It has been running well for over 18 years.....although today there are large gantry plasma bevelers that can cover a lot more area more effectively, and with better accuracy.

Jim Colt

I still say that for 99% of flat sheet cutting, which is what most gantry machines are doing, a gantry machine will be cheaper, more reliable, easier to program, and use off the shelf parts.

So I guess what I am trying to say is that Milacron will not become a Millionaire by finding a market niche nobody else thought of, and selling cheap used robots to companies that otherwise would have bought a gantry machine.

Yes, you could kludge one into working, but my guess is that if you are that good at software, systems integration, and robotics, you can get paid a LOT more money than you would save doing this by just staying at your day job.

I mentioned this to one of my instructors and his thoughts are similar to mine. Once you're dealing with the model in CAD, the software for the robot control will take over for the practical application.

The increased operating cost issue isn't without merit, but there's a functional envelope threshold you're also nearly obliterating by using the robot - you're no longer locked into 2D.

Whether or not your application warrants the robot is likely the deciding factor here. I also doubt the robot could compete with the rapid ability of a well built gantry, and I'm not sure about the sustainable travel speeds either (even spray welding is pretty slow going compared to plasma cutting thin stock). However, there's nothing you can do to a gantry to make it follow a circular interpolation around X or Y.

All said, I wouldn't just write it off.

Jim Shaper said:

I mentioned this to one of my instructors and his thoughts are similar to mine. Once you're dealing with the model in CAD, the software for the robot control will take over for the practical application.

The increased operating cost issue isn't without merit, but there's a functional envelope threshold you're also nearly obliterating by using the robot - you're no longer locked into 2D.

Whether or not your application warrants the robot is likely the deciding factor here. I also doubt the robot could compete with the rapid ability of a well built gantry, and I'm not sure about the sustainable travel speeds either (even spray welding is pretty slow going compared to plasma cutting thin stock). However, there's nothing you can do to a gantry to make it follow a circular interpolation around X or Y.

All said, I wouldn't just write it off.

Click to expand...

Travel speeds on newer robotic systems can be well over 100 IPM and well over 180 IPM in rapid modes. with the tight tolerances and hybrid processes it's not uncommon to see welding in the 40-80+ IPM range.

although I still don't think it would be practical, there already is a large precedent for robotic cutting applications in the automotive industry and they generally use laser cutters for excess material removal and for removal of flash or projections.

The real issue here is the difficulty in calibration and tool path programming for 3d cutting, the auto industry has long production runs and the available hardware and expertise to justify robotic cutting.

I think the programming standpoint still exists for robotic cutting in 2d applications compared to similar gantry systems which were designed specifically for 2d shape cutting operations. I have never heard of any software developed for this specific purpose but I'm sure it could, but most robotic systems can only maintain a 1-5 thousandths tolerance which is far surpassed by most gantry systems.

some of the newer cutting machines are also going to a 3-5 axis system with tilt heads to follow part profiles that are not flat. http://www.youtube.com/watch?v=DoB_qQEeRW0

but again it adds more degrees of difficulty in programming and maintaining tolerances. Really it boils down to the fact that there's not that it's not economical to produce 3d cuts on a "one off" or small production run basis. Although technology has a way of overcoming those types of things

Plasma Cutting Robot

Here's a system that married a Hypertherm robot to industrial robot. Beauty of it is not having to program the robot . . . feed it a dstv file (standard file format for structural steel detailing) and , I think, Autocad file, and it programs itself.

Little pricey though . . .

visit plasma-cutting beam drill line

It seems that Space and Angles might be the advantage for a robot.

A big gantry machine is, well, big. And when it's not cutting, it's still big.

A robot could, one imagines, be designed to fold up into a corner.

Of course this presumes cutting over the concrete floor like I do by hand...

The Python arrangement looks appealing because of its "it can do anything, almost" attributes.

A robot would work for this, provided the tolerances are within the robot's capabilities. Repeatability on a FS10E Kawasaki is 0.1 mm at 1480 mm reach WFO. However, the tolerance gets worse the bigger you go on robot size.
Circular interpolation is pretty much standard stuff on anything 10 years old or newer.
2D cutting would not be an issue as you have linear interpolation available.
As for programming the path, the controller can't do it for you, but CAM systems should be able to make a post for you (I would think, no CAM experience in many years here).
What you do is set up a fixed reference point, not unlike a CNC machine.
All program positions are referenced to this reference point (just like a CNC) and are done with a "shift" to the reference position.

Example:
ref = Reference position (stored, taught fixed position within the work envelope of the robot)
For this exercise, we will assume the reference point is the front left corner of your table. The variable tcht = torch height and comes from a calculation program that uses whatever is required to achieve the correct torch height needed.
To cut a X= 150 mm x Y= 160 mm square 300 mm in X and 400 mm in Y from the left front corner of the table you would:
.PROGRAM torch ();
CALL calc
POINT #ref = ref
POINT point1 = SHIFT(ref by 300,400,tcht) ; First corner - start
POINT point2 = SHIFT(ref by 450,400,tcht) ; Second corner
POINT point3 = SHIFT(ref by 450,560,tcht) ; Third corner
POINT point4 = SHIFT(ref by 300,560,tcht) ; Fourth corner
HOME 1
JMOVE #approach
JMOVE point1
SIGNAL 10 ; Turn torch on
DELAY X ; Delay for heating
SIGNAL 11 ; Turn gas on
ACCURACY 0.1 ALWAYS
SPEED 80 MM/MIN ALWAYS
LMOVE point2 ; Cut first side
LMOVE point3 ; Cut 2nd side
LMOVE point4 ; Cut 3rd side
LMOVE point1 ; Cut 4th side
SIGNAL -11 ; Turn gas off
SIGNAL -10 ; Turn torch off
SPEED 100,100 ALWAYS
ACCURACY 150
JMOVE #approach
HOME 1
.END

Programmed in Kawasaki AS Language - which is based on C++ (And many, many C++ commands actually work in it, even if they aren't listed in the manuals)

Now, the shifted dimensions can be variable names instead of hard numbers, and those can be pulled from a separate file, calculated, or manipulated in many ways, including direct input via several methods.
To do this, you'd basically write a macro for the robot main program and feed it variables.

The best solution with a robotic application would be a standard G-Code solution with tool tip path generated by something like MasterCam or equivalent.

The biggest challenge here is to solve the kinematic relationship between the robot joints / linkages and the Cartesian / Polar coordinate geometry. This solution needs to include both the forward kinematics (for path generation) and reverse kinematics (for homing / power up position awareness).

We have done this with a number of robot types - mostly SCARA robots where end effector position is defined with X, Y, Z and orientation is defined using A, B, C angles.

This solution involves abandoning the original controller and going with a 3rd party motion controller and then using standard G-Code to define the tool tip / end-effector path.

Toolpath velocity can be programmed and is only "clipped" when a robot joint velocity or acceleration limit is reached.

Accuracy is treated from two different perspectives - path accuracy with respect to programmed trajectory - this has to do with interpolation limits and is typically when look-ahead is parameterized to define 3 dimensional motion from a limited set of specified pathway points. Next is actual robot mechanical accuracy - defined by feedback resolution + servo loop performance + mechanical stiffness. Most often, you can set up a laser tracker with a target in the tool holder and sweep through the operating envelope of the robot and put together a 3D compensation table - given this comp table, you can typically achieve decent accuracy, well within the cut tolerance of the typical plasma cutting torch.

I’m resurrecting an old thread, but one which is now meaningful to me since recently acquiring an old, but in like new condition, Motoman K10S with two external rotary tables. I was in the market for a small footprint (4x4) CNC plasma system when I stumbled upon the robot at an attractive price point. Currently I’m exploring options to calibrate the machine and hoping someone here has done something similar during the last six years since the original posting.

Tonytn36 has described exactly what I’m planning with Inventor + InventorCAM (SolidCAM). I envision a two part post processor where the InventorCAM post is nothing more than a trace log of all called functions (@arc, @line, etc.). The custom processor will parse the log file and use the external calibration data (per Motion Guru post) to generate the necessary Motoman INFORM program.

The unit came with a 6-axis adjustment card (3 for path correction / 3 for shift adjustment) which can be used for either arc length (TIG) or height control (plasma). I’m not sure how this will work with the default controller since all the calibration data for driving the robot is external to the interpolation routines used by the Motoman controller. My first guess would be to add additional teach points when generating the program. For instance when processing the @line command to move from (X1,Y1) to (X2,Y2) I would use a maximum step to generate many secondary points each with (X,Y,Z,Tx,Ty,Tz) based on the calibration data.

Please let the ideas flow as I’m looking forward to the challenges this project is sure to create.

I think its been done for 25 years at least. The first application I saw was from James Camerons aliens where they used a fanuc robot to plasma cut the door of a space ship in the very first 5mns of the movie.