20 X 40 ft gantry robot

We had a unique request a while back to build an inexpensive gantry style robot. The unique part was that the work envelope was to be 20 feet by 40 feet!

Our customer (who does not wished to be named here) had a requirement to etch large sheets of glass. They are basically abrasive blasting cooling channels in huge sheets of glass that are later laminated to one another. The key to the project was that it needed to be reasonably priced. This was one of those very rare applications where the accuracy of the system was not all that great. Generally, if we were within about 0.0625" (1/16") we were good.

All of the systems that they had looked at previously were extremely expensive. Most in the several hundreds of thousands of dollars range. Many of the systems boasted of 0.0001 inch resolutions and 0.0005 accuracies. Way overkill for what they wanted.

Most of the mechanics of the proposals were also very expensive. On many of the systems, the gantry was driven from only one side. The mechanics of overcoming the moments of inertia for moving the gantry (remember it is 20 feet long!) were absurd.

After spending some time with the customer and listening to what they wanted instead of trying to sell them what we had, we came up with a pretty simple system that more than met thier needs.

The first requirement was that this be cost effective. There may be more than one of these systems and the industry that they represent is very competitive so they were very sensitive to cost.

A big key to the success of the system was that we heard the customer say that they only needed about 1/16" accuracies. There was no reason to try to make a machine that was stiff enough to achive 0.0005" accuracies.

Instead of trying to find precision linear bearings 40+ feet long (at $250 a foot) we chose some heavy duty roller track at $25 a foot. We needed about 120 feet of track total. This was a $27000 savings alone!

Because the span of the machine was so wide we built an aluminum truss frame for the gantry. The overall width of the frame was 22 feet. For the prototype, we decided to build it out of extruded aluminum framing (80/20). This allowed for lots of fine tuning. The plan was that for subsequent machines, they would be welded aluminum. But in the end, the profiles were a better option even for subsequent machines.

To get all of this moving we used an inexpensive servo on both sides of the gantry to move it in the 'X' direction. On the reference side, we pinned the gantry to the truck that it sat on. On the other end, we mounted the gantry on a short linear bearing and pinned that as well. This allowed the gantry to be a few degrees out of square without binding. (We found this to be really handy in fast traverses.)

When the servo motors were mounted to the machine, we made a point of having the index pulses from the motors very near each other when the carriage was perpendicular. We let the software do the last bit of fine tuning as to exactly where the index pulses should be. While technically not needed after the machine was homed, this gave us a sanity check every inch of travel. We also installed a pair of small prox switches that would indicate if the carriage was grossly out of square.

All of the motors are brush type PM motors, with off the shelf amplifiers. The whole control system in a PC computer.

There are a total of 4 motors (two X's, a Y and a Z) and 20 IO points. All of this is put together with commercial off the shelf components.

The software that runs the machine is written in C and very straight forward. To define the pattern that will be etched in the glass, an ascii file that is based on the HPGL plotter language is created.

Another file was created that has corrections that compensate for the reference track not being perfectly straight. Once the machine was assembled in it's final place, we mounted a laser at one end of the table and shot it down the length of the table. We moved the cutting head around the table and noted how far off it was at various points on the table. This was captured in the correction file and used to correct the positions during normal operation.

When we were finished, we were able to move to any position on the table to within about 0.005", still about 12 times more accurate than we needed!

One of the surprising benefits of our mechanical design, was that during high speed traverses (not cutting) we were able to loosen the following accuracy of the slave side of the X axis to its master. Depending on where the Y axis is at the time the X axis wants to move, the inertia on that axis changes quite substantially. Instead if trying to keep every thing perfectly square during a traverse, the carriage is allowed to be as much as 1 inch over 20 feet out of square. As soon as it reaches it's destination, it squares itself back up. We were able to bring the traverse speeds and accelerations up to the mechanical limits of the machine.

After we built the first machine, the customer went on to build several more.

The overall accuracy of the machine was about 0.020" still about 3 times than what was required for well below half the cost of other proposals.

At last check the machines have been running for several years without interruption.