I recently came across an application where the engineer wanted to automate the positioning of a blade that cuts a part as it comes down a conveyor at 140 ft/min. Thankfully, the blade and its mounting block already had a round rail bearing for guide support. See Figure 1 for an illustration. Several conditions made this application difficult:
- No current way of tracking the part; part position is unknown.
- No consistency as to when the part would come.
- Harsh environment; very dusty and oily.
- We had 14” to make the cut.
- Blade weight (includes mounting block) = 50 lbs.
- ±0.001” repeatability. Quality of the cut was extremely important.
- Part had to be cut on the fly.
Figure 1: Front View
Figure 2: Top View
To recap, the part position was unknown, we had to cut the part on the fly, and do it all with ±0.001” repeatability! Of course, the customer wanted the application to be as cost effective as possible. Talk about one of those solutions where it has to do everything for nothing!
Two possible solutions were eliminated instantly. First, vision was going to be way too expensive and time consuming to even consider. Lighting would have to be setup for the machine and adjusted often due to the dusty, oily environment. Second, pneumatics would not provide a solution that was either fast enough or have enough control.
The first route that we went down involved tracking the part with sensors and an encoder coupled to the conveyor. The theory was that we could sense the part with an inexpensive sensor that has a very narrow window. Once the part had been detected, we would cam a ballscrew actuator to the encoder. This is a classic flying shear example. The actuator would be required to move the payload at a maximum velocity of 1.42 m/s and a maximum acceleration of 5.69 m/s^2. However, there were several problems with this.
- If the part slipped on the conveyor, even in the slightest, the blade position would be off.
- There was too much latency in the system. The time that it would take for the sensor to detect the part, send the signal to its amplifier, send the signal to the controller, have the controller (with a cycle time of 5 ms) update the input and calculate the move, and then move the actuator was too long. It sounds like a lot, but all of this would happen in approximately 5.1 ms. Due to the latency in the system, we would have a worst case scenario of ± 0.075” repeatability.
- Only a select few manufacturers build an actuator that can position a 50 lb load with ±0.001” repeatability.
So, it was back to the drawing board! Our second solution hit home. It was determined from our first try that we would have to use mechanical contact to achieve this kind of repeatability on the fly. So, how do we track the part mechanically without leaving a mark on the part, slamming into end stops, and not move the part on the conveyor?
We had the idea that we could let the part push the blade into position. However, this is not possible with a ballscrew because a ballscrew actuator is not backdriveable. The conclusion was that we needed a linear motor instead of a ballscrew actuator. This way we could actually let the part come into contact with a piece of metal that would be attached to the mounting block of the blade (See Fig 5)
This way we would have the position of the part and the blade. Basically, we would put the linear motor into freedrive (disable the drive), let the part come into contact with the contact piece and let it push the blade into position mechanically. However, there was a problem.
We couldn’t just let the part slam into the contact piece. The difference in inertia between the part and the static load (blade mounting plate and blade) was too much. At 140 ft/min, when the moving part came into contact with the static load of the blade, it would damage the part. In order to solve this problem, we determined that we would have to begin moving the blade before the part got there and let the part run into the contact piece (essentially matching the velocity of the two objects).
Another problem arises in this situation. How can we maintain control of a linear motor, but at the same time, let the part push against the actuator without faulting the drive? The answer was to increase the maximum amount of following error on that axis. As long as the part was in contact with the contact piece, we did not care about following error. Almost done!
- Linear motors are expensive.
- The environment was very harsh with a lot of dust and oil.
- The stage of the linear motor would potentially experience very large moments due to the high acceleration and weight of the load.
Our solution was Parker Hannifin’s ETT tubular motor. In principle, it operates the same as a linear motor, but in this case the magnets are embedded in the rod and the stator is the body of the actuator. This application was perfect for the ETT for two reasons. The first is that the ETT is IP67 rated. This is beneficial because this actuator could be dropped into this environment as is. The second is that the ETT could push the load from the side. With the support of the linear bearings (already there) the rod would see little to no moment.
In the end, we had a solution that works as follows:
- Part is sensed with a through beam.
- Blade begins moving (custom trap move) with calculated acceleration and max velocity.
- Part runs into contact piece.
- Blade cuts the piece as it is moving.
- Contact piece is mechanically moved so that part can go by and actuator begins decelerating.
- Blade is moved back to its home position, ready for the next part.
This was a really cool solution because it involved some creative thinking about how we can manipulate the components that we carry at Cross Company and use them to solve problems. It would have been easy to walk away from this application because we did not have an easy solution right off the bat, but, then again, the best projects never do. This is just one application that proves at Cross Company, ‘We Make Motion Work!’