[Cliff567]

We recently had a retrofit done on a Horizontal Milling Machine. I do much more electrical work in the plant, I’m not a mechanical person, so please excuse any mistakes. Here’s the best description of the machine I can give:

The machine has an opposing torque anti-backlash Rack & Pinion system. There are 2 motors (about 25KW each), 2 gearboxes, and 2 pinions driving 1 rack attached to the table. The motors drive in equal but opposite directions to create a tension torque on the rack at stand still. As I understand it, the theory is if both start out with equal torque in opposite directions and then you apply more torque to one and remove an equal amount of torque on the other the table will move in one direction. If you reverse the roles of the motors the table will drive in the other direction. Because of the opposing torque, there should be little or no backlash in the system.

One question I have is how should a system like this take into account the weight of the table, friction, and things like that?

Is there any literature available about this type of system?

Does anyone have a system like this that operates correctly?


Thanks,
Cliff

[steve323]

Sounds horribly inefficient. But, I am not a mechanical engineer either.

Steve

[harryn]

This is a good way to reduce backlash in a rack and pinion system. It also helps hold things "still" when machining as there is a substantial holding torque. Pretty creative really. I have no idea how common of concept this is.

Another way to obtain similar results is to put two pinions on one rack, with each sort of "sprung" apart in opposite directions. The challenge of this approach is that rack often isn't all that thick, so getting two pinions on there is not all that easy. It also limits you to the tooth strength of one "tooth".

Is this a commercial machine or "custom". Probably a brand name will help you find more info on line. I have good luck using dogpile.com but there are a lot of search engines out there.

[Cliff567]

This is a good way to reduce backlash in a rack and pinion system. It also helps hold things "still" when machining as there is a substantial holding torque. Pretty creative really. I have no idea how common of concept this is.

Another way to obtain similar results is to put two pinions on one rack, with each sort of "sprung" apart in opposite directions. The challenge of this approach is that rack often isn't all that thick, so getting two pinions on there is not all that easy. It also limits you to the tooth strength of one "tooth".

Is this a commercial machine or "custom". Probably a brand name will help you find more info on line. I have good luck using dogpile.com but there are a lot of search engines out there.

This is a fairly large 1950's vintage G.A. Gray Mill that has been retrofitted with a new control. Someone at work found this link which is a similar type system to what we've got but a different manufacturer:

http://www.danahermotion.com/website...tibacklash.pdf

We've got the information from the controls manufacturer of the equipment we've got installed. The manufacturer says we have problems with friction and "stick-slip" and I'd like to find out a little more about it.


Thanks,
Cliff

[harryn]

Hi Cliff

There might be a bit more info on a machine like this in the metal working part of the forum. If the challenge really is "stick slip", then perhaps the linear motion part of the table needs some work. Is it based on "ways" ?

[Cliff567]

Hi Harryn,

The machine has V ways and the table rides on micarta. There is a pressurized, continuous flow way oil system.

We did some servo traces yesterday and I decided to try running very slow. We ran .001 ipm and I set the servo trace time to 500 seconds. We were monitoring the table scale and the current of both motors. In the 500 second trace, the scale made 28 steps of exactly .0003”. The leading edge of each step was slightly rounded with the machine settings that were left in the machine. Changing some gain settings made the steps look much closer to a square leading edge on the second trace we did. Everyone from the retrofit company and controls manufacturer has done tests with much faster speed. I was expecting a large jump in position on the scale and then a return to compensate, but it didn’t happen. I would be interested to know how you and other forum members that have experience in this type problem would test for this?

The symptoms we see that made the manufacturer conclude the problem was “stick-slip” is that when the gain is set where they think it should be, the table oscillates back and forth about 8-10 thousandths of an inch at about 5 second intervals when it is supposed to be at a standstill. There are also problems with that axis completing a circularity test properly.

Like I said, if anybody has any ideas I would love to hear them.

Thanks,
Cliff

[harryn]

Hi Cliff - I am afraid you are way out of my league. I am more of a wood hacker type.

I am just guessing, but it seems like stick / slip should happen going from 0 to something, or perhaps a reversal, not a constant slow motion.

From a simplistic controls engineering perspective, if there is oscillation "sometimes" then in a conventional PID loop, you would normally assume that the gain is too high, and the integral setting is too low.

The fact that you are having trouble with that axis from a dimension perspective (out of round circle) is a possible indication that "something" is causing the controller to have an offset from the setpoint. This is also an "indication" that the PID setpoints are off.

Of course, PID is just one control method, and I am 25 years away from this kind of work.

Like I said before, you are way ahead of me in terms of experience and equipment, but here is a link to PID control on wikipedia. I read through it, and it looks "correct" - at least as I remember.

http://en.wikipedia.org/wiki/PID_controller

The fact that you are seeing a problem at the 5 seconds mark is telling us something also. Often, controllers have an additional factor over PID - measurement timing (my term - I don't know the correct term).

True PID is a continuous, analog function, but no one actually does this. Instead, measurements are taken "periodically". Perhaps there is some kind of situation where a timing constant of the machines natural harmonic and the control network timing go in and out of phase with each other sometimes ?

I don't normally recommend to post the same question in 2 different forums, but when you are really stuck, sometimes it helps. Perhaps consider to post a link to this thread in the machinist area with a short comment. Maybe more people will see it and come over to post.

[WGoyer]

Cliff,
I'm curious - whose motors/drives/CNC did the retrofitter use?


We're struggling with an even more complex issue than yours: Our customer has this arrangement (two motors on one rack), but has them on both sides of a very long Ekstrom Carlson Gantry. So we have four motors controlled by a Fanuc 15MB in Tandem/Sychronous mode (four motors PLUS external rack and pinion encoders on each side with anti backlash torque motors for them too).

It sounds like you must also have external feedback (probably scales) too if you're experiencing hunting. Hunting is insidious because people assume the scale should take care of it, but it can't.

Warren
www.uptimecorp.com

[Cliff567]

Thanks for the responses

Warren,
The retrofitter used Siemens motors, drives, and 840D control. They used a Heidenhain external scale on this axis. I’m sure they were trying to use good quality equipment for the project. The table travel is 26 feet and the table fully loaded can weigh as much as 90 tons. After seeing the problems with dealing with 2 motors, I don’t think I’d want to deal with 4. I hope you can get your system running properly.

Do you have any experience with 2 motor systems of this type? If you do, did you ever get one working properly?



Harryn,
A wood hacker? What ever you want to call yourself, I appreciate your ideas.

I agree about the stick-slip happening on the start of movement. When I did the test, I was trying to gradually bring the force up on the rack by entering the .001 ipm setting. I was expecting to develop a breakaway force and then have the table lurch forward, then try to correct and drive back. It didn’t happen.

I think this is only a PI control. There’s no derivative function available. There are several PI loops for different purposes and different drives. That’s part of what I’m trying to sort out in my mind right now. If I was guessing, I would say this control has at least 1000 different settings. Thanks for the Wiki link, maybe it’ll have some helpful tuning information.

Since the control manufacturer feels the problem is “stick slip”, our mechanical people are working on changing the way pump to one that will supply more flow. They’ve also found a better type of way oil for what we are doing. I guess we’ll see if this helps the problem.

[WGoyer]

Cliff,

Yes, we've done machines with various configurations from tandem like yours to parallel as the old Cincinnati's were. The best way to proceed is what you're doing: remove the scales from the equation first, solve the mechanical issues next, THEN try to re-introduce scale feedback. I'm on my third project this year where the scales were masking all kinds of mechanical problems. Once we were back on motor feedback for positioning, the mechanical problems became apparent.

Siemens and Heidenhain both make excellent products. Hang in there.

Warren
www.uptimecorp.com

[harryn]

Hi Cliff - I did a little more reading.

The P setting will affect how quickly you reach setpoint - as well as how much oscillation you have. This would be one place to look for the 5 second problem.

The I settings affect settling - and correct for offset. Too low of an I setting will let the system keep "ringing", as well as allow a control offset from setpoint.

Your system has too many settings for my basic knowledge set.

Since the mill rack is driven from one side, there is always the chance of "racking", which could add in some stick / slip type effects. Even if the racking is very small, the electronics might be picking it up as noise. Just guessing here of course.

[Cliff567]

Sorry it took so long to get back here.

Warren,
We’re trying to get the engineer from the manufacturer back in to set this up after we complete the pump/oil modifications. I don’t know how much I’ll be able to play with the system before we get him in. When you’re talking about running off the motor encoders, the retrofitter made some PLC modifications to do that when he wanted to run that way. Unfortunately, there is no switch or setting in the control that will completely allow it. When you run off encoders, do you still look at the scale to see what the machine is actually doing? When he did it, he monitored the encoder. It made a circle that was in tolerance when doing a circularity test in the control, but the machine really didn’t perform that way.

I was also wondering if you had any idea what percentage of motor full load current is tolerable for stick-slip since it’ll be impossible to get rid of it all? Right now I don’t think our total start current exceeds 25% of motor full load current.


Harryn,
Have you looked at that Ziegler-Nichols tuning method? That looks really interesting. There’s some valuable information in that Wikipedia article. I’m kind of wondering if there are 2 oscillations that might happen, one fast and one slower (5 seconds). This control has a PI loop for position and another for the tension torque. I’m sure they affect each other; I’m going to pay close attention to where the engineer that’s coming starts with this.

The mechanical people have checked for racking, table lift, and anything else they could think of. They definitely don’t want a problem like this to fall on their shoulders at this time. Our mechanical crew is Really Good at what they do. I’ll double check with them to make sure though.



Once again, I want to thank both of you for the help.
Cliff

[WGoyer]

Cliff,

No, the scale should be out of the picture. I would ask the retrofitter to unplug the scale and give you a procedure to change over to pure motor feedback. Someday, when the scale gets dirty or bumped by the operator, you will want a quick way to isolate problems.

I spent the day with our problem child today. I think we have a bad drive (Fanuc calls them Velocity control units). The machine runs fine most of the time then gets a small jolt at any speed. This eventually progresses into a violent shake felt in the office 100' away. Then the control, sensing loss of sync shuts it down.

We swapped out all four drives in various combinations and it was running when we left. With any luck, the bad unit is now in either the Y or Z axis.

Our motors run about 25% also. The leading motor shows this amount and the trailing motor reads near zero at around 200 IPM.

Stick Slip: I used to work on a Hydrotel with lots of Stick Slip. Before I came to tune the hydraulic servo, the plant engineer would swab the machine with STP. Not really the right answer, but very useful in identifying the problem.

Warren
www.uptimecorp.com

[harryn]

Interesting - if STP worked, then this would indicate that synthetic motor oil would be a near ideal lubricant for the ways. It has very high pressure tolerance and much lower friction than conventional oil.

Cliff - I will look over the Ziegler-Nichols method. I will tell you that I mostly did my (much simpler) tuning manually, and it was a long slow process. Most of my work was in fluid flow and temperature control.

An oscilloscope with memory is really a must.

The tuning optimum is often different for a small forced setpoint change vs a large one. Pretty exciting.

[Cliff567]

Hi ,

Our mechanical group got the pump changed, and changed the way oil to Mobil Vacuoline 1419 which is supposed to be a very good oil to remedy the situation we have. I was able to do a couple of comparison servo traces and there seems to be a difference. I hope I get some more time with the machine tomorrow.

Harryn - We don’t have a storage scope, but the control has some very nice diagnostic functions built in where you can do a servo trace or circularity test. I hope I get a chance to “play with” the tuning myself. If I get a chance , I’ll try the Ziegler Nichols tuning and let you know how it worked.


Warren – I hope you got your problem fixed by changing the drives. Was there any mechanical damage done when that violent shaking happened? We haven’t had anything that violent happen, but we noticed some metal slivers actually peeling off the rack when the gain was high and tension torque was around 30%.

I’ll ask the retrofitter to make it possible to run off motor encoders for troubleshooting purposes. I also would like them to make it so we can run off 1 motor on the X axis.


I’ll keep posting updates as we find out more.

Thanks,
Cliff