RS5000 and M02AS hydraulic servo motor

[Robb B]

Hello everyone, I've been reading through the forums on motion, servos, etc. I'm looking for some help in determining the cause of some oscillation in our machine.

The machine is a tie stacker, it uses an SSI encoder and a proportional valve (I don't know the brand/model at the moment) to run a hydraulic motor which pulls a chain to move the trolley on a rail. There's two operating modes, automatic and manual. When in manual mode, MAM and MAS commands are used to move the stacker. Sometimes (repeatable, but random) the stacker will oscillate around a commanded stop position. In the M02AS the tolerance is set at 1", typically it will home between 0.01 and 0.001 inches from commanded position, reasonably precise. When the oscillation occurs, the commanded position is set exactly, but the actual position varies approximately 1.4" to either side. I am new to motion control, but I believe the issue is to do with the trapezoidal profile vs the s-curve, and the accel/decel rates. The controller is attempting to recalculate the speed and position, but I think the requirement is to reach full acceleration/deceleration before a stop can be achieved. Scan time on the motion loop is about 40 ms, which is probably faster than the valve can respond to the positioning.

I will attach the MAS and MAM files to this post. If you want the whole logic file, I can try.

[Peter Nachtwey]

If the motor oscillates back and forth with stopped it could be from excessive static friction or a dead band in the valve. In these cases the actual position can not reach the command position so the integrator winds up. When the motor does move either the dynamic friction is very low so there is over shoot or the integrator shifted the valve across the dead band. Motors will start to leak as their seals get old. This will result in the motor not moving even if there is a little flow.

What is the minimum voltage required to start the motor moving?
What is the valve? Post a link to the valve.

The whole file is not necessary. A trend showing the control signal, actual position, target position and command position would be much better.
The speed, acceleration and deceleration parameters would be good too. Provide the units.

[Robb B]

Peter, how can i get a graph showing this information? It is control logix, I don't have any experience extracting trending, but I will attempt on Tuesday, as well as valve information. The pictures of the MAS MAM don't show the relevant units? I will try to get more information again, seems I am a bit shy in that department...

The local millwright says the motor and valve are well sealed, but I don't know how that was verified since the motor has not been bench tested that I know of.

[Peter Nachtwey]

You need to ask others about how to get a trend. I know it can be done because I have seen it done in person. I haven't touched RSLogix software in 15 years.

The screen display says units whatever units are. Radians, degrees, probably inches. if so the speeds and accelerations are not that fast. The MAM and MAS are not the problem.
BTW, you shouldn't need to use the MAS unless there is a need to stop the motor for emergency or some unforeseen location. The MAM should stop at the command position.

[Robb B]

Originally Posted by Peter Nachtwey

The screen display says units whatever units are. Radians, degrees, probably inches. if so the speeds and accelerations are not that fast. The MAM and MAS are not the problem.
BTW, you shouldn't need to use the MAS unless there is a need to stop the motor for emergency or some unforeseen location. The MAM should stop at the command position.

Yes, I believe it is inches, calculated via the encoder. We use the MAS to stop the stacker while in manual mode. In auto it does just run from home to drop-off and back.

I will try to get a trend for you by Tuesday end, long weekend here! Willxfmr pdf seems to be perfect information.

[willxfmr]

Robb,
Page 70 of this document will get you started with making a trend. There are also several YouTube videos on the subject.

https://www.rockwellautomation.com/r...Studio5000.pdf

[Robb B]

Position units are inches, Velocity timebase is 0.1s.

Valve for proportional control of servo motor is a Rexroth 4WRAE
10 e60-22/G24N9KK31/A1V
Manual is here https://dc-us.resource.bosch.com/med..._4/re29055.pdf

Attached are two trends showing normal movement in manual mode and the oscillation. I think it's clear which is which It looks to me like the servo output level starts to spike even before there is a significant position error calculated, but what's causing that? Could there be issues with the fact that it's a pushbutton controlling the input as a jog function? Intermittent contact, or something...?

Please let me know if there is some other values I should be trending. I'm not exact on what each tag is defined for, but it seems reasonable to me.

[Peter Nachtwey]

I don't have a lot of time to look at it right now but the valve is garbage. If there is no Bode plot, the valve should not be used for servo applications. More on this later.

I can't make out what the green line is but I think it is the control output. If so it is changing in huge steps. The feedback resolution is not fine enough.

What is the yellow line? It looks like it might be the acceleration but the polarity is the opposite of the velocity ( purple and blue )

You are fighting an uphill battle with what you have.

So what is the resolution of the feed back in inches per count?

[Robb B]

Peter, thanks for your input. I will try to get clearer pictures tomorrow, I've only just started working with trends. I can get more specs on the controller tomorrow, resolution, signals, etc. I believe it is 40.1 counts per inch, but will confirm.

[Plastic]

Originally Posted by Robb B

Peter, thanks for your input. I will try to get clearer pictures tomorrow, I've only just started working with trends. I can get more specs on the controller tomorrow, resolution, signals, etc. I believe it is 40.1 counts per inch, but will confirm.

Robb,
Two tips for your trends:
Right click on trend, PROPERTIES, Change background to white. This will be of great help if you print.
The default background was chosen to be black to emulate old-school CRT oscilloscope. You will have to visit PENS, the one pen that is white, needs to change to black, after you invert the background display color.


You may also increase the pen width from 1 to 2or3 for to highlight one or more of the pens to increase readability.

Plastc

[Peter Nachtwey]

Robb, the Rexroth valve you now have ( Bosch Rexroth is now Rexroth ) isn't satisfactory for motion because there is no spool feed back. The flow forces move the spool around.
Page5/16 shows the valve is rated at 10 bar not the usual 70 bar. If you try to use this valve at higher pressures and flow there is a good chance you are trying to push too much oil through the valve. Look at the hysteresis. It is 10 times worse than a normal Rexroth servo solenoid valves. The reversal and sensitivity values are too high.

Look at page 8/16. The valve has a ramp generator that can be slowing the response even more. This is an indication that the ramp generator was intended to ramp the valve velocities, not a hydraulic motion controller. Turn off the ramp generator if you can.

Look at page 9/16. There is a 20% dead band and after than the response is not very linear. The dead band is what cause the slow oscillation around the set point.

Look at page 10/16. The valve has a slow response to control signals when opening. The bottom graphs show that the flow will actually go down at higher differential pressures with the same command signal. This is due to the lack of spool feed back to maintain spool position as a function of the control signal.

Don't trust hydraulic people. Most do not know how to design a hydraulic servo system and will sell you what ever they have on the shelf. I know!!!!

Rexroth does make some good valves. A good NG10 that I like is this one.
http://deltamotion.com/peter/Hydraul...o_ng10_obe.pdf
We use this valve in our hydraulic motion control training class.
There are faster Rexroth valves. Ask forum member Norm Dziedic about these. See this
https://forum.deltamotion.com/viewto...t=524&start=10
Norm mentions a faster valve. Norm sells Rexroth.

[Robb B]

I've attached pictures of the oscillation and of movement without the oscillation. When it does oscillate, it is fairly violent, approximately 3" of travel back and forth, and rapid.

I don't really understand how we get reasonably smooth operation in automatic mode, but when we run manually, we get this oscillation randomly. If there was a problem with the deadband of the valve wouldn't this oscillation show up in any movement/stop action? Why is the output fluctuating so much when the command and actual positions are not relatively far apart?

I've plotted the Command position, actual position, servo output and position error on the charts. I can try to capture more recent data this afternoon if needed.

[Peter Nachtwey]

I looked at the new data. Look at the control output ( green ) notice how it changes in steps? This is due to course feed back but then I see you have a 22 bit SSI encoder that should provide very good resolution. What is the scale factor in inches/count?

[Robb B]

Peter, Thanks for the help!

I've found what I think is the scale you are looking for, 41.7 counts / inch. Rockwell calls it a "Conversion Constant".

There is also a tab called "Offset" on the properties of the axis, which has a "Friction/Deadband Compensation" field, could that be used to compensate for the valves deadband of 20%?

I've attached screenshots of each, unfortunately the website scales them down somehow, despite uploading as .png, apologies for the poor quality.

I've had another thought, is it possible we are simply using the wrong command for the manual movement? If the function is to be basically a jog, shouldn't we use a MAJ command? Currently it is using a MAM, but stopping that MAM with a MAS when the button is released (only travels forward and back). I can use the existing limit switch inputs to interrupt the jog when it gets to home and unload positions. Would using the MAM and interrupting it cause the oscillating due to controller recalculating? The valves response time is about 20ms, our scan time is about 30-32.

[Peter Nachtwey]

Originally Posted by Robb B

Peter, Thanks for the help!

I've found what I think is the scale you are looking for, 41.7 counts / inch. Rockwell calls it a "Conversion Constant".

That is 0.024 inches/count. That is too high. If the resolution was 0.001 inches/count the output would change in steps that are 24 time smaller. The output would be smoother.

There is also a tab called "Offset" on the properties of the axis, which has a "Friction/Deadband Compensation" field, could that be used to compensate for the valves deadband of 20%?

That was one of my big contributions to the M02AE, M02AS and HYD02.
Rockwell wanted to patent the dead band eliminator. I was not pleased.

I've had another thought, is it possible we are simply using the wrong command for the manual movement? If the function is to be basically a jog, shouldn't we use a MAJ command?

You can try the job command for manual mode.

Currently it is using a MAM, but stopping that MAM with a MAS when the button is released (only travels forward and back).

Both ways work. I prefer the MAM and MAS way but we call the commands something different. When stopped the controller is still in closed loop and will not drift.

Would using the MAM and interrupting it cause the oscillating due to controller recalculating? The valves response time is about 20ms, our scan time is about 30-32.

The valves response time is much longer than 20ms according to the pdf file.
A scan time of 30-32 ms is too long. Look to see if the target position is hunting. The closed loop in still done inside the motion controller.
The target position should be constant when stopped in closed loop.