How many of you have used cascaded loops?

[JHarbin]

I use them in process control. An example is DO cascaded to flow controlled by valve position. I tune manually by trial and error. Sensitivity and difficulty for operators to tune is one reason I don't use PID.

Anytime "operators" and "tune" are used in the same sentence someone is about to get a paid service call...

 

[ndzied1]

The winder process mentioned here:
http://www.automationworld.com/feature-4348
...uses a cascaded loop.

The pressure wheel on the winder is driven by a non-linear mechanism and could not be tuned for the entire stroke of the mechanism with one loop. I don't know the specifics and I'm not sure they would give them to me but I witnessed the startup and the 2nd loop made all the difference.

 

[robo77]

ya J I can't recall how many times the previous programmer gave "tune" data on the HMI to the pid, too many to count. And they can't figure out why it won't work.

Although last DO project worked great pid with 3x VFD blowers and multiple DO's in each basin to control.

 

[maverick33]

While I agree that cascaded loops can sometimes cause more problems that they solve, there are situations where they are helpful/necessary. I have used them on several occasions to control both air and liquid flow that is volume, pressure and ratio dependent simultaneously ... while mathematics gave a good starting point in terms of tuning, these applications required some manual trial and error type tuning to get everything working within the tolerances desired.

 

[Peter Nachtwey]

Boilers are a poor example

There are some non-linear things that happen in them due to shrink and swell.

One system I did had an inner and outer loop to control a big steam flow valve which the DCS system controlled which was yet another outer loop. There was no way this could work any other way. The inner loop controlled a valve spool which ported air to a big cylinder that in turn moved a big steam control spool.

Controlling the inner loop was simple as the spool responded almost instantly and certainly instantly compare to the DCS. The pneumatic system was only difficult because it was moved by air. The spec called for full travel in 1 second which is still fast compared to the DCS. These kinds of systems are easy to control in the cascaded loop sense because the inner loops are so much faster.

However, this isn't always case and that is where the real problems arise when the inner loop can't be that much faster. For instance on a hydraulic system the servo valve isn't usually that much faster than the force or acceleration loop if it exists and the acceleration/force loop must be much higher that the position loop because the acceleration loop is two derivatives away from the position whereas a velocity loop is only a single derivative from the position.

Cascaded loops have an advantage over single loops when there are disturbances. The inner loop can also make the system look more linear to the outer loop and easier to tune.

 

[nettogrisen]

I've used cascaded loop once to roll my own position loop on a kinetix2k system because we needed to switch back and forth between velocity and position loop. After trying to tune it for a couple of days with a rockwell tech it turns out that you can't, as the gain for the controllers own velocityloop can't be switched off.

I would read an article like that, but I would probably not understand everything. I'm sure to learn something though.

 

[Peter Nachtwey]

I've used cascaded loop once to roll my own position loop on a kinetix2k system because we needed to switch back and forth between velocity and position loop.

I don't understand this. The velocity loop should be able to work on its own with its own Ki_i and Kp_i and the outer loop has own Ki_o and Kp_o

What you do is simply by pass the the outer position loop.

After trying to tune it for a couple of days with a rockwell tech it turns out that you can't, as the gain for the controllers own velocityloop can't be switched off.

Why would you want to turn off the inner loop gains? Those are the gains that are required. It is the outer loop gains you don't need.

I would read an article like that, but I would probably not understand everything. I'm sure to learn something though.

Do you understand Laplace transforms? Can you wade through the pdf file posted above?

I am truly interested because I would like to write an article about this but it may get shot down for being too technical. I am usually happy if 5 people understand but magazines don't like to scare away the rest.

 

[kamenges]

Originally posted by Peter Nachtwey:

I don't understand this.

I didn't quite get that either. My guess is that nettogrisen is equating using a position loop with the requirement for a torque command. I dodn't think the Knietix drives allow the user to toggle the command mode while the system is in operation. While there may be some very good reasons to use torque mode drives for positioning it is oftewn not a requirement.

Originally posted by Peter Nachtwey:

I am truly interested because I would like to write an article about this but it may get shot down for being too technical.

That is unfortunate. I would rather have the article more technical and less "fluffy". I can always do more research (or ask you) if I don't understand something. If an important technical point isn't made in an article I wouldn't even know to ask. But "fluffy" is what puts a vendor's name into readers' minds.

Keith

 

[Peter Nachtwey]

I didn't quite get that either. My guess is that nettogrisen is equating using a position loop with the requirement for a torque command. I dodn't think the Knietix drives allow the user to toggle the command mode while the system is in operation. While there may be some very good reasons to use torque mode drives for positioning it is oftewn not a requirement.

I am still confused.
We can change from position to velocity mode on-the-fly. It isn't hard to do.
I do all my motor testing in torque mode. Our controller outputs +/- 10 volt but I don't like depending on the drive to close the velocity loop. We can do the velocity control better than many and maybe drives. I prefer to use the drive as a dumb transconductance ( voltage to current ) amplifier so our control out is controlling the acceleration and not the velocity. However we do not usually get a separate velocity feed back so there is simply a PID with feed forwards. We do a good job at this but it isn't as good as a cascaded loop for disturbance rejection. For cascaded loops to work well and have good disturbance rejection there needs to be a separate feed back device for the inner loop.

That is unfortunate. I would rather have the article more technical and less "fluffy".

You know I can get deep, very deep. Every once in a while a magazine will permit a "deep dive" article that is deeper than the normal article. I would like to write articles that people remember and keep the articles. Dan Simon wrote an article on Kalman filters. I kept the article until I had read his book and thoroughly analyzed Kalman filters.

I can always do more research (or ask you) if I don't understand something.

Yes, and I do get asked by people that read my articles. I have had three inquiries in the last two weeks. Two inquiries about the article I wrote on the www.controlguru.com which isn't very deep.

If an important technical point isn't made in an article I wouldn't even know to ask.

The key thing about cascaded loops is controlling the derivatives. Acceleration leads velocity and velocity leads position. To get the best control one must monitor or control the derivatives.

 

[nettogrisen]

I don't understand this. The velocity loop should be able to work on its own with its own Ki_i and Kp_i and the outer loop has own Ki_o and Kp_o

What you do is simply by pass the the outer position loop.

I'm sorry for the confusion, I ment switching between _torque_ and position loop on the fly. Like Keith wrote, it's not possible with kinetix2k

Why would you want to turn off the inner loop gains? Those are the gains that are required. It is the outer loop gains you don't need.

This was my first servo application, so I had a RA tech help me (You might recall helping me with my airplane door application). He (the RA tech) found some code that was circulating inside rockwell. It was supposed make the system switch between position and torque control on the fly. I can post it here if you want. We tested it and it seemed to work fine, (but that was without any load), so we wrote the whole program around this piece of code and had the whole system installed in Sweeden. Then under commisioning we had a really hard time tuning it, so we started to track the person that made the code. When we found him in Holland he told us that the code should have never been released as it will not work.
Along with the code was a powerpoint presenation on how to implement it. It showed that you should put all the systems internal gains to 0, and use the codes gains instead. And that's what he told us could not be done.

Do you understand Laplace transforms? Can you wade through the pdf file posted above?

I have come across that word before, but no, I don't know what it is. I can wade through alot if I feel that I'm learning something. I always read your posts even if I don't understand them.

I plan on going back to school sometime soon as I would really like to learn to be able to follow your and others calculations. Not knowing enough math is what I feel is my biggest handicap.

I am truly interested because I would like to write an article about this but it may get shot down for being too technical. I am usually happy if 5 people understand but magazines don't like to scare away the rest.

You don't scare me!

 

[kamenges]

Originally posted by Peter Nachtwey:

We can change from position to velocity mode on-the-fly. It isn't hard to do.

It's not between positon mode and velocity mode. It is between velocity mode and torque mode att he drive. As you said, the way to do it is keep the drive in torque mode and switch between control modes at the controller. Then again, I don't know if the CLX motion stsyem will let you do that either.

Keith

 

[mordred]

I for one would look forward to another article written by Yu Peter I've read a couple and have never failed to learn something from them. When I hit something I don't understand I usually look in the internet and get the answers I need. Often including some of the mathematics

 

[Peter Nachtwey]

It's not between positon mode and velocity mode. It is between velocity mode and torque mode att he drive.

This is still easy, now. Torque mode is just the rotational equivalent of force mode. We can go from position or velocity mode to force mode easily. It is press control 101. However, we need extra current/torque feedback to control torque just like we need pressure or force feedback to control force on a press.

As you said, the way to do it is keep the drive in torque mode and switch between control modes at the controller.

Yes, obviously we can't control the torque either if the drive is controlling the velocity loop. Again, feed back is required if you are going to control it. However, if we are controlling the torque proportionial to our voltage output, it is it easy to get an open loop torque limit simply by limiting our voltage output and no feed back is required. If closed loop torque control is required this gets to be more tricky as you can't use current sensors due to the fact the current is often PWM or chopped up in some way. In these cases it is necessary to get the effective torque or current used by the model inside the drive but in these cases we are at the mercy of the drive again. It really is best if the drive can do all the control since it knows the position and the current/torque and velocity, the derivatives. The problem is that so many drives are simply made to move conveyors and aren't very sophisticated.

Then again, I don't know if the CLX motion stsyem will let you do that either.

If you are talking about the M02AE class of controllers, they have only position feedback. I know nothing to little about the drives that probably have velocity and current feedback too.

I know the CLX controllers have code in them for an inner velocity loop. The only advantage to having a velocity loop when there is no velocity feed back is that there is an option to control velocity without regard to what is happening to the position which is justification enough if this mode gets used but there is not performance increase like can be obtained by using a true cascaded loop with a separate inner loop feedback.

Think about this, the integrator gain for the velocity loop does the same thing as the proportional gain in the position loop. When tuning for position mode adjusting either the velocity loop integrator or the position loop proportional gain will have the same effect except that adjust the velocity integrator affects the velocty loop too which is why it is best to tune the inner loops first. Integrating velocity error yields a position error.

 

[harryting]

I have used them many time in process control. The one I can think of is vessel temperature control via steam jackets. It make the tuning easier for a dummie like me. I am quite sure I learned the rule on whether to use casacade loop or not in college but I can't remember them now.

OT: anyone went to the ISA West Expo in Portland today?

 

[Roy Matson]

I have used them many time in process control. The one I can think of is vessel temperature control via steam jackets. .

Yes this is a good example of where cascade works well
The slow temperature loop provides the setpoint for the fast flow control loop.
First of all I would tune the flow loop to have a reasonable response, it doesn't need to be super fast.
Then I would tune the temperature loop.
Without cascade in a case like this the temperature loop driving the steam valve directly will be very hard to tume. Any hysteresis in the valve will cause the temperature to slowly oscilate forever seeking setpoint.
I should say I don't understand the math involved but feel I can do a reasonable job of tuning most loops.