PIDE: the autotuning feature

[buhnen]

Hi!

I'm trying to understand how does work the PIDE autotuning feature and I've got good documented resources for it from the Rockwell library but when testing its functionality for real, on the work field, I'm yet finding some troubles.

For instance, I have to tune a PID for a level control process. I must keep a setpoint value, say 160 cm, for a the liquid level in a tank (the liquid is sewage waters). I've got a running flow into the tank which is not steady at all, but waving (not a big waving though) and causing disturbances to the system. So when trying to use the autotuning, I'd need to know first of all, WHEN, is the right moment for starting its execution...may it be in the exact moment when the level line is going up and it's near to the setpoint? for instance, when the level line is 1,40 going up getting closer to my setpoint of 160 cm?...or is it rather the opposite to this, this is, when the level line is going down?

I've made several attempts but a displayed error message popped up several times saying that the PV wasn't steady enough to keep on with the autotuning process, and of course, everytime I got this message, the autotuning execution got interrupted by itself....but I wonder, on the other hand, how is this possible because the flowing in of liquid will always be waving and there's nothing I can do about...

In a couple of occassions I didn't get that error message but the autotuning process seemed to last too long and it never got to be completed, just before the level line had reached the top of the tank.

I don't know either very well what should I do about these two parametres "PV change limits" and " CV Step Size". I've let the first one to zero and the second one to 10% as I could understand from reading the Rockwell manuals that it is well advised to start from small values ( I mean the CV Step Size) for the tuning.

Could I get any clues for you folks?

Thanks in advance.

 

[Peter Nachtwey]

I am assuming you are controlling the out flow.

I doubt the Rockwell auto tune will be able to tune your system. Auto tuning software generally assumes there are no disturbances. If the auto tuning software can take disturbances in to account then the auto tuning software must know what the disturbances are and you have not indicated that you have a flow meter.

The good news is that these kinds of systems should work well with just a proportional controller if you can tolerate some error. For instance, you may set the gain so you get 100% out flow when there is a 10 cm error. As long as the flow doesn't exceed the out flow capabilities the level will never go over 170 cm.

If you can't tolerate error then you must use the integrator.

Different types of systems require different techniques and some are much harder to tune than others. Level control systems are easy.

 

[buhnen]

Thanks for the advices Peter,

I've tuned manually the PID and I think it is working fine now, more or less.But still I need to acquire more experience tuning PIDs ...I've heard of any helpful softwares like RSTune from Rockwell and I think I'll keep researching this way.

 

[Peter Nachtwey]

You shouldn't need tuning software for what you are doing

As I said you can usually get by with a proportional gain set so the control signal is at 100% after an allowable amount of error. The only other thing to play with is the integrator time constant. That is easy to adjust. If you want the error to be reduced to within 1% of the original error 1 minute you set the integrator time constant to 1/5 of a minute. The problem is that the flow, disturbance, will probably change faster than the integrator can respond. There is nothing you can do about that.

 

[Plastic]

PIDE - Electric Resistance Heat direct contact with thermal load

We have implemented PIDE with electric heat attached to fixed mechanical thermal loads.

It works great. The pre-requisite is, prior to invoking the autoutune, is to have the thermal load idle, at a steady state, typically room temperature.

Then we invoke the Auto-Tuning feature of the PIDE. The loop will sit idle for about 20 seconds, to ensure that the thermal load is truly at idle. Then the AT will kick the CV output to your pre-selected level 0-100%. Once the AT kicks the CV, it starts an internal stopwatch, to determine the lag time between heating output and measured response. The AT also measures the resultant rate-of-change due to the static CV output.

After the AT is complete, the AT function will return (3) sets of PID values that it deems applicable to the installation. We typically use the most aggressive of the (3) calculated PID value sets, and then graphically plot a "cold-start" with those values in the PIDE. We have seen good response (short time) without over-shoot.

Our applications can have up to 60 seconds or more of thermal lag time between CV output, and measured PV response.

We feel that this is the best approach, RE: mathimatical measurments are to determine optimal values for the loop(s) and not trial-and-adjust manual values.

 

[Peter Nachtwey]

A temperature control system is not the same as a level control system

The PIDE auto tune was designed to tune temperature control systems like yours but not level control systems like Buhnen's, especially when there are disturbances.

We feel that this is the best approach, RE: mathimatical measurments are to determine optimal values for the loop(s) and not trial-and-adjust manual values.

I agree and you are fortunate to have tuning tools that work on your system.

 

[Tom Jenkins]

Don't trust autotune. I agree with Peter that an experienced engineer or operator can do better.

 

[Peter Nachtwey]

It isn't a matter of trusting.

Don't trust autotune.

A few years back I posted a Scilab program that would auto tune temperature systems. It did a reasonable job given the input data. However, it is not capable of auto tuning a position control system or a tank level for that matter. A generic auto tune system would need to have many different models and be able to choose between them. Most auto tune systems can't do that. For instance, our auto tune for our motion controller can only auto tune position systems. It isn't designed to tune temperature and level control.

What bothers me is that I often see people recommending gain that work on their system and they ignorantly think that these gains will work for everything when they won't.

Buhnen's system is simple to tune manually so why make it complicated?

BTW, Tom, have you ever tuned an under damped system with a very low damping factor and low natural frequency. It is times like that where you want a auto tuning system that is designed to tune those types of systems.

 

[Tom Jenkins]

BTW, Tom, have you ever tuned an under damped system with a very low damping factor and low natural frequency. It is times like that where you want a auto tuning system that is designed to tune those types of systems.

Thankfully, no. My systems are all process control. Their challenge is the very long process response time and transport time. An additional challenge in some of the applications is that the signal fluctuation (real fluctuation, not noise) is about ±25% of setpoint with a period of about half the response time.

But I'd still rather tune my systems than yours!