Allen-Bradley PIDE

[mordred]

I'll have to remember the dpendency good site you posted by the way helps to clarify some misconceptions I had

 

[Roy Matson]

Looking at your screen shot I see both Proportional and intergral are now at 1.0 that's quite high (it was 5.0 which is extremely high).
My simple explanation of Proportional gain is the setting that adjusts for thre valve gain e.g if you move the valve 20% and the process changes 40% your valve has a gain of 2 so the controller gain would be 0.5, I don't try and tune that way, it's just a way of explaining
Have you tried Ziegler and Nichols? I find that works very well for me.
http://en.wikipedia.org/wiki/Ziegler%E2%80%93Nichols_method
Don't even think about using D, it shouldn't be nessesary in your case.
I looked at my application, my controllers are also "preriodic". My execution rate is every 10mS which is very fast and I'll probably slow it down at commissioning. Some large DCS systems I have seen are 250 - 1,000 mS.
On your EUs/Limits tab you had the Cv Max at 90, that should be 100, the Cv Max at 0 should be 0. If you want to limit the valve stroke you do that with the Cv Limits box

http://en.wikipedia.org/wiki/Ziegler–Nichols_method

 

[Peter Nachtwey]

Voodoo control advice

Looking at your screen shot I see both Proportional and intergral are now at 1.0 that's quite high (it was 5.0 which is extremely high).

It may it may not. It depends on what the plant gain is. If the plant gain is low then the controller gain will be higher.

My simple explanation of Proportional gain is the setting that adjusts for thre valve gain e.g if you move the valve 20% and the process changes 40% your valve has a gain of 2 so the controller gain would be 0.5, I don't try and tune that way, it's just a way of explaining

That isn't quite right but you do understand the basics. So why say a gain of 5 is high. You forgot to take into account the desired response.

Have you tried Ziegler and Nichols? I find that works very well for me.
http://en.wikipedia.org/wiki/Ziegler%E2%80%93Nichols_method

Witch craft. The ZN methods were developed empirically but without math.
The response that ZN provides is very underdamp and oscillates. The IMC techniques on www.controlguru.com are much better. One can prove the IMC techniques work mathematically. One needs the plant model and a desired response.

Don't even think about using D, it shouldn't be nessesary in your case.

How can you say that? We haven't been told anything about the system. The derivative gain might be very necessary. It depends on the number of poles the Michal_dk's system has. We come across systems that require a second derivative gain on the derivative of the derivate a few time each year. Like I said, it depends on the number of poles the plant has and we have not been given that information.

 

[Roy Matson]

It may it may not. It depends on what the plant gain is. If the plant gain is low then the controller gain will be higher.


That isn't quite right but you do understand the basics. So why say a gain of 5 is high. You forgot to take into account the desired response.


Witch craft. The ZN methods were developed empirically but without math.
The response that ZN provides is very underdamp and oscillates. The IMC techniques on www.controlguru.com are much better. One can prove the IMC techniques work mathematically. One needs the plant model and a desired response.


How can you say that? We haven't been told anything about the system. The derivative gain might be very necessary. It depends on the number of poles the Michal_dk's system has. We come across systems that require a second derivative gain on the derivative of the derivate a few time each year. Like I said, it depends on the number of poles the plant has and we have not been given that information.

Peter, Luckily "what upsets me would kill most people".
You are right, neither of us know much about the process because the poster hasn't given us much to go on. I assumed (perhaps incorrectly) that since Michal_dk was having difficulty with the PIDE that he was fairly new to process control.
The gain went from 5 down to 1, in my book that's quite a dramatic change.
Ziegler and Nichols would be pleased to hear that all their hard work is now regarded as "witch-craft". Perhaps I'm old fashioned but "One can prove the IMC techniques work mathematically" doesn't do it for me I'm afraid.
Because I have only ever seen D applied successfully to temperature loops.
Anyway, lets not turn this into an argument over which method is best, you use your computer I'll use my slide rule.
Regards
Roy

 

[Michal_dk]

OK, I am trying to get thru all the informations that have already been shared - Thanks!

To Roy: I'm not new to process control, if only I was that far already! No I'm way below that, I too a class at the university some 12 years ago and haven't looked back since And now it comes back to haunt me.

I most often deal with conveyor systems and packaging systems.

This system is designed to test the forces the wind will excert on roofing materials.
What we use is:

1. One large wooden box (app. 3,5x4,5x0,5 m)
2. One 22 kW motor/ventilator (that's not the right word! I'm looking for that thingy that sucks air out of a room/box)
3. One servomotor attached to a shaft with wings

The servo is then working as my valve, by positioning the servo in 0 degrees I can almost make my box airtight and thereby creating a vaccuum that will work as the wind forces om my roofing. By positioning it at 90 degrees the vacuum is 0.

The vaccuum I get from a differential pressure meassurement sensor and using some math I convert the value to kN/m2, as this is the value the user desires.

My "valve" can move at speeds up to 500 degrees/s, and this gives me an almost instantaneously force of 2 kN/m2 if I run the venting system at 50 Hz and then close the Valve. This of course is too much for the roofing and it will pop very quickly afterwards.
Now the system needs to go from 0 to say 0,22 kN/m2 in a maximum af 0,7 s (very little overshoot is allowed) and keep that for 2 s before returning to 0 and a rest period of about 2-3 s before starting over. And this is to be repeated 500 times for this value.
Next the value is encreased and the cycle is repeated 200 times. This goes on for quite some time, we expect the test to run over a couple of days every time we have something to test.

The guy who is to use the system, has tested it and found that if we start by 15 Hz on the venting system and close to 23 degrees at 70 degrees/s then after 0,7 s we change to 20 degrees at 500 degrees/s we will get the result needed (0,50 kN/m2)
But that's a lot of testing and work to be done, as this is not the only configuration we need to test. Therefore I would like to be able to use some form of automated regulating and just watch the value fall into shape.

So any advice is still appreciated.

regards,

Michal_dk

 

[Peter Nachtwey]

The servo is then working as my valve, by positioning the servo in 0 degrees I can almost make my box airtight and thereby creating a vaccuum that will work as the wind forces om my roofing. By positioning it at 90 degrees the vacuum is 0.

What about in between? Do you have a table to correct for the non-linear way pneumatics react?

My "valve" can move at speeds up to 500 degrees/s, and this gives me an almost instantaneously force of 2 kN/m2 if I run the venting system at 50 Hz and then close the Valve. This of course is too much for the roofing and it will pop very quickly afterwards.

It looks like you know how the valve reacts. Do you know the time constant. If you graphed the 'valve' velocity as a function of time you will find there is an exponential rise from 0/degrees per second to 500 degrees/sec. If you assume it takes 5 time constants to get within one percent you can estimate the time constant by dividing the rise time by 5.

So does the motor over shoot the position? You really seem to have a dual loop system. One to control the position of the 'valve' and the other to control the vacuum.

Here is another questions. Does the vacuum fan run at a constant velocity as the load changes? I bet not. I would measure the blower speed to see if its speed is constant as the 'valve' opened or closed.

But that's a lot of testing and work to be done, as this is not the only configuration we need to test. Therefore I would like to be able to use some form of automated regulating and just watch the value fall into shape.

You do have a lot of testing to do to build the tables or formulas to compensate for non-linearities will take much more effort than programming a few simple loops to count cycles.

At different blower speeds I would move the valve to different positions and measure the pressure drop to build the tables or determine the equation. It would also be nice to know if the valve position changes from 5 to 10 degrees is the pressure change the same as a valve position change from 40 to 45 degrees.

It would also be good to know how long it takes the vacuum to reach steady state after the 'valve' position reaches steady state.

 

[Michal_dk]

The system seems to be very stable, the vaccuum has a slight overshoot but becomes stable in about 10 ms. Run at 15 Hz and angle of 20 at 100 degrees/s there is no overshoot and it is stable at once.

This is without the PID, I command the angle and velocity.

Oh, darn!

I have to run, a breakdown on another machine.

I'll be back

Michal_dk

 

[Roy Matson]

Congratulations, it sounds like you have a good handle on your system.
Without PID eh!
What do you think causes the overshoot, perhaps the servo positioner travels too far or the blower is on the point of surging.

 

[Michal_dk]

Oh, how I have wanted to come bask here and give an update on my system. But work takes too much of my free time

We did a lot of manual testing and I build an Excel worksheet that holds the parameters. We then download the values for a specific configuration of roofing attachment (e.g. 60 cm between fasteners) and the test can run. Doing the same test 3 times gives almost the same results, so repeatability is proven.
But there are a lot of parameters, we have the speed of the fan, there are 2 angle values at 2 different speeds (variable) and we have 3 time values too.

Now my Customer tells me that he has discovered that the surface on which the roofing material is mounted has a great deal of influence on the vaccuum.

So I think that I will have to ask Rockwell to come by and help with the PID/servo setup.
Oh, yes! They have already said they would come to my rescue.

If and when the customer lets me at the system again, I'll report what the result was.

Thanks for all the input,

Michal_dk