Automatic and Manual controller mode and it's PID form

Pandiani

Lifetime Supporting Member
Join Date
Apr 2005
Location
Tz
Posts
718
Hello people,
I must admit I don't understand story about manual and automatic mode of operation completely.
As far as I know when process is switch to automatic control that means that system works in closed loop. There is a set point that is set by operator, controller, actuator and finally proces variable. For example set point is with some kind of electronic device converted to to 4-20 mA signal that goes into PLC. PID controller is implemented in PLC, and mA signal goest to actuator (for example it's some kind of regulating valve that is used to control process temperature). In feedback path is usually temperature transmitter that converts PV (temperature) to mA signal that is led to PLC.
If mode of operation is automatic then whole system works in a closed loop with all parts I've just stated, right?
When operator switch to manual, does that mean that only feedback path is open (closed loop is broken), and control system works in open loop:
Set point --> PID controller -->Actuator-->Process?
Is whole story about manual/automatic control tailored to some kind of software switch that break feedback path, or there's more to it.

In my modest experience, I have worked with Siemens FB41 continuous PID block and in that scheme I can see that when MAN is ON (bool) output from controller is just "Manual" variable (real) and all PID structure is simply bypassed.

There is one more thing.
In school I learned that PID form is:
CO = K*e(t)+Ki*integral(e(t))+Kd*de(t)/dt
and in Laplace domain that is:
Gpid = K+Ki/s+Kds, where Gpid is defined as ration of Laplace transforms of CO(t) and e(t)
where CO(t) represent controller output (in time) and e(t) is error (in time) which is defined as SP(t)-PV(t).

Recently, after I have read one article about PID I saw the following form:
CO = K*e(t)+Ki*integral(e(t))+Kd*de(t)/dt + CObias
where CObias is defined as controller output when error is zero and disturbance is not present and needed to maintain PV depending of SP.
It also stated that CObias must exist because when error signal is zero there must be some controller output to maintain output PV.
I can understand that but how to obtain Laplace transfer function then?
Why bias is not included in most textboox I have read?
Why many software for simulation like Matlab and Simulink doesn't include this bias value in PID block parameters?
I tried to explain this to myself:
CO(t) = K*e(t)+Ki*integral(e(t))+Kd*de(t)/dt + CObias
Since, we're interested into process dynamics and that means we want to test how system reacts to variable change that is exactly why Laplace tranform is used.
What does that means?
First of all, definition of transfer function(TF):
TF is defined as ratio of Laplace transforms of output and input respectively when all initial conditions are zero.
Hmm, this requirement zero initial conditions is somewhat unrealistic, but soon I realised that this does not means that all changes are from 0 to some final value, it can be from from some initial value IV to some final value FV, linear system will behave same just like change from 0 to some new value (FV-IV).
So in my opinion, to obtain transfer function of PID we need to introduce new variable, called deviation varibale which is defined as current value minus steady state value:
co(t) = CO(t)-CObias, so new form will be:
co(t) = K*e(t)+Ki*integral(e(t))+Kd*de(t)/dt
CO(t) has initial condition CObias, but co(t) has initial condition zero, so Laplace transform can be applied.
Simulation software such as Simulink is good for analyse of process dynamics, that's why systems are often test to step change.
Can you, please, comment this?
Oki, that was theory, what about practice?
I have worked with Siemens FB41 PID ant that PID block doesn't have CObias and that is where this story begins to confuse me.
I feel I'm pretty close to completely understand this but need your help.
I have read in that same article that most tuning software has CObias value, but in FB41 I cannot see one.
I also tried to simulate CObias in Simulink (adding constant to CO output of PID block) but system behaves like that CObias is disturbance and finaly reach steady state after some time.
So I'm not really sure if this CObias is needed, because if PI is used, integral part represent continuous sum and hence that is CObias, if P-only is used, process will most likely have offset in it's output (response to step input), so controller will have some output value different from zero.
Did you have to manually set up this CObias or not?
This thing really bothers me.

If you made up this far, I thank you for your time and patience.
I hope your answers will help me to understand and overcome this issue.
I have set up only one PID controller in my life so far, and it worked well but at that time I didn't know that CObias exists. In Matlab and in my textboox there were only Kp, Ki and Kd.

Thank you very much and I'm looking forward to your replies.
 
Pandiani,

The terms Manual and Automatic can into existance from the old days when PID controllers were stand-alone instruments. Many times the instruments were not ideally tuned to the process. It was sometimes necessary, during testing and during process upset conditions, for the operator to be able to manually set the output, with the PID functions being bypassed. An Automatic-Manual selector switch was placed on the controller instrument. In Manual Mode, the setpoint value (0 to 100%) was sent directly to the output device in the form of a miliampere current or voltage signal.

In most PLCs, the old Auto-Manual switch is simulated with a bit mode switch that causes the PID function to be bypassed and the Setpoint copied directly to the Control Variable output.

I have not used LaPlace Transforms in over 40 years, and have forgot about all that stuff, thanks be to God! I will say that the LaPlace equations are for theoretical perfect systems, but the PID controllers have to work with real-world imperfect systems. There is a large difference. In the real world, there are variables that do not exist in the perfect theoretical world of the LaPlace Transforms.
 
Pandiani said:
Hello people,
I must admit I don't understand story about manual and automatic mode of operation completely.
Automatic is closed loop control where the feedback is used to compute the output Manual is open loop control where the output is controlled manually without regard to the feedback.

Is whole story about manual/automatic control tailored to some kind of software switch that break feedback path, or there's more to it.
It is a software switch.

Recently, after I have read one article about PID I saw the following form:
CO = K*e(t)+Ki*integral(e(t))+Kd*de(t)/dt + CObias
where CObias is defined as controller output when error is zero and disturbance is not present and needed to maintain PV depending of SP.
It also stated that CObias must exist because when error signal is zero there must be some controller output to maintain output PV.
I can understand that but how to obtain Laplace transfer function then?
The same way you always have. The CObias is what I would call a feed forward and on the S7 I think it is called the DISV or disturbance variable.

Why bias is not included in most textboox I have read?
Why many software for simulation like Matlab and Simulink doesn't include this bias value in PID block parameters?
I have know idea. Feed forwards are obvious to me. I wrote my first control algorithm using feed forwards before I even knew what they were called. I would include feed forwards or bias in my Peter Ponders PID book. ( don't hold your breath ).


Simulation software such as Simulink is good for analyse of process dynamics, that's why systems are often test to step change.
Can you, please, comment this?
Oki, that was theory, what about practice?
1. Simulink is good simulation and getting answers. However, what you need to know is how Simulink is getting the answers.
2. In real life you don't want to do step changes on 4000 Kg loads. You would be the center of attention and not in a good way.

I have worked with Siemens FB41 PID ant that PID block doesn't have CObias and that is where this story begins to confuse me.
It does I think. I haven't used FB41 but I think it is called DISV or the disturbance variable. That is a non-standard name.

[/quote]
I also tried to simulate CObias in Simulink (adding constant to CO output of PID block) but system behaves like that CObias is disturbance and finaly reach steady state after some time.
[/quote]
Yes, but you haven't explained what you are trying to do. It would really help so I can use examples that you can relate too. I do motion control so I have lots of motion control examples.

So I'm not really sure if this CObias is needed, because if PI is used, integral part represent continuous sum and hence that is CObias, if P-only is used, process will most likely have offset in it's output (response to step input), so controller will have some output value different from zero.
Did you have to manually set up this CObias or not?
This thing really bothers me.
Feed forwards ( bias ) are an essential part of motion control but not for temperature control and most other slow processes because there is plenty of time for the integrator to wind up to whatever value is necessary to reduce the error to 0. That is not the case for a motion controller that may be moving very quickly from point to point.

Lancie1 said:
I will say that the LaPlace equations are for theoretical perfect systems, but the PID controllers have to work with real-world imperfect systems.
There is a large difference. In the real world, there are variables that do not exist in the perfect theoretical world of the LaPlace Transforms.
I hear this a lot. I think I hear it most often from people that don't understand the theory. Part of understanding the theory is understanding what it can and cannot do. The theory works. I use it. In a imperfect world the theory may not work perfectly but it can be use to reduce the effects of the imperfections.

If the world was perfect then I can predict how exactly how much output is required and no feedback would be required.
 
Hello
Thanks for replies.
If I understood you correctly switching from Automatic to manual means bypassing
PID controller. I think that it must also break feedback path, otherwise system will
continue to work as closed loop system without controller.
Please can you see in attachment and correct me if I'm wrong. I think that (in principle)
switch from Automatic to Manual mode bypass PID controller and break feedback path.
Please can you confirm this?
Lancie1 said:
I have not used LaPlace Transforms in over 40 years, and have forgot about all that stuff, thanks be to God! I will say that the LaPlace equations are for theoretical perfect systems, but the PID controllers have to work with real-world imperfect systems. There is a large difference. In the real world, there are variables that do not exist in the perfect theoretical world of the LaPlace Transforms.
Uhh, this means that I have lost one year of my life for nothing :( !?!

I hope I'll somehow manage to make bridge between perfect theory and imperfect practice...
 
Re your diagram.

I would put a label between PID and Actuator called OUT. Then the Auto-Manual switch would be between the PID and OUT. The choices of source for OUT would be PID or User input. Often, the PID in not calculated when the loop is in Manual.

If I think about single loop controllers (eg Honeywell 3000, Watlow 988, etc.) there is a button for auto-manual. When the controller is in 'Manual' mode the user has to switch the display to 'out' and use arrow keys to increase/decrease the output. The setpoint is usually not affected and unused in %Out calculation, (although deviation alarms may still be active).

When switched to 'Auto' mode the behavior usually depends on setup. Often times the SP is set equal to the PV. And sometimes the internal PID variables are set so that the OUT from the PID is also equal to the OUT when the 'Auto-Manual' switch is made.

When implementing PIDs in practice it is VERY important to check the documentation of the particular controller you are using. WIll it use bumpless transfer? what does bumpless transfer mean? (E.g. AutomationDirect PLCs have at least 2 Bumpless tranfer modes. ( http://web1.automationdirect.com/static/manuals/d2user/ch8.pdf p.9,26 (see p22-24 for info on man-auto) ) If you're lucky, they will even include the actual calculations.
 
Peter Nachtwey said:
Pandiani, that isn't right. Yes, the PID is by passed but the signal comes from a manual input not from the error between the SP and PV.
Ok, so the scheme should look like in attachment (although that switch in feedback isn't necessary now).

However I'd like to make some things clear regading that CObias.
I'd like to post my models and simulation results, but don't know how many people here are interested, and maybe I can send these by email to those who are interested and discuss with them.
Thank you hanziou, I'll read that pdf file.
Thanks guys
 
Still not right

Your drawing indicates that the manual mode still has its signal go through the PID. Move the switch to the right side of the PID.

When going to auto you will need to to do as hanziou said. Set the integrator to the current control ouput minus what ever bias or feed forward you have and set the SP to the PV.
 
I'd like to post my models and simulation results, but don't know how many people here are interested,

many people will be interested ... even if not NOW then at some time in the future ... please post as much as you can ... that's how the forum grows - and provides help to those who need it ...

thanks ...

PS Edit: even though the brand is different, the FORMAT of the figures on page 14-40 of the Allen-Bradley PLC-5 Instruction Set Reference Manual might be interesting to you ...
 
Last edited:
I agree with Ron.

Test subject will be interested. He just said he would be diving into the s and z domains.

I think you should start a series of new threads labeled 'Advanced Control - subject' That way those that see Advanced Control will know it is not the normal PLC stuff. The 'subject' should fit the subject of the thread like 'Advanced Control - PID' 'Advanced Control - Feed Forward' 'Advanced Control - Tuning' 'Advanced Control - Feedback' etc. This way the thread can be found. I do this in my bit tricks threads. They are all marked with the word tidbits so I can find them again. The problem with this forum is that much good information is lost so the same topics are raised over and over and over again and instead of the forum saving time, it wastes time because people must repeat their answers. The 'Advanced Control' will also indicate to those that don't care that they can skip the thread.

So go for it.

Also, can you use Scilab instead of Simulink? Scilab is free so many can participate.
 

Similar Topics

Hello, I know this might seem simple but however I cant seem to get my head around it! What is the best way to have a manual and automatic mode...
Replies
4
Views
2,450
I have a question concerning automatic reset vs. manual reset of a safety relay. I'm wondering if configuring the relay for automatic reset...
Replies
14
Views
5,765
Dear All, Consider a selector switch is located at the electrical panel and it is wired to the PLC at I0.5 (When I0.5 is low = Manual , when...
Replies
1
Views
2,318
Hi guys, i was trying to make a program in simatic step 7 with which i could with the useing of a switch choose between automatic, manual and step...
Replies
0
Views
4,638
Hello everyone, I'm having trouble solving this one and was hoping someone could help. I have a AB CompactLogix L16ER PLC connected to a AB...
Replies
2
Views
551
Back
Top Bottom