A first order model looks like this. The blue curve represents the response of the process variable for a step change in the manipulated variable.

Time Delay Td – The time it takes to for the response to start , after the change in the manipulated variable

Time Constant – The time it takes for the process variable to reach 63 % of the final value

Gain – K Although the technically the gain means the change in the process variable for a unit change in the manipulated variable , here we are interested in % Change in the PV range / % change in the MV Range PV refers to the process variable and MV refers to the manipulated variable.

Once we have the model parameters , we can calculate the tuning parameters using any of the following methods. .

1. Zieglar- Nichols Method
   

   
   

2. Shinskey Tuning parameters
   

   
   

3. Lambda tuning
   

The Lambda tuning method is based on the desired closed loop time constant. Lambda tuning works better for set point response but the disturbance rejection is poor.

Where is the desired closed loop time constant.

Following diagrams show poorly tuned loops , the first is the sluggish tuning and second is the aggressive tuning.

Picure :1 Example of poorly tuned loops ( sluggish )

Picture 2 – Example of poorly tuned loops (aggressive)

The difference between sluggish tuning and aggressive tuning is that , in sluggish tuning , there are variations , but not much oscillatory , but in aggressive tuning , the variations are oscillatory.

Let us assume that , as a process control engineer, you are have been asked to retune all the 200 PID loops in the plant by the operations department.

Now where to begin ?

How to identify , if a loop needs tuning or not ?

If tuning is required, what is the Control objective ?

Which tuning method to follow ?

First step is to prepare the list of loops and note down the existing tuning parameters .

Then discuss with the operators and prepare a table as follows

Note that different vendors use different terminologies for the P, I , D parameters

Refer to their manuals to understand the terms first before you start tuning.

Next step is to

and look for the loops that are in manual mode due to tuning problems

Discuss with the operator, why the loops are kept in manual mode.

Some times , engineers come to conclusion that all the oscillating loops are poorly tuned. But it is not so .

Oscillation can be either due to cyclic disturbance or valve stick-slip ( what is known as limit cycles) or poor tuning.

To check , whether the tuning is the culprit for the oscillation, simply take the loop in manual mode.

If oscillation stops or reduces , then tuning is the problem. Else , it is not.

From the control point of view , these above loops (or processes ) can be classified into two categories.

• Stable (or Self regulating) loops
• Integrating loops

Stable loops are those, which reach a steady stage for a given change in the manipulated variable.

Stable Loops

Generally speaking , Flow , analysers, temperatures are stable loops.

Integrating loops don’t reach steady state Eg level and some pressure loops

Figure 2 : Integrating loops

Flow loops have faster response time , where as the response time of the temperature loops may vary from the order of few seconds to few hours.

Much has been said and written about the PID tuning.

Some people argue that tuning is best done manually and can never be done by computer programs.

For the beginners, in process control , one of the basic questions will be “can PID loops be really tuned by softwares?” ( well, atleast , I had this question in my mind, during my initial years in this industry)

Is it true ?

Well it is both true and false .

True in the sense, for a given control objective , it is possible to “calculate” the tuning constants for the specific PID algorithms by computer programs

And False in the sense, it is the user who has to decide the control objective , because the software can not.

What is meant by control objectives ?

Well we know , that a loop can be tuned “slowely” to get a smooth response or tuned aggressively to quickly reject the disturbances. So there exists a trade off between the “robustness” and “aggressiveness” . 

So this “degree of robustness or aggressiveness” must be measureable by some parameter , like “Closed loop time constant ” or “Integrated absolute error” or % overshoot etc. If we know this , we know our control objectives then we can calculate the tuning constants.

So the main challenge in tuning is to decide the control objective and not the tuning itself.

In the next post, we will see , how to decide about the tuning objectives and select the appropriate tuning methods.