PID Tuning Step by Step - Part 5

Quick refresh with previous parts: (1), (2), (3) & (4) - And we could continue ..

Q22. What are the precautions to be taken before tuning?

We could sum up the most important precautions in the following:

1 - Be prepared to document the "tuning process" work including: original PID controller setting, attempted PID setting values during tuning or any screenshot images for the trends. It is always useful to utilize a recording pre-prepared templates for the tuning process documentation.

2 - Identify the process control strategy which includes the PID controller to be tuned (Cascade, Ratio, Feedforward .. etc.) as each strategy may need a special treatment (eg. Cascade loop will need to tune the secondary/slave PID controller before primary/master one)

3 - Check with operators if the operation is in a “normal” condition or not before attempting to improve its performance. (check for equipment shutdown, flow re-routing, different production rates, etc.)

4 - Check with operators the process/system hazards and SD limits .. So you could estimate your working range and what to avoid during tuning.

5 - Check with operators the process needs, like: accepted error range, accepted response speed to SP/Load change, accepted overshoot/undershoot/oscillation ranges.

As we can see .. Discussion/Investigation with Operation is critical important before start tuning .. as it may shorten a lot of time and efforts.

Will stop the questions here and to continue in normal writing ..

Before getting into the tuning steps, it is better to indicate that putting a procedure for the tuning is meant only for simplicity. Just to make the tuning has a general plan. But tuner shouldn't mindlessly follow procedure steps without having a good understanding for what they are doing and why they are doing. Tuner needs to have a good understanding and analysis for the process type and its dynamic characteristics.

We can categorize the most famous tuning methods into the following:

1 - "Ziegler-Nichols" tuning methods:

It was a paper published in 1942 by two engineers named Ziegler and Nichols, included the recommendations for PID controller tuning. Still this method could be found in some modern control references. It led directly to exact numerical values to be used for P, I & D in PID controller, but practically .. it has a serious implementation issues.

As per the above figure "Ziegler-Nichols" tuning method has 2 main techniques - We could discuss the "Closed Loop" one in the following:

a. Closed Loop Tuning:

We could sum up the tuning steps in the following:

1. Control loop will be tuned while the controller in Automatic and the process is live.
2. Disable "I" & "D" actions of the controller.
3. Raise "P" action just far enough self-sustaining oscillation in the PV curve occurs - as per below graph:

This "P" value now - which caused this oscillation - is recorded and it is known as "Ultimate Gain" or "Ku".

4. To calculate the time period between 2 successive oscillation peaks which is "Ultimate Period" or "Pu".

5. In case of P only controller, the "P" or "Kp" value needed for tuning is as per the following:

6. In case of PI controller, (P & I) needed values will be as per the following:

7. In case of PID controller, (P,I & D) needed values will be as per the following:

A very good illustrating videos series is indicating "Ziegler-Nichols" closed-loop tuning method using "Simulink" simulation software - I recommend watching this series here (length of all 3 parts are less than 10 minutes) to have a good illustration about this method.

Final to say about "Ziegler-Nichols/Closed Loop" method .. that it is somehow impractical. As to determine the "Ultimate Gain", it is requested to make the process to oscillate .. Or in simple words, the system must be brought to the brink of total instability - which may trouble the process and cause potential equipment hazards. This is why "Ziegler-Nichols/Closed Loop" method has many practical limitations.

Now switching to the second method of "Ziegler-Nichols" which is "Open Loop":

b. "Open Loop" Tuning:

We could sum up the tuning steps for this tuning method in the following:

1. The investigation this time will be done in manual mode, and a "Step-Change" will be introduced to the controller output to the valve for process analysis.
2. Get "Dead Time - L" & "Reaction Rate - R" from PV curve as per the following graph:

3. In case of P only controller, the "P" value needed for tuning is as per the following:

4. In case of PI controller, (P & I) needed values will be as per the following:

5. In case of PID controller, (P,I & D) needed values will be as per the following:

"Open Loop" as a tuning method is less disruptive to the process than "Closed Loop", but it relies mainly on the presence of "Dead Time" in the process, which may be insubstantial.

Besides .. both "Open Loop" & "Closed Loop" are impractical with other process factors like process noise.

However, both methods should be taken as a starting points rather than definite answer to (P, I & D) tuning values.

In the next part we will continue with the rest of techniques to tune ..