PID Tuning Step by Step - Part 6

Quick refresh with previous parts: (1), (2), (3), (4) & (5), And we could continue ..

Last part we discussed "Ziegler-Nichols" methods, and now will continue with the second one ..

2 - "Trial & Error" Tuning method:

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

1. Put I and D actions to minimum, and put P action near to or at 1.
2. Bumping setpoint value up/down and try to change P until self-sustaining oscillations are achieved.
3. When the ultimate gain "Ku" is determined, reduce the aggressiveness of proportional action by a factor of two.

Till now these first three steps are identical to the "Ziegler & Nichols/Closed Loop" method.

4. Repeat steps 2 and 3, this time adjusting I action instead of P.

5. Repeat steps 2 and 3, this time adjusting D action instead of P.

Suppose we follow the above steps till the loop is stabilized, but this method is considered poor and it is not recommended!

The principle behind "trial-and-error" tuning technique that we can tune a PID controller by incrementally adjusting the aggressiveness of a controller’s P, I, and/or D actions until we see oscillations then start reducing the aggressiveness of the actions until stable control is achieved.

Some other practices (variation on the above steps that is untexted and also not recommended) are following the first 2 steps only without achieving full oscillation .. just manipulate P for some loop improvement, then switching to I manipulation and maybe D also till the loop is finally stabilized.

The main issue with this "trial & error" tuning that we need to know which "control action" to focus on and which one to avoid while tuning .. or in simple words which of (P, I or D) is to be aggressive/not aggressive. To get the answer for such a question, we need to recognize the suitability of P, I, and D actions to the process characteristics.

Random experimentation with P, I, and D parameters could be harmful or dangerous to the process. So the key is to understand the role, applicability and limitation of each action.

The improved technique for this "Trial & Error" Tuning is in the next method ..

3 - Recommended Tuning method:

After taking all the precautions mentioned in the previous part 5, go on with the following steps:

1. Perform open-loop test for the process - So put the controller in manual and introduce a step-change to its O/P signal. From PV curve response try to determine the process characteristics (e.g. self-regulating versus integrating, steady-state gain, noisy versus calm, dead time, time constant, lag order) and also try to identify any existing process problems (e.g. control valve with excessive friction, noisy instrument need to be damped, large dead time). Correct all problems before proceeding, this is critical as there is no "tuning" that will fix an issue with a component of the process loop.
2. Try to identify any controller actions that may be problematic (e.g. derivative action on a noisy process), and check if to minimize or completely turn it off.
3. Try to identify the controller’s “dominant” action to be tuned for loop stability. Consult the following chart for PV curve resulted from open-loop test:

Note: Aggressiveness mentioned in the table is limited by process tendency for oscillation.

4. Start tuning with all terms of the controller set for minimal response (minimal P, minimal I, no D).

5. Set the dominant action to some safe value (e.g. gain less than 1, integral time much longer than the time constant of the process) and check the loop’s response to setpoint and/or load changes in automatic mode.

6. Increase aggressiveness of this action until a point is reached where any more causes excessive overshoot or oscillation.

7. Increase aggressiveness of the other action(s) as needed to achieve the best compromise between stability and quick response.

8. If the loop ever shows signs of being too aggressive (e.g. oscillations), apply steps detailed in the following below section (Recognizing 2 famous process behaviours while tuning).

9. Repeat (6, 7 & 8) steps above as often as needed.

Recognizing 2 famous process behaviours while tuning:

a. To identify "Over-Tuned" Controller using "Phase Shift" sign:

While tuning we may have a case when one or more of “actions” (P, I or D) is not configured properly, which will lead to "sinusoidal oscillations": Damped, Undamped or something in between - as per the following graphs:

If we have such a case .. so it is a sign of "Over-Tuned" Controller, and one of (P, I or D) is needed to be reconfigured - The solution is summed up in the following:

• Compare "PV curve" vs Controller "O/P curve".
• Examine the phase shift between PV & O/P to identify the dominant action - This could be captured from the following:

• I & D actions introduce a phase shift between the PV & O/P. This phase shift direction will reveal which time-based action (either I or D) dominates the controller’s response and is therefore most likely the cause of this oscillation.
• So if O/P is in phase with PV (There is no phase-Shift) .. This is a sign of a dominant P controller action, and P here is too aggressive and need to be reduced for stability.
• If O/P is phase-shifted 90o behind PV (lagging) .. This is a sign of a dominant I controller action, and I here is too aggressive and need to be reduced for stability.
• If O/P is phase-shifted 90o after PV (leading) .. This is a sign of a dominant D controller action, and D here is too aggressive and need to be reduced for stability.
• For some cases, multiple PID terms are set too aggressively, and this way the phase shift will be splitted somewhere between 0o and ± 90o. In such cases, one must “round up/down” the phase shift to the nearest value of ?90o, 0o, or +90o in order to determine which of the three actions (I, P, or D, respectively) is the dominant and most responsible for the oscillations.
• If the PID controller is "Reverse Action" so we will have the following curves:

b. To identify "Porpoising" Controller:

Another interesting case of "Poor-Tuned" controller is when we have the following response of the process loop:

The loop in this case is known as "Porpoising" which is a poor loop behaviour, combining "Over-Tuning" (instability) and "Under-Tuning" (delay achieving SP).

P&D are the only accused controller actions here .. as they are the only parameters capable of producing a "directional change" in PV curve prior reaching SP, while I action is spared from this as it always drives PV in one direction towards SP. So this "Porpoising" curve is a good sign for (P or D or both) are needed to be reconfigured.

Some "General Rules" to be in the background while tuning:

a. Using D action while tuning allows a bit more P and I actions to be used (sometimes more aggressive P/I) since derivative (D) will act to limit the overshoot.

b. In a purely integrating process, using I action will guarantee overshoot as a response to "setpoint change", but I action is needed anyway to compensate for "load change".

c. With Slow process (Large Lag Time):

• Less I action is to be used, as aggressive I could result in oscillation due to wind-up.
• D action works well with slow process.

d. With Fast process (Small Lag Time):

• I action works well with fast-acting process.
• D action could be used with caution, avoid causing oscillation.

e. If the process is “noisy”, try to reduce the noise - and after all, if still noisy:

• Use P action sparingly, as P reproduces process noise onto the output signal.
• Use no D action, as it amplifies the process noise.
• I action has the ability to ignore process noise.

Final word to say .. "Learning curve" of PID controller tuning is continuously enhanced by practising. Hope we had in this series some general rules and basic criteria of what to do.