Advanced
Tuning from scratch, systematically
Before you start
No presets. A repeatable procedure, and a trade-off to defend.
Everything so far has been one idea at a time. This is the whole job: an untuned loop, a disturbance you did not choose, and a result you have to justify.
Use a procedure rather than intuition. Establish a repeatable experiment so two runs can be compared at all. Set Ki and Kd to zero. Raise Kp until the response is fast but still calm. Add just enough Ki to remove the offset. Add Kd only if the measurement is clean and the response genuinely needs damping.
Judge the result on more than error. A controller that tracks beautifully by moving the valve constantly is not better; it is trading equipment life for a prettier trace. Read tracking error and valve travel together, and be able to say which one you chose to favour and why.
Published tuning rules (Ziegler-Nichols, Cohen-Coon, relay autotuning) give you a starting point from a measured response, not a final answer. They are opening bids.
Your task
- Both loops start with proportional action only, and a low gain.
- The oil valve loses 40% of its capacity at 45 seconds. You are not told this during the run; the level will tell you.
- Tune both loops to hold their setpoints through the restriction, without spending the valves to do it.
Configure both controllers before starting. Gains are locked during the attempt.
SEP-101 horizontal separator with gas, oil, and water. True total level 0.0 percent; measured total level 0.0 percent. True interface 0.0 percent; measured interface 0.0 percent. Total inlet 0.0 cubic metres per hour; oil outlet 0.0 cubic metres per hour; water outlet 0.0 cubic metres per hour. Actual oil valve 0.0 percent; actual water valve 0.0 percent. Safety: Filling: low-level protection bypassed.
SEP-101 · engineering fallback
Separator process cutaway
- Total true
- 0.0%
- Total measured
- 0.0%
- Total SP
- 50.0%
- Interface true
- 0.0%
- Interface measured
- 0.0%
- Interface SP
- 20.0%
- Total inlet
- Stopped / 0.0 m³/h
- Oil outlet
- Stopped / 0.0 m³/h
- Water outlet
- Stopped / 0.0 m³/h
- Oil command
- 0.0%
- Oil actual
- 0.0%
- Water command
- 0.0%
- Water actual
- 0.0%
- Active safety
- Filling: low-level protection bypassed
2D engineering fallback · process values and safety logic remain fully active.
Total level / oil valve
More valve opening → more oil out → lower total level
- Total level process value
- 0.0%
- Total level measured value
- 0.0%
- Total level setpoint
- 50.0%
- Oil valve command
- 0.0%
- Oil valve actual position
- 0.0%
- P / I / D
- 0.0 / 0.0 / 0.0
Advanced settings
Interface / water valve
More valve opening → more water out → lower interface
- Interface process value
- 0.0%
- Interface measured value
- 0.0%
- Interface setpoint
- 20.0%
- Water valve command
- 0.0%
- Water valve actual position
- 0.0%
- P / I / D
- 0.0 / 0.0 / 0.0
Advanced settings
Synchronized historian
Live control trends
Oil/water interface
Truth, LT-102, SP and protection limits · %Phase flows
Feed and phase flows · m³/h; water cut · %Valve response
Commands and actual positions · %Four synchronized trends compare both controlled levels, phase flows, water cut, and both valve responses.
Debrief
Run the exercise to see how you did.