Interacting control loops
Two valves on one process do not fail independently. One loop's output is the other's load. A level and a flow off the same vessel, two temperatures on coupled exchangers, a pressure and a flow on the same header — retuning the loop that is complaining often makes the quiet one start. The complaint then moves.
The manual test still works, but you have to freeze both. Put one in manual and watch the other. If the hunt stops, you were looking at interaction, not at a single set of constants.
Freeze one, watch the other
If loop A sits when loop B is in manual, B was driving A. If both still move with one frozen, look for a third player — a supply that cycles, a sticking valve, a recycled stream. Do not retune A while B is still in auto and fighting it.
Structure beats constants
Relative gain, decoupling, ratio and cascade exist because interaction is a pairing problem. Two equally fast PIDs on strongly coupled variables will fight at any reasonable tune. Make one slow (the less important inventory) and one fast (the flow that should absorb the swing), or put them in a cascade or ratio so only one PID sees the shared process.
What causes it, most likely first
Two loops of similar speed on a shared process
most common tuning can fix this
Each controller treats the other's moves as a disturbance and corrects them, which is the disturbance for the first.
- How to confirm it
- The hunt period matches neither loop in isolation. Freezing one stops or dramatically slows the other.
- What to do
- Slow the less critical loop — often a level or a pressure that can float — and leave the flow tight. Or restructure as cascade or ratio.
Wrong pairing
common tuning cannot fix this
Each valve has more effect on the other loop's PV than on its own. No amount of detuning fixes a pairing that works against the process.
- How to confirm it
- A bump on valve A moves PV-B more than PV-A. Relative-gain thinking, even informal, shows the pairing is inverted.
- What to do
- Swap the pairing or add a decouple. Constants will not.
A mechanical problem blamed on the neighbour
common tuning cannot fix this
One valve is sticking. The other loop sees a cycling load and is retuned for it.
- How to confirm it
- The constant-amplitude cycle survives with the neighbour in manual.
- What to do
- Fix the sticking valve. Stop retuning the victim.
Questions that come up
Should I tune them one at a time with the other in auto?
No. Tune the fast one with the slow one in manual, then put the slow one in auto with conservative constants. Tuning both in auto is how the fight is maintained.
Related
Last reviewed 2026-09-10.