Control valve hunts in AUTO but runs smoothly in MANUAL
If a valve cycles in automatic and goes quiet in manual, the feedback path is generating the cycle. What that rules out, what it leaves, and how to finish the analysis.
Read a control loop from its symptom
Start from what the loop is actually doing. Each page works through the causes in the order they usually turn out to be the answer, and says plainly which ones tuning cannot fix.
Valve position oscillates but the process variable barely moves
When the output swings and the process does not follow, the loop is working hard for nothing. The four causes, and how to tell them apart from a trend.
Control valve hunting
A hunting control valve is usually mechanical, not badly tuned. How to tell stiction, oversizing and aggressive tuning apart from the trend, and what to check first.
PID loop oscillating
PID oscillation tells you the cause through whether the cycle decays. How to tell decaying, growing and constant-amplitude oscillation apart.
Valve stiction symptoms
Stiction produces a distinctive limit cycle: stepped PV against a smoothly ramping output, constant amplitude, gone in manual. How to confirm it.
Control loop will not hold setpoint
A persistent gap between PV and setpoint has a small number of causes: no integral action, output saturation, or a loop that has lost the selection.
Controller output saturated
A PID output stuck at 0% or 100% means the controller has no remaining authority. Why it happens, why windup follows, and what to check first.
Flow loop unstable
Flow loops are fast and noisy, which makes them intolerant of derivative action and of reset set faster than the loop's dead time. What to check.
Integral windup symptoms
The output stays on its limit after the process variable has crossed setpoint, then overshoots hard. How to tell windup from aggressive tuning.
Overshoot after a setpoint change
Overshoot on setpoint changes but not on load changes points at controller structure rather than tuning. How to tell windup and derivative kick apart.
Sluggish control loop
A slow loop is not always under-tuned. A filter, a drive ramp or the tuning itself can all produce sluggishness, and only one of those yields to retuning.
Cascade loop hunting
A cascade that hunts is usually two loops being moved at once, or a slave that is not actually faster. How to tell which loop is generating the cycle.
Control valve passing
The controller output is at 0% and the process variable still will not come down. A passing valve is a leak, not a PID.
Derivative kick on setpoint change
A setpoint step produces an output spike when derivative acts on error. The fix is derivative on PV, not less gain.
Interacting control loops
Two loops sharing a process will hunt when each treats the other as a disturbance. How to tell interaction from a single bad tune.
Noisy PV and valve chattering
A chattering valve is usually derivative action amplifying measurement noise. How to tell that apart from real process noise and from an instrument fault.
Pump speed loop hunting
On a VFD-driven loop the drive's own accel and decel ramps are often the dominant lag. What to check before re-tuning a hunting speed loop.
Valve deadband and backlash
Backlash shows as a flat spot on the PV every time the output reverses. How to measure the deadband and what causes it.
Won't transfer to AUTO
Switching to automatic makes the output step. Bumpless transfer has failed. What to check before you blame the tune.