Flow Loop Hunting After Being Put in Auto
A flow control loop has a magnetic flowmeter, a DCS PID controller, and a 4-in. air-to-open globe valve with a smart positioner. After commissioning, in AUTO the measured flow swings plus or minus 12% with a regular period of about 90 seconds. In MANUAL, holding the same valve position, the flow holds within plus or minus 1%. The valve strokes full travel in 45 seconds. Current tuning is gain 0.25 (proportional band 400%), integral time 0.02 seconds, derivative 0. The operator has been reducing the gain further and shortening the integral time, and the oscillation is getting worse. The positioner is in relay mode with no input filter, and the DCS has no PV filter.
1. Use the evidence to locate the cause of the oscillation and explain why the operator's tuning direction made things worse.
Read the worked answer
Manual versus auto is the decisive test. In MANUAL the PV holds within plus or minus 1%, so the transmitter, the wiring, and the process are not generating the swing. The loop is being driven unstable by the controller/valve combination.
The tuning is backwards for a flow loop.
- An integral time of 0.02 s is essentially no integral time at all. The controller behaves like a near-pure integrator with almost no proportional damping.
- A gain of 0.25 (PB 400%) is very weak proportional action - it cannot damp anything.
With the proportional term that small, the only things setting the loop's character are the integral action and the valve's 45 s stroke. A limit cycle settling at roughly 90 s - twice the valve stroke time - is consistent with valve/actuator travel dominating the period.
Why the operator made it worse: reducing gain further removed the last of the damping, and shortening the integral time made the output move even harder for the same error. More aggressive is not better here. The loop needs slower integral and more proportional action, not the reverse.
2. Propose a practical tuning and bring-up procedure, with starting values.
Read the worked answer
Bring the loop up systematically, starting in manual:
- Process reaction test. With the process steady in manual, step the controller output about 5% and record PV against time. From the dead time and the initial rate of change, estimate gain and the dominant time constant. This takes a few minutes and tells you more than any rule of thumb.
- Set derivative to zero. Flow loops are noisy and fast; derivative usually buys nothing and amplifies measurement noise.
- Start modest: gain around 0.5-1.0 (PB 100-200%) and integral time around 10-20 s, then place the loop in AUTO.
- Disturb and observe. Make a 5% setpoint change. Shorten the integral in steps (halve it each time) until the response shows roughly quarter-amplitude damping, then adjust gain to trade speed against overshoot.
- Stop when the response is repeatable and the valve is not being driven hard. Do not tune to one perfect step - tune for repeatability across a few disturbances.
Document final gain, integral, and derivative, and confirm no PV filter is left fighting the tuning. If the loop is later fitted with a filter, retune.
3. What mechanical and valve checks should be made before and after tuning, and what findings would change your conclusion?
Read the worked answer
Before tuning, prove the valve can respond smoothly:
- Step the output in 5% increments with the controller in manual and watch the valve position feedback. A deadband greater than about 1% followed by a jump is stiction, not a tuning problem. Tuning cannot cure stiction - it only changes the period of the limit cycle.
- Check air supply pressure against the positioner requirement; look for moisture in the air; verify the filter regulator.
- Verify positioner zero and span and its internal tuning. Confirm it is in the intended mode and not stuck in a partial-stroke test or manual mode.
- Check packing tightness, actuator linkage for lost motion, and whether the valve is oversized or running below 10% or above 90% output.
Findings that would change the conclusion:
- PV oscillates with the controller in MANUAL. The source is then the measurement or the process - air entrainment or electrode fouling on the mag meter, an upstream pump cycling, or interaction with another loop.
- A manual step produces a staircase response. It is stiction; the valve needs packing/linkage work, not new tuning numbers.
- The oscillation period changes when the positioner is bypassed. The positioner is the culprit.
Any of these means stop tuning and fix the hardware first.