Flow Loop Cycling After a Control Valve Replacement
A hydrocracker charge-flow loop uses a 6 in equal-percentage ball valve, a 4–20 mA positioner, and a 0–120 t/h magnetic flowmeter. After the original butterfly valve was replaced with the ball valve, the flow oscillates ±12% at about 0.25 Hz. The controller is P=0.8, I=15 s, D=0. A bump test shows 1.5 s dead time and a 4 s time constant from controller output to flow. A valve friction test shows 3% hysteresis, and the positioner output is hunting by about 2%. At normal flow the valve runs at 58% open. The operator wants to halve the gain and double the integral time. The replacement was justified because the old butterfly valve eroded, but the new valve has a different installed characteristic and is one size larger than the line. Production is pushing to keep the unit online; the loop is the feed control to a critical reactor.
1. Diagnose the likely cause of the oscillation and evaluate the proposed tuning change.
Read the worked answer
The oscillation period is about 4 s. That is on the same order as the 1.5 s dead time and 4 s process time constant, so it is too fast to be a classic integral windup cycle. The valve friction test shows 3% hysteresis and the positioner output is hunting at 2%, which points to a valve/positioner limit cycle rather than a controller that is simply too aggressive. An oversized ball valve running at 58% open also has a high installed gain, so small output changes produce large flow changes and aggravate the cycle.
Halving the gain and doubling the integral time may reduce the amplitude, but it treats the symptom. It would slow the loop, degrade disturbance rejection, and leave a mechanical defect in service on a critical reactor feed. The better sequence is to confirm the valve signature, check air supply and positioner gain, inspect packing friction and actuator sizing, and compare the valve Cv with the required range. If the valve is oversized, a characterized trim or a smaller trim may be needed. Temporary detuning is acceptable only as a documented online measure with a defined repair date.
2. What tests and calculations would you use to separate valve problems from tuning problems?
Read the worked answer
I would run a closed-loop step test at the normal operating point and record controller output, valve position, and flow on the same time base. From the bump test, the process gain Kp, dead time θ = 1.5 s, and time constant τ = 4 s give a starting point for an IMC or lambda tuning calculation, for example λ = max(θ, 0.5τ) = 2 s, with Kc = τ/[Kp(λ+θ)] and Ti = τ. But I would first run an open-loop valve signature: command 10% steps up and down and measure stem position versus command. Hysteresis of 3% with 2% positioner hunting confirms a mechanical or positioner loop problem.
I would also check whether the flowmeter damping and DCS scan rate are appropriate, whether the controller output is hitting limits, and whether the valve is operating in a high-gain part of its characteristic. If the valve signature is clean, then retuning is justified. If it is not, the tuning calculation is secondary because the valve is not delivering the commanded position. The decision should be based on the signature, not on the operator’s preference for softer tuning.
3. What would you recommend to operations and management while the unit remains online?
Read the worked answer
I would recommend a controlled interim mode: reduce the loop gain modestly, add a small output filter if the DCS supports it, and place a high-high/low-low flow alarm with a clear operating limit. I would also ask operations to avoid large setpoint moves and to monitor valve position for hunting. The interim settings should be recorded in the management-of-change system with a repair date and a named owner, not left as an undocumented tuning change.
To management I would present the risk plainly: the loop is cycling because the replacement valve is oversized and has friction/hunting, and detuning only masks it. I would propose a valve signature and a trim or positioner repair at the next available window, with a fallback of a smaller characterized trim. If the reactor feed is safety-critical, I would also review the consequence of a flow excursion and confirm that independent alarms and interlocks are functional. The recommendation should protect production without pretending the mechanical issue has been solved.