Reflux Drum Level Cycling Caused by Valve Stiction
A reflux drum level controller on a distillation column uses a PID with gain 2, integral time 10 min, and derivative 0. The level has cycled with a period of 22 min and a peak-to-peak amplitude of 0.35 m for three days. The controller output moves slowly upward for about 18 min, then drops by 6% in less than 10 s. The valve is a 4-20 mA globe valve with a pneumatic positioner. A portable stem-position test shows that a 5% change in controller output produces no stem movement, followed by a sudden jump of about 5%. When the controller is placed in manual at the current output, the level stabilizes within 10 minutes. The drum is 2.4 m in diameter. Normal distillate flow is 80 m3/h. Operations want to retune the PID to stop the cycling.
1. Diagnose the likely cause using the evidence. Distinguish it from a tuning problem, and check the amplitude with a simple estimate.
Read the worked answer
The sawtooth output, the slow ramp followed by a fast drop, and the stem test showing a 5% deadband with a sudden 5% jump are the classic signature of valve stiction. In manual, the level stabilizes, which rules out a sustained process disturbance and points to the controller-valve loop. A tuning problem alone would not produce a stem that ignores a 5% output change. Estimate the level effect: 5% of 80 m3/h is 4 m3/h. Drum area = pi/4 x 2.4^2 = 4.52 m2, so the level rate is 4/4.52 = 0.88 m/h. Over a 10-20 minute half-cycle, that gives 0.15-0.30 m, consistent with the observed 0.35 m peak-to-peak amplitude. The loop is limit-cycling because the controller integrates error until the valve breaks free, then the flow overshoots and the level reverses.
2. Explain why retuning cannot eliminate the cycle, and propose temporary and permanent fixes.
Read the worked answer
Retuning changes the frequency and amplitude of a stiction limit cycle but cannot remove the deadband in the valve. Lowering gain and increasing integral time will make the output ramp more slowly, so the valve jumps less often and the level excursion may be smaller, but the loop will still cycle and level control will be sluggish. A temporary fix is to detune the controller, or run the level in manual with an operator adjusting the valve, while the valve is repaired. A better temporary measure is to install or repair a positioner with friction compensation and verify the stem test again. The permanent fix is to overhaul the valve, replace the packing and guide, and size the trim for the normal flow so the valve does not operate near the seat. After repair, repeat a stem-friction test and then retune the level loop with a proper averaging-level objective.
3. If a positioner with friction compensation is installed, what else should be checked in the loop before returning it to automatic?
Read the worked answer
Confirm the valve strokes smoothly through its full range and that the positioner zero and span match the 4-20 mA signal. Check for mechanical issues such as a bent stem, tight packing, or a sticking plug, because compensation can mask rather than cure these. Verify the controller output limits and anti-windup so the integral term does not wind up during manual operation. Check that the level transmitter is calibrated and not affected by density changes. Review the valve sizing: if the normal flow requires less than about 20% of travel, the valve will be too sensitive and may cycle again. Finally, perform a bump test in manual, identify the process time constant, and set the level controller for averaging control rather than tight level control.