Lean Amine Flow Loop Cycles After a Valve Trim Change
Lean amine flow to an absorber is controlled at 85 m³/h by a 3-inch globe valve with a pneumatic piston actuator and a smart positioner. The flow is measured by an orifice plate and a DP transmitter. The DCS uses a PI flow controller with gain 0.8 and integral time 30 s. After the valve trim was replaced with a larger trim to reduce pressure drop, the flow began cycling between 76 and 93 m³/h with a period of about 45 s. The controller output moves as a sawtooth.
The positioner stem-position feedback shows a 4–6% deadband. The absorber outlet H2S has ranged from 2 to 9 ppm against a 4 ppm target. Operators reduced the controller gain to 0.3 and increased the integral time to 120 s; the cycle amplitude fell to about ±4 m³/h, but H2S excursions continue. With the controller in manual at a fixed output, the flow is steady. A 5% step change in controller output produces no flow change for 12 s, then a sudden jump.
1. Diagnose whether this is a tuning problem, a transmitter problem, or a valve problem. Explain why reducing the controller gain did not solve it.
Read the worked answer
The step test is decisive: a 5% output change with no flow change for 12 s and then a jump is classic stiction or deadband, not tuning. The sawtooth controller output is the integrator winding up until the valve breaks free, then overshooting; the 45 s period is set by the valve, not by the process. In manual the flow is steady because the valve is parked at one position. The larger trim likely made the valve oversized, so it operates at low lift where stiction and nonlinearity are worst.
Reducing gain to 0.3 and increasing integral time to 120 s reduced the amplitude but slowed the correction, so the H2S excursions continued. The root cause is mechanical, not controller tuning. A transmitter problem would show a steady offset or noise, not a command-dependent deadband and a sawtooth output.
2. Describe the tests you would run to confirm the diagnosis and quantify the valve problem.
Read the worked answer
Run a valve signature test: stroke the valve in 5% steps and record command versus stem position and flow. A 4–6% hysteresis and a deadband confirm stiction. Compare the positioner setpoint and stem feedback; if the positioner is in pressure-control mode, it can mask friction. With the controller in manual, make a 5% step and record the flow response; no response for 12 s confirms the deadband. Check the valve Cv and normal opening: if it is below 20%, the larger trim is oversized and should be replaced. Inspect the packing, stem, and actuator; sticky packing or a worn plug can produce the same signature. Also verify that the orifice DP transmitter is not plugged and that the flow signal is stable in manual.
3. Recommend a permanent fix and a safe temporary measure.
Read the worked answer
Fix the mechanical cause: overhaul the valve, replace the trim with a characterized trim sized for 30–70% opening at normal flow, and install a smart positioner with automatic tuning and stiction diagnostics. If the valve is oversized, a smaller trim or a V-ball valve will improve control. After repair, re-tune the flow loop at normal and minimum flow and test the H2S response.
Temporarily, keep the loop in auto with a low gain but add flow high and low alarms and use the H2S analyzer to trim the flow setpoint. Do not add a dither to the positioner without checking packing wear. A cascade from the H2S analyzer to the flow controller is not a substitute for a functioning valve, and it must be designed with the analyzer lag in mind.