Drum Level Excursion After Feedwater Valve Swap
A 45,000 kg/h packaged water-tube boiler runs on a single-element feedwater control loop (drum level transmitter to controller to feedwater control valve). During a maintenance outage the 3-inch equal-percentage feedwater control valve is replaced with a same-size linear-characteristic valve; the pneumatic actuator and positioner are reused without recalibration. On restart at 70% load the operator observes the drum level cycling roughly ±40 mm about setpoint with a period near 90 seconds, even though steam demand and feedwater pressure are steady. At low load (30%) the level is stable, and at 90% load the swings grow to about ±70 mm. The level transmitter was calibrated before the outage and reads correctly against the gauge glass, and the feed pump discharge pressure is steady. Drum level low-low trips at -150 mm.
1. Identify the most probable root cause of the level cycling and explain the mechanism by which the valve change produced load-dependent instability.
Read the worked answer
The valve characteristic change is the prime suspect. The loop was tuned around an equal-percentage valve, whose gain rises steeply as it opens; the linear valve has nearly constant gain across its stroke. With the original valve, loop gain at high load was moderate because the valve was in its lower (low-gain) region and the controller output was matched to it. With a linear valve, the same controller output change produces a much larger flow change at high load, so total loop gain increases with load.
A feedback loop becomes oscillatory when loop gain exceeds roughly the point where the controller's proportional and integral action over-correct. At 30% load the valve sits low in its stroke, flow gain is small, and the loop is stable. At 90% load the linear valve's flow gain is at its maximum while the level process gain (flow error to level change) is also large, so the product exceeds the stability margin and the loop hunts. The ~90 s period is consistent with integral action winding up and unwinding against the drum's integrating level response.
The transmitter and pump are excluded because level reads true and supply pressure is steady, so this is a control-loop gain problem, not a measurement or supply problem.
2. State the immediate corrective actions you would take before returning the boiler to automatic control, and the longer-term fix.
Read the worked answer
Immediate: place the feedwater valve in manual and hold drum level at setpoint by modulating the valve by hand, watching the gauge glass and trend. Do not leave the loop in auto with ±70 mm swings against a -150 mm trip; margin is only about 80 mm. Confirm the positioner is calibrated to the new valve's stroke and that the valve fails in a safe direction. Then retune the PID: reduce proportional gain and increase integral time so the loop is stable at the highest expected load, and verify stability at 30%, 70%, and 90% load. If a single tuning cannot cover the range, schedule gain scheduling on load.
Longer term: either reinstall an equal-percentage trim (or a characterized cage) so valve gain matches the original design, or formally retune and document the loop for the linear characteristic, including a load-swing test. Also verify the single-element loop is adequate; at high load with varying demand, a three-element (steam flow, feedwater flow, level) scheme would reduce sensitivity to valve gain. Record the change in the maintenance and management-of-change file.