Level Loop Hunting After a Transmitter Change
A 4–20 mA level transmitter on a 2.5 m storage tank was replaced during a shutdown. The new transmitter is ranged 0–2.5 m, but the PLC analogue input block was left at the previous 0–2.0 m range. The level controller drives a 50 mm equal-percentage control valve on the make-up line. The make-up pump develops 35 m at design flow, and the valve drops 5 m when fully open at design flow. The operator reports the level cycling ±0.3 m with a period of about 40 s, and the trend shows the valve output swinging between 10% and 90%. The overflow nozzle is at 2.1 m, the level setpoint is 1.5 m, the controller gain is 4 and the integral time is 20 s.
1. Quantify the effect of the mis-scaled input and its consequences.
Read the worked answer
The error
At an actual level of 2.0 m the transmitter sends 2.0/2.5 = 80% of span, i.e. 16.8 mA.
The PLC interprets 80% of a 0–2.0 m range and displays 1.6 m.
So indicated level = 0.8 × actual level, everywhere on the scale.
Consequences
- Overflow. With a 1.5 m setpoint the true level is 1.875 m - only 225 mm below the 2.1 m overflow nozzle. With a ±0.3 m cycle, the tank is overflowing on every peak. The controller is not just reading wrong; it is holding the tank at the wrong place, and its own setpoint now sits closer to the nozzle than the cycle amplitude.
- Alarms and interlocks. Any high-level alarm at, say, 1.9 m indicated actually trips at 2.38 m - above the nozzle, i.e. never. A low-level trip at 0.3 m indicated actually trips at 0.375 m, which runs the make-up pump dry or trips the transfer duty later than intended.
- Loop gain. The wider span reduces the %-signal produced by a given real level change to 80% of what it was, so the effective loop gain drops by 20%. That makes the loop slightly more sluggish - it does not explain the oscillation. Do not let the range error take the blame for the cycling; it is a separate fault that makes the cycling dangerous rather than merely annoying.
Action: correct the PLC range, then verify with a simulated 4/12/20 mA and an independent dip-tape at two points, and re-check every alarm and interlock that uses the signal.
2. Explain why the valve is causing the limit cycle, using the installed authority.
Read the worked answer
Valve authority
β = ΔP across the valve when fully open at design flow / total system ΔP
= 5 / 35 = 0.14
A practical target is β ≥ 0.3, and 0.3–0.5 is the usual band for a well-behaved installed characteristic. At 0.14 the valve is oversized for the pressure drop it has been given.
What that does at part load
At 50% flow the fixed system losses fall with the square of flow: 30 × 0.25 = 7.5 m. The pump still delivers about 35 m, so the valve now has to absorb 35 − 7.5 = 27.5 m - 5.5 times its full-open drop. To modulate in just that region, the plug moves a small amount and the flow changes a lot: the installed gain is very high near the closed end and very low near the open end. An equal-percentage trim makes the distortion worse, not better.
Now put controller gain 4 on top. A gain that is stable against the low-gain end of the characteristic is far too aggressive at the high-gain end, so the loop overshoots, the integral term winds, the valve slams, and it repeats - a classic limit cycle, and exactly the 10–90% swing seen on the trend, with a period set by the valve stroke and the 20 s integral time. The 40 s period is a fingerprint of the valve and the integral setting, not of the tank.
Fixes, in order: reduce the trim or fit a smaller valve so that it drops about 0.35 × 35 = 12.25 m at design flow; or run the make-up pump on a variable speed with a fixed valve drop; or fit a characterised segmented V-port plug. Then retune - a valve change alters the process gain and invalidates the existing tuning.
3. Write the commissioning and handover plan that would stop this recurring.
Read the worked answer
Instrument and signal
- Correct the PLC input range to 0–2.5 m and record the as-found and as-left values.
- Verify with a loop calibrator at 4, 12 and 20 mA, then cross-check against an independent dip tape at two levels. Sign the calibration sheet.
- Audit every alarm, interlock, totaliser and operator display that consumes the signal - the scaling error propagates into all of them.
- Fit an independent high-high level switch on the overflow protection; protection should never rely on the same transmitter that the controller uses.
Valve and actuator
- Verify the actuator stroke and the position feedback against the command, and record the fail-safe position. For a make-up line, fail-closed is normally what protects the tank.
- Confirm the installed characteristic by stepping the output 0, 10, 25, 50, 75, 100% and logging the resulting flow. Compare it with the authority calculation. Record the result.
Controller
- Do a step test at the normal operating level: a 5% output step, record the dead time and the time constant, and fit the model. Then tune - IMC or lambda tuning with a robustness factor - rather than copying gains from the previous installation.
- After the valve fix, repeat the step test. Document the gain, integral, derivative, filter and the model the tuning assumes.
Handover
- Trend for 24 hours at low, normal and high demand, and check the level stays inside the alarm band with the setpoint undisturbed.
- Hand over an as-found/as-left record, the loop drawing, the range table and the tuning rationale. A loop that cannot be reproduced cannot be maintained.