Tank Level Loop: Real Oscillation or Sensor Artefact?
A 3 m² water tank supplies a process at a nearly constant 10 m³/h. A centrifugal pump, driven by a variable-frequency drive, can deliver 12 m³/h at 100% speed. A 4–20 mA level transmitter is scaled 0–2 m and feeds a PLC PID loop that sets pump speed. The setpoint is 1.20 m. Operators report a slow oscillation of about ±0.18 m with a period near 90 s; the pump speed swings between 35% and 100%. A manual step test gave a dead time of 8 s and a time constant of 25 s. The current PID settings are proportional gain 4, integral rate 0.05 s⁻¹ and derivative 0. The level signal has about 2% noise and shows spikes whenever the agitator starts. The pump has a minimum controllable speed of 30% and a check valve. High-level alarm is at 1.8 m; the overflow is at 2.1 m. The process cannot tolerate the pump running dry.
1. Check whether the reported 90 s, ±0.18 m level oscillation is physically consistent with the tank area and pump capacity. What does the result imply?
Read the worked answer
The tank area is 3 m². With the pump off and the draw at 10 m³/h, the maximum falling rate is (10/3600)/3 = 9.26×10⁻⁴ m/s, or 0.926 mm/s. To fall 0.18 m would take 0.18/9.26×10⁻⁴ = 194 s. With the pump at full speed, the maximum net filling rate is only 2 m³/h, giving 0.185 mm/s and a much longer time.
A 90 s period means a half-period of about 45 s. Moving 0.18 m in 45 s would require a level rate of 4 mm/s, equivalent to about 43 m³/h of net flow in a 3 m² tank. That is far beyond the 12 m³/h pump capacity. The reported level oscillation cannot be a real hydraulic oscillation of the whole tank. It is more likely a sensor or signal artefact: agitator interference, a grounding problem, PLC scan aliasing, or a transmitter spike being interpreted as level. Investigate signal wiring, shielding, filtering and the agitator interlock before tuning the PID.
2. Assuming a genuine loop oscillation is later found, propose PI settings and signal conditioning. Explain why derivative action is not appropriate here.
Read the worked answer
The process has dead time 8 s and time constant 25 s, so it is dominated by dead time and a long lag. A reasonable starting point is proportional gain about 1, integral time about 25–35 s (reset rate about 0.03–0.04 s⁻¹), and derivative off. Gain 4 with an integral rate of 0.05 s⁻¹ is aggressive for an 8 s dead time and can produce a sustained cycle, especially with a pump that has a 30% minimum speed.
Add a first-order filter of a few seconds, spike rejection for agitator starts, and a deadband or minimum-speed strategy so the pump does not hunt around its lower limit. If the pump cannot modulate below 30%, use a recirculation valve or an on/off pump with a small buffer instead of forcing the VFD to chase every small error. Derivative action would amplify the 2% noise and the agitator spikes, causing valve or pump chatter; it should remain zero unless the signal is heavily filtered and the dead time is reduced.
3. Design a fail-safe response to transmitter failure and to a high-level condition. Include the PLC logic and independent protection.
Read the worked answer
The PLC should treat a transmitter current below about 3.6 mA or above about 21 mA, and a rate-of-change that is physically impossible, as a fault. On a level-signal fault, stop the fill pump and close the fill valve, raise an alarm, and latch the fault until reset. This is fail-safe for overflow, but the process draw must then be managed separately. Add dry-run protection for the pump, such as a low-level switch or current monitoring, so stopping the pump does not create a different hazard.
Do not rely only on the analog transmitter for high level. Fit an independent high-high float switch at 1.85–1.9 m that directly removes power from the pump, hard-wired outside the PLC. Test the trip path periodically. The overflow at 2.1 m and the high-level alarm at 1.8 m should be clearly separated, and the check valve should prevent backflow when the pump stops.