A pump feeds a transparent tank while a bottom valve drains it. You follow the level-sensor signal into a proportional controller and back to pump flow, then compare open-loop drive with feedback, sensor bias and an extra drain load.
• A process tank with a variable-speed pump, a level transmitter, a comparator and proportional controller, and an outlet valve with a disturbance branch. • Controls for pump command (closed-loop proportional or open-loop fixed drive), target level (0.2-0.8 m), proportional gain (0-4 drive/m), sensor bias (-0.2 to 0.2 m), drain coefficient and an extra drain that opens after 10 s. • Readouts of actual water level, sensor reading, control error, pump command, pump flow and drain flow. • Experiments: closed-loop disturbance rejection, open-loop disturbance, and a biased transmitter.
The tank area is 0.25 m² and the level follows A dh/dt = Qin − Qout − overflow, with Qin = 0.04u and Qout = k√h plus an optional 0.01 m³/s branch. The controller computes u = clamp[0.4 + Kp(r − (h + bias)), 0, 1], or holds u = 0.4 in open-loop mode.
This is a nonlinear lumped tank with an ideal pump and sensor and deterministic fixed integration. Because the controller is proportional, a steady offset generally remains. There is no pump curve, fluid inertia or sensor noise, and overflow is recorded in the model rather than allowing impossible level growth.
No. A sustained correction often requires a nonzero error, so with proportional control only a steady offset generally remains.
No. Feedback acts on the measurement, including its error. In the lab a sensor bias can produce a small indicated error while the actual level is wrong.
In open-loop mode the pump command is fixed at 0.4 and cannot respond to the measured level. In closed-loop mode the command is 0.4 + Kp times the error between target and measured level, limited to 0-1.
An optional extra drain branch that adds 10 L/s after ten seconds. The closed-loop pump increases its drive after the added drain lowers the level; the open-loop pump cannot.