Simple control system design

An open-loop control system is the simplest possible arrangement. You have a controller, an actuator, and a plant — the thing being controlled — and the controller simply sends a command to the actuator and trusts that the actuator did its job. There is no measurement of what actually happened and no way for the system to know whether the output matched what was asked for.

A household toaster is the classic example: you set a timer for two minutes, the heating element runs for two minutes, and whatever the toast looks like at the end is the toast you get. If the bread was thicker than usual, or the room was colder, or the heating element has aged, the toaster has no way to compensate. The same is true of a sprinkler on a timer, or a washing machine that runs a fixed cycle regardless of how dirty the clothes were.

Open-loop systems are cheap, simple, and totally blind to disturbances. They work fine when the relationship between command and result is predictable and consistent. The moment that relationship can vary — friction, wear, load, temperature, leakage — the output drifts away from what you wanted and the system will never know.

A closed-loop control system fixes the blindness of open loop by adding a sensor that watches the output and feeds the measurement back to the controller. The controller subtracts the measurement from the setpoint to compute the error — how far off the actual value is from the target — and then adjusts the actuator to shrink that error. The loop is “closed” in the sense that information now flows all the way around: setpoint in, command to actuator, action on the plant, measurement back to the controller.

A room thermostat is the everyday example. You set the dial to twenty degrees. The thermometer inside reads the room temperature. If the room is at eighteen, the controller turns the heater on. As the room warms toward twenty, the error shrinks, and at twenty the controller switches the heater off. If a window opens and cold air rushes in, the thermometer notices, the error grows again, and the heater comes back on without you doing anything.

This is the structure behind cruise control, motor speed controllers, tank level controllers, and almost every industrial process loop. Disturbances — load changes, leaks, wear — show up as error in the measurement, and the controller continuously corrects for them. The cost is a sensor, a controller, and the engineering effort to tune the loop, but in return you get a system that actually holds the value you asked for.

Every basic feedback control loop is built out of the same four pieces, and you can spot all four on almost any piece of process kit once you know what to look for.

The setpoint is the target — where you want the controlled variable to sit. It is usually entered by an operator or set by a higher-level supervisory system, and is expressed in the units of the process (degrees, bar, litres per second, rpm).

The sensor, also called the transducer, sits in the process and converts the physical quantity into an electrical signal the controller can understand. Pressure becomes millivolts, temperature becomes a resistance change, flow becomes a pulse rate. The quality of the loop is fundamentally limited by the quality of this measurement.

The controller is the brain. It compares the measurement to the setpoint, calculates the error, and decides how hard to push the actuator. Simple controllers do on/off switching at the setpoint; more capable ones use proportional, integral, and derivative terms — a PID controller — to make the correction smoother and faster.

The actuator is the muscle. It is the part that physically acts on the plant: a valve that opens or closes, a motor that spins, a heater that warms, a pump that moves fluid. The actuator turns the controller's command into a real physical change, and the loop starts again.