Cybernetics and system control

Cybernetics, feedback and system control

How feedback turns measurements into corrective action

Cybernetics studies communication, control and feedback across systems, including machines, living organisms and social organisations. Control engineering develops practical methods for making a dynamic system meet an objective: maintaining motor speed, regulating an oven or controlling a tank level. The fields overlap, but cybernetics has a broader scope than industrial controller design.

The central idea is that information about the result influences the next action. If a motor is commanded to run at 1,500 rpm and its measured speed falls to 1,420 rpm under load, a controller can respond to that difference. How strongly and how quickly it responds depends on the machine, drive and operating limits.

The parts of a feedback loop

The reference or setpoint specifies the desired result. A sensor provides information about the actual condition. The controller determines an action, and an actuator produces its physical effect. The controller may run in a PLC, a temperature controller or a variable speed drive. The motor, oven or tank being controlled is called the plant or process.

Control error = reference − feedback value. In the speed example, 1,500 − 1,420 = 80 rpm. This tracking error is different from an error in the sensor itself. The signals being compared must use compatible units and scaling.

Negative feedback loop Target r Σ Compare Controller Plant Sensor Disturbance Measured value: b Output: y e u + −
The sensor returns information to the comparison point. Control error: e = r − b.

Open-loop and closed-loop operation

Open-loop control does not use feedback from the controlled output to correct that output. Running a heater for a fixed period is a simple example. A change in ambient conditions can produce a different final temperature. An open-loop system can still contain monitoring sensors; the distinction is whether their measurements close the control loop.

Closed-loop control adjusts its action using information about the result. A temperature sensor feeding a temperature control loop is one example. The open-loop/closed-loop distinction applies to a particular loop, rather than necessarily to the whole machine. One machine can contain both arrangements.

Negative feedback is not a stability guarantee

Negative feedback aims to counteract a deviation; positive feedback reinforces a change. The words do not mean “bad” and “good.” Delays, an incorrect sign or excessive gain can cause a negative-feedback system to oscillate. Achieving stable control takes more than connecting a feedback cable.

Illustrative example: An oven has an 80 °C setpoint, but its sensor reads 3 °C below the actual temperature. In an idealised case, a displayed value of 80 °C corresponds to an actual 83 °C. Holding the measured value at the setpoint does not prove that the physical value is correct.

Where PID and motor control fit

A PID controller is one method for deciding the control action. P responds to the current error, I to accumulated error and D to the rate of change of the error or measurement signal used by the implementation. The P, I and D terms are not always used together; some loops use PI control. Controller tuning must reflect the process dynamics.

A servo system can contain nested position, velocity and current loops. Its speed loop regulates velocity, while encoder or resolver feedback provides motion information. Under increased load, a motor speed controller can change the command it produces. Depending on the drive architecture, that may be a torque/current reference or another actuation variable. It is not necessarily a direct frequency command.

Why a loop misses its target or starts oscillating

Delay: A slow measurement or process response means the controller acts on information that is already old. Excessive correction can cause overshoot and oscillation. Time delays and system dynamics matter when choosing control settings; there is no single sampling interval suitable for every process.

Actuator limits: A heater at full power or a drive at its current limit cannot produce more physical capacity merely because the controller asks for it. If the integral term continues accumulating, recovery can be delayed after the output leaves its limit. This is integrator windup. Appropriate anti-windup and output limits belong in an assessment of PID limitations.

Measurement and mechanical problems: Noise, incorrect scaling, backlash and friction can affect the response. Filtering may reduce noise while adding delay. Investigating control-loop problems starts with checking the measurements, control direction and physical equipment.

What does good control look like?

Assess rise time, overshoot, settling time and steady-state error together. Also examine recovery after a load change and unnecessary actuator movement. A data logger or SCADA trend showing the reference, measurement and controller output together provides more insight than the final value alone.

Model predictive control (MPC) uses a model to predict future behaviour and choose actions while accounting for constraints. Adaptive control updates selected controller or model parameters as conditions change. Robust control focuses on preserving performance and stability within specified uncertainty bounds. None of these labels means better than PID for every task.

AI and machine learning can support modelling or prediction, but cybernetics does not require AI. A simple thermostat can use feedback without learning from data. Cybersecurity concerns protection against unauthorised access and attacks; it is a different field.

Technical references: ASC — the field of cybernetics · MathWorks — feedback connections · University of Michigan — PID control · MathWorks — anti-windup · MathWorks — MPC