Feedback in control loops is the basic concept of control engineering on which everything else is based. We explain this concept without mathematics, only using illustrative examples. Gain an intuitive understanding of why control loops can become unstable. And learn what is crucial for good control design.
A control loop always consists of at least two components:
As an example, let's take a car as process and a driver as controller. The driver uses steering wheel, accelerator pedal and brake to control the car. However,he does not just do this somehow – otherwise the journey would quickly come to an end – rather he or she reacts to certain signals. These are primarily visual information, for example how far the car is from the right edge of the road.
In other words, the driver detects positions of his car and edge of the road, processes it and reacts with steering interventions. And the driver does this not just once, but continuously throughout the entire journey. Again and again, the driver reacts to the position of the car and the car reacts to the steering interventions of the driver.
We can represent this chain of effects as a signal flow and obtain this picture:
And this is exactly the basic structure of a closed control loop. The basic principle of control engineering is to feed back an output signal of the controlled system (process variable) to an input signal of the controlled system (manipulated variable) via a controller. The aim of this feedback is to bring the process variable to a desired value (set-point) and to keep it there. The overall system of controller and controlled system is then the closed control loop.
Control systems can easily be imagined as a chain of smaller components connected in series. The individual steps in this chain do not run infinitely fast, they each need some time. This means the controlled process variable always reacts delayed to control actions. Each controlled process has its own characteristic dynamics, which can be completely different depending on the process.
A car reacts comparatively quickly to its driver's steering wheel movements. But there are still significant differences between specific models. When sport cars are described with "very direct steering" or "go-kart feeling", translated into terms of control technology, it means nothing other than process with low delay and lag, i.e. fast inherent dynamics.
Here it is quite different. It may take a while before the skipper's steering interventions via the rudder result in a visible change in the boat's direction.The process boat has a comparatively high time delay, i.e. slow inherent dynamics.
The problem with this slow process dynamic is that the skipper receives visual feedback on his steering actions only after a very long delay. With rather hectic contemporaries with little boating experience, this leads to a vicious circle:Before the effect of one intervention is visible, and it is clear whether further interventions are necessary, the next intervention is already made. Gladly much more extreme, so that finally something happens. Or even in the opposite direction, maybe it was the wrong one before.
This unavoidable leads to a zigzag course, which, in the worst case, builds up and ends on the shore or in another watercraft.
And this is exactly what can happen in principle in all control loops. It always happens when a too aggressive controller meets a comparatively sluggish process. Just like a hectic skipper, this controller simply reacts too strongly to small signals from the controlled system. This overreaction causes oscillations of the control system. With increasing aggressiveness of the controller or increasing process sluggishness, these oscillations show damped decaying behavior up to exponentially growing amplitudes, i.e. unstable behavior.
To design the best possible control system, it is important to understand that both parts of the control loop – controller and process – have an impact on performance.And it is always worth looking for improvement in the process first before trying to improve the controller.
This principle is best explained by the examples of car and boat already used. A boat reacts very slowly and sluggishly. This makes the control loop directly slower because changes of the manipulated variable take a long time until changes of the controlled variable become observable. But even worse is the indirect effect: To avoid oscillations, the skipper of a boat must react much more carefully and cautiously than a car driver. Or expressed in terms of control technology: The controller for the process boat must be set much slower than that for a car.
This means process is slower and controller is slower. Due to the control loop feedback, these effects also reinforce each other. Therefore, the rule of thumb is:
Reduce lag and time delay of the controlled system wherever possible!
Every process acceleration is returned to you several times over as a performance gain of the closed loop.
Accelerating a given process is easier said than done. Therefore, here is a list of typical examples of potential improvements that we have often encountered in practice:
If an experimental investigation is too time-consuming and expensive, the potential of such improvements for a specific application can be evaluated very well by system simulation of the closed-loop system.
In another article, we describe step by step a field-tested method to tune PID controllers for a given process.