Scribd
Upload
82 views

« Introduction to Biomedical Engineering»: Сourse

The document discusses the PID controller, which is used in 90% of control systems. It explains each component of the PID controller - proportional, integral, and derivative - and how they work together to improve system response. Specifically, the proportional control reduces steady state error, the derivative control improves transient response without increasing steady state error, and the integral control eliminates steady state error. Finally, it notes that tuning the three gain parameters is important for optimal system performance but varies between applications.

Uploaded by

Ebrahim Abd El Hady
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
82 views

« Introduction to Biomedical Engineering»: Сourse

The document discusses the PID controller, which is used in 90% of control systems. It explains each component of the PID controller - proportional, integral, and derivative - and how they work together to improve system response. Specifically, the proportional control reduces steady state error, the derivative control improves transient response without increasing steady state error, and the integral control eliminates steady state error. Finally, it notes that tuning the three gain parameters is important for optimal system performance but varies between applications.

Uploaded by

Ebrahim Abd El Hady
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Сourse

« Introduction to Biomedical
Engineering»

Dr. Kirill Aristovich

Section 2: Control theory


Lecture 2.3: PID controller
PID controller
Hello, in this lecture we will talk about the most commonly used controller out there.
It is called PID controller by the first letters of individual control components: Proportional,
Integral, and Derivative. Rumor goes that they account for nearly 90% of all use cases,
which is a fine example of totally made up statistics, but I would not be surprised if it is
true.
To see how it works and why it is used, let’s start by considering each component
independently. Note that we have transferred the controller from a feedback path to a
forward path, to illustrate the point of manipulating with error, but this still IS the feedback
controller.
So the proportional controller basically amplifies the position error. Let’s see how
it works. Formally, we can easily compute the new transfer function. If you evaluate the
step response and track the performance, you can see that Low values of Kp give stable
but slow responses, and high steady state velocity error. High values reduce steady state
velocity error but response overshoots considerably. To understand why these
overshoots occur, let’s look at the error term over time for a purely proportional controller
for, say, a motor. With proportional control the error and thus the control signal does not
reach zero until the motor is at the desired position. However, due to the inertia of the
system, the motor continues to move even without a control signal, resulting in an
overshoot. Controlling the servo in this manner is like NOT applying the brakes until you
reach your target position when driving!

Figure 1 - PID Control

In reality when driving you apply the brakes beforehand, and also dependent upon
the speed you are travelling. You break earlier when you are driving faster (I hope!).
We can replicate this by considering the derivative of the error. This is negative as
the servo approaches the target, so offers a way of restraining the motor forward motion.
Further, if we consider the integral or total error over time: This gradually increases over
time, and can be used to magnify the control signal for small errors, and improve the
steady state error.

2
Figure 2 - Steady State Error

In order to understand this formally, let’s consider a derivative controller with unity
proportional gain. Its open loop transfer function is the following, and the closed loop
transfer function is on your screens. The closed-loop system is still second order, but now
there is a zero in the numerator. The effect of this is subtle compared to the change in
the poles. More importantly, the steady state velocity error of this system can be
computed and we can see that, unlike for example, pure velocity feedback, the value is
unchanged by derivative error. So overall we can improve transient response without
compromising steady state error, which is great. The effect on the step response is
following: The poles of the system become less oscillatory and decay quicker. This,
combined with a reduced proportional gain, gives an improved transient response.

Figure 3 - Open Loop gain and The closed loop transfer function

This is clear when looking at the step response, and the relative contributions of
the two error terms. At the start, the derivative term is significant and in the opposite
direction to the proportional error term, but becomes negligible as the system settles.
The integral of the error is used for correcting steady state errors, as it adds a pole
at zero in the transfer function. We can see that by computing closed-loop transfer
function and noting that now we have s^3 in denominator.
Computing the steady state velocity error, we can see that it is zero. YAY! If you
look at the s-plane, you can see that the effect of an extra pole close to origin – which
would make our system very slow - is largely cancelled out by the nearby zero. This
means the system is broadly similar to a proportional controller, with the exception of
improved steady state performance.
Looking at ramp response it is clear how the integral term has enabled proper
tracking of a ramp input! This is particularly useful for overcoming steady state errors
arising from nonlinearities not included in the model which prevent proper tracking in
reality.

3
Putting all three of these controllers together gives the complete PID. Here it is in
all its glory! The closed-loop transfer function is rather big, but shows that it is possible to
design a controller with improved transient response and decreased or zero steady state
error. In theory. Just need to choose Kp, Ki, and Kd appropriately.
There is no unique solution for the settings for the three gains, and so the values
are dependent upon the specific system and the application. Selecting these parameters
is known as tuning, and it was (and still is in machine learning circles) a very active area
of research.

Figure 4 - PID parameters

Sometimes overshoot reduction is more important than settling time (like in heaters
or ventilators). In other cases reducing settling time may be crucial and some overshoot
is an acceptable trade off: like in camera focus. Many methods for “Tuning” these
parameters exist, and often form the first estimate, before refinement based on testing
real system.
Finally, I must repeat, that these controllers are used everywhere, and can even
be applied with little or no knowledge of the system being controlled, often producing
acceptable results!

Some graphic material used in the course was taken from publicly available
online resources that do not contain references to the authors and any restrictions on
material reproduction.

4
This course was developed with the support of
the "Open Polytech" educational project

Online courses from the top instructors of SPbPU

You might also like