Control in Nonlinear Systems

Diego Iván Perea Montealegre (2185751) diego.perea@uao.edu.co

Faculty of Engineering, Universidad Autónoma de Occidente
Cali, Valle del Cauca

Abstract—This document contains the process to obtain

a nonlinear system control design, where different types
IV. THEORETICAL FRAMEWORK
of controllers have to be designed and analyzed for the
automation of the process to be controlled, such as In the 1950s, new tools capable of solving truly interesting
regulators, trackers, smith predictor, and slider method. problems from the control point of view began to emerge,
nonlinear tools that attempted to extend results from linear
Keywords; linear; system ; Mat Lab ; control; Control in
systems theory to nonlinear systems, results such as those
Nonlinear Systems.
referring to controllability and observability of the system,
II. .INTRODUCTION thus becoming an important alternative to classical linear
control. The methods for the design and analysis of control
Control systems theory deals with the analysis and design of
systems for nonlinear processes arise from the need to
interacting components of a system in a configuration that
broaden the knowledge of the behavior of real dynamic
provides the desired behavior. The essential configuration
systems, since, when obtaining a model that describes the
used in control systems theory is based on the fundamental
dynamics of the system, some aspects are often omitted. in
concept of feedback, which consists of the process of
order to reduce inconveniences presented when applying
measuring the variables of interest in the system and using
some classic control technique, either to analyze a process or
that information to control its behavior. Control theory and
to design a system that controls it. On the path of designing
practice have a wide range of applications in the fields of
control systems, we are faced in the first instance with how
aeronautical, chemical, mechanical, environmental, civil, and
we should mathematically represent the dynamics of the
electrical engineering, as well as in many other non-
process to be controlled, this is where the first obstacle could
engineering disciplines. The benefits of efficient control in
arise, as the vast majority of systems in the industry already
the industry are immense, including improvements in product
know. real-life have non-linear characteristics, opting for
quality, reduction in energy consumption, minimization of
linearization of the system around an equilibrium point would
waste material, higher levels of safety, and pollution
be the most obvious solution, but we would obtain limited
reduction. [1]
knowledge of the real system since we would only be
analyzing the system around an operating point, but it turns
out that the nonlinear systems present multiple isolated
III. OBJECTIVES
equilibrium points.[2]
GENERAL OBJECTIVES:
A solution was proposed by Alexander Mikhailovich
*Control a plant or non-linear system. Lyapunov in Lyapunov's Stability Theory, based on this
theory nonlinear control techniques arise for linear and
SPECIFIC OBJECTIVES. nonlinear systems, Sliding Mode Control and Adaptive
* Design and analyze different types of controllers for the Control. are two of the most relevant control techniques, and
nonlinear system. with them, it is possible to start from a nonlinear model of the
process to design a nonlinear controller. [Khalil, H. K.],
*Design continuous and discrete regulator controller for [Slotine, J. J. E., 1991] In addition to these techniques, in the
nonlinear plant 1980s Isidori demonstrated that it was not only possible to
*Design continuous and discrete tracker controller for the extend the results of linear systems theory to nonlinear
nonlinear plant systems, but also managed to demonstrate the possibility of
extending the geometric control theory of linear systems
* Design continuous and discrete predictor Smith controller under the use of a set of useful tools in the treatment of linear
for nonlinear plant systems, one of these most important tools is Feedback
* Design of control with slider method Linearization. [Isidori, A.][2]
The Jacobian matrix collects all first-order partial derivatives
of a multivariate function. Specifically, consider first a
function that maps u real inputs, to a single real output:
 For this case to be analyzed, it was analyzed with three tanks
arranged in cascade, with selected or assumed input and
selected or assumed constants and the output to be found, the
Then, for an input vector, x, of length, u, the Jacobian vector
third tank was selected, so that the equation of the liquid level
of size, u × 1, can be defined as follows:
system in a set of three tanks arranged in a shell is as follows:

Now, consider another function that maps u real inputs, to v

real outputs:

Figure 2. Model of liquid level system in a set of three tanks

Then, for the same input vector, x, of length, u, the Jacobian arranged in cascade
is now a v × u matrix, J ∈ ℝv×u, that is defined as follows:

VI. RESULT AND ANALYSIS

To observe the nonlinearity of the plant, we proceed to make
the plant in the Simulink program, using the corresponding
block that allows us to do it with the fcn block and other
blocks such as constat, mux, integrator, etc. Taking as
constants A=10 and C=5.
Reframing the Jacobian matrix into the machine learning
problem considered earlier, while retaining the same number
of u real inputs and v real outputs, we find that this matrix
would contain the following partial derivatives [4]:

Figure 3. Model of liquid level system in a set of three tanks

arranged in cascade made in Simulink

V. DESCRIPTION OF THE PROBLEM

The Plant or Nonlinear Liquid Level System in a set of tanks
arranged in a cascade as shown in the following figure 1.

Figure 4. Graphic of the liquid level system model in a set of

Figure 1. Model of liquid level system in a set of tanks three tanks arranged in cascade made in Simulink
arranged in cascade [4]
Observing the previous figure 4 at the moment of giving it an
input value u=1, this non-linear plant at its output does not
result in the input value, but it returns a value of 0.04, so this Define the equilibrium points to be treated using the Matlab
non-linear plant needs to perform control. 'solve' function makes the function equal to 0 and thus clears
the unknown to be found of x1, x2, and x3.
But to confirm its non-linearity, its input is a sine signal

Linearization of the plant is performed through the Jacobian

Figure 5. Model of liquid level system in a set of three tanks arrays in which the previously found equilibrium parts were
arranged in cascade made in Simulink with sine input evaluated.

In this part, it is necessary to know if the plant to be treated

is controllable or not and if it is observable or not, these two
factors are very important in the controllers to know what
type of controller is needed or not.

Figure 6. Graphic of the liquid level system model in a set of Specify the appropriate poles for the regulator to improve the
three tanks arranged in cascade made in Simulink with sine system reaction time; these poles must be negative and far
input away from the center. Finally, the algorithm generates the
proportional gain variable k.
VII. CODE IN MATLAB

The constants to be used in the liquid level system are defined

in a set of three tanks arranged in cascade, which within these
are the variable A and the variable C.
The augmented matrix is produced, and the necessary poles,
as well as another pole that must be increased because the
system moved up in order, are given. Finally, the program
provides us with the integral and proportional gains in the
variable kt.
The observer employs the same mechanism as the regulator,
except that the poles are moved further away from 0 to Figure 7. Block diagram of Regulator with continuous plant,
increase the observer's response time and guarantee it is faster observe, and predictor smith.
than the plant's. The sample time and discretization approach
must be carefully chosen in the discretization portion of the
procedure; for the time of sampling is 0.1 and the
discretization method Zero-order hold because is one method
very appropriate for this case of discretization.

It is the same process that was carried out for a continuous

time, unlike the fact that it is carried out for a discrete-time, Figure 8. Graph of Regulator with continuous plant, observe,
so it takes the plant in discrete in the cases of the regulator and predictor smith.
and tracker, the commented part is a second way to carry out
the tracker.

Because the plant is taken independently, the discrete

observer is computed, which is identical to the continuous
Figure 9. Block diagram of Discrete regulator with
observer in its computation. And the time of retarded is of 1.
non_linear plant and with the observer in discrete
VIII. CONTROLLER DESIGN
For said non-linear plant, the respective controllers were
carried out
DESING OF REGULATING CONTROLLERS
The regulator controller is responsible for sending the signal
to the system's equilibrium point, which is typically zero.
This controller must set the starting circumstances because it
lacks an input. To do this, we will design many types of
regulatory controllers for this non-linear system.
 Figure 13. Block diagram of Continuous plant with regulator
discrete

Figure 10. Graph of Discrete regulator with non_linear

plant and with the observer in discrete

Figure 14. Graph of Continuous plant with regulator

Figure 11. Block diagram of Continuous Plant Regulator discrete

Figure 15. Block diagram of Continuous plant regulator with

Figure 12. Graph of Continuous Plant Regulator observer
 Figure 18. Block diagram of Plant with retard

Figure 16. Graph of Continuous plant regulator with

observer

Figure 19. Graph of Plant with retard

Now having this clear, the design of trackers controllers
begins.

Figure 17. Graph of all regulating controllers

As mentioned above, the regulator controllers are responsible
for sending the signal to the balance point of the system,
observing graph 17 shows the operation of the regulators for Figure 20. Block diagram of Tracker in discrete with linear
plants or non-linear systems, and for this case in the liquid plant and with the observer in discrete
level system in a set of three tanks arranged in a cascade it is
possible to use regulators in different ways depending on the
desired stabilization time.
DESING TRACKERS CONTROLLERS
Tracking controllers are well-known and widely used around
the world, as well as in a range of industrial applications, to
set a system on a value and, as the name implies, to make the
system follow an input. But first, it should be mentioned that
at the time of taking data from the plant there may be a delay
and this must be defined, so if that delay is not taken the
following could happen.

Figure 21. Graph of Tracker in discrete with linear plant and

with the observer in discrete
Figure 22. Block diagram of Tracker in discrete with linear
plant in discrete with observer in discrete

Figure 26. Block diagram of Tracker with linear plant and

observer

Figure 23. Graph of Tracker in discrete with linear plant in

discrete with observer in discrete
Figure 27. Graph of Tracker with linear plant and observer

Figure 24. Block diagram of Tracker with linear plant

Figure 28. Block diagram of Tracker with retard smith

Figure 25. Graph of Tracker with linear plant

Figure 29. Graph of Tracker with retard smith
Figure 30. Block diagram of Tracker observer smith

Figure 34. Graph of all trackers controllers

As mentioned above, the trackers' controllers are responsible
for the value of the input is equal to the value of the output of
the system, observing figure 34 of the graph, it is evident that
the trackers work for plants or non-linear systems and this
case in the system. of liquid level in a set of three tanks
arranged in cascade, it is possible to use trackers in different
ways depending on the desired stabilization time.
SLIDER METHOD
Figure 31. Graph of Tracker observer smith For greater ease in the calculations to demonstrate the slider
In this figure 31 It is very peculiar because you can see the method, it is chosen to put y=x1 at the output. The first thing
different responses that are being given to the states that are is to find the relative degree to which a derivative is
in turn controlled performed at the output
𝑦 = 𝑥1
𝑦̇ = 𝑥̇1
𝐶 1
𝑦̇ = 𝑥̇1 = − √𝑥1 + ∗ 𝑢
𝐴 𝐴
Since only one derivative is used to find where u was in the
equation, then r=1, α1=1. After this process, S(e) must be
found
Figure 32. Block diagram of Tracker observer smith with
external perturbation 𝑖 −1

𝑆(𝑒) = ∑ 𝛼𝐶𝑖 + 1) ∗ 𝑒 𝑖
𝑖=−1
0

𝑆(𝑒) = ∑ 𝛼𝐶𝑖 + 1) ∗ 𝑒 𝑖
𝑖=−1

𝑆(𝑒) = 𝛼0 𝑒 −1 + 𝛼1 𝑒 0
Taking into account the rules
∞
−1
𝑒 = ∫ 𝑒(𝑡) ∗ 𝑑𝑡
0
0
𝑒 = 𝑒(𝑡)
Figure 33. Graph of Tracker observer smith with external
perturbation 𝑑𝑒(𝑡)
𝑒1 =
𝑑𝑡
 𝑆(𝑒) = 𝛼0 ∫ 𝑒(𝑡) + 𝑒(𝑡)

With which it is observed that S(e) is a sum of an integral and

a proportional, that is, a PI
For the following, calculate the derivative of S(e)

𝑆̇(𝑒) = 𝛼0 𝑒 + 𝑒̇
With 𝑒 = 𝑦 − 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒
𝑒̇ = 𝑦̇
𝐶 1
𝑒̇ = 𝑦̇ = 𝑥̇1 = − √𝑥1 + ∗ 𝑢
𝐴 𝐴
It is replaced in the 𝑆̇(𝑒)
𝐶 1 Figure 36. Graph of slider method with y=x1
𝑆̇(𝑒) = 𝛼0 𝑒 + − √𝑥1 + ∗𝑢 Previously it was said that α0 n η can be any positive value
𝐴 𝐴 for this, we will change α0=1 to see its result
We know that 𝑆̇(𝑒) = −𝜂 ∗ 𝑠𝑖𝑔𝑛(𝑠)
𝐶 1
𝑆̇(𝑒) = 𝛼0 𝑒 + − √𝑥1 + ∗𝑢 = −𝜂 ∗ 𝑠𝑖𝑔𝑛(𝑠)
𝐴 𝐴
𝐶 1
𝛼0 𝑒 + − √𝑥1 + ∗𝑢= −𝜂 ∗ 𝑠𝑖𝑔𝑛(𝑠)
𝐴 𝐴
We clear the value of u

𝐶
𝑢 = (−𝛼0 𝑒 + √𝑥1 − 𝜂 ∗ 𝑠𝑖𝑔𝑛(𝑠)) ∗ 𝐴
𝐴

So the block diagram is implemented in Simulink with

Figure 37. Graph of slider method α0=1
𝛼0 = 30 and 𝜂 = 10 , 𝛼0 𝑎𝑛𝑑 𝜂 can be any positive value
With α0=100

Figure 35. Block diagram of slider method with y=x1

Figure 38. Graph of slider method α0=100

When analyzing the different values of α0 , it can be seen that
the value that is put on α0 modifies the shadow impulse that
the output will have.
Now changing the value of η=1, leaving α0=30 * The regulator controllers as a control option towards the
stability state for the situation that needs to be used, as is the
case of satellites, for example.
* The delay time in non-linear plants is important data for its
control, so if it is not taken into account, it can be difficult to
carry out control.
*Sampling time is vital for systems where controllers are
performed in the discrete time since when the time is varied,
the control result will be different and may not be controlled.
* Trackers and regulators as effective and efficient when
performed in non-linear plants and linear plants.
*In slider method the value set to η modifies the stabilization
value that the output will have.
* In slider method the value that is put on α0 modifies the
Figure 39. Graph of slider method η= 1 shadow impulse that the output will have.

With η= 100
REFERENCES
[1] Eng.newcastle.edu.au, 2022. [Online]. Available:
https://www.eng.newcastle.edu.au/~jhb519/teaching/snolin/
material/cap01.pdf.
[2] Biblioteca.utb.edu.co, 2022. [Online]. Available:
https://biblioteca.utb.edu.co/notas/tesis/0063144.pdf.
[3] S. Cristina, "A Gentle Introduction to the Jacobian",
Machine Learning Mastery, 2022. [Online]. Available:
https://machinelearningmastery.com/a-gentle-introduction-
to-the-jacobian/. [Accessed: 22- May- 2022]
[4]2022. [Online]. Available:
https://www.researchgate.net/profile/Hebertt-Sira-
Ramirez/publication/327929276_Control_de_Sistemas_No_
Lineales_Linealizacion_aproximada_extendida_exacta/links
Figure 40. Graph of slider method η= 100 /5c009b67a6fdcc1b8d4a9371/Control-de-Sistemas-No-
When analyzing the different values of η, it can be seen that Lineales-Linealizacion-aproximada-extendida-exacta.pdf.
the value set to η modifies the stabilization value that the [5]J. Leite, "Ingenieria-de-Control-Moderna-Ogata-5ed",
output will have. Academia.edu, 2022. [Online]. Available:
CONCLUTIONS https://www.academia.edu/9814191/Ingenieria_de_Control_
Moderna_Ogata_5ed.
* For a non-linear system plant, it is possible to make
regulators and trackers depending on it.
* There are different ways of making trackers for non-linear
plants, the choice of the tracker that we choose will depend
on the needs we have, such as the stabilization time and
external disturbances that may occur.
* Realizing trackers controllers in continuous and discrete-
time gives the possibility of implementation being a form of
choice towards the costs that may occur in the manufacture
of these controllers.
* When control is carried out towards the non-linear plant,
the energy that is consumed must be taken into account, since
this energy factor represents very important costs towards the
implemented place.