Control of A DC Motor Using Feedback Linearization and Gray Wolf Optimization Algorithm
Control of A DC Motor Using Feedback Linearization and Gray Wolf Optimization Algorithm
Control of A DC Motor Using Feedback Linearization and Gray Wolf Optimization Algorithm
Abstract
The aim of this study is to investigate nonlinear DC motor behavior and to control velocity as output variable. The linear
model is designed, but as it is experimentally verified that it does not describe the system well enough it is replaced by
the nonlinear one. System’s model has been obtained taking into account Coulomb and viscous friction in the firmly non-
linear environment. In the frame of the paper the dynamical analysis of the nonlinear feedback linearizing control is car-
ried out. Furthermore, a metaheuristic optimization algorithm is set up for finding the coefficient of the proportional-
integral feedback linearizing controller. For this purpose Gray wolf optimization technique is used. Moreover, after the
introduction of the control law, analysis of the pole placement and stability of the system is establish. Optimized non-
linear control signal has been applied to the real object with simulated white noise and step signal as disturbances.
Finally, for several desired output signals, responses with and without disruption are compared to illustrate the approach
proposed in the paper. Experimental results obtained on the given system are provided and they verify nonlinear control
robustness.
Keywords
Feedback linearization technique, gray wolf optimization, nonlinear friction modeling, nonlinear systems and control, sto-
chastic disturbance robustness
Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work
without further permission provided the original work is attributed as specified on the SAGE and Open Access pages
(https://us.sagepub.com/en-us/nam/open-access-at-sage).
2 Advances in Mechanical Engineering
Many different techniques can be applied to control equations it is senseless examining system’s stability,
the velocity of the output series DC motor shaft. but one can check the stability of the separate equili-
Frequently used are, surely, conventional methods such briums. System which has two equilibrium points can
as traditional feedback control: proportional-integral- initially be in equilibrium number one, which is
derivative (PID) like controllers. They have low prices unstable. When the disturbance is applied, it can hap-
(compared to more complicated control systems), they pen that the second equilibrium attracts trajectory and
are simple and different variants of this control systems the state can converge in the second equilibrium.11
(proportional-integral PI or proportional-derivative Therefore it is necessary to check whether the global
PD) manage to keep the output of the system well minimum time of motion is actually realized by the
matched with the set value within the error limits. On obtained solution.12 Finding the optimal solutions rep-
the other hand, they suffer due to lack of robustness.4 resent the basic and challenging task, that is widely
Other than conventional, there are many nonlinear studied for decades. Nature has been a major source of
controllers. Some of them use adaptive control tech- inspiration for researchers in the field of optimization.13
nique, because the estimated velocity is heavily con- The implementation of metaheuristic algorithms can
taminated by noises from the switching signals. This be solved with nonconvex, nonlinear, and multimodal
direct method nullifies the merit of the sliding mode problems subjected to linear or nonlinear constraints
observer.5 In order to overcome the boundaries of by continuous or discrete decision variables, in the
model reference adaptive control several are built from form of global optimization algorithms. For example,
Artificial Neural Network (ANN), like it was proposed differential evolution and genetic algorithms have been
in the papers.6,7 ANNs enable estimating and control- used to perform an optimal design of a phase controller
ling velocity for a separately excited DC machine and it to track the trajectory of moving robots.14–16 Some of
is one of the most important modern techniques. The these algorithma include the genetic algorithm (GA),17
rotor speed of the machine can be made to follow an particle swarm optimization (PSO),18 whale optimiza-
arbitrarily selected trajectory, especially when the tion algorithm (WOA),19 gray wolf optimization
motor and load parameters are unknown. In (GWO),20 etc. Based on20 GWO technique illustrates
Alhanjouri8 these two neural networks are trained by its supremacy with an improved version of GWO tech-
Levenberg-Marquardt back-propagation algorithm. nique named as IGWO. This controller was proposed
Simulation results indicates to the advantages, effec- and demonstrated for step load disturbances, stochastic
tiveness, good performance of the artificial neural net- load disturbances, and varied conditions. In combina-
work controller, which is illustrated through the tion with other nonlinear control systems, these Park
comparison obtain by the system. In nonlinear system, et al.21 as well as the other techniques such as ant col-
self-tuning ANNs technique is related to linearization ony (AC)22 and a novel method called the cuckoo
of model at operating time interval. This may produce search (CS)23 can be applied.
error because of linearization of nonlinear model and Although modeling a nonlinearity is often a very
cannot control the speed accurately.1 complicated challenge, one of the first steps in the
In general, the control of system is difficult due to synthesis of a control system is to create a mathemati-
high nonlinearity properties. To overcome this diffi- cal model. This will save time and bring the cost-effec-
culty, another technique, which include Fuzzy Logic tiveness.24 On the other hand, an exact mathematical
Controller (FLC) can be developed.9 FLC is just one of models can not be easily obtained. One way to battle
the intelligent controllers and represents a widespread this kind of problem is given in the paper which used
control technique as it has satisfactory performance for real coded genetic algorithm with a goal to estimate the
nonlinear and complex systems. Basic character and unknown system parameters. In order to integrate the
the aim of this control strategy is to take advantages of resulted system model and the feedback linearizing con-
knowledge and the control experience of the operator troller (FBL), the nonlinear robotic system can be
for intuitive synthesis of the control system. Analysis transferred to a linear model with a nonlinear bounded
and comparison with classical PID control, sliding con- time-varying uncertainty.25
trol, adaptive control, etc., that is, by conventional con- Feedback linearization is the nonlinear control tech-
trol techniques, have led to the results which were also nique which has attracted increasing attention in past
used for the analysis of stability and quality of dynamic decades.26 By a synergy of a nonlinear transformation
behavior.10 The fuzzy design can be considered as an and linearization with feedback loop, the nonlinear
optimization problem, where the structure, antecedent, control design allows the creation of linear control
and consequent parameters are required to be law.27 The central idea of the approach is to algebrai-
identified. cally transform a nonlinear system dynamics into a
Global optimization problems are difficult to be (fully or partly) linear one, so that linear control tech-
solved efficiently because of their high nonlinearity and niques can be applied. This differs entirely from con-
multiple local optima. Nonetheless, with nonlinear ventional linearization in that feedback linearization is
Vesović et al. 3
achieved by exact state transformations and feedback, In this paper, the FBL method was used to control
rather than by linear approximations of the dynamics.28 the velocity of the DC motor. The feedback lineariza-
This technique has been successfully implemented in tion is a powerful nonlinear method based on the princi-
many applications of control, such as industrial robots, ple of canceling the nonlinearities of the system model.
high performance aircraft, helicopters, and biomedical In most papers dealing with the similar topics of FBL
dispositifs; more tasks which used this methodology are technique application for DC motor control (see Mehta
being now well advanced in industry.29 In Cambera and and Chiasson,27 Moradi et al.,37 Shirvani Boroujeni
Feliu-Batlle30 the problem of tracking the trajectory of et al.38) the nonlinear model is made on the basis of flux
the flexible end effector was solved using FBL method, and motor current nonlinearities. For the purposes of
where the force of gravity and the force of joint friction this research, feedback linearization was performed
are taken into account. As relatively simple and easy using a mathematical model that takes into account the
understandable technique FBL is suitable to be used in nonlinearity resulting from friction. Therefore, a non-
nonlinear systems: Aerial Package Delivery Robot,31 linear mathematical model of a DC motor with previ-
manipulators,32 and even for a high-DOF robots.33 ously determined friction, in the form of a Tustin
An optimal control strategy for a bearingless perma- model, was adopted. Moreover, a new model was
nent magnet synchronous machine drive is proposed in applied in which the discontinuous nonlinearity was
Sun et al.34 The state feedback control (SFC) based on approximated by a differentiable nonlinearity of the
the GWO algorithm is applied. The discretized state hyperbolic tangent, which ensured the conditions for
model with the augmented integrals of the displacement the application of FBL. After the feedback linearization
error and angular speed error is obtained. Then, the method has been successfully utilized to change the non-
weighting matrices are obtained by employing the linear states of the system to their linear forms, classical
GWO algorithm. The results show the superiority of PI controller has been implemented for DC motor velo-
the proposed method reflecting in faster response and city control. Determining controller gains has been con-
no overshoot compared with the PI controllers. On the sidered as an optimization problem and solved using
other hand, paper35 presents an optimal control strat- GWO optimization algorithm, as one of recent meta-
egy for a permanent-magnet synchronous hub motor heuristics swarm intelligence methods. GWO has been
(PMSHM) drive using the state feedback control widely tailored for a wide variety of optimization prob-
method plus the gray wolf optimization (GWO) algo- lems due to its impressive characteristics over other
rithm. To acquire satisfactory dynamics of speed swarm intelligence methods: it has very few parameters,
response, the discretized state space model of the and no derivation information is required in the initial
PMSHM is augmented with the integral of rotor speed search. Also it is simple, easy to use, flexible, scalable,
error and integral of current error. Then, the GWO and has a special capability to strike the right balance
algorithm is employed to acquire the weighting matrices between the exploration and exploitation during the
Q and R in liner quadratic regulator optimization pro- search which leads to favorable convergence.39 The last
cess. Finally, comparisons among the GWO-based state contribution of the paper is the demonstration of
feedback controller, the conventional state feedback robustness and good control performances of nonlinear
controller, and the genetic algorithm enhanced PI con- system control against external disturbance and in the
trollers are conducted in both simulations and experi- presence of noise, through experimental results.
ments. The comparison results show the superiority of
the proposed state feedback controller with the penalty
term in fast response. Finally,36 proposes an improved Description of the system
deadbeat predictive controller for PMSM drive systems. In this Section from the electrical, mechanical, and
It can eliminate the influence of parameter mismatch of combined equations we obtain and investigate linear
inductance, resistance, and flux linkage. A composite model of the DC motor . Mathematical model and dis-
sliding mode disturbance observer (SMDO) based on cussion of the model benefit is conducted. Here the
stator current and lumped disturbance is proposed, experimental verification of the obtained mathematical
which can simultaneously estimate the future current representation is provided.
value and lumped disturbance caused by the parameter The first prominent pace in control design is obtain-
mismatch of inductance, resistance, and flux linkage. ing the most precise model possible, because it reduces
Both simulation and experimental performances of the time for task performing. Furthermore, it brings the
proposed method have been validated and compared labor saving, cost-effectiveness, and makes work easier
with the conventional control methods under different and faster. That is why establishing the valid model is a
conditions. The comparison results show the superiority crucial stage in the practical control problems.
of the proposed predictive current control method The main aim of modeling DC motor is to find the
based on the composite SMDO. applied voltage, torque, current, or speed related
4 Advances in Mechanical Engineering
d
Jeq y(t) + Beq, v y(t) = Am u(t): ð1Þ
dt
where: Jeq is total moment of inertia calculated as: Figure 3. Comparison between real and model data for
sinusoidal input, amplitude 1 and frequency 0.5 rad/s.
Jeq = hg Kg2 Jm + Jl (hg and Kg are, respectively, the
gearbox efficiency and the total gear ratio, Jm is
Experimental confirmation of the acquired
moment of inertia of the motor shaft and Jl of the
load), the equivalent damping term is given by
mathematical representation
h K 2 h kt km + (h K 2 Bm + Bl )Rm Responses of the system are given in the following:
Beq, v = g g m Rm
g g
(hm is the motor effi-
Figures 2 and 3. Comparisons were made with the
ciency, kt is the current-torque constant, Bm and Bl are
responses obtained by simulations of the linear model,
viscous frictions acting on the motor shaft and on the
for step and sinusoidal inputs. The responses of the real
load shaft), and finally the actuator gain which equals
h K h k
object and the linear model to the step and sinusoidal
Am = g Rgm m t , Table 1. excitations do not match well. The model does not fol-
A DC series motor can serve as a vehicle for the eva- low the actual behavior of the system. Decreasing the
luation of the performance of the various controllers.42 amplitude changes only the sign of the deviation. This
Vesović et al. 5
is because the system is not linear and homogeneity where f(x), g(x), and h(x) are sufficiently smooth in a
principle could not be applied. domain D Rn , and x_ = ½x1 , x2 , x3 , :::xn T is a state vec-
The nonlinearity in the form of the dead zone can tor. It is necessary to find a state feedback control u,
also be seen in the sine wave. This nonlinearity repre- that transforms the nonlinear system into an analogous
sents the effect of the friction. It is special expressed in linear form. Undoubtedly, generalization of this idea is
low-frequency sinusoidal functions (and with change of not always realizable – there must be a special systemic
the rotation direction), because then the effect of the belongings, which will allow cancelation.
friction is most noticeable. It is easy to see that the lin- There are four constraints that must be fulfilled in
ear model fails to replicate the response of the system. order to achieve this kind of control.
Friction phenomenon is the cause of many failures
in mechanical parts of mobile systems, and compensa- State equation of the system requires the non-
tion can be encouraged by various constructive solu- linear state equation in the following form:
tions, but they do not eliminate nonlinearities at low
speeds. In order to obtain the most accurate model of a x_ = Ax + Bg(x)½u a(x) ð3Þ
DC motor and to enable a good synthesis of the control
where A is n 3 n and B is n 3 p matrix. The functions
system later, it is necessary to consider and take into
a : Rn ! Rp and g : Rn ! Rp 3 p are defined on domain
account the nonlinearity of friction.
D Rn and represent potential nonlinearities in the sys-
tem. It is not difficult to note that to cancel a nonlinear
Feedback theory derivation and relative part by subtraction: a(x), the control signal u, and the
degree of the system nonlinearity a(x) must appear together as the sum:
u + a(x). On the other hand, to undo the nonlinear
In the following two sections, we will look at the theoreti-
member g(x) by division, control u and nonlinearity
cal derivations related to the subject: FBL and GWO algo-
g(x) must appear as product g(x)u. If the system has
rithms. We will going to need those derivations in order to
different shape than one in the equation (3), in some
better understand behavior of the control synthesis, which
occasions it could be modified, because the model of
will be capable of rejection the disturbances in machines.
the system in space is not unique, but depends on the
System described by linear differential equations in
choice of state variables.
the state space is always possible to solve analytically.
On the other hand, when an engineer comes across g(x) has to be non-singular for any x in the
nonlinear equations it is almost regularly impossible to
domain of interest x 2 D or if it is a scalar value
reach an analytical solution. Then, the solution can
then it has to be non-zero value (the reason for
only be sought numerically.
this is clear from the previous constraint);
However, very often, especially when designing, it is Pair (A, B) must be controllable, that is there has
convenient to have an analytical solution or at least some
to be a controllability matrix U whose rank is
analytical guarantees about what the solution looks like.
equal to the order of the system: n = rank(U);
This is important primarily when choosing parameters, All functions has to be differentiable.
because then one can judge in advance, without some
complex computational operations, how changing a para-
With these conditions satisfied control law could be
meter affects the behavior of the whole system.
obtained in the following form: u = a(x) + g(x)v which
Linearization is usually done around the desired
will provide a new control signal v.
equilibrium point and uses approximation with Taylor
expansion. The concept of feedback linearization is fun-
damentally different. It does not use approximation. Input-output feedback linearization
Feedback linearization requires exactness in mea-
Sometimes, it is very cost-effective to perform lineariza-
surements in order to eliminate the nonlinearities from
tion only from input to output, even if it means that
the system.43,44
one part of the state equation will remain nonlinear.
In this section, the theoretical basis for the implemen-
The only pitfall with this type of linearization is that it
tation of the suggested algorithm will be introduced.
does not always take into account the whole dynamics
Theory derivation rely on Khalil.11 Of particular rele-
of the system.
vance will be designing the control signal with the feed-
It is necessary to determine the relative degree of the
back linearization law which will cancel the nonlinearity.
system in order to find out will the linearization all of
Consider the class nonlinear system11:
the states be possible, or will the existence of internal
dynamics occur. When relative degree of the system is
x_ = f(x) + g(x)u
ð2Þ equal to the order of the system Input-Output
y = h(x) Feedback Linearization is feasible and full state
6 Advances in Mechanical Engineering
linearization can be performed. Otherwise, internal There are four different ranks of the wolf in a pack: a,
dynamics requires further analysis. b, d, and v wolf. The leader of the pack is the a wolf
The relative degree of a linear system is defined as and the rest of the pack members follow it.
the difference between the poles and zeros.24 To Furthermore, all of the wolves are involved in the main
broaden this idea to nonlinear systems the following activity, the hunting of the prey, which is given in the
statement is given in Khalil11 and repeated here for the main steps: searching for the pray and attack.
wholeness: So as to achieve a mathematical model of encircling of
Relative degree r of the system equation (2), at a the prey, the following equations of the distance vector D
point x0 , is defined if: and a vector for position updating X(t + 1) are ensued45:
Lg Lkf h(x) = 0, for all x in an eighbourhood of x0 D = C Xp (t) X(t), X(t + 1) = Xp (t) A D,
and all k\r 1 ð7Þ
Lg Lf r1 h(x) 6¼ 0,
where A, C are the coefficient vectors and can be calcu-
where Lie derivatives of the function h(x) are noted lated as: A = 2ar1 a and C = 2r2 . r1 and r2 are ran-
with Lg and Lf k k-times respectively. Relative degree dom vectors in the range ½0, 1. Component a
will be useful in creating control signal later; simply decreasing from 2 to 0 over the course of iterations, so
put, r requires the number of output signal derivatives it is clear that the above equations are used to update
to obtain its explicit control u dependence. the wolf position according to the position of pray. The
In order to obtain control law the equation (2) is coefficient vectors are the main reason why the GWO
reconsidered in the single-input-single-output SISO is considered to be a stochastic algorithm. Finally, t is
case, when u and y are scalar values. First derivative of the current iteration, Xp is the position of the prey and
the output in general case is: X is the position vector of the gray wolf/agent.
To mathematically simulate the hunting behavior of the
x_ gray wolves, hunting process is guided by a and it is
∂h zfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflffl{ def assumed that the a, b, and d have a finer knowledge about
y_ = ( f (x) + g(x)u) = Lf h(x) + Lg h(x)u ð4Þ
∂x the potential location of the prey (i.e. the optimal solution).
where Other wolves in the pack will update their position accord-
ing to the position of a, b, and d. All of the above can be
∂h(x) expressed in accordance with the given equations:
Lf h(x) = f,
∂x
Da = jC1 Xa X(t)j, Db = C2 Xb X(t),
and analogously for Lg h(x).
If Lg h(x) = 0 ) y_ = Lf h(x), first derivative of the Dd = jC3 Xd X(t)j, X1 = jXa A1 Da j, ð8Þ
output signal will not depend on the control u, so the X2 = Xb A2 Db , X3 = jXd A3 Dd j,
search for derivatives will continue all the way until
input shows up thta is until some y(r) which will contain Xa , Xb , and Xd are the position vectors of the a, b, and
a non-zero coefficient before control signal: d wolf. A1 , A2 , A3 , C1 , C2 , C3 , are the elements of the
coefficient vectors, which are written in column.
y(r) = Lrf h(x) + Lg Lr1
f h(x) u ð5Þ
X1 + X2 + X3
X(t + 1) = : ð9Þ
By introducing the following equation, the system is lin- 3
earized by feedback from input to output: Last step is killing the prey: a wolf will finish the
hunt by attacking the prey. Then according to the wolf
1
u= ½Lrf h(x) + v ð6Þ rank in the pack they will eat. So basically, this will be
Lg Lr1
f h(x) modeled as: 1. when the prey stopped moving wolves
attack it to finish the hunting process. 2. This is mod-
eled by decreasing the value of a from 2 to 0 during the
Gray wolf optimization GWO
iterations. 3. As the value of a decreases A also
The gray wolf optimization algorithm (GWO) imitates decreases. 4. When the value of A is less than 1, the
the hunting procedure, together with the highly orga- wolf is forced to attack toward the prey and if the value
nized pecking order and social scale of the gray wolves of A is greater, it leaves and finds a better pray. C vec-
in environment. In nature, they are mutually loyal and tor is random value in interval [0, 2]. It helps with put-
respect the established hierarchy. This hierarchy is ting some extra weight on the pray. If C.1 algorithm
important for their survival because the ability to work will emphasize and if C\1 it will be turned on the de-
together increases the chance of success during hunt. emphasize (reducing importance).
Vesović et al. 7
the model and the actual object is too large and notice-
able. The main cause of machines nonlinear behavior is
the friction and that is why the new model will be con-
sidering the velocity-friction dependency.
In this work Tustins friction nonlinear model was
adopted as follows:
Figure 5. Approximation of the friction characteristic. Figure 6. Comparison between real and nonlinear model data
for step input, amplitude 5.
the approximation is performed in a different way-
using the tangent hyperbolic function. In this way only
Coulomb and viscous friction is modeled, the exponen-
tial part of the Stribek curve (static friction) is
neglected,24 which is shown on the Figure 5.
Satisfactory parameters were found. The general for-
mula of the approximation function is:
2
tanh(x) = l1 ( 1), where : ð13Þ
1 + el2 x
l1 = 0:0173607 ð14Þ
l2 = 2500 ð15Þ
Beq, n Am
x_ = Jeq x f (x) + Jeq u ð16Þ nonlinearity. With the nonlinear system that is not the
y =x case. Dead zone curve is well modeled.
Conclusion from the Figure 8:
With this choice of variables the system still remained It follows that the nonlinear model quite comparable
of the first order and the state variable is a scalar representation of the original system, for various kind
quantity. of input signals.
Having the experience with the inaccuracy of the lin-
ear model and to check the correctness of the nonlinear
model an experiment was again conducted. In the Design of the control law
experiment the actual operation of the object was again
compared with the nonlinear model. Checking the fulfillment of the conditions
Conclusion from the Figure 6: Taking into account equations (3)–(16) it could be
With the nonlinear mathematical model response of obtained:
the system to the step functions of high amplitude is
now good modeled. Increased amplitude does not Beq, n Am
A = , B= ,
increases the deviation of the model. Jeq Jeq
ð17Þ
Conclusion from the Figure 7: Jeq
The response of the real object and the linear model g(x) = 1, a(x) = f (x):
Am
to the sinusoidal excitation was not match well, because
linear model was not following the existing
Vesović et al. 9
Jeq
u= ½f (x) + v ð21Þ
Am
The signal v, that becomes the new control signal is
selected as the proportional-integral PI controller:
ðt
v = Kp e + Ki edt, ð22Þ
0
Figure 8. Comparison between real and nonlinear model data where error e = w y represents the difference between
for sinusoidal input, amplitude 5 and frequency 0.5 rad/s.
the desired output variable w and the real output angu-
lar velocity y.
State equation linearized by feedback is:
Coordinate transformation is not necessary, sys-
tem is already in the suitable form for the feed- ðt
Beq, n
back linearization; x_ = x + Kp e + Ki edt ð23Þ
Jeq 0
g(x) is scalar value which is different from zero;
Pair (A, B) is controllable when rank of the con-
trollability matrix is equal to the system order; Optimization of FBC using the GWO algorithm
In this paper, for proper operation it is necessary to set
Controllability matrix is defined as: the PI controller parameters and optimize them to get
satisfactory dynamic behavior. Further, for the design
U = ½B AB A2 B ::: An1 B ð18Þ of the optimal PI controller the metaheuristic GWO
According to the velocity, system order is equal to optimization algorithm was used. Moreover, the men-
n = 1, so controllability matrix has the form: tioned parameters are all coded into one wolf, that is
one agent, that is presented with a vector which con-
Am tains, in our case, two parameters. For the objective
U =B= ; ð19Þ function performance criterion theÐ integral of absolute
Jeq
errors (IAE) is utilized, as: IAE = jejdt.
The number of the non-controllable states is equal to In the suggested GWO algorithm the number of the
the zero. search agents is set to 30, while the maximum number
of iterations is set to 500. Additionally, one agent rep-
All functions are smooth because earlier the resents one potential optimal PI controller. All of the
approximation with the differentiable tangent parameter values that were used in the application of
hyperbolic was made; the GWO were taken from the original paper.45 After
the optimization the obtained parameters for the scal-
It is obvious that the condition rank U = n is fulfilled, ing factors are:
so it is possible to introduce a nonlinear control that
Kp = 3:8906 Ki = 130:0000 ð24Þ
will nullify the existing nonlinearity of the system itself
and perform linearization.
Stability check and experimental results
Control signal synthesis for the application of the From the new state equation (23) and PI gains equa-
feedback control method tion (24) characteristic polynomial poles of the system
are easily obtained as: p1 = 2:2261 + 11:1823i and
According to the equations (4)–(17) first derivative of p2 = 2:2261 11:1823i. As the poles are in the left
the output is: half of the s-plane the system is stable. Besides analyti-
cally, this statement is also experimentally confirmed.
Beq, n Am In the following, the comparison between the real and
y = x ! y_ = x_ = x f (x) + u ð20Þ
Jeq Jeq the desired velocity is shown.
10 Advances in Mechanical Engineering
Figure 9. Velocity tracking of step signal. Figure 11. Velocity tracking of sinusoidal signal with small
frequency of 0.1 rad/s.
Figure 13. Error signal for velocity tracking of sinusoidal signal Figure 16. Error signal for velocity tracking of chirp signal.
with small frequency of 0.1 rad/s.
Figure 18. Velocity tracking of step signal when white noise is Figure 21. Error signal for velocity tracking of sinusoidal signal
applied as disturbance to the system. with small frequency of 0.1 rad/s when white noise is applied as
disturbance to the system.
Figure 24. Velocity tracking of selected function when white Figure 26. Error signal for velocity tracking of step signal with
noise is applied as disturbance to the system. step disturbance starting in the fifth second.
Figure 29. Velocity tracking of chirp signal with step Figure 31. Velocity tracking of selected function with
disturbance starting in the fifth second. disturbance starting in the fifth second.
Declaration of conflicting interests 11. Khalil H. Nonlinear systems. 3rd ed. Hoboken, NJ: Pre-
The author(s) declared no potential conflicts of interest with ntice-Hall, 2002.
respect to the research, authorship, and/or publication of this 12. Radulović R, Jeremić B, Šalinić S, et al. A new approach for
article. the determination of the global minimum time for the bra-
chistochrone with preselected interval for the normal reac-
tion force value. Int J Non Linear Mech 2018; 101: 26–35.
Funding
13. Mirjalili S, Mirjalili SM, Saremi S, et al. Whale optimiza-
The author(s) disclosed receipt of the following financial sup- tion algorithm: Theory, literature review, and application
port for the research, authorship, and/or publication of this in designing photonic crystal filters. Nat-Inspired Optimi-
article: The research of the first two authors was supported zers 2019; 811: 219–238.
by the Science Fund of the Republic of Serbia, grant No. 14. Pishkenari HN, Mahboobi SH and Alasty A. Optimum
6523109, AI- MISSION4.0, 2020–2022. Their work was as synthesis of fuzzy logic controller for trajectory tracking
well financially supported by the Ministry of Education, by differential evolution. Sci Iran 2011; 18: 261–267.
Science and Technological Development of the Serbian 15. da Silva RM, de Souza Leite Cuadros MA and Gamarra
Government, MPNTR RS under contract 451 to 03-9/2021- DFT. Comparison of a backstepping and a fuzzy con-
14/200105, from date 05.02.2021. Also, this research was per- troller for tracking a trajectory with a mobile robot. Intell
formed within the TR 35011 and ON 174001, projects sup- Syst Des Appl 2019; 941: 212–221.
ported by Ministry of Science and Technological 16. Zhao J, Han L, Wang L, et al. The fuzzy PID control
Development, Republic of Serbia, whose funding is gratefully optimized by genetic algorithm for trajectory tracking of
acknowledged and COST Action: CA18203 – Optimising robot arm. In: 12th world congress on intelligent control
Design for Inspection. and automation, Guilin, China, 2016, pp.556–559.
17. Cordon O and Herrera F. A two-stage evolutionary pro-
ORCID iD cess for designing TSK fuzzy rule-based systems. IEEE
Mitra Vesović https://orcid.org/0000-0003-0457-1874 Trans Syst Man Cybern Part B Cybern 1999; 29: 703–715.
18. Tsai SH and Chen YW. A novel identification method
References for Takagi–Sugeno fuzzy model. Fuzzy Sets Syst 2018;
338: 117–135.
1. George M. Speed control of separately excited DC 19. Mirjalili S and Lewis A. The whale optimization algo-
motor. Am J Appl Sci 2008; 5: 227–233. rithm. Adv Eng Softw 2016; 95: 51–67.
2. Li W, Chen L, Trisovic N, et al. First passage of stochas- 20. Sahoo BP and Panda S. Improved grey wolf optimization
tic fractional derivative systems with power-form restor- technique for fuzzy aided PID controller design for power
ing force. Int J Non Linear Mech 2015; 71: 83–88. system frequency control. Sustain Energy Grid Netw
3. Orošnjak M, Jocanović M, Čavić M, et al. Industrial 2018; 16: 278–299.
maintenance 4(.0) horizon Europe: consequences of the 21. Park D, Kandel A and Langholz G. Genetic-based new
iron curtain and energy-based maintenance. J Clean Prod fuzzy reasoning models with application to fuzzy control.
2021; 314: 128034. IEEE Trans Syst Man Cybern 1994; 24: 39–47.
4. Sarkar SK and Das SK. High performance nonlinear 22. Kamali MZ, Kumaresan N and Ratnavelu K. Takagi–
controller design for AC and DC machines: partial feed- Sugeno fuzzy modelling of some nonlinear problems
back linearization approach. Int J Dyn Control 2018; 6: using ant colony programming. Appl Math Model 2017;
679–693. 48: 635–654.
5. Furuhashi T, Sangwongwanich S and Okuma S. A 23. Turki M and Sakly A. Extracting T–S fuzzy models using
position-and-velocity sensorless control for brushless DC the cuckoo search algorithm. Comput Intell Neurosci
motors using an adaptive sliding mode observer. IEEE 2017; 2017: 8942394.
Trans Ind Electron 1992; 39: 89–95. 24. Vesović M, Jovanović R, Laban L, et al. Modelling and
6. Valluru S, Singh N and Kumar M. Implementation of control of a series direct current (DC) machines using
NARMA-L2 neuro controller for speed regulation of series feedback linearization approach. In: IcEtran conference,
connected DC motor. In: IEEE 5th India international con- Belgrade, Serbia, 2020, pp.191–197.
ference on power electronics (IICPE), Delhi, India, 2012, 25. Chen JL and Chang WD. Feedback linearization control
pp.1–7. of a two-link robot using a multi-crossover genetic algo-
7. Azar AT and Vaidyanathan S. Handbook of research on rithm. Expert Syst Appl 2009; 36: 4154–4159.
advanced intelligent control engineering and automation. 26. Vesović M, Jovanović R, Zarić V, et al. Modelling and
Advances in computational intelligence and robotics speed control in a series direct current (DC) machines
(ACIR) book series, New York: IGI Global, 2014. using feedback linearization approach. In: The Fifth inter-
8. Alhanjouri M. Speed control of DC motor using artifi- national conference mechanical engineering in XXI century
cial neural network. Int J Sci Res 2017; 6: 2140–2148. – MASING 2020, Niš, Serbia, 2020, pp.207–213.
9. Ismail NL, Zakaria KA, Moh Nazar N S, et al. DC motor 27. Mehta S and Chiasson J. Nonlinear control of a series
speed control using fuzzy logic controller. In: AIP conference DC motor: theory and experiment. IEEE Trans Ind Elec-
proceedings, Maharashtra, India, 1930, 2018, p.020026. tron 1998; 45: 134–141.
10. Uddin MN and Rahman MA. High-speed control of 28. Slotine JE and Li W. Feedback linearization. In:
IPMSM drives using improved fuzzy logic algorithms. Wenzel J (ed.) Applied nonlinear control. Englewood
IEEE Trans Ind Electron 2007; 54: 190–199. Cliffs, NJ: Prentice Hall, 1991, p.207.
16 Advances in Mechanical Engineering
29. Ghozlane W and Knani J. Nonlinear control via input- 42. Kandel A and Langolz G. Fuzzy control systems. Boca
output feedback linearization of a robot manipulator. Raton, FL: CRC Press, 1993.
Adv Sci Technol Eng Syst J 2018; 3: 374–381. 43. Piltan F, Yarmahmoudi M, Mirzaie M, et al. Design
30. Cambera JC and Feliu-Batlle V. Input-state feedback lin- novel fuzzy robust feedback linearization control with
earization control of a single-link flexible robot arm mov- application to robot manipulator. Int J Intell Syst Appl
ing under gravity and joint friction. Robot Auton Syst 2013; 5: 1–10.
2017; 88: 24–36. 44. Chen YT, Yu CS and Chen PN. Feedback linearization
31. Mehndiratta M, Kayacan E and Kayacan E. A simple based robust control for linear permanent magnet syn-
learning strategy for feedback linearization control of chronous motors. Energies 2020; 13: 5242.
aerial package delivery robot. In: IEEE conference on 45. Mirjalili S, Mirjalili SM and Lewis A. Grey wolf optimi-
control technology and applications, Copenhagen, Den- zer. Adv Eng Softw 2014; 69: 46–61.
mark, 2018, pp.361–367. 46. Gruyitch LT, Bučevac ZM, Jovanović RŽ, et al. Structu-
32. Kali Y, Saad M and Benjelloun K. Optimal super- rally variable control of Lurie systems. Int J Control
twisting algorithm with time delay estimation for robot 2020; 93: 2960–2972.
manipulators based on feedback linearization. Robot 47. Mao WL and Hung CW. Adaptive neural network-based
Auton Syst 2018; 108: 87–99. synchronized control of dual-axis linear actuators. Adv
33. Bagheri M, Naseradinmousavi P and Krstić M. Feedback Mech Eng 2016; 8: 1687814016654603.
linearization based predictor for time delay control of a high- 48. Spanos PD and Evangelatos GI. Response of a non-
DOF robot manipulator. Automatica 2019; 108: 108485. linear system with restoring forces governed by fractional
34. Sun X, Jin Z, Cai Y, et al. Grey wolf optimization algo- derivatives—time domain simulation and statistical line-
rithm based state feedback control for a bearingless per- arization solution. Soil Dyn Earthq Eng 2010; 30:
manent magnet synchronous machine. IEEE Trans 811–821.
Power Electron 2020; 35: 13631–13640. 49. Pan W, Ling L, Qu H, et al. Nonlinear response analysis
35. Sun X, Hu C, Lei G, et al. State feedback control for a of aero-engine rotor bearing rub-impact system caused by
PM hub motor based on gray wolf optimization algo- horizontal yawing maneuver load. Int J Non Linear Mech
rithm. IEEE Trans Power Electron 2020; 35: 1136–1146. 2021; 137: 103800.
36. Sun X, Cao J, Lei G, et al. A robust deadbeat predictive 50. Spanos PD, Sofi A and Di Paola M. Nonstationary
controller with delay compensation based on composite response envelope probability densities of nonlinear oscil-
sliding-mode observer for PMSMs. IEEE Trans Power lators. J Appl Mech 2007; 74: 315–324.
Electron 2021; 36: 10742–10752. 51. Čavić M, Penčić M, Oros D, et al. High-capacity stacking
37. Moradi M, Ahmadi A and Abhari S. Optimal control apparatus for thermoforming machine – Part I: Synthesis
based feedback linearization for position control of DC of intermittent mechanisms as stacker driving units. Adv
motor. In: IEEE 2nd international conference on advanced Mech Eng 2021; 13: 16878140211040892.
computer control 2010, Shenyang, China, 2010, vol. 4, 52. Penčić M, Čavić M, Oros D, et al. High-capacity stacking
pp.312–316. apparatus for thermoforming machine – Part II: Struc-
38. Shirvani Boroujeni M, Markadeh GRA and Soltani J. tural design of the adjustable stacker driving mechanism.
Torque ripple reduction of brushless DC motor based on Adv Mech Eng 2021; 13: 16878140211052458.
adaptive input-output feedback linearization. ISA Trans 53. Wang T, Wang H, Hu H, et al. LQR optimized BP
2017; 70: 502–511. neural network PI controller for speed control of
39. Faris H, Aljarah I, Al-Betar MA, et al. Grey wolf optimi- brushless DC motor. Adv Mech Eng 2020; 12:
zer: a review of recent variants and applications. Neural 1687814020968980.
Comput Appl 2018; 30: 413–435. 54. Medjghou A, Ghanai M and Chafaa K. Improved feed-
40. Haq H and Okumusx HI. A simple approach to DC motor back linearization control based on PSO optimization of
modeling and comparative implementation of PI and an extended Kalman filter. Optim Control Appl Methods
FLC for its speed controlling. Int J Innov Res Sci Eng 2018; 39: 1871–1886.
Technol 2016; 5: 98–105. 55. Grozdanović I, Todorović K, Vasović N, et al. Interplay
41. Apkarian J, Levis M and Gurocak H. Student workbook: between internal delays and coherent oscillations in
SRV02 base unit experiment for Matlab/Simulink users delayed coupled noisy excitable systems. Int J Non Linear
Markham, ON: Quanser Inc., 2012. Mech 2015; 73: 121–127.