Sie sind auf Seite 1von 6

The International

Proceedings Federation
of the 20th Worldof Congress
Automatic Control
Proceedings
Toulouse,
The of the
France,
International 20th9-14,
July World
Federation of Congress
2017
Automatic Control
The International Available online at www.sciencedirect.com
Toulouse, France,Federation of Automatic
July 9-14, 2017 Control
Toulouse, France, July 9-14, 2017

ScienceDirect
IFAC PapersOnLine 50-1 (2017) 13498–13503
Modern Control Systems via Virtual and
Modern Control Systems via Virtual and
RemoteControl
Modern Laboratory Systems Based viaon Matlab
Virtual and
Remote Laboratory Based on Matlab
Remote Laboratory ∗
Based

on ∗Matlab
P. Bisták M. Halás M. Huba
P. Bisták ∗∗ M. Halás ∗∗ M. Huba ∗∗
∗ P. Bisták M. Halás M. Huba
Slovak University of Technology in Bratislava, Slovakia

Slovak
(e-mail:

University mikulas.huba,
pavol.bistak, of Technology miroslav.halas@stuba.sk)
in Bratislava, Slovakia
Slovak
(e-mail: University mikulas.huba,
pavol.bistak, of Technology miroslav.halas@stuba.sk)
in Bratislava, Slovakia
(e-mail: pavol.bistak, mikulas.huba, miroslav.halas@stuba.sk)
Abstract: Controller design for hydraulic plants represents a task carried out frequently to
Abstract: Controller
demonstrate and compare design for hydraulic
properties plants represents
of different nonlinear acontrol task carried
approaches. out frequently
This paper to
Abstract:
demonstrate
discusses theController
and
problem compare design
of for hydraulic
properties
applying plants solution
of different
the theoretical represents
nonlinear to theacontrol
task carried
approaches.
trajectory out frequently
tracking This paper
problem, to
demonstrate
discusses
which is the
based and oncompare
problem theoffeedbackproperties
applying of different
the theoretical
linearization nonlinear
solution
techniques, tointhecontrol approaches.
trajectory
practice. It istracking
shown This paper
problem,
that the
discusses
which is the
theoretical based problem
solutionon theis ofnot applying
feedback
sufficient theintheoretical
linearization ofsolution
the casetechniques,
parameter tointhe trajectory
practice.
or model It istracking
shown problem,
uncertainties, that
and thean
which
additional is based
theoretical control on
solutionloop the
is not feedback linearization
sufficient The
is necessary. in the techniques,
case of modifications
respective in
parameter or are practice.
model It is shown
uncertainties,
proposed, and the that
and the
an
results
theoretical
additional
are solutionloop
control
demonstrated byisthenot sufficient
is necessary.
Matlab in the
The
application caseon of
respective parameter
themodifications
real laboratory or are
model uncertainties,
proposed,
system of and the and
coupled an
results
tanks.
additional
are demonstrated
The control
application canbyrun loop is necessary.
thelocally
Matlab The
as application respective
a simulationon tool themodifications
real laboratory
designed are proposed,
system of coupled
in Matlab/Simulink and the
graphical results
tanks.
user
are
Thedemonstrated
application
interface but alsocan byrunthelocally
through Matlab as application
a simulation
the Internet on
tool
as a virtual the
orreal
designed
remotelaboratory
inlaboratory. system of coupled
Matlab/Simulink graphical tanks.
user
The application
interface but alsocan run locally
through as a simulation
the Internet tool designed
as a virtual or remoteinlaboratory.
Matlab/Simulink graphical user
© 2017, IFAC
interface but (International
also throughFederationthe Internet of Automatic
as a virtualControl) Hostinglaboratory.
or remote by Elsevier Ltd. All rights reserved.
Keywords: nonlinear systems, trajectory tracking, feedback linearization, coupled tanks,
Keywords: nonlinear
Matlab/Simulink, systems, tool,
interactive trajectory
virtualtracking,
and remote feedback linearization, coupled tanks,
laboratory
Keywords: nonlinear
Matlab/Simulink, systems, tool,
interactive trajectory
virtualtracking,
and remote feedback linearization, coupled tanks,
laboratory
Matlab/Simulink, interactive tool, virtual and remote laboratory
1. INTRODUCTION In this paper we continue in this direction and focus on
1. INTRODUCTION In
thethis paper we
trajectory continue
tracking in thisi.e.direction
problem, a problem andof focus
tracking on
1. INTRODUCTION aIn given
the this paper
trajectory
reference we continue
tracking
signal, in thisi.e.
problem,
which direction
a problem
represents and
an of focus
tracking
important on
thegiven
trajectory
apractical reference tracking
signal, problem,
which i.e. a problem
represents
control problem employed in various areas using an of tracking
important
Feedback linearization techniques represent one of the a given reference signal, employed
which
Feedback linearization
most fundamental techniques
approaches to deal represent one of con-
with nonlinear the practical control statements
various problem problem and represents
in various
solutions, an areas
see important
e.g. using
Conte
Feedback linearization techniques represent one of the practical
various
et al. control
problem
(2007); Aguiar problem
statements
and employed
and
Hespanha in various
solutions,
(2007). see
The areas
e.g. using
Conte
theoretical
most fundamental
trol systems. Theyapproaches
do not play to deal
a key with nonlinear
role only in con- the various problem statements and solutions, see e.g. Conte
mostsystems.
trol fundamental
(input-to-output They orapproaches
do not play
input-to-state) to deal with nonlinear
a linearization
key role the et
in con-
onlyproblem al. (2007);
solution to the
et al. (2007);
Aguiar
Aguiar
and Hespanha
trajectory
and Hespanha
(2007). The
tracking problem,
(2007). The
theoretical
as discussed
theoretical
in
trol systems.
(input-to-output They or do not play
input-to-state) a key role
linearization only in the
problem solution
Conte et to
al.the trajectory
(2007), is trackingasproblem,
considered another as discussed
application in
of
itself, but they are essential also in solving various decou- solution to the trajectory tracking problem, as discussed in
Conte et al. (2007), is considered as another application of
(input-to-output
itself,
pling and they areorproblems
but tracking input-to-state)
essential also in
(e.g. linearization
solving
trajectory various
tracking prob- the
problem
decou- Conte
feedback
et than
linearization
al. (2007),
technique.
is considered
However,
as another
it is more
application of
the feedback linearization technique. However, it problem
is more
itself,disturbance
pling
lem, but tracking
and they are essential (e.g.
problems
decoupling, also in solving various
trajectory
noninteracting control,decou-
tracking etc.), complex
prob- the feedback
those to
linearization
the disturbance
technique.
decoupling
However, it is more
pling
lem,
as and summarized
trackingdecoupling,
disturbance
nicely problems (e.g.
for instance trajectory
in Contetracking
noninteracting control, etc.), complex
prob-
et al. (2007). complex
than those control,
and noninteracting
than those
to the disturbance
to the
as it involves
disturbance
decoupling
also the problem
decoupling
require-
lem,
as disturbance
nicely summarized decoupling,
for noninteracting
instance in Conte control,
et al. etc.),
(2007). and
ment noninteracting
that the outputcontrol,
y of as itclosed-loop
the involves also the problem
system require-
tracks
However, main drawback of the feedback linearization and noninteracting control, as it involves also the require-
as nicely summarized
However,
techniques main drawback
is, in general, for that
instancethe in
of they Conteonly
feedback
provide et al. (2007). ment
linearization
a (static)
that thesignal
the reference outputyry asymptotically.
of the closed-loop Undersystem tracks
parameter
However, main drawback ofandthedoprovide
feedback linearization ment
the
or modelthat uncertainties
reference thesignaloutputyry asymptotically.
aof corresponding
the closed-loop Undersystem tracks
parameter
modification of
techniques
state feedbackis, in
togeneral,
the system that they not reallyonly a (static)
control the the reference signal y asymptotically. Under parameter
control the or model uncertainties a corresponding modification of
or model uncertainties a corresponding modification
techniques is, in general, that they provide only a (static) the solution to the r
trajectory tracking problem is needed
state
system. feedback
Therefore,to the in system
practice, and do work
they not really
as satisfactorily of
the solution to inthe trajectory tracking problem isofneeded
state
as thefeedback
system. Therefore,
mathematicalto the system
inmodel
practice, and
(e.g. do work
they not really
differential control the
asequations)
satisfactorily re- and the
suggested
solution to the
this paper. As
trajectory
a representative
tracking problem is
a real
needed
and suggested
which inthe this paper. Astracking a representative of solved
a real
system.
as the
ally Therefore,
models the real in
mathematical practice,
model
process (e.g.
under they work asequations)
differential
consideration. satisfactorily
If therere-is process,
and suggested inthe
trajectory
this paper. Astracking
problem is
a representative of solved
a real
as the
ally mathematical
models thefor real model (e.g.
processbetween differential
under consideration. equations)
If there re- process,
for, the which
coupled tankstrajectory
were chosen.
is process, which the trajectory tracking problem is solved The problem
main is
motivation
any difference instance the real process output
ally difference
any models theforreal process
instance under consideration. If there for, the coupled tanks were chosen.
is is that this system is one of the most illustrative nonlinear The main motivation
and its required value (duebetween
to the real process
the inaccuracies between output
the for, thethis
coupled tanks were chosen.
is that system isand
one of the mostThe mainandmotivation
illustrative nonlinear
real process and its model) the feedback linearization is practical
any
and difference
its required for instance
value (due between
to the the real
inaccuracies process
between output
the is that this
processes
system is one
many
of the
laboratories
most illustrative
institutes
nonlinear
and
real its required
process and value (due
its model) to the inaccuracies
the feedbackare, between
linearization the practical
whose processes
activities are and many
dedicated
is practical processes and many laboratories and instituteslaboratories
to control and institutes
engineering pos-
not sufficient. However, such inaccuracies in principle,
real
not process
sufficient. and its model) the feedback linearization is whose
sess activities
one, see e.g. are
Hou dedicated
and Zhu to control
(2013); engineering
Králik and pos-
Žáková
always presentHowever,
in all controlsuch problems,
inaccuracies are,
as it is in
notprinciple,
possible whose activities are dedicated to control engineering pos-
not sufficient.
always present However,
in all such
control inaccuracies
problems, as are,
it is in
not principle,
possible sess one,
(2015) andseethe e.g.references
Hou and therein.Zhu (2013); Králik and Žáková
to model a process with absolute precision. Therefore, sess one, see e.g. Hou and Zhu (2013); Králik and Žáková
always
to
one canpresent
model incertain
a process
expect all control
with problems,
absolute
problems as itwhen
precision.
appear is not possible (2015) and the references therein.
Therefore,
applying This paper
(2015) and the introduces
references a Matlab/Simulink
therein. interactive ap-
to
one model
the feedback a
can expect process with
certain problems
linearization absolute
techniques precision.
appear on awhen Therefore,
real applying
process. This paperappropriate
introduces for a Matlab/Simulink interactive ap-
plication the experimentation with the
one feedback
the
For can expect
example, thatcertain
linearization problems
the feedback techniques appear on awhen
linearization real applying
process. This
techniques paper
plication introduces
appropriate a Matlab/Simulink
for the experimentation interactive
withcom- ap-
the
coupled tanks. The recently developed and described
the
For feedback
example, linearization
that the techniques
feedback on
linearization
are not robust under parameter uncertainties. This aspect coupled a real process.
techniques plication appropriate
tanks. tool The recentlyfor the experimentation
developed control
and described with the
puter support for corresponding design com-(Bis-
For not
are
was example,
robust that
discussed under the
in details feedback
parameter linearization
uncertainties.
on practical applications techniques
This two coupled
aspect
of puter tanks. The recently developed and described com-
tak etsupport
al. (2015);tool for
Bistakcorresponding
and Hubacontrol
(2016)) design (Bis-
integrates
are not robust
controlunder parameter uncertainties.
tanks, This aspect
the puter etsupport tool for
for corresponding Hubacontrol
was discussed
specific inproblems
details onforpractical
coupled applications namely of two tak
control al. (2015);
algorithms Bistakseveral andfrequent plant design
(2016)) (Bis-
integrates
and control
was discussed
specific
disturbancecontrol inproblems
detailsproblem
decoupling onforpractical
coupled applications
tanks, namely
and noninteracting of con-
two
the tak controlet al. (2015);
algorithms Bistak andfrequent
Huba (2016)) integrates
configurations. Thisfor paperseveraldescribes tool plant and
extension control
by the
specific
see control problems for coupled tanks, namely the control algorithms for several frequent
disturbance
trol, Žilkadecoupling
and Halásproblem
(2010); and
Halás noninteracting
and Žilka con-
(2011). configurations.
trajectory tracking Thiscontrol
paper describes
algorithm for plant
tool and control
extension
the coupled by the
tanks
disturbance
trol,
It wasseeshown
Žilkadecoupling
and an problem
thatHalás (2010); and
additional Halás noninteracting
and Žilka
controller con- configurations.
(2011).
is required trajectory tracking This paper
control describes
algorithm tool extension
coupledby
for thealgorithm the
tanks
system. The proposed nonlinear control with
trol,
It
in wasseeshown
order Žilka and
thatHalás
to satisfactorily (2010);
control Halás
an additional the and
controller
real Žilka
process (2011).
and to trajectory
is required system. Thetracking
proposed control algorithm
nonlinear for
control the coupled
algorithm tanks
with
input and output disturbances is accessible for students
It order
was shown that an additional controller is required
and to system. The proposed nonlinear is control
in
achieve to satisfactorily
the disturbance control
decoupling the real
and process
noninteracting input and
via Internet outputin thedisturbances
form of virtual and algorithm
accessible for students
remote with
labora-
in order
achieve to
the satisfactorily
disturbance control
decoupling the real
and
control respectively. One can think of these solutions as via process and
noninteracting to input and output disturbances is accessible for students
tory.Internet
Virtual in andtheremote
form laboratories
of virtual and have remote
became labora-
very
achieve
control
the robust theequivalences
disturbance
respectively. One of decoupling
can think ofand
the disturbance thesenoninteracting
solutionsand
decoupling as via tory. Internet
Virtual in
andthe form
remote of virtual
laboratories andhave remote
became labora-
very
popular and one can see many different approaches and
control
the robustrespectively.
noninteracting controlOne
equivalences of can think of these
the disturbance
respectively. solutionsand
decoupling as tory.popular Virtual
and one and canremotesee manylaboratories
different have became very
approaches and
the robust equivalences
noninteracting of the disturbance decoupling and popular and one can see many different approaches and
control respectively.
noninteracting control respectively.
Copyright © 2017 IFAC 14040
2405-8963 © 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright © 2017 IFAC 14040
Peer review©under
Copyright 2017 responsibility
IFAC of International Federation of Automatic
14040Control.
10.1016/j.ifacol.2017.08.2335
Proceedings of the 20th IFAC World Congress
Toulouse, France, July 9-14, 2017 P. Bisták et al. / IFAC PapersOnLine 50-1 (2017) 13498–13503 13499

technologies how to build them (Duro et al. (2008); Chacón


et al. (2015); Sáenz et al. (2015); Kalúz et al. (2015)). The
novelty of the described virtual laboratory consists in the
fact that it is based exclusively on the Matlab/Simulink
software.
The paper is organized as follows. After Introduction
the section 2 discusses mathematical setting and problem
statement. The next chapter introduces the coupled tanks
system. Section 4 and 5 are describing the design of trajec-
tory tracking and PI controllers whereas the section 6 dis-
cusses their application to the real system. The developed
application is described as an interactive Matlab/Simulink
tool in the section 7 that is followed by Conclusions.

2. MATHEMATICAL SETTING AND PROBLEM


STATEMENT

Generally, the trajectory tracking problem can be ex-


pressed for the output y of a given system and the desired
reference signal yr by requirement of an asymptotic de- Fig. 1. Coupled tanks laboratory system
crease of the control error e = yr − y. Mathematically, this the asymptotic behavior can again be easily reached by
can be formulated as follows. prescribing p differential equations in the form of (3)
Let us consider a system of the form i=ρj −1

(ρj ) (i)
ej =− λj,i ej , j = 1, . . . , p (5)
ẋ = f (x) + g(x)u i=0

y = h(x) (1) where ρj is relative degree of the output component yj ,


sρj + λj,ρj −1 sρj −1 + · · · + λj,1 s + λj,0 , j = 1, . . . , p are
where x ∈ R , u ∈ R , and y ∈ R are state, input,
n m p
Hurwitz polynomials, and all λj,i are in R.
and output of the system, respectively. The functions f , (ρ )
g, and h are considered to be real analytic functions on After substituting (4) into (5) one can isolate yj j
some suitable open and dense subset, and the desired i=ρj −1

(ρj ) (ρ ) (i) (i)
output trajectory, denoted by yr , is considered to be a yj = yrj j + λj,i [yrj − yj ] (6)
real analytic function of time t, for t ≥ 0. i=0
(ρj )
The aim is to find, if possible, a feedback uF such that it From (1), which is affine in u, implies that also yj (x, u)
achieves the trajectory tracking. That is, it guaranties the is affine in u
control error dynamics (ρ )
yj j = aj (x) + bj (x)u , j = 1, . . . , p (7)
e = yr − y (2)
for some vector functions aj (x) and bj (x). Then if the
goes asymptotically to 0 for t → ∞. Under this condition assumption
the output y tracks the desired trajectory yr asymptoti-  
(ρ ) (ρ )
cally. ∂ y1 1 , . . . , yp p
rank =p (8)
In the case the system (1) is a single-input single-output ∂u
system, and has a relative degree holds by substituting (7) into (6) u can be solved and
  denoted as the searched feedback control law uF consisting
∂y (k)
ρ = min k > 0 ; = 0 of elements uF j of the form
∂u
finite, the asymptotic behavior can easily be assured by uF j = Fj (x, t) , j = 1, . . . , p (9)
prescribing the differential equation of the control error
dynamics in the following form where Fj is a real meromorphic function.
i=ρ−1

e(ρ) = − λi e(i) (3) 3. COUPLED TANKS SYSTEM
i=0
The coupled tanks system (Fig. 1) is built of two tanks,
where all λi ∈ R are such that s +λρ−1 sρ−1 +· · ·+λ1 s+λ0
ρ
three valves and two pumps. Its variables and parameters
is a Hurwitz polynomial.
are denoted according to Fig. 2. The liquid level in the first
tank is denoted as x1 , with the liquid inflow u1 , tank cross-
For the multi-input multi-output system (1) it is necessary section area A1 and the valve outflow coefficient c1 . Similar
to assure that each of the output components yj , j = denotation is valid also for the second tank. Both tanks are
1, . . . , p, will track asymptotically the respective desired coupled through the valve characterized by the coefficient
trajectory yrj , j = 1, . . . , p. After introducing p control c12 . The system can be modeled by differential equations
errors dynamics derived via the application of the mass conservation law.
ej = yrj − yj (4) That is

14041
Proceedings of the 20th IFAC World Congress
13500
Toulouse, France, July 9-14, 2017 P. Bisták et al. / IFAC PapersOnLine 50-1 (2017) 13498–13503

√ √
uF 1 = A1 (c1 x1 + c12 ± x1 − x2 + ẏr1 + λ1,0 (yr1 − x1 )) =
√ √
= A1 (c1 x1 + c12 ± x1 − x2 + M1 ω1 cos ω1 t
+λ1,0 (M1 sin ω1 t + w1 − x1 ))
√ √
uF 2 = A2 (c2 x2 − c12 ± x1 − x2 + ẏr2 + λ2,0 (yr2 − x2 )) =
√ √
= A2 (c2 x2 − c12 ± x1 − x2 −
−M2 ω2 sin ω2 t + λ2,0 (M2 cos ω2 t + w2 − x2 ))
(15)
Such a feedback control causes the outputs of the coupled
tanks y1 and y2 track the desired trajectories yr1 and yr2
Fig. 2. Schematic drawing of coupled tanks respectively. This could be true only in the ideal case when
the mathematical model of the coupled tanks (10) exactly
di do matches with the real system of the coupled tanks. In the
case of parameter or model uncertainties, which is always
Trajectory + + Coupled + + y the case in practice, the proposed feedback control law (15)
yr tracking uF = u tanks does not precisely fulfill the requirement of an asymptotic
x1 x2 tracking, and even a permanent control error appears.
Fig. 4 shows the time responses of the simulated cou-
pled tanks trajectory tracking control when the model
Fig. 3. Trajectory tracking control loop parameters of (10) (c1 , c2 , c12 , A1 and A2 ) have been
decreased by 10% while the trajectory tracking control
1 √ √
ẋ1 = u1 − c1 x1 − c12 ± x1 − x2 (15) was calculated with the original values. From the
A1 output signals y1 and y2 one can see they differ from
1 √ √ the reference trajectories yr1 and yr2 permanently. The
ẋ2 = u2 − c2 x2 + c12 ± x1 − x2
A2 transients responses show also the reaction to the input
y1 = x1 and output disturbances. The output disturbances do1
and do2 were applied in times 100s and 200s respec-
y2 = x2 (10)
√ tively and they did not increase the error permanently
where the symbol · stands for
±
whereas the input disturbances di1 = −6 · 10−6 m3 s−1 and
√ di2 = 6 · 10−6 m3 s−1 applied in times 300s and 400s re-
√ z if z ≥ 0
±
z= √ (11) spectively increased the error permanently. The other
− −z if z < 0
parameters were: M1 = 0.01m , M2 = 0.005m , ω1 =
From the system description (10) one can see that for both π/40 rad s−1 , ω2 = π/80 rad s−1 , w1 = 0.23m , w2 =
outputs y1 and y2 the relative degrees ρ1 and ρ2 are equal 0.21m , λ1,0 = 0.5 , λ2,0 = 2.
to 1 √ √
ẏ1 = −c1 x1 − c12 ± x1 − x2 + A11 u1 5. TRAJECTORY TRACKING AND PI CONTROL
√ √ (12)
ẏ2 = −c2 x2 + c12 ± x1 − x2 + A12 u2 OF COUPLED TANKS
and (8) holds.
There are several possibilities how to get rid of a perma-
nent control error. From the internal model principle it
4. TRAJECTORY TRACKING FOR COUPLED follows that also a simple linear PI controller could reject
TANKS constant disturbances where, of course, the I part plays
the significant role. Therefore a parallel PI-controller has
Fig. 3 displays the principal block diagram of the trajec- been designed in the form
tory tracking for the coupled tanks (10) where also the 1
input and output disturbances are presented as di and do . Cj (s) = Gj (1 + ) , j = 1, 2 (16)
Assume now that the desired trajectories yr are given for Tj s
instance as harmonic signals Then the resulting control loop consists of the trajectory
tracking control loop (it produces uF j signal) and the PI
yr1 = M1 sin ω1 t + w1 control loop (uP Ij signal) as depicted in the Fig. 5 that
for the control law uj yields the sum
yr2 = M2 cos ω2 t + w2 (13) uj = uF j + uP Ij , j = 1, 2 (17)
Because relative degrees of (10) ρ1 = ρ2 = 1 it is required Fig. 6 shows time responses of the control loop according
according to (9) to evaluate first derivatives of the desired to Fig. 5 considering also the output do and input di
trajectories disturbances. These are acting at the same times and with
the same values as in the previous case (Fig. 4). The
ẏr1 = M1 ω1 cos ω1 t parameters are also the same with four additional ones:
G1 = 0.0002s, G2 = 0.0005s, T1 = 2s and T2 = 0.5s. From
ẏr2 = −M2 ω2 sin ω2 t (14) the output responses one can notice there is no permanent
Then after using (10),(12),(13),(14) for solving (9) one can control error. Due to the PI loop this control structure is
derive the resulting feedback also able to cope with the input and output disturbances,

14042
Proceedings of the 20th IFAC World Congress
Toulouse, France, July 9-14, 2017

P. Bisták et al. / IFAC PapersOnLine 50-1 (2017) 13498–13503 13501

Output y1 and reference trajectory yr1 Output y1 and reference trajectory yr1
0.28 0.28
y y y y
1 r1 1 r1
0.26 0.26
y1,yr1

y1,yr1
0.24 0.24

0.22 0.22

0.2 0.2
0 100 200 300 400 500 0 100 200 300 400 500
t[s] t[s]
Output y and reference trajectory y Output y and reference trajectory y
2 r2 2 r2
0.26 0.26
y2 yr2 y2 yr2
0.24 0.24
y2,yr2

y2,yr2
0.22 0.22

0.2 0.2

0.18 0.18
0 100 200 300 400 500 0 100 200 300 400 500
t[s] t[s]
−5 Control uF1 −5 Control u1,uF1,uPI1
x 10 x 10
2
2

PI1
1 uF1

u ,u ,u
1 u1 uF1 uPI1
uF1

F1
1
0 0

−1 −1
0 100 200 300 400 500 0 100 200 300 400 500
t[s] t[s]
−5 Control u −5 Control u ,u ,u
x 10 F2 x 10 2 F2 PI2

2
uF2 2 u2 uF2 uPI2
PI2

1
u ,u ,u

1
uF2

F2
2

0 0

−1 −1
0 100 200 300 400 500 0 100 200 300 400 500
t[s] t[s]

Fig. 4. Trajectory tracking control of coupled tanks with Fig. 6. Trajectory tracking control of coupled tanks com-
model mismatch considering input and output distur- bined with PI control considering input and out-
bances - outputs y1 and y2 , reference trajectories yr1 put disturbances - outputs y1 and y2 , reference tra-
and yr2 , and control time responses u1 = uF 1 and jectories yr1 and yr2 , and control time responses
u2 = uF 2 . (Time responses of local simulation) u1 , uF 1 , uP I1 and u2 , uF 2 , uP I2 . (Time responses of
client application as an example of virtual laboratory)
e PI uPI
controller
di do
+
yr Trajectory + + + Coupled + + y
tracking uF u tanks c2 = 0.0261m 1/2 −1
s . The constraints of control signal were
 
x1 x2 uj ∈ 3.16 · 10−6 m3 s−1 ; 4.76 · 10−5 m3 s−1 , j = 1, 2.
The time responses of the real system can be seen in Fig. 7.
The parameters of the desired trajectories and control pa-
Fig. 5. Control structure consisting of Trajectory tracking rameters are identical with the simulations in the previous
and PI controller chapter (Fig. 6). The most important result is that the
contrary to the trajectory tracking loop itself which alone outputs y1 and y2 track the desired trajectories yr1 and yr2
is able to eliminate only output disturbances. given by (13) without the steady state error. The control
law is calculated according to (17). The trajectory tracking
6. APPLICATION TO THE REAL COUPLED TANKS control parts (uF 1 , uF 2 ) are almost identical with the sim-
SYSTEM ulations (Fig. 6) except of the noise naturally presented
in the real process. The PI control parts (uP I1 , uP I2 )
The performance of the designed trajectory tracking con- are different with respect to the simulations that means
trol combined with the PI controller is demonstrated on a model mismatch and non-modeled dynamics are much
the example of the real coupled tanks system (10). The higher than expected in the simulations. The noise, par-
system was identified with the following parameters: A1 = tially caused by the measurement, has been reduced to an
A2 = 0.001m2 , c1 = 0.0271m1/2 s−1 , c12 = 0.0251m1/2 s−1 , acceptable level using a FIR filter.

14043
Proceedings of the 20th IFAC World Congress
13502
Toulouse, France, July 9-14, 2017 P. Bisták et al. / IFAC PapersOnLine 50-1 (2017) 13498–13503

Output y and reference trajectory y


1 r1
0.28
y y
1 r1
0.26
1 r1
y ,y

0.24

0.22

0.2
0 100 200 300 400 500
t[s]
Output y and reference trajectory y
2 r2
0.26
y2 yr2
0.24
2 r2
y ,y

0.22

0.2

0.18
0 100 200 300 400 500
t[s]
−5 Control u1,uF1,uPI1
x 10
3
PI1

2
u ,u ,u
F1

1 Fig. 8. GUI of client application


1

0 u1 uF1 uPI1

−1
0 100 200 300 400 500
t[s]
−5 Control u2,uF2,uPI2
x 10
3
PI2

2
u ,u ,u
F2

1
2

0 u u u
2 F2 PI2
−1
0 100 200 300 400 500
t[s]

Fig. 7. Trajectory tracking control of real coupled tanks


combined with PI control considering input and out-
put disturbances - outputs y1 and y2 , reference tra-
jectories yr1 and yr2 , and control time responses
u1 , uF 1 , uP I1 and u2 , uF 2 , uP I2 . (Time responses of
client application as an example of remote laboratory)

7. MATLAB/SIMULINK APPLICATION
Fig. 9. Client block diagram in Simulink
The control algorithms described in the previous sections
are implemented in the Matlab/Simulink software envi- feedforward and Trajectory tracking controllers are imple-
ronment and together with the developed graphical user mented. After choosing the controller the corresponding
interface (GUI) they create an interactive tool that can run block diagram is opened in Simulink (Fig 9). For the case of
locally or through the Internet. The presented application Internet Experiment the GUI offers additional parameters
is an extension of the application described in Bistak and related with the Internet connection (Server settings) and
Zakova (2013); Huba et al. (2014); Bistak et al. (2015); the choice between the virtual and remote laboratory.
Bistak and Huba (2016). The Control of hydraulic system
GUI shown in the Fig. 8 has been designed in Matlab The application is based on a client-server architecture
using GUIDE. The results of experiments are displayed that is completely realized in the Matlab/Simulink en-
during the experiment using Simulink scopes and after the vironment. Through the command channel Matlab com-
experiment in the form of Matlab graphs. As an example mands are transfered using Matlab scripts on the client
of local simulation the responses of trajectory tracking and server side. The experiment output data are transfered
control without PI controllers are displayed in the Fig. 4. through the data channel that is realized using UDP Send
In this case the GUI has been set to local experiment. The (Fig 10) and UDP Receive (Fig 9) Simulink blocks coming
GUI enables to choose controller and systems parameters from the Instrument Control Toolbox library. See Bistak
(basic and advanced). Up to now the Feedback, Dynamical and Zakova (2013); Huba et al. (2014) for more details.

14044
Proceedings of the 20th IFAC World Congress
Toulouse, France, July 9-14, 2017 P. Bisták et al. / IFAC PapersOnLine 50-1 (2017) 13498–13503 13503

ACKNOWLEDGEMENTS

This work has been partially supported by the Slovak


Grant Agency VEGA, grant No. 1/0276/14 and grant
No.1/0937/14.

REFERENCES
Aguiar, A.P. and Hespanha, J.P. (2007). Trajectory-
tracking and path-following of underactuated au-
tonomous vehicles with parametric modeling uncer-
tainty. IEEE Transactions on Automatic Control, 52,
1362–1379.
Bistak, P. and Huba, M. (2016). Three-tank virtual
laboratory for input saturation control based on matlab.
IFAC-PapersOnLine, 49(6), 207 – 212.
Bistak, P., Huba, M., and Belai, I. (2015). Comparing
constrained controllers for nonlinear hydraulic plant
using matlab interface. IFAC-PapersOnLine, 48(29),
283–288.
Bistak, P. and Zakova, K. (2013). Rapid design of simple
remote laboratory using matlab. In Emerging eLearning
Technologies and Applications (ICETA), 2013 IEEE
Fig. 10. Server block diagram in Simulink 11th International Conference on, 41–45.
Chacón, J., Vargas, H., Farias, G., Sánchez, J., and
The remote user must first connect to the server and set Dormido, S. (2015). Ejs, jil server, and labview: An
parameters. Then the corresponding Simulink block dia- architecture for rapid development of remote labs. IEEE
gram is opened on the server (Fig. 10) and the experiment Transactions on Learning Technologies, 8(4), 393–401.
could start. Experiment data are received and visualised Conte, G., Moog, C., and Perdon, A. (2007). Algebraic
continuously using blocks of the corresponding client block Methods for Nonlinear Control Systems. Theory and
diagram (Fig 9). After stopping the experiment data are Applications. Communications and Control Engineer-
saved to the Matlab workspace on the client side to be ing. Springer-Verlag, London, 2nd edition.
ready for post-processing (e.g. for plotting using Plot time Duro, N., Dormido, R., Vargas, H., Dormido-Canto, S.,
responses block of the client block diagram as it has been Snchez, J., Farias, G., Dormido, S., and Esquembre, F.
done for the results of virtual and real laboratory Fig. 6, (2008). An integrated virtual and remote control lab
Fig. 7). the three tank system as a case study. Computing in
Science and Engineering, 10(4), 50–59.
8. CONCLUSIONS Halás, M. and Žilka, V. (2011). Noninteracting control
of coupled tanks: from theory to practice. In Eurocast.
This paper discussed the problem of applying the theoreti- Gran Canaria, Spain.
cal solution to the trajectory tracking problem in practice, Hou, Z. and Zhu, Y. (2013). Controller-dynamic-
namely the implementation on the real coupled tanks sys- linearization-based model free adaptive control for
tem was studied. The necessity of an additional controller discrete-time nonlinear systems. IEEE Transactions on
was shown to cope with the model mismatch, other non- Industrial Informatics, 9(4), 2301–2309.
modeled dynamics, nonlinearities or input disturbances. Huba, T., Huba, M., Ťapák, P., and Bisták, P. (2014). New
Therefore the trajectory tracking control was combined Thermo-Optical Plants for Laboratory Experiments. In
with the PI control so the real system tracked desired IFAC World Congress. Cape Town, South Africa.
trajectories without a permanent control error. Kalúz, M., Garcı́a-Zubı́a, J., Fikar, M., and Čirka, L.
(2015). A flexible and configurable architecture for au-
For the purposes of the trajectory tracking control demon- tomatic control remote laboratories. IEEE Transactions
stration the Matlab/Simulink interactive application was on Learning Technologies, 8(3), 299–310.
developed. Together with the applications developed be- Králik, M. and Žáková, K. (2015). Interactive webgl model
fore (Bistak et al. (2015); Bistak and Huba (2016)) it of hydraulic plant. IFAC-PapersOnLine, 48(29), 146–
can be used to study various control engineering problems 151.
and compare different control designs. The application can Sáenz, J., Chacón, J., De La Torre, L., Visioli, A., and
run locally or remotely in the form of virtual or remote Dormido, S. (2015). Open and low-cost virtual and
laboratory. After small modifications in the client and remote labs on control engineering. IEEE Access, 3,
server block diagrams the application can also serve as a 805–814.
networked control system. The main advantage of the de- Žilka, V. and Halás, M. (2010). Disturbance decoupling
signed application consists in the fact that it is exclusively of coupled tanks: from theory to practice. In IFAC
developed in the Matlab/Simulink software and therefore Symposium on System, Structure and Control. Ancona,
users from the control area can easily modify it for their Italy.
own purposes. At this time the developed simulation tool is
suitable for advanced level of control engineering courses.

14045

Das könnte Ihnen auch gefallen