Available online at www.sciencedirect.

com

ScienceDirect
IFAC PapersOnLine 53-2 (2020) 3657–3662
Design
Design of aa Robust Fractional Order Controller for Burning Zone Temperature
Design of
of a Robust
Robust Fractional
Fractional
Control in an
Order
Order Controller
Controller
Industrial Cement
for
for Burning
Burning
Rotary
Zone
Zone
Kiln
Temperature
Temperature
Design
Design of
of a
a Robust
Robust Fractional
Control in an
Fractional Order Controller
Industrial
Order Cement
Controller for
for Burning
Rotary Zone
Kiln
Burning Zone Temperature
Control
Design of a Robust in anOrder
Fractional Industrial Cement
Controller forRotary Zone Temperature
Kiln
Burning Temperature
Control
Control in
in an
an Industrial
Industrial Cement
Cement Rotary
Rotary Kiln
Kiln
Control in an*V.Industrial
*V. Feliu-Batlle*, Cement
Feliu-Batlle*, R. Rotary Kiln
R. Rivas-Perez**
Rivas-Perez** *V. Feliu-Batlle*, R. Rivas-Perez**
*V.
*V. Feliu-Batlle*, R. Rivas-Perez**
*Escuela
*Escuela Técnica
Técnica Superior
Superior de
de Ingenieros
Ingenieros *V. Feliu-Batlle*,
Industriales
Industriales and
Feliu-Batlle*,
and
R.
R. Rivas-Perez**
Instituto
Instituto de Investigaciones Energéticas
Rivas-Perez**
de yy Aplicaciones Industriales,
*Escuela
Universidad Técnica de Superior deMancha,
Castilla-La Ingenieros Industriales
Campus Universitarioand Institutos/n, de Investigaciones
Ciudad Investigaciones
Real, 13071,
Energéticas
Energéticas
Spain (e-mail: y Aplicaciones
Aplicaciones Industriales,
Industriales,
vicente.feliu@uclm.es).
Universidad
*Escuela
Universidad Técnica de
de Castilla-La
Superior
Castilla-La de Mancha,
Ingenieros
Mancha, Campus
Campus Universitario
Industriales
Universitarioand s/n,
Instituto
s/n, Ciudad
de
Ciudad Real,
Investigaciones
Real, 13071,
13071, Spain
Energéticas
Spain (e-mail:
(e-mail: y vicente.feliu@uclm.es).
Aplicaciones Industriales,
vicente.feliu@uclm.es).
*Escuela
**Department
*Escuela Técnica
Técnica ofSuperior
Automatica
Superior de Ingenieros
and
deMancha,
IngenierosComputer Industriales
Science,
Industriales and
and Instituto
Universidad
Instituto de
de Investigaciones
Tecnológica
Investigaciones de la Energéticas
Habana
Energéticas CUJAE, y Aplicaciones
Calle
y Aplicaciones 114 Industriales,
No 11901,
Industriales,
**Department
Universidad
**Department de of
of Automatica
Castilla-La
Automatica and
and Computer
Campus
Computer Science,
Universitario
Science, Universidad
Universidads/n, Tecnológica
Ciudad Real,
Tecnológica de
13071,
de la
la Habana
Spain
Habana CUJAE,
(e-mail:
CUJAE, Calle 114 No
No 11901,
vicente.feliu@uclm.es).
Calle 114 11901,
Universidad
Universidad de
de Castilla-La
Castilla-La Mancha,
Marianao,
Mancha, Campus
Habana,
Campus Universitario
19390,
UniversitarioCuba s/n, Ciudad Real, 13071,
(e-mail:rivas@automatica.cujae.edu.cu)
s/n, Ciudad Real, 13071, Spain
Spain (e-mail:
(e-mail: vicente.feliu@uclm.es).
vicente.feliu@uclm.es).
**Department of Automatica Marianao,
Marianao, and Habana,
Computer
Habana, Science,19390,
19390, Cuba Cuba (e-mail:rivas@automatica.cujae.edu.cu)
Universidad Tecnológica
(e-mail:rivas@automatica.cujae.edu.cu)de la Habana CUJAE, Calle 114 No 11901,
**Department of
**Department of Automatica Automatica and Computer
and Computer Science,
Science, Universidad
Universidad Tecnológica
Tecnológica de la Habana CUJAE, Calle 114 No 11901,
de la Habana CUJAE, Calle 114 No 11901,
Marianao,
Marianao, Habana, 19390, Cuba (e-mail:rivas@automatica.cujae.edu.cu)
Marianao, Habana, Habana, 19390, 19390, Cuba Cuba (e-mail:rivas@automatica.cujae.edu.cu)
(e-mail:rivas@automatica.cujae.edu.cu)
Abstract: The
Abstract:
Abstract: The control
The control of
control of the
of the temperature
the temperature of
temperature of the
of the burning
the burning zone
burning zone of
zone of an
of an industrial
an industrial cement
industrial cement rotary
cement rotary kiln
rotary kiln is
kiln is
is
addressed
addressed in
in this
this paper.
paper. An
An experimental
experimental identification
identification of
of the
the process
process was
was carried
carried out,
out, which
which yielded
yielded aa
Abstract:
addressed
Abstract:
second order The
in
The this control
paper.
control
transfer of
Anthe
of
function the temperature
experimental
temperature
with no zeros of
of
but the
identification
thewithburning
burning
a of the
very zone
zone
large of
process
of
time an
an industrial
was carried
industrial
delay. cement
out, which
cement
Moreover, it rotary
rotary
was kiln
yielded
kiln
detected is
a
is
Abstract:
second order order The control
transfer of
function the temperature
with no zeros
zeros ofbutthewithburning zone oftime an was
industrial cement itrotary kiln isa
addressed
second
addressed
that this timein
in this
this
delay paper.
transfer
paper.could An
function
An experimental
with
experimental
change no
between identification
but of with
identification
±8% its aanominal
very
of
very
of the
the
large
process
large timeThen
process
value.
delay.
delay.
was a
Moreover,
carried
Moreover,
carried
robust out,
out, which
it
which
controller
wasyielded
was had
detected
detected
yieldedto be a
addressed
that thisorder
timein this delay paper.could An experimental
change between identification
±8% of of the
its anominal
nominal process
value. was
Then carried out, which yielded bea
second
that
second
designed this time
orderfor transfer
delay
transfer
this function
could
function
process. change
A with
with
standard no
between
no zeros
zeros
PI ±8%but
but with
of
with
controller, its a avery
veryPI large
large time
value.
time
controller delay.
Then
delay. aa robust
embedded
robust
Moreover,
Moreover,
in
controller
a it
controller
it
Smithwas
was
had
had to be
detected
to
detected
Predictor
second
designed orderfor transfer
this process. function
process. A with
standard no zeros but of with
controller, anominal
avery large timeThen
controller delay.
embedded Moreover, it washad
incontroller detected
that
designed
that this
this time
for
time aadelay
this
delay could
could change
A
change between
standard
between
PI ±8%
PI controller,
±8% of its
its
PI controller
aaa Smith
nominal PI value.
value. embedded
Then aa robust
robust in aa Smith
Smith
controller
Predictor
to
Predictor
had to be
scheme,
that
scheme,
designed this and
time
andfor
fractional-order
delay could
fractional-order
this process. change
A
controller
between
controller
standard PI
embedded
±8%
embedded of
controller,
in
its
in nominal
a Smith
PI
Predictor
value.
Predictor
controller Thenhave
have
embedded a been
robust
been studied.
controller
studied.
in a Smith
A
A method
had
method to be
Predictor
to
be
to
scheme,
designed
design and
the for a fractional-order
this process.
fractional-order A controller
standard
controller is PIembedded
controller,
developed in
in a a
this Smith
PI Predictor
controller
paper that have
embedded
yields been
better studied.
in a
results SmithA
than method
Predictor
the otherto
designed
design
scheme, the
andfor a this process. controller
fractional-order
fractional-order A standard
controller is PI controller,
developed
embedded in
in a a Smith
this controller
PIpaper that
Predictor embedded
yields
have better
been in a Smith
results
studied. than
A Predictor
the
method otherto
design
scheme, the
and fractional-order
aa fractional-order controller
controller is developed
embedded in
in this
aa Smith paper that yields
Predictor have better
been results
studied. than
A the other
studied
scheme,the
studied
design
controllers.
andfractional-order
controllers. Simulated
fractional-order
Simulated results
controller
results
controller
are
are
is
presented.
embedded in
presented.
developed in this Smithpaper Predictor
that havebetter
yields been results
studied.than A method
method
the other
to
to
studied
design controllers.
the fractional-order Simulated results are
controller is presented. in this paper the
developed that CCyields
BY-NC-ND better results
license than the other
Copyright
design
studied the © 2020
controllers. The
fractional-order Authors.
Simulated This
controller
results is an
is
are open
developed
presented. access inarticle
this under
paper that yields better results than the other
Keywords:
studied Fractional
controllers. order controller,
order licenses/by-nc-nd/4.0)
Simulated results are robust
robust control,control, time-varying
time-varying time time delay,delay, cementcement rotary rotary kiln,
(http://creativecommons.org/
Keywords:
studied
Keywords:
burning zone
Fractional
controllers.
Fractional
temperature,Simulated
order controller,
results
controller,
control are presented.
oriented presented.
robust
model control, time-varying time delay, cement rotary kiln,
kiln,
burning
Keywords: zone
burning zoneFractional temperature,
Fractional
temperature, order control
control oriented
controller,
orientedrobust model
robust control, time-varying time
modelcontrol, time-varying time delay, cement rotary kiln, delay, cement rotary kiln,
Keywords:
Keywords: Fractional order
order controller,
controller, robust control, time-varying time delay, cement rotary kiln,
burning
burning zone temperature, control oriented model
burning zone zone temperature,
temperature, control control oriented
oriented model model
1. INTRODUCTION cement
cement industry
industry and
and develop
develop four
four different
different functions:
functions: 1)
1)
1. INTRODUCTION
1. INTRODUCTION cement
chemical industry
reactor, and
2) heat develop
generator four 3) different
heat functions:
exchanger, and 1)
4)
chemical
cement
chemical reactor,
industry
reactor, 2)
and
2) heat
heat generator
develop
generator four 3) heat
different
3) different exchanger,
heat 2014). functions:
exchanger, and
and 4)
1)
4)
The cement industry 1.
1. INTRODUCTION
is currently
INTRODUCTION a fundamental factor for cement
conveyor
cement industry
of
industry gasesand and
andheat develop
solids
develop four
(Alsop,
four different functions:
The
functions: main1)
1)
The cement
The cementand industry
industry 1. is currently
INTRODUCTION
isdevelopment a
currently a worldwide, fundamental
fundamental because factor
factor for for conveyor
chemical
conveyor of
reactor,
of gases
gases2) and
and solids
generator
solids (Alsop,
3)
(Alsop,heat 2014).
exchanger,
2014). The
The andmain
main4)
economic social it chemical
objective
chemical reactor,
of these
reactor, 2)
2) heat
kilns
heat generator
is to
generator produce3)
3) heat
clinker
heat exchanger,
of
exchanger, and
consistent
and 4)
4)
economic
The cement and social
industry is development
currently worldwide, because it objective
it conveyor of
of these
gases kilns
and is to
issolids produce
(Alsop, clinker
2014). of
of Theconsistent
main
economic
The cement
represents
The cementand
represents
and
the
the
social
industry is
cornerstone
industry
development
currently
isdevelopment
cornerstone currently of aaathe
of
fundamental
theworldwide,
fundamental
construction
fundamental
construction
factor
because
factor
factor
for
for
sector
for
sector
objective
conveyor
quality
conveyor
quality from
from
of
of
ofraw
these
raw
gases
gases
kilns
meal
meal
and
andwith
with
to produce
solids
minimum
solids
minimum
(Alsop, clinker
energy
(Alsop,
energy
2014). consistent
The
consumption.
2014). The
consumption.
main
main
economic
represents
economic the
and social
cornerstone
social development of the worldwide,
construction
worldwide, because
because sector it
it objective
quality
objective from of
of these
raw
these kilns
meal
kilns withis
is to
to produce
minimum
produce clinker
energy
clinker of
of consistent
consumption.
consistent
(Schneider,
economic
(Schneider, and2019).
2019). social The
The cement
development
cement is
is the most
worldwide,
the most used
used building
because
building it objective of these kilns is to produce clinker of consistent
represents quality from raw meal with minimum energy consumption.
(Schneider,
represents
material on the2019).
the cornerstone
The
cornerstone
aaa global
global scale, andof
cement of itis the
the construction
istheaa key
mostelement
constructionused building sector
sector
for the In
quality
In the
the dry
from
dry cement
raw
raw meal
cement process
process with lines,
minimum
lines, the
the rotary
energy
rotary kilns are
consumption.
kilns are divided
representson
material
(Schneider,
material on
the
2019). cornerstone
global Thescale,
scale, and
cement
and
ofitis
it
the
is
isthea
construction
key
most
key element
used
element for
building
for the quality
sector
the In
into from
thethree
dry cement
zones: meal
process
heating with zone,
minimum
lines, the
burningenergy
rotary kilns
zone, are divided
consumption.
and divided
cooling
(Schneider,
development
(Schneider, 2019).
and
2019). The
economic
The cement
cement growthis the
is istheaof most
any
most used
country.
used building
building The into
In thethree
dry zones:
cement heating
process zone,
lines, burning
the rotary zone,
kilns and
are cooling
divided
development
material
development on a and
global
and economic
scale,
economic and growth
it
growth of key
ofkey any country.
element
anyelement
country. for The
the
The into
In
zone. thethree
dry
In the zones:
cement
burning heating
process
zone, zone,
atlines,
a burning
the
nominal rotary zone,
operatingkilns and
are cooling
divided
temperature
material
cement
material on
production
on a
a global
global scale,
involves
scale, and
several
and it
it is
is a
complex
a key processes
element for
for the
that
the In
zone.
into the dry
In
three the cement
burning
zones: process
zone,
heating atlines,
a
zone, the
nominal
burningrotary
operatingkilns
zone, are divided
temperature
and cooling
cement
development
cement production
production and involves
economic
involves several
growth
several complex
of
complex any processes
country.
processes that
The
that into zone.
into
of In
three
1450ºC, the burning
zones:
a chemical zone,
heating at a
zone,
reaction nominal
burning
takes operating
zone,
place thattemperature
and cooling
makes it
development
take place
development under and
and economic
different
economic raised growth
growth of
temperatures.
of any
any country.
Consequently,
country. The
The of
zone. three
1450ºC,
In the zones:
a
burning heating
chemical zone, zone,
reaction
at a burning
nominaltakes zone,that
place
operating and cooling
makes
temperature it
take
cement place
takecement
place under
production
under different
involves
different raised temperatures.
several complex Consequently,
processes that of
zone.
possible 1450ºC,
In the
to a
obtain chemical
burning the zone, reaction
at
tricalciuma nominaltakes
silicate, place
operating
which that makes
temperature
is the main it
cement
the
cement
the cement
production
industry
production
industry oneraised
involves
is one
involves
is of thetemperatures.
ofseveral
several
the
complex
industries
complex
industries withConsequently,
processes
the highest
processes
with the highestthat
that zone.
possible
of In the
1450ºC, to burning
obtain
a the
chemical zone, at a nominal
tricalcium
reaction silicate,
takes operating
which
place thattemperature
is the
makes main it
take
the place
cement under
industry different
is oneraisedraised
of in temperatures.
thetemperatures.
industries withConsequently,
the highest possible
of
component1450ºC, to obtain
a
of the
chemical
clinker tricalcium
reaction
and which silicate,
takes
confers which
place thethatis the
makes
mechanicalmain it
take
thermal
take place
place under
energy
under different
consumption
different raised the world, represents
temperatures. Consequently,
represents
Consequently, about component of 1450ºC,
possible to a chemical
of
obtain clinker
the reaction
and
tricalcium which takes
confers
silicate, place the
which that makes
mechanical
is the main it
thermal
the cement
thermal energy
industry
energy consumption
is
consumptionone of in
the
in the
the world,
industries
world, with
representsthe about
highest
about component
possible
resistance to of
obtain clinker
the
characteristic. and
tricalcium
The which
burning confers
silicate,
zone the
which
temperaturemechanical
is the main
(BZT)
the
5% cement
of global industry is
anthropogenic one of COthe industries
emissions, with
and the
affectshighest the possible
resistance to obtain the
characteristic. tricalcium
The burning silicate,
zone which is
temperature the main
(BZT)
the
5% cement
of global industry is one ofCO
anthropogenic the2the
2 industries
emissions, with
and the affectshighest the component
resistance
component of clinker and
characteristic.
of which
The burning confers
zone the
temperaturemechanical
(BZT)
thermal
5%
environment,
thermal
energy
of global
thermal energy
energy the consumption
anthropogenic
consumption
depletion of
consumption of CO in
in 2the
natural
in the
world,
emissions,
world,
resources,
world,
represents
and
represents affects
as well
represents well about
about
as
about
the
the is aa key
component
is
resistance key of clinker
clinker and
technological
technological
characteristic.
variable
and
variable
The
which
which
burning
that
that
confers
must
confers
must
zone
the
be
be
mechanical
thecontrolled
mechanical
controlled
temperature (BZT)
to
to
environment,
5% of global
environment, the depletion
anthropogenic
the depletion of natural
CO resources,
2 emissions,
natural resources, andas
as affects
well as
as the
the is
ensure a
resistance key thetechnological
characteristic.
clinker variable
The
quality burning
withthat must
zone
the be
minimumcontrolled
temperature (BZT)
energy to
5%
healthof
5% of and global
and safety
global anthropogenic
of people
anthropogenic CO
(Atmaca
CO emissions,
and
2 emissions, Yumrutas, and
and affects
2014).
affects the ensure resistance
ensure
is a key the characteristic.
clinker
technological The
quality
variable burning
withthat zone
the
must temperature
minimum
be controlled (BZT)
energy to
health
environment, safety the of people
depletion (Atmaca
of 2
natural and Yumrutas,
resources, as 2014).
well as the is a key
consumption. the clinker
technological
Therefore, quality
variable
it is withthat
vital the
must minimum
be controlled energy to
health and safety
environment,
environment, the
the
of people of
depletion
depletion
(Atmaca
of natural
natural
and Yumrutas,
resources,
resources, as
as
2014).
well
well as
as the
the is a keythe
consumption.
ensure technological
Therefore,
clinker variable
it
quality is thatto
vital
with to the
maintain
must
maintain
minimum
the
be controlled
the nominal
nominal
energy to
health
The and
cement safety
is of
obtained people from (Atmaca
the clinkerand Yumrutas,
pulverizing, 2014).
which is consumption.
ensure
operating theBZT Therefore,
clinker
and avoid it
quality
its is vital
with
fluctuations to themaintain
(Chenminimum
et the
al., nominal
energy
2016).
health and safety of people (Atmaca and Yumrutas, 2014). ensure
operating the clinker
and quality with the minimum energy
The
health
Themaincement
and
cement is
is obtained
safety of people
obtained andfrom
from the clinker
it (Atmaca
is the clinkerand pulverizing,
Yumrutas,
pulverizing, which
2014).
which is
is consumption.
operating
consumption.
BZT and avoid
BZT Therefore, avoid its
Therefore, it
its
it
fluctuations
is vital to
fluctuations
is
(Chen
(Chen et
maintain al.,
al., 2016).
et the nominal
2016).
its
its
The main
cement
component
component
is obtained and it
from is attained
attained
the clinker
from
from a burned
a burned
pulverizing,
mixture
mixture
which is consumption.
operating
However, BZT
the Therefore,
and
controlavoid
of its
the is vital
it BZT vital
fluctuations
is
to
toamaintain
not maintain
(Chen
simple et
task
the
the
al.,
nominal
nominal
2016).
due to the
its clay
The
of main
cementcomponent
and is obtained
limestone and it is the
from
materials attained
clinker
(raw from aadding
burneditwhich
pulverizing,
meal), mixture
grinded is operating
However, BZT
the and
controlavoid
of its
the fluctuations
BZT is not a (Chen
simple et
task al., 2016).
due to the
The
of
its cement
clay
main and is
component obtained
limestone and from
materials
it is the clinker
(raw
attained pulverizing,
meal),
from a adding
burned it which
grinded
mixture is operating
However,
complex BZT
the
dynamic and
controlavoid
of
behavior its
the fluctuations
BZT
of this is not
processa (Chen
simple
that et
task
is al., 2016).
due to
characterized the
of
its clay
main
gypsum and limestone
component
(Alsop, 2014). andmaterials
Theit isclinker(rawproduction
attained meal),
from aaadding
burned
processit mixture
grinded
is the
the However,
complex dynamic behavior of this process that is characterized
its
of main
gypsum
clay component
(Alsop,
and limestone 2014). and Theit
materials is attained
clinker(raw from
production
meal), burned
process
adding it mixture
is
grinded complex
However,
by exhibiting the
dynamic
the control
control
non-linearof
behavior
of the
theandBZT
of
BZT this is
is not
process
not
distributed a
a simple
that
simple task
is
parameters, due to
characterized
task due as to the
the
well
gypsum
of clay
stage (Alsop,
ofand limestone
highest 2014).
energy The
materials clinker
consumption (raw production
meal),the adding
in the process
cement it is the However,
grinded
industry, by exhibiting the control
non-linearof the andBZT isprocess
not a simple
distributed task
parameters, dueastowellthe
of clay
stage
gypsum ofand limestone
highest
(Alsop, energy
2014). materials
consumption
The clinker(raw meal),
in
production adding
cement
processit grinded
industry,
is the complex
by
complex
as large dynamic
exhibiting
dynamic
time-varying behavior
non-linear
behavior time of
and
of this
distributed
this
delays process
due to that is
parameters,
that is
changing characterized
raw as
characterized well
meal
stage
gypsum of
representing highest
(Alsop, energy
2014).
approximately consumption
The clinker
80% of in
the the
production
total cement
process
thermal industry,
is
energy the complex
as large dynamic
time-varying behavior time of this
delays process
due to that is
changing characterized
raw meal
gypsum
representing (Alsop, 2014).
approximately The clinker production
80% processenergy is the by exhibiting non-linear and distributed to parameters, as well
stage of
representing
stage
consumed of highest
highest energy
approximately
(Atmaca energy 80% of
consumption
consumption
and Yumrutas,
Yumrutas,
the
of2014).in
the
in
total
the
total
the
thermal
cement
thermalindustry,
cement
Therefore, energy as
industry,
research
by
by
large
characteristics, time-varying
exhibiting
exhibiting
characteristics,
non-linear
and both
non-linear
and both
time and
and
delays
physical
physical
duechemical
distributed
and
distributed
and
changing
parameters,
parameters,
chemical
raw
reactions
reactions
as meal
well
that
as meal
well
that
stage
consumed of
representing highest
(Atmaca energy
and
approximately consumption
80% of in
2014).
the the
total cement
Therefore,
thermal industry,
research
energy as
as
happen large
characteristics,
large intime-varying
time-varying
the and
burning both time
time
zone delays
physical
delays
(Ravi etdue
and
due
al., to changing
chemical
to changing
2016). raw
reactions
raw that
meal
consumed
representing
aimed at (Atmaca
ensuring and
approximately
the Yumrutas,
quality 80% clinkerof 2014).
the Therefore,
total
production thermal
and research
energy
reducing as
happen large intime-varying
the burning time
zone delays
(Ravi etdue
al., to changing
2016). raw meal
representing
aimed
consumed at ensuringapproximately
(Atmaca the
and quality 80%
Yumrutas, clinkerof2014).
the total
production thermal
Therefore, and energy happen
reducing
research characteristics,
in
characteristics, the and
burning
and both
both zonephysical
(Ravi
physical and
et
andal.,chemical
2016).
chemical reactions
reactions that
that
aimed
consumed
energy atconsumption
ensuring
(Atmaca the
and quality
and Yumrutas,
CO clinker
2 emissions
production
2014). Therefore,
to the and reducing
research
environment characteristics, and both physical and chemical reactions that
consumed
energy
aimed (Atmaca
consumption
at ensuring and
the and Yumrutas,
quality CO 2014).
emissions
2clinker Therefore,
to
production the research
environment
and reducing happen
Currently,
happen in
in the
the
the burning
most
burning zone
commonly
zone (Ravi
(Ravi et
used
et al.,
al., 2016).
procedures
2016). to control of
energy
aimed
by consumption
at
applying ensuring
modern the and
quality CO2clinker
manufacturing emissions
and to themethods
production
control environment
and reducing is of Currently,
happen
Currently, in the
the most
burning
the1)most commonly
zone
commonly (Ravi used
et
used procedures
al., 2016).
procedures to control
to control of
of
aimed
by atconsumption
applying
energy ensuring
modern themanufacturing
quality
and CO clinker production
and
emissions control
to the and
methods reducing
environment is of the BZT are: manual control by means of human operators
by
greatapplying
energy modern manufacturing
consumption
scientific-technical and CO 2
2 emissions
interest and
and control
to the
represents methods
environment
a huge the
is of Currently,
the BZT
BZT are:
are:the 1)
1) manual
most
manual control
commonly
control by
used
by means of
procedures
means of human
human to operators
control
operators of
energy
great
by consumption
scientific-technical
applying modern and CO
manufacturing interest
2 emissionsand
and to the
represents
control environment
methods aa huge is of Currently,
with
Currently, expert the
the most
knowledge
most commonly
commonlyof the used
process,
used procedures
and
procedures 2) to control
conventional
to control of
of
great
by
challenge scientific-technical
applying for modern
the manufacturing
international interest
scientific andcommunity.
and represents
control methods huge
is of with
the
withBZT expert
BZT are:
expert knowledge
1) manual
knowledge of
control
of the
thewellprocess,
by means
process, and
of
and 2)
human conventional
2) these operators
conventional
by applying
challenge
great for modern
the manufacturing
international
scientific-technical scientific
interest and
and control
community.
represents methods a is
huge of the
PID
the BZT are:
control.
are: 1)
1) manual
However,
manual control
it is
control by
by means
known
means of
of human
that
human operators
control
operators
challenge
great for the international
scientific-technical scientific
interest and community.
represents aa huge PID control. However, it is well known that these control
greatclinker
challenge
The scientific-technical
for the
productioninternationalprocessinterest
scientific andcommunity.
(clinkerization) represents is performed huge with PID
with
procedures
with
expert
control.
expert knowledge
However,
expert knowledge
are
knowledge not of
it is
of the
the
sufficient
of thewell
process,
well known
process,
when
process,
and
and
the
and
2)
that
2) conventional
these
2)processes
control
conventional
conventional are
challenge
The clinker for the
productioninternationalprocess scientific community.
(clinkerization) is performed procedures
PID
procedures control. are
are not
However,
not sufficient
it is
sufficient when
known
when the
that
the processes
these
processes controlare
are
challenge
The
in theclinker
rotaryfor the
production
kilns, international
which process
are thescientific community.
(clinkerization)
most important is performed
units of the PID
PID control.
characterized
control. However,
by distributed
However, it
it is
is well
dynamic
well known
known that
parameters
that these
and
these control
a time-
control
in the rotary kilns, which are the most important units of the characterized
procedures
characterized by
are
by distributed
not
distributedsufficientdynamic
dynamic when parameters
the
parameters and
processes
and a
a time-
are
time-
The
in
The theclinker
rotary production
clinker kilns, whichprocess
production are the (clinkerization)
process most important units
(clinkerization) is
is performed
of the
performed procedures
varying
procedures time are
delay
are not
(Astrom
not sufficient
and
sufficient when
Hagglund,
when the
2009;
the processes
Feliu-Batlle
processes are
are
The
*This clinker
work hasproduction
been supported process in (clinkerization)
part by the University is performed
of Castilla- varying
characterized
varying time
time delay
by
delay (Astrom
distributed
(Astrom and
and Hagglund,
dynamic
Hagglund, 2009;
parameters
2009; Feliu-Batlle
and a time-
Feliu-Batlle
in
in the
the
*This rotary
rotary
work has kilns,
kilns,
been which
which
supported are
are inthe
the most
most
part by important
important
the University units
units
of of
of the
the
Castilla- characterized
and Rivas-Perez,
characterized by
by distributed
2019).
distributed dynamic
Therefore,
dynamic the parameters
quality
parameters of and
the
and a
a time-
obtained
time-
in
La the
*This rotaryunder
work
Mancha has kilns,Project
been which 2019-GRIN-26969
supported are the most important
in part by the University
and in units
partofbythethe
of Castilla- and
varying
and Rivas-Perez,
time
Rivas-Perez, delay 2019).
(Astrom
2019). Therefore,
and
Therefore, the
Hagglund,
the quality
quality of
2009; the obtained
Feliu-Batlle
of international
the obtained
La Mancha under Project 2019-GRIN-26969 and varying
clinker time
is delay
unstable, (Astrom
it does not and Hagglund,
andreach the 2009;
required Feliu-Batlle
*This
La
European
*This work
Mancha
work has
under
Social
has been
Fund
been supported
Project
(FEDER,
supported in part
2019-GRIN-26969
inEU).part by
by the
the and in
University
University in part
of
part
of
by
by the
Castilla-
Castilla- the varying
clinker
and
clinker
time
is
Rivas-Perez,
is
delay2019).
unstable,
unstable,
(Astrom
it
it does
does not
Therefore,
not
Hagglund,
reach
reach the
the
the quality 2009;
required
required of Feliu-Batlle
international
the obtained
international
European
*This
La work
Mancha Social
has
under Fund
been (FEDER,
supported
Project EU).
inEU).part by the University
2019-GRIN-26969 and in of Castilla-
part by the and
standards
and Rivas-Perez,
and
Rivas-Perez, the 2019).
energy
2019). Therefore,
consumption
Therefore, the
the quality
is high.
quality of
To
of the
the obtained
solve these
obtained
European
La Mancha Social
under Fund (FEDER,
Project 2019-GRIN-26969 and in part by the standards
clinker
standards is and
unstable,
and the
the energy
it
energydoes consumption
not reach
consumption the is
is high.
required
high. To
To solve these
international
solve these
La Mancha
European under
Social FundProject
(FEDER, 2019-GRIN-26969
EU). and in part by the clinker
clinker is unstable,
is unstable, it does not
it doesconsumption reach the
not reach theisrequired required international
international
European Social Fund (FEDER, EU). standards and the energy high. To solve these
European Social Fund (FEDER, EU).
2405-8963 Copyright © 2020 The Authors. This is an open access article under standards
standards the CCandand the
BY-NC-ND energy
the energy consumption
license .
consumption is
is high. To solve these
high. To solve these
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2020.12.2048
3658 V. Feliu-Batlle et al. / IFAC PapersOnLine 53-2 (2020) 3657–3662
drawbacks, different control strategies have been proposed in
the last two decades to control the BZT, which include: model
predictive control (Salcedo et al., 2018), fuzzy control
(Zhang, 2017), artificial neural network control (Ying, 2014),
and neuro-fuzzy control (Fallahpour et al., 2008). However,
these controllers require a very precise knowledge of the
dynamic behavior of the cement rotary kiln, which is usually
something challenging to be obtained due to the dynamic
complexity of the physical and chemical reactions involved.
Moreover, they do not consider the large time-varying time
delay of this process that makes significantly difficult the
effective control of the BZT. Then robust controllers are
required for the effective control of the BZT, in order to
achieve the established international standard of clinker
quality with the minimum energy consumption. This issue
has been insufficiently addressed in the scientific literature.
Fractional order controllers have been proposed for robust
control of processes with complex dynamical behaviors, e.g.
Monje et al., 2010; Sayyaf and Tavazoei, 2018. These
controllers have more tuning parameters than standard
controllers. It enables them to increase their robustness against
time-varying dynamic parameters of the processes. Moreover,
their structure is simpler than the one of other robust
controllers, and their implementation is relatively easy.
Based on these considerations, our objective is to design a
robust fractional order controller for effective control of the
BZT in an industrial cement rotary kiln. The main
contributions of this paper are the development of a new
mathematical model of a BZT with a time-varying time delay,
and a new methodology to design a fractional order controller
for robust control of the BZT in cement rotary kilns.
The organization of the paper is as follows. This section has
provided the introduction to BZT control in cement rotary
kilns. Section 2 introduces the cement rotary kiln under study
and, based on a system identification procedure, obtains a
linear model with time-varying time delay of a BZT. Section
3 develops the fractional-order controller and compares it
with other controllers. Section 4 gives some conclusions.
2. SYSTEM IDENTIFICATION OF BURNING ZONE
TEMPERATURE
In the current study, the René Arcay cement factory of the
Mixed Company Cementos Curazao S.A. at the municipality
of Mariel, Cuba, is considered. The cement factory operates
on a dry cement process line. The rotary kiln is a refractory
steel cylindrical tunnel with a length of 93.6 meters and a
diameter of 4.6 meters. The kiln has a slight slope of 4° and
slowly rotates on its axis at 2.0-3.0 rpm by an electrical motor
with a working power of 283 kW. At the lower end of the
kiln, a fuel burner is installed to ensure temperatures over the
1400°C in the burning zone. The average clinker production
capacity of the kiln is about 2200 t/day. The kiln has a four-
stage cyclone preheater to heat the raw meal before it enters
the kiln, by using the hot gases from the burning zone.
Raw meal is fed into the upper end of the kiln and the
rotation of the kiln causes it to move gradually towards the
outfeed side with a variable residence time between 40 and
60 minutes. Upon reaching the combustion zone, the raw
meal attains the nominal operating temperature of about
1450°C, causing the chemical reactions that transform it into
clinker. After leaving the combustion zone, the clinker is
quickly cooled to 1000°C in order to avoid the reversion of
the produced chemical reactions. This abrupt cooling is also
necessary to improve clinker quality and for heat recovery. A
scheme of the studied cement rotary kiln is shown in Fig. 1.
Fig. 1. Schematic of the cement rotary kiln under study.
This cement rotary kiln can operate manually or
automatically, has a medium level of automation, the quality
of the clinker obtained is unstable, and has high fuel
consumption. Therefore, its modernization is planned in order
to improve its performance.
The kiln is equipped with the following instruments: wireless
temperature transmitter (WTT) and receiver (WTR), fuel flow
sensor (FFS), fuel control valve, a PLC SIMATIC S7-1200, an
industrial personal computer (PC), etc. The WTT is installed
on the external surface of the rotary kiln and it obtains the
temperature data by means of a thermocouple installed inside
the kiln. It has a mechanical protection cover that attenuates
the heat irradiation of the kiln surface. The WTT sends real
time data to a WTR installed in the control room of the rotary
kiln. The industrial PC has a SCADA application installed
that performs the supervision of the kiln operation.
Despite the cement rotary kiln is well known as a complex
distributed parameter system (Zhou and Meng, 2014), it is
not necessary to know the temperature variations along the
whole cement rotary kiln in order to carry out the BZT control,
but only at a specific measurement point (Salcedo et al.,
2018). Thus, a linear model with lumped parameters and a
time delay can adequately characterize the dynamical behavior
of a BZT at a specific point (Chen et al., 2016). This model
can therefore be obtained by using system identification
techniques, which are experimental procedures based on real-
time input and output operation data derived from the process
under study. Because the kiln object of this study is a real
industrial cement rotary kiln subject to operation and safety
restrictions, the system identification experiments to obtain
the operation data cannot be developed based on the response
to a persistently exciting binary signal input (e.g. a PRBS).
Therefore, with this purpose a step like input signal was used.
This signal excites the dynamics of interest and can be easily
generated. By limiting its amplitude, it is possible to keep the
process within its nominal operation regime allowing the
 V. Feliu-Batlle et al. / IFAC PapersOnLine 53-2 (2020) 3657–3662 3659
attainment of simple control-oriented models with an
acceptable degree of accuracy (Ljung, 1999).
In our experimental identification of the process based on a
step like input, the fuel flowrate ( Q f (t ) ) was considered the
input variable u (t ) , and the burning zone temperature the
output variable y (t ) . The most important disturbances that
influence this process are the variation in the raw meal load,
and its water and material contents, that cause great
fluctuations in the burner flame and originate abrupt
variations in the temperature of the burning zone (Chen et al.,
2016). Fig. 1 shows the experimental arrangement made for
the collection of input/output data under the nominal
operating regime of the studied kiln. The experiment based
on the response to a step like input consisted of increasing the
fuel flowrate entering the kiln burner by 11.62% (from 8600
to 9600 l/h) and measuring the temperature variation in the
combustion zone of the kiln. The experimental response
obtained is shown in Fig. 2.
Fig. 2. Experimental response of BZT to a step like input.
From Fig. 2 it is observed that the BZT varies from 1150ºC to
its nominal operating value of 1450ºC in an overall time period
of 800 s. The shape of the response suggests that the nominal
dynamic behavior of the BZT can be represented by a second
order transfer function with a dominant time delay given by:
K −τ ⋅s
P( s ) = e= G ( s ) ⋅ e −τ ⋅s ,
(T1 ⋅ s + 1)(T2 ⋅ s + 1)
where K is the static gain, T1 and T2 are time constants,
is the time delay, and G ( s ) is the rational part of P ( s ) . The
estimated parameters of the nominal model P0 ( s ) are:
K 0 = 0.3ºC·h/l, T10 = 55.39 s, T20 = 43.29 s, and τ 0 = 500 s.
The cross-validation result of P0 ( s ) was quantified by a FIT
index (Ljung, 1999) of 90.47%. Since this index is higher
than 80%, which is the minimum index value considered
valid, the obtained model P0 ( s ) can be regarded as accurate
enough to be used in the controller design (Ljung, 1999). The
settling time of the step response of G0 ( s ) is 235 s and the
settling time of P0 ( s ) is ts 0,open = 235 + τ 0 = 735 s.
The experiments developed in the BZT of our cement rotary
kiln proved that the higher the water content in the raw meal
is, the greater the BZT time delay would be. It is due to the
fact that for a given raw meal load, if its water content rises,
more heating time is required to evaporate this water and,
therefore, to observe growths in the BZT. These experiments
evidenced that model (1) of the BZT has a time-varying time
delay included in the following range:
τ ∈ [τ − ,τ + ] =
[ 460, 540] . (2)
Consequently, for an effective control of our BZT, the design
of a robust controller is needed.
3. DESIGN OF A ROBUST FRACTIONAL-ORDER
CONTROLLER FOR OUR BZT
3.1 Partial cancellation of the dynamics
It is well known that G0 ( s ) can be approximated by:
K0 K0
G0 =
( s) ≈ = Gc ( s )
(T10 ⋅ s + 1)(T20 ⋅ s + 1) (T10 + T20 ) ⋅ s + 1
in a certain range ω ∈ [ 0, ωmax ] of the frequency response.
Let us consider this approximation valid in the frequency
range in which the magnitude of the Bode diagram of
G0 ( s ) / Gc ( s ) is less than ±1 db, and the phase diagram is
less than ±5 . Fig. 3 shows these diagrams and these

accuracy bands. It shows that the magnitude has the required
accuracy in range [0,0.0093] rad/s and the phase in range
[0,0.0081] rad/s. The intersection of these intervals is the
range of validity of Gc ( s ) , which is ωmax = 0.0081 rad/s.
(1)
τ
Fig. 3. Frequency response of G0 ( s ) / Gc ( s ) .
The controller proposed in this paper has the structure:
3660 V. Feliu-Batlle et al. / IFAC PapersOnLine 53-2 (2020) 3657–3662

Kb ⋅ s + Ka π
j⋅ ⋅α
0≤α ≤ 2, ( j ⋅ ωc 0 )
α
Cα ( s ) = , (4) account that =e 2
⋅ ωcα0 .
sα
3.3. Design of a PI controller in the standard control scheme
where α is the fractional order, which is hereafter denoted
Fig. 4 shows the standard control scheme in which a PI
the PI α controller. Note that the PD and PI controllers are
controller is used (hereafter denoted the PI control scheme).
particular cases of (4) when α = 0 and α = 1 respectively.
In order to facilitate the controller design, we impose to The optimization problem (6) is solved for the integral order
cancel the dynamics of Gc ( s ) on Cα ( s ) : α = 1 . The only tunable parameter in (5) is Kα , which is
found by a pure search procedure that takes into account
Kα ⋅ ( (T10 + T20 ) ⋅ s + 1) Kα ⋅ ( 98.68 ⋅ s + 1) constraint 1). It yields an optimum value ωc 0 = 0.00081 rad/s
Cα ( s ) = = , (5)
sα sα and a minimum cost of 2450 s. The optimum controller is
C1 (s) = 0.0027 ⋅ (98.68 ⋅ s+ 1) / s . Fig. 5 shows the step
3.2 Control specifications responses of that control system for the minimum, nominal
The controller of the BZT is designed according to the and maximum time delays. It is mentioned that in this figure,
simultaneous verification of the following specifications: and the following ones, all the variables are incremental
around the operating set point and have been normalized.
1) A nominal damping condition. The step response of the
nominal closed-loop system (with P0 ( s ) ) must do not
have overshoot.
2) A robust settling time condition. Let us denote ts 0 , ts −
and ts + the closed-loop settling times of P ( s ) with
respectively (2) τ = τ 0 , τ = τ − and τ = τ + . It is desired
to minimize the maximum value of these three settling Fig. 4. Standard control scheme.
times (we reasonably assume that this maximum will be
1.2
the highest settling time among all the processes included
in the range (2)).
1
3) A precision condition. Zero steady state error to a step
command, which implies that the controller must include 0.8
an integral term (of integer or fractional order).
response y(t)

Then the optimization problem given by Specification 2 has 0.6

to be solved subject to the constraint given by Specification

1. This is expressed as: 0.4 *
y (t)

( max ( t , ts − , ts + ) ) ,
y(t) with L = 460 s

min s0
(6) 0.2 y(t) with L = 500 s
0≤α ≤ 2 y(t) with L = 540 s
0≤ωc 0 ≤ωmax
5% band
0
subject to Specification 1, where ωc 0 is the gain crossover 0 1000 2000 3000 4000 5000

time (s)
frequency of the nominal open-loop transfer function
L=
0 ( s) P0 ( s ) ⋅ C(s) , which must be included in the validity
Fig. 5. Step responses of the closed-loop system using a PI.
interval of Gc ( s ) . This last constraint is required in order to
guarantee the validity of the approximation (3), that is used to 3.3. Design of a PI controller embedded in a SP scheme.
define the controller structure (5). Taking (3) into account, it
is verified that: Fig. 6 shows a Smith Predictor control scheme in which a PI
controller is used (hereafter denoted SP-PI).
K 0 ⋅ Kα −τ 0 ⋅s
L0 (s) ≈ ⋅e (7)
sα
and the gain Kα is tuned for each couple of values (α , ωc 0 )
using the expression:

ωcα0
Kα = (8)
K0
obtained from the condition L0 ( j ⋅ ωc 0 ) =
1 and taking into Fig. 6. Smith Predictor based control scheme.
 V. Feliu-Batlle et al. / IFAC PapersOnLine 53-2 (2020) 3657–3662 3661

The closed-loop transfer function of our process is: half their values. In the case of the nominal process P0 ( s ) ,
y(s) the nominal closed-loop transfer function (10) becomes
= M= ( s ) Mˆ ( s ) ⋅ e −τ ⋅s (9)
y* ( s)
K 0 ⋅ Kα
Mˆ 0(α ) ( s ) = (11)
C ( s ) ⋅ G0 ( s ) α
s + K 0 ⋅ Kα
Mˆ ( s ) = (10)
1 + C ( s ) ⋅ G0 ( s )(1 + e −τ ⋅s − e −τ 0 ⋅s )
This transfer function has a slow convergence to its final
In the case of the nominal process, τ = τ 0 and expression value if α < 1 . Then, in order to achieve the same nominal
behaviour as the one obtained using the SP − PI , the reference
(10) becomes Mˆ 0 ( s ) = C ( s ) ⋅ G0 ( s ) / (1 + C ( s ) ⋅ G0 ( s ) ) .
y * (t ) is passed through a prefilter which has the form:
The optimization problem (6) is solved using a fixed integral
order α = 1 , in which the controller is designed taking into Mˆ 0(1) ( s ) K1 sα + K 0 ⋅ Kα
account that now L0 (s) ≈ K 0 ⋅ Kα / sα . It yields an = F ( s) = (12)
Mˆ 0(α ) ( s ) Kα s + K 0 ⋅ K1
optimum value ωc 0 = 0.0081 rad/s, and a minimum cost of
1373 s. The optimum controller is where we recall that K1 is Kα for the case α = 1 . Then a
C1 (s) = 0.027 ⋅ (98.68 ⋅ s+ 1) / s . Fig. 7 shows the step search procedure is carried out in which the pair of values
responses of that control system for the minimum, nominal, (α , ωc 0 ) are tuned to fulfil objectives 2 and 3 (using again
and maximum time delays. This figure shows that the
maximum settling time (the worst case) has been significantly L0 (s) ≈ K 0 ⋅ Kα / sα ). The result of this search is
reduced compared to the settling time of the PI. Moreover, α = 0.85 , ωc 0 = 0.0027 rad/s, which yields a controller:
the response in the case of the nominal process is considered
adequate since it does not have overshoot and its settling time 0.0219 ⋅ (98.68 ⋅ s+ 1)
is 878 s, which is relatively close to the one of the open-loop C0.85 ( s ) = . (13)
response. However, an undesirable 15.7% overshoot appears s 0.85
in the case of the maximum time delay and an undesirable The closed-loop responses to step commands using prefilter
16.7% undershoot in the case of the minimum time delay. (12) and controller (13) are shown in Fig. 8. This figure
shows that the maximum settling time is 1370 s (slightly
1.2
lower than 1373 s), the overshoot obtained with the maximum
time delay is 7.5% and the undershoot with the minimum
1
time delay is 8.4%, which are both about the half of the
values obtained with the SP − PI controller. Moreover, these
0.8
responses are smoother than the ones shown in Fig. 7.
response y(t)

0.6
1.2

*
0.4 y (t)
1
y(t) with L = 460 s
0.2 y(t) with L = 500 s
0.8
y(t) with L = 540 s
response y(t)

5% band
0 0.6
0 1000 2000 3000 4000 5000
time (s) *
0.4 y (t)
y(t) with L = 460 s
Fig. 7. Step responses of the SP-PI. y(t) with L = 500 s
0.2
y(t) with L = 540 s
3.4. Design of a PI α controller embedded in a SP scheme 0
5% band

0 1000 2000 3000 4000 5000

In order to reduce these over and undershoots, the Smith time (s)
Predictor control scheme is combined with the fractional
order controller (4), (hereafter denoted SP − PI α ). The design Fig. 8. Step responses of the SP- PI α .
of this controller has therefore the following objectives: 1)
3.5. Performance analysis.
maintain the nominal response (when the process is P0 ( s ) )
obtained with the previous SP − PI , 2) the cost index The performances of the three designed controllers are
exhibited in Table I. It shows: 1) the output index
max ( ts 0 , ts − , ts + ) must be equal or smaller than the ∞

optimum cost of 1373 s attained with the SP − PI , and 3) the IAE= ∫ y (t ) − y (t ) ⋅ dt ,

*
2) the maximum amplitude
overshoot and undershoot obtained in the previous subsection 0

in the cases of the extreme time delays must be reduced to

3662 V. Feliu-Batlle et al. / IFAC PapersOnLine 53-2 (2020) 3657–3662

(MU) of the control signal, and 3) the settling time t s of

p REFERENCES

y (t ) when a step disturbance is applied to the input of the Alsop, P.A. (2014). The Cement Plant Operations Handbook,
process. This settling time is the time needed by the output to 6th ed. Tradeship Publication LTD, Surrey, UK.
reduce its error to the 5% of the steady state error that the Astrom, K., and Hagglund, T. (2009). Advanced PID Control.
open-loop system would have as consequence of the step Pearson Education S.A., Madrid, Spain.
disturbance (i.e, the error caused by the disturbance if a Atmaca, A., and Yumrutas, R. (2014). Analysis of the
closed-loop control had not been implemented). parameters affecting energy consumption of a rotary kiln
in cement industry. Applied Thermal Engineering, 66,
4. CONCLUSIONS 435–444.
Chen, H., Zhang, X., Hong, P., Hu, H., and Yin, X. (2016).
Three schemes have been designed for the temperature
Recognition of the temperature condition of a rotary kiln
control of one of the processes of an industrial cement rotary
using dynamic features of a series of blurry flame images.
kiln with a large time delay that experiences moderate changes
IEEE Transactions on Industrial Informatics, 12(1), 148–
of up to ±8% of its nominal value. The robustness of these
157.
three controllers has been assessed and the proposed SP − PI α Fallahpour, M., Fatehi, A., Araabi, B.N., and Azizi, M. (2008).
controller (combined with a prefilter) showed the best A Neuro-fuzzy controller for rotary cement kilns. IFAC
performance in terms of settling time and over-undershoot. It Proceedings Volumes, 41(2), 13259–13264.
is noted that prefilter (12) does not add any degree of freedom Feliu-Batlle, V., and Rivas-Perez, R. (2019). Smith predictor
to the design of the control system since it is uniquely based fractional-order integral controller for robust
defined once the parameters α and Kα of the closed-loop temperature control in a steel slab reheating furnace.
controller have been tuned. Table I shows the best IAE index Transactions of the Institute of Measurement and Control,
of the response of the SP − PI α in all the time delay range 1–14.
and that it is achieved with less control effort (less maximum Feliu-Batlle, V., Rivas-Perez, R., and Castillo-Garcia, F.J.
amplitude of the control signal) than the SP − PI . (2009). Fractional order controller robust to time delay
variations for water distribution in an irrigation main
canal pool. Computers and Electronics in Agriculture,
Table I. Performance indexes
69, 185–197.
Control system Process IAE MU Ljung, L. (1999). System Identification: Theory for de User.
tsp Prentice Hall, New Jersey, USA.
Monje, C.A., Chen, Y.Q., Vinagre, B.M., Xue, D., and Feliu,
τ− 748.2 3.3 3014
V. (2010). Fractional-order systems and controls:
PI τ0 718.3 3.3 2910 fundamentals and applications. Springer-Verlag, London,
UK.
τ+ 693.7 3.4 2834
Ravi, T., Sridhar, P., and Guruprasath, M. (2016). Control
τ− 151.9 3.5 1851 and optimization of a triple string rotary cement kiln using
model predictive control. IFAC-PapersOnLine, 49(1),
SP-PI τ0 115.6 3.3 1499 748–753.
τ+ 149.1 4.2 2031 Salcedo, J., Rivas, R., and Sotomayor, J.J. (2018). Design of
a generalized predictive controller for temperature control
τ− 140 3.3 3524 in a cement rotary kiln. IEEE Latin America Transactions,
16(4), 1015–1021.
SP- PI α τ0 115.4 3.3 3423
Schneider, M. (2019). The cement industry on the way to a
τ+ 136.5 3.7 3308 low-carbon future. Cement and Concrete Research, 124,
105792.
Sayyaf, N., Tavazoei, M.S. (2018). Robust fractional-order
compensation in the presence of uncertainty in a
However, the last column of Table I shows that the SP − PI α pole/zero of the plant. IEEE Transactions on Control
has the worst performance when rejecting step disturbances Systems Technology, 26(3), 797–812.
at the input. This feature is caused by having a controller with Ying, K. (2014). Research on the burning zone temperature
an integrator of order less than 1, which slows down the control of cement rotary kiln based on CMAC-PID
convergence of the response to its final value. There are at algorithm. Journal of Chemical and Pharmaceutical
least two ways of improving this: one is to design a controller Research, 6(3), 813–817.
(5) with a higher α at the cost of reducing the performance Zhang, H. (2017). Application of improved fuzzy-Smith
and robustness to a step command, the other is to substitute controller in the control system of cement rotary kiln.
the term 1 / sα by ( s + µ ) / s1+α in (5), choosing a µ value Advances in Engineering Research, 153, 108–112.
small enough to do not change the gain crossover frequencies Zhou, P., and Meng, Y. (2014). Intelligent dynamic modeling
for online estimation of burning zone temperature in
ωc of the system in the range of time delays ( µ is one tenth cement rotary kiln. In: Proceeding of the 11th World
of the minimum of these ωc , e.g. Feliu-Batlle et al., 2009). Congress on Intelligent Control and Automation, 6167–
6171, Shenyang, China.