Available online at www.sciencedirect.

com

ScienceDirect
IFAC PapersOnLine 55-19 (2022) 31–36
Integrating
Integrating PHM into production scheduling through a Digital Twin-based
Integrating PHM
PHM intointo production
production scheduling
scheduling
framework
through
through a
a Digital
Digital Twin-based
Twin-based
Integrating PHM into production
Integrating PHM into production scheduling
scheduling
framework through
through a
a Digital
Digital Twin-based
Twin-based
Integrating PHM into production framework
scheduling through a Digital Twin-based
framework
framework
Elisa Negri*, Laura Cattaneo**, Vibhor Pandhare***, Marco Macchi*, Jay Lee***
framework
Elisa Negri*, Laura Cattaneo**, Vibhor Pandhare***, Marco Macchi*, Jay Lee***
Elisa Negri*, Laura Cattaneo**, Vibhor Pandhare***, Marco Macchi*, Jay Lee***
Elisa Negri*, Laura Cattaneo**, Vibhor Pandhare***, Marco Macchi*, Jay Lee***
*Department Elisa
Elisa of Negri*, Laura
Laura Cattaneo**,
Management,
Negri*, Politecnico diVibhor
Cattaneo**, Milano,
Vibhor Pandhare***,
Economics and
Pandhare***, Marco
Marco Macchi*,
Industrial
Macchi*, Jay
Jay Lee***
Engineering, Lee***Milan, Italy
*Department Elisa of Management,
Negri*, Laura Politecnico
Cattaneo**, di Milano,
Vibhor Economics
Pandhare***, andMarcoIndustrial
Macchi*, Engineering,
Jay Lee***Milan, Italy
*Department
*Department of
(Tel:
of Management,
+390223993969;
Management, Politecnico
e-mail:
Politecnico di Milano, Economics
elisa.negri@polimi.it; and Industrial Engineering,
marco.macchi@polimi.it) Milan, Italy
*Department(Tel:
*Department of
of +390223993969;
Management,
Management, e-mail:di
Politecnico
Politecnico di
di
Milano,
Milano,
Milano,
Economics
elisa.negri@polimi.it;
Economics
Economics
and
and
and
Industrial
Industrial
Industrial
Engineering,
marco.macchi@polimi.it)
Engineering,
Engineering,
Milan,
Milan,
Milan,
Italy
Italy
Italy
(Tel:
(Tel: +390223993969;
+390223993969; e-mail:
e-mail: elisa.negri@polimi.it;
elisa.negri@polimi.it; marco.macchi@polimi.it)
marco.macchi@polimi.it)
** School(Tel:
*Department of
of Industrial
Management,
+390223993969;Engineering,
Politecnico
e-mail: Università
di Milano, Carlo Cattaneo
Economics
elisa.negri@polimi.it; and - LIUC,
Industrial C.so Matteotti
Engineering,
marco.macchi@polimi.it) 22, 21053
Milan, Italy
** School(Tel: +390223993969;
of Industrial Engineering, e-mail: elisa.negri@polimi.it;
Università Carlo Cattaneomarco.macchi@polimi.it)
- LIUC, C.so Matteotti 22, 21053
**
** School
School of
(Tel:
of Industrial
+390223993969;
Industrial Engineering,
Engineering,Castellanza,
e-mail: Università Italy Carlo Cattaneo
(lcattaneo@liuc.it)
elisa.negri@polimi.it;
Università Carlo Cattaneo -- LIUC, C.so
marco.macchi@polimi.it)
LIUC, C.so Matteotti
Matteotti 22,
22, 21053
21053
**
** School
School of of Industrial
Industrial Engineering,
Engineering,Castellanza,
Castellanza, Università
Università Italy
Italy Carlo Cattaneo
Cattaneo -- LIUC,
(lcattaneo@liuc.it)
Carlo
(lcattaneo@liuc.it) LIUC, C.soC.so Matteotti
Matteotti 22, 22, 21053
21053
*****Mechanical and Materials
School of Industrial Engineering,Castellanza,
Engineering,
Castellanza, Università Italy
College
Italy (lcattaneo@liuc.it)
of
Carlo Engineering
Cattaneo
(lcattaneo@liuc.it) - and
LIUC, Applied
C.so Science,
Matteotti University
22, 21053 of
***
*** Mechanical and Materials Castellanza,
Engineering, Italy
College (lcattaneo@liuc.it)
of Engineering and Applied Science, University of
*** Mechanical
Cincinnati,and
Mechanical
Cincinnati, and Materials
Cincinnati,
MaterialsUnited
Cincinnati,
Engineering,
United
Castellanza,
Engineering,
States
College
States (e-mail: of
of Engineering
Engineering and
pandhavr@mail.uc.edu;
Italy (lcattaneo@liuc.it)
College
(e-mail: pandhavr@mail.uc.edu; and Applied
Applied Science,
Science, University
lj2@ucmail.uc.edu)
lj2@ucmail.uc.edu) University of of
*** Mechanical
Cincinnati,
*** Mechanical and Materials
Cincinnati,
and MaterialsUnited Engineering,
States
Engineering, College
(e-mail: of Engineering
Collegepandhavr@mail.uc.edu; and
of Engineering and Applied Applied Science,
Science, University of
lj2@ucmail.uc.edu) University of
Cincinnati,
*** Mechanical
Cincinnati, Cincinnati,
and MaterialsUnited
Cincinnati, United States
Engineering,
States (e-mail: pandhavr@mail.uc.edu;
Collegepandhavr@mail.uc.edu;
(e-mail: of Engineering and Applied lj2@ucmail.uc.edu)
Science, University of
lj2@ucmail.uc.edu)
Abstract: Cincinnati,
Production Cincinnati,
scheduling United States
has aStates (e-mail:
long (e-mail:
history of pandhavr@mail.uc.edu;
research but still presents lj2@ucmail.uc.edu)
open challenges, when
Abstract: Cincinnati,
Production Cincinnati,
scheduling United
has aa long history pandhavr@mail.uc.edu;
of research but still lj2@ucmail.uc.edu)
presents open challenges, when
Abstract:
considering
Abstract: Production
productionscheduling
Production systems with
scheduling has
has a long
uncertainty.
long history
history of
The
of research
digitization,
research but
but still
and
still inpresents
particular
presents open
open challenges,
Digital Twins,when
challenges, may
when
considering
Abstract:
considering production
Production systems
scheduling with uncertainty.
has uncertainty.
aafield.
long The history The digitization,
of digitization,
research but but and in particular
stillinpresents
presents openDigital Twins,
challenges, may
when
Abstract:
play
considering inproduction
a role Production
progressing
production systems
scheduling
the research
systems with
has
with long
uncertainty. history The
of
paper
The research
proposes
digitization, and
still
a framework
and in particular open
to exploit
particular Digital Twins,
challenges,
the Digital
Digital Twins, may
when
Twin
may
Abstract:
play
play a
considering
a
consideringrole
role Production
in
in progressing
production
progressing
production scheduling
the
systems
the
systems has
research
with
research
with a long
field.
uncertainty.
field.
uncertainty. history
The
The of
paper
The
paper
The research
proposes
digitization,
proposes
digitization, but
a
a still
framework
and
framework
and in
inpresents
to
particular
to
particular open
exploit challenges,
the
Digital
exploit the
Digital Digital
Twins,
Digital
Twins, when
Twin
may
Twin
may
synchronization
play a role in with the field
progressing the to includefield.
research healthThe assessment
paper models
proposes a into the simulation-based
framework to exploit the optimization
Digital Twin
considering
synchronization
play
play a
a role
role in
in production
with
progressing
progressingthe systems
field
the
the to with
include
research
research uncertainty.
health
field.
field. The
The The
paper
paper digitization,
assessment models
proposes
proposes a
a and
into
framework
framework in
the particular
simulation-based
to
to Digital
exploit
exploit the
the Twins,
optimization
Digital
Digital may
Twin
Twin
synchronization
of the production
synchronization with
with the
system
the field
field to
scheduling.
to include
includeThe health
health
health assessment
assessment
assessment models
is based
models into
intoon the
a
the simulation-based
modelling of
simulation-based failure optimization
modes
optimization and
play
of the a role
synchronization in
production
synchronization progressing
with the
system
with the the
field research
to
scheduling.
field to include
includeThe field.
healthThe
health
health paper proposes
assessment
assessment
assessment models
is a
based
models framework
into
intoon the
a
the to
modelling exploit
simulation-based
of
simulation-based the
failureDigital
optimization
modes
optimization Twin
and
of
of the
monitoring
the production
signals
production system
to
system scheduling.
connect the
scheduling. The
physical
The health
health assessment
production resources
assessment is
is based
to
based on
health
on a modelling
and operating
aa modelling of
of failure
states
failure modes
in
modesorder and
andto
synchronization
of
of the
monitoring
the production
signals
production withto the
system
system field
connect to
scheduling.
the
scheduling. includeThe
physical
The health
health assessment
assessment
production
health models
resources
assessment is
is based
to
based intoon
health
on the
a simulation-based
modelling
and operating
modelling of
of failure
states
failure optimization
modes
in
modesorder and
andto
monitoring
have a
monitoring more signals
accurate
signals to
to connect
prediction
connect the
the physical
of the
physical production
makespan
production withresources
respect
resources to
to
to health
the
healthactualand
and operating
makespan
operating states
of the
states in order
production
in order to
to
of
havethe
monitoring
a
monitoringproduction
more signals system
accurate
signals to
to scheduling.
connect
prediction
connect the
the of The
physical
the
physical health
makespan assessment
production
production withresourcesis
respect
resources based
to
to
to the on
health
health a modelling
actualand
and operating
makespan
operating of failure
states
of the
states modes
in order
production
in order andto
to
have
system.
have a
a more
The
more accurate
scheduling
accurate prediction
framework
prediction of
of the
is
the makespan
validated
makespan with
through
with arespect to
laboratory
respect to the
the actual
application
actual makespan
in
makespan the of the
Industry
of the production
4.0 Lab
production at
monitoring
have
system.
have a more
The
a more signals
accurate
scheduling
accurate to connect
prediction
framework
prediction the physical
of the
is
of is
thevalidated production
makespan
makespan with
through resources
arespect
laboratory
with aarespect to
to health
the actualand
application
to theapplication operating
makespan
actual makespanin the states
of the
Industry in order
production
4.0
of the production Lab to
at
system.
the School
system. The
The ofscheduling
Management
scheduling framework
of Politecnico
framework is validated
di Milano.
validated through
through laboratory
laboratory application in
in the
the Industry
Industry 4.0
4.0 Lab
Lab at
at
have
system.
the a
School
system. more
The of
Theof accurate
scheduling
Management
scheduling prediction
framework
of
framework of the
is
Politecnico
is makespan
validated
di
validated Milano. with
through
through a
arespect to
laboratory
laboratory the actual
application
application makespan
in the of the
Industry production
4.0 Lab at
the School Management of Politecnico di Milano. CC BY-NC-NDthe Industry 4.0 Lab at
under the application in the Industry
license 4.0 Lab
Copyright © 2022 The Authors. an
Thisisisvalidated open Fault article
access
Keywords:
the School
system. The Digital
of Twin,
Management
scheduling Equipment
of
framework Politecnico Health,
di Fault
Milano.
throughDetection
a laboratory and Diagnosis, Production Scheduling,
the School
Keywords:
the School of
of Management
Digital Twin,
Management of Politecnico
Politecnico
Equipment
of di Milano.
Health,
di Milano. Detection and Diagnosis,inProduction Scheduling, at
(https://creativecommons.org
Keywords:
Prognostics
Keywords:
the School Digital
and
Digital
of Twin,
Health
Twin,
Management Management/licenses/by-nc-nd/4.0/)
Equipment
Equipment
of Politecnico PHM.Health,
Health,
di Fault
Fault
Milano. Detection
Detection and
and Diagnosis,
Diagnosis, Production
Production Scheduling,
Scheduling,
Keywords:
Prognostics
Keywords:
Prognostics Digital
and
Digital
and Twin,
Health
Twin,
Health Equipment
Management
EquipmentPHM.
Management PHM.Health,
Health, Fault Fault Detection
Detection and and Diagnosis,
Diagnosis, Production
Production Scheduling,
Scheduling,
Prognostics
Keywords:
Prognostics and
Digital
and Health
Twin,
Health Management
Equipment
Management PHM.
PHM.Health, Fault Detection and Diagnosis, Production Scheduling,
Prognostics and Health Management PHM.
Prognostics and Health Management PHM.
1. INTRODUCTION AND MOTIVATION resources perturbations in real time, but at the same time it
resources
resources perturbations
perturbations in
in real time, but at the same time it
1.
1.
1.
INTRODUCTION
INTRODUCTION
INTRODUCTION
AND
AND
AND
MOTIVATION
MOTIVATION
MOTIVATION
bases
resources
bases
the optimization
the perturbations on
optimization onin aareal
real time,
time, but
prediction
prediction butof at
of
the
the same
atuncertainty
uncertaintysame and time
timeon
and
it
it
on
Production 1. scheduling
INTRODUCTION is one of
ANDthe most
MOTIVATION important activities resources
bases
resourcesthe
quasi-real perturbations
optimization
perturbations
time simulations onin
in areal
real time,
prediction
time,
ofprediction but
butof at
at the
uncertainty
the
scenarios,oftouncertainty same
same
make resultsand time
and
time
more it
on
it
Production 1. INTRODUCTION
scheduling is one AND
of the MOTIVATION
most important activities bases
resourcesthe
quasi-real optimization
perturbations
time simulations onin areal
of time,
scenarios, but toat the
make same
results time
moreon
it
Production scheduling is one bases the optimization on a prediction of uncertainty and on
for manufacturing
Production
for
1. INTRODUCTION
scheduling
manufacturing is one of
operations.
operations.
AND
of the
This
the most
MOTIVATION
most important
is reflected
important also activities
in the bases
activities robust. the
quasi-real
quasi-real Thisoptimization
time
time simulations
research on aims
idea
simulations aof
ofprediction
scenarios, ofto
at integrating
scenarios, touncertainty
make
make resultsand
results
Prognostics moreon
and
more
Production
for
importance
for
scheduling
manufacturing
Production
manufacturing
is
is one
that thisoperations.
scheduling topic
operations. ofThis
of
onecovers the
This
the in
This
is
most
is
most
is
reflected
important
reflected
important
industrial
reflected
also
also
also
in
activities
in
activities
engineering
in
the
the
the
bases
robust. the
quasi-real
quasi-real
robust.
Health
robust.
This
This
This
optimization
time
timeresearch
simulations
research
Management
research
on aims
idea
simulations
idea
(PHM)
aof
of
aims
idea aims
aims
prediction
scenarios,
of at
the
at
ofto
at integrating
scenarios, to
integrating
production
integrating
uncertainty
make
makePrognostics
results
results
Prognostics
resources
Prognostics
and moreon
and
more
and
into
and
Production
importance
for
importance scheduling
for manufacturingthat
manufacturing
that this
this is
topic
topic one
operations. of
covers
coversthe
This most
in
in is important
industrial
reflected
industrial activities
engineering
also in
engineering the quasi-real
Health
robust. time
Management
This simulations
research (PHM)
idea of scenarios,
of the
at to
production
integrating make results
resources
Prognostics more
into
and
research
importance
for
since
manufacturing thisoperations.
that long time. covers
topic Thisinisindustrial
Production reflected
scheduling alsoisin “the
engineering the robust.
Health
the
Health This research
Management
production
Management idea aims
(PHM)
scheduling
(PHM) of
of at
theintegrating
the production
optimization
production Prognostics
based resources
on Digital
resources and
into
into
research
importance
research
importance
allocation
since
since
that
of thisoperations.
long
thatshared
long
this time.
topic
time.
topic Thisin
Production
covers
Production
covers
resources inisindustrial
over
reflected
scheduling
industrial
scheduling
time
alsois isin “the
engineering
to engineering
competing
the Health
“the robust.
the This research
production
Management idea aims
scheduling
(PHM) at integrating
optimization
of the production Prognostics
based on Digital
resources and
into
research since long time. Production scheduling is “the Health
the
Twins.
the Management
production
This
production paper (PHM)
scheduling
proposes
scheduling of
a the production
optimization
progress
optimization withbased
based resources
on
respect into
Digital
to
on Digital
Digitalthe
importance
allocation
research
allocation
research thatshared
of
since
of
since this topic
shared
long
long covers
resources
time. Production
resources
time. in industrial
over
over
Production time
time to engineering
scheduling
to
scheduling competing
is
competing
is “the
“the Health
Twins.
the
Twins. Management
This
production
This paper
paper (PHM)
proposes
scheduling
proposes of
a the production
progress
optimization
a progress with
withbased resources
respect
on
respect to
to into
the
the
activities”
allocation (Hopp
of shared and resources
Spearman,over 1996)time withto the aim of the
competing Twins. production
preliminary This results
paper scheduling
of this
proposes optimization
research
a collaboration
progress withbased on Digital
(Negri
respect et
to al.,
the
research
activities”
allocation
activities”
allocation since
(Hopp
of
(Hopp
of long
shared
shared andtime.
and Production
Spearman,
resources
Spearman,
resources 1996)time
over
1996)
over scheduling
timewithtoto the is
aim“the
competing
competing of Twins.
the production
preliminary This paper
results scheduling
proposes
of this optimization
a
research progress with
collaboration based on
respect
(Negri Digital
to
et the
al.,
optimizing
activities” certain
(Hopp criteria,
and such
Spearman, as times
1996) and with
costs,
with the
by
the aim
deciding
aim of
of Twins.
preliminary
2019, Thisresults
2021).
preliminary paper
Some
results proposes
of
of this
background
this a progress
research
research with respect
collaboration
work is presented
collaboration (Negri
in
(Negri to
et
Section
et the
al.,
al.,
allocation
activities”
optimizing of shared
(Hopp
certain and resources
criteria,Spearman,
such as over
1996)
times andtimewith to the
costs, bycompeting
aim
deciding of Twins.
2019,
of preliminary Thisresults
2021). paper
Some proposes
background
of this a progress
research work is with respect
presented
collaboration in
(Negri to
Section
et the
activities”
optimizing
the sequence
optimizing (Hopp
certain
of orders
certain and to
criteria,
criteria,Spearman,
such
be
such as
as 1996)
times
produced,
times and
when
and with
costs,
costs, the
by
to produce
by aim
deciding
them
deciding 2019,
2, gaps
2019, 2021).
preliminaryand
2021). Some
results
objectives
Some background
of this
to be
backgroundresearch
pursued work is presented
collaboration
workin this
is paper
presented in
(Negri
are in
in et al.,
Section
Sectional.,
activities”
the
the sequence
optimizing
sequence (Hopp
of
certain
of and
orders to
criteria,Spearman,
be produced,
such as 1996)
times when
and with
to
costs, the
produce
by aim them
deciding of preliminary
2, gaps
2019, and
2021). results
objectives
Some of this
to be
backgroundresearch
pursued collaboration
workin this
is paper
presented (Negri
are in
in et
Sectional.,
optimizing
and
the with which
sequence
optimizing
and of orders
certain
certain
with which
which
resources
orders to
to be
criteria,
criteria,
resources be produced,
such as times
(Baker,
produced,
such as times
(Baker,
when
1974).
when
1974).
Itto
and costs,
Itto
and costs,
isproduce
by deciding
clear
isproduce
by deciding
clear
them
that
that
2,
3, gaps
themaa 2019,
2, and
2021).
in which
gaps objectives
itSome
will be
and objectives
objectives to be
background
clear
to pursued
be pursued
pursued thein
how work this
inpaper
this paper
is presented
paper are
are in
is a progress in
in Section
with
Section
the
and
the sequence
with
sequence
certain optimal of orders
resources
of scheduling
orders to
to be
be produced,
(Baker,
produced,
solution when
1974).
may when
changeItto
toisproduce
produce
as clear
a resultthem
them
that ofa 2019,
3,
2, 2021).
in which
gaps and itSome
will be background
clear
to be how work
thein is presented
paper
this is a progress
paper are in
in Section
with
and with which resources (Baker, 1974). It is clear that a 2,
3, gaps
in
respect
3, in and
which
to
which objectives
it
the
it will be
previous
will be to be
clear
clear pursued
how
works;
how the
in
the in this
paper
Section
paper paper
is
is a
4
a are inproposed
progress
the
progress Section
with
with
the
and sequence
certain
certain
and optimal
with which
optimal
with whichof orders
scheduling
resources
scheduling to be produced,
solution
(Baker,
solution
resources solution
(Baker, may
may when
1974).
1974). change
changeIttoisproduce
as
as a
clear
a result
resultthem
that of
ofa 2, gaps
respect
3, in and
to
which objectives
the
it previous
will be to be
clear pursued
works;
how in
the in this
Section
paper paper
is 4
a are
the in
progress Section
proposed
with
disruptions
certain optimal andscheduling
perturbations that may
occur inIt the
change is asclear
real
a that ofa 3,
world,
result in
respect which
framework
respect to
to it
the
and
the will be
previous
its clear
novelties
previous how
works;
works;are the
in paper
Section
presented;
in Section isina
4
4 progress
the
Section
the with
proposed
5
proposed the
and with
disruptions
certain which
optimal and resources
perturbations
scheduling (Baker,
that
solution 1974).
occur inIt is
the clear
real that
world, a
disruptions
certain
many optimal
of
disruptions and
which
and perturbations
scheduling
are generated
perturbations by may
that
solution
that occur
may change
the health
occur in
change
in the
the
as
asreal
status aa result
real world,
result
of the
world,
of 3,
of respect in which
framework
respect
framework
experimental
framework
to
to it will
the
and
the
and its be
previous
previous
setupits
and previous
clear
novelties
novelties
and the
its novelties
novelties
how
works;
works;are
are
first
the
are in
in
in paper isin44a Section
Section
presented;
Section
presented;
results are
presented; in
shown;
progress
the
the
Section 55with
proposed
proposed
conclusions
in4 Section
Section
the
the
5 the
the
certain
many
many optimal
of
disruptions
of
disruptions which
and
which
and scheduling
are
are generated
perturbations
generated
perturbations solutionby
that
by
that may
the
occur
the
occur change
health
in
health
in the
the asreal
status
status a result
real of
world,
of
world, of framework
the
the respect
experimentalto the
andsetupits and the works;are
first Section
presented;
results are in
shown; the proposed
conclusions5
production
many of resources.
which are For this reason,
generated by the scheduling
health approaches
status of the framework
experimental
are reported
experimental and
insetupits
Section
setup novelties
and
and the
6.
the are
first
first presented;
results
results are
are in
shown;
shown; Section
conclusions
conclusions 5 the
disruptions
production
many of and perturbations
resources.
which are For this
generated that
reason,
by occur
the in
scheduling
health the real
status world,
approaches
of the framework
ofin the experimental andsetupits and
novelties
the areresults
presented; in Section 5 the
production
many
started of
production resources.
which
to embed
resources. For
are generated
stochasticity
For this
this reason,
and by
have
reason, thescheduling
health
been
scheduling developingapproaches
status
approaches are reported
experimental
are
are reported
reported
in
in
in
Section
setup
Section
Sectionand the first
6.
6.
6. first results are
are shown;
shown; conclusions
conclusions
many
started of
production which
to embed are generated
stochasticity
resources. For
For this and by
have
reason, thescheduling
health
been status
developing ofin
approachesin the experimental
production
started
two to
main
started to resources.
embed
streams
embed stochasticity
of robust
stochasticity thisand
and
and have
reason,
reactive been
have scheduling developing
scheduling
been developingapproaches
(Aytugin the et are
the are reported
reported
2. insetup
in
2. BACKGROUND
BACKGROUND
Section
Sectionand 6.the ON
6. firstSIMULATION-BASED
ON
results are shown; conclusions
SIMULATION-BASED
production
two main
started
started to resources.
streams
embed
to embed of For
robust
stochasticity
stochasticity this
and reason,
and
and reactive
have
have scheduling
scheduling
been
been developing
developingapproaches
(Aytug
(Aytugin
inand et
the
the are 2. BACKGROUND
reported in
2. BACKGROUND Section 6. ON
SCHEDULING SIMULATION-BASED
two
al., main
2005; streams
Duenas of robust
and and
Petrovic, reactive
2008). scheduling
Digitalization et ON
two
al.,
two main
started
2005;
main streams
to embed
Duenas
streams of
of robust
stochasticity
and
robust and
and
Petrovic,
and reactive
have
2008).
reactive scheduling
been developing
Digitalization
scheduling
(Aytug
(Aytuginand et
the
et 2. BACKGROUND
2. BACKGROUND ON SIMULATION-BASED
SCHEDULING
SCHEDULING
ON SIMULATION-BASED
SIMULATION-BASED
two
al., main
2005;
Industry
al., 2005; streams
4.0 Duenas of
technologies
Duenas robust
and
and and
Petrovic,
open reactive
new
Petrovic, 2008).
ways
2008). scheduling
to Digitalization
take decisions
Digitalization (Aytug and
andet SCHEDULING
two
al., main
Industry
2005; streams
4.0 of
technologies
Duenas robust
and and
open reactive
new
Petrovic, ways
2008). scheduling
to take decisions
Digitalization (Aytug andet Most of the productionSCHEDULING
2. BACKGROUND scheduling
SCHEDULING problems have been long
ON SIMULATION-BASED
al., 2005;
Industry
elaborate
Industry 4.0
4.0 Duenas
technologies
solutions:
technologiesand Petrovic,
open
production
open new
new 2008).
ways
scheduling
ways to
to Digitalization
take
take decisions
may
decisions also and be Most
and Most of
of
recognized the
the production
production
as NP-hard scheduling
scheduling
(Neufeld,
SCHEDULING problems
problems
Gupta
have been long
have
and Buscher, been long
2016)
al., 2005;
Industry 4.0 Duenas and Petrovic, 2008). toDigitalization and Most of the production scheduling problems have been long
elaborate
Industry
elaborate
impacted.
elaborate Intechnologies
4.0solutions:
technologies
solutions:
fact, robust
solutions:
open
production
open
production new
production
production
ways
newscheduling
scheduling
to
waysscheduling
scheduling take
takemaydecisions
decisions
may
may
also
also
has
also beenbe
and
be
be recognized
Most of
recognized
Most
and of
the the
the
effort
as
as NP-hard
production
NP-hard
production
to compute
(Neufeld,
scheduling
(Neufeld,
scheduling
an
Gupta
problems
Gupta
optimal problems and
and
solution
Buscher,
have
Buscher,
havemay been
been 2016)
long
2016)
long
increase
Industry
impacted.
elaborate
impacted.
elaborate 4.0 In
Intechnologies
fact,
solutions:
fact,
solutions: robust
robust open
production new
production
production
production ways
scheduling
scheduling to take
scheduling
scheduling decisions
may
may has
also
has
alsobeenand
been
beenbe recognized
Most
and
be recognized of
the the
effortas NP-hard
production
as to compute
NP-hard (Neufeld,
scheduling
an
(Neufeld, Gupta
optimal problems
Gupta and
solution
and Buscher,
havemay
Buscher, been 2016)
long
increase
2016)
traditionally
impacted. In studied
fact, robustwithproduction
simulation-based scheduling optimization
has and the
recognized
even effort
more as
than to compute
NP-hard an
(Neufeld,
exponentially optimal
with Gupta
thesolution
and
size of may
Buscher,
the increase
2016)
problem.
elaborate
traditionally
impacted.
traditionally solutions:
In studied
fact,
studied production
robustwith
with productionscheduling
simulation-based
simulation-based scheduling may also
optimization
has
optimization beenbe and the
recognized
even effort
more as
than to compute
NP-hard an
(Neufeld,
exponentially optimal
with Gupta
thesolution
and
size may
Buscher,
of the increase
2016)
problem.
impacted.
approaches.
traditionally In fact,
With
studied robust
the Fourth production
Industrial
withproduction scheduling
Revolution,
simulation-based has been
simulation and
optimization and the effort toexponentially
computescheduling
an optimal
optimal solution may increase
impacted.
approaches.
traditionally
approaches. InWith
fact,the
studied
With therobust
Fourth
with Industrial
with
Fourth Industrial scheduling
Revolution,
simulation-based
Revolution, has been even
simulation
optimization
simulation
Research
even
and
more
the
more
the
onthan
effort
than
effort
to compute
production
toexponentially
an
computescheduling
with
with the
an optimal the size
size of
solution
proposed
solution
the
of may
to
theuse
may
problem.
increase
meta-
problem.
increase
traditionally
becomes
approaches. studied
field-synchronized
With the Fourth simulation-based
and connected
Industrial two-ways
Revolution, in real- Research
optimization
simulation even
Research
even more
more
heuristics
on
onthan
than production
exponentially
production
algorithms exponentially with
scheduling
thatschedulingwith proposed
the size
proposed
the
search andproposed size of
of to
the
to
find sufficientlythe use
use meta-
problem.
meta-
problem.
good
traditionally
becomes
approaches. studied
field-synchronized
With the with
Fourth simulation-based
and
and connected
Industrial
connected two-ways
Revolution,
two-ways optimization
in
simulation
in real-
real- Research
even more
heuristics onthan production
algorithms exponentially
that with
search and the size
find of to
the
sufficiently use meta-
problem.
good
approaches.
becomes
time
becomes with With the
field-synchronized
the
field-synchronized Fourth
physical Industrial
production
and connected Revolution,
resources.
two-ways simulation
This in new
real- Research
Research
heuristics on
on production
production
algorithms that scheduling
scheduling
search and proposed
proposed
find to
sufficiently use
use meta-
meta-
good
approaches.
becomes
time with With the
field-synchronized
the Fourth
physical Industrial
and connected
production Revolution,
two-ways
resources. simulation
This in real-
new solutions,
heuristics that
algorithmsdo not necessarily
that search correspond
and find to
sufficientlythe global
good
becomes
time
time with
simulation
withfield-synchronized
the
concept
the physical
physicalis and connected
production
known
productionas two-ways
resources.
Digital
resources. Twin This
This in real-
new
(Negri,
new Research
solutions,
heuristics
heuristics on
that production
algorithms
algorithmsdo not that
that scheduling
necessarily
search
search and
and proposed
correspond
find
find to
sufficiently
sufficiently use
the meta-
global
good
good
becomes
simulation
time withfield-synchronized
concept
the physicalis and
known connected
productionas Digital two-ways
resources. Twin This in real-
(Negri,
new solutions,
optima,
solutions, but that
that
that do
have
do not
the
not necessarily
advantage
necessarily correspond
to be found
correspond into
toa the
the global
reasonable
global
time with
simulation
Fumagalli
simulation andthe
conceptphysical
Macchi,
concept is
is production
known
2017;
known as
Kritzinger
as resources.
Digital
et
Digital al., Twin
2018).
Twin This
The(Negri,
new
real-
(Negri, heuristics
optima,
solutions, butalgorithms
that
that have
do not that
the search
advantage
necessarily and
to befind
found
correspond sufficiently
intoa reasonable
the good
global
time withand
Fumagalli
simulation the physical
Macchi,
concept is2017; production
knownKritzinger
as resources.
Digital 2018).
et al., 2018).
Twin This
The solutions,
optima,
new optima,
real-
(Negri, but
computational
but that
that
that do
have
time
have not
the
the necessarily
(Osman advantage
and
advantage correspond
to
to be
Laporte,
be found
1996).
found in
intoa the global
reasonable
Among
aa reasonablethese
simulation
Fumagalli
time andconcept
capabilities
Fumagalli and Macchi,
and
Macchi, is
the known
2017; as
Kritzinger
simulation Digital
et al.,
potentials Twin
of
2018). the The
The real-
(Negri,
Digital
real- solutions,
computational
optima, but that
that do
time
have not
the necessarily
(Osman and
advantage correspond
Laporte,
to be 1996).
found into the
Among global
reasonablethese
simulation
time
time capabilities
Fumagalli andconcept and is2017;
the knownKritzinger
simulation et
as potentials
Digital al., Twin
of
of the (Negri, optima,
real- computational
Digital but thattime
computational
meta-heuristics,
but thattime
have
one theofadvantage
(Osman
(Osman and
the
and to beused
Laporte,
most
Laporte, found
1996).
1996). inAmong
for a production
inAmong
reasonablethese
these
Twins
time capabilities
Fumagalli
enableand
capabilitiesaMacchi,
Macchi,
newand the
the
2017;
andscheduling
2017; Kritzinger
simulation
Kritzinger
approach
simulation
et al., 2018).
al.,
potentials
et that
potentials 2018).
brings
of theThe
the The real-
Digital
together
Digital optima,
meta-heuristics,
computational
meta-heuristics,
computational
scheduling is
have
time
time
the
one
one theof
(Osman
of
(Osman
Genetic
advantage
the
and
the
and
Algorithm
to beused
most
Laporte,
most
Laporte, found
used
(GA) 1996).
1996).
(Ruiz
for
for a production
reasonable
Among
production
Among
and these
these
Maroto,
Fumagalli
Twins
time
Twins enable
enableand
capabilitiesa
aMacchi,
new
newand 2017;
scheduling
the Kritzinger
approach
simulation
scheduling approach et al.,
that
potentials
that 2018).
brings
of
brings the The real-
together
Digital
together meta-heuristics,
computational
scheduling is time
the one of
(Osman
Genetic the
and
Algorithm most
Laporte, used
(GA) 1996).
(Ruizfor production
Among
and these
Maroto,
time capabilities
robustness
Twins enable newandscheduling
andareactivity thetosimulation
disruptions
approach potentials
andthat of thetogether
perturbations.
brings Digital
This meta-heuristics,
scheduling
meta-heuristics, is the one
Genetic
one of
of the
Algorithm
the most
most used
(GA)
used (Ruizfor
for production
and Maroto,
production
time capabilities
robustness
Twins enable and a and
reactivity
new the simulation
to
schedulingdisruptions
approach potentials
and of
perturbations.
that brings the Digital
This
together 2005;
schedulingZandieh,
is theMozaffari
Genetic and Gholami,
Algorithm (GA) 2010).
(Ruiz Typically,
and Maroto, the
robustness
Twins
new enable
approach
robustness and a reactivity
new
managesto to
andareactivity to
schedulingdisruptions
approach
react toand
disruptions and perturbations.
that
physical brings
perturbations. This
together
production meta-heuristics,
2005;
scheduling
This scheduling
2005; Zandieh, one
Mozaffari
is algorithms
Zandieh, theMozaffari of
Geneticare the
and
Algorithm
and most
Gholami,
Gholami, used
(GA) 2010). for production
Typically,
(RuizTypically,
2010). and Maroto,
Maroto, the
the
Twins
new enable
approach
robustness and new
managesscheduling
reactivity to to approach
react
disruptions to and that
physical brings
perturbations. together
production This optimization
2005; is
Zandieh, the Genetic
Mozaffari Algorithm
andused
Gholami, (GA)
together (Ruiz
with
2010). and
simulation,
Typically, in
the
new
new approach
robustness
approachand manages
reactivity
manages to to react
disruptions
to react to
to and physical
perturbations.
physical production
production This scheduling
optimization
2005; is
Zandieh,
optimization the Genetic
algorithms
Mozaffari
algorithms Algorithm
are
and
are used
Gholami,
used (GA)
together
together (Ruiz
with
2010).
with and
Typically,Maroto,
simulation,
simulation, in
the
in
robustness
new approachand reactivity
manages to disruptions
to react to and perturbations.
physical production This 2005; Zandieh,
optimization Mozaffari
algorithms and
are Gholami,
used together 2010).
with Typically,
simulation, the
in
new approach manages to react to physical production 2005; Zandieh,
optimization Mozaffari
algorithms and
are Gholami,
used together 2010).
with Typically,
simulation, the
in
2405-8963 Copyright
new approach © 2022 The
manages to Authors.
react to Thisphysical access article under the CC BY-NC-ND license.
is an openproduction optimization algorithms are used together with simulation, in
Peer review under responsibility of International Federation of Automatic Control.
optimization algorithms are used together with simulation, in
10.1016/j.ifacol.2022.09.180
32 Elisa Negri et al. / IFAC PapersOnLine 55-19 (2022) 31–36

a so-called “simheuristics” approach to production scheduling GAP 1: The works proposed in literature deal with simulation-
(Juan et al., 2015; Gonzalez-Neira et al., 2017; Hatami et al., based scheduling optimization, with traditional simulation:
2018). In particular, the role of the metaheuristics is to search Digital Twins are not fully exploited as synchronized
for alternative solutions, that are then evaluated within a simulations to describe the real-time conditions of the
simulation model, that represents the dynamics and the production asset/system that is scheduled.
interactions in time of the production resources, evaluates the OBJECTIVE 1: The proposed framework integrates the
effects of modifications on the production resources and Digital Twin simulation providing real-time information for
schedules, before implementing them in the physical world, the degradation and failure statistical models for a fully
and embeds uncertainty prediction through stochastic dynamic rescheduling.
modelling. The outputs of the simulations are sent back to the
metaheuristic algorithm that is then able to compute the fitness GAP 2: The health model of production resources within
function based on them. This work uses a simheuristic simulations for scheduling is typically framing a healthy state
approach to production scheduling, based on GA and Digital (i.e. resource able to operate) and unhealthy state (i.e. resource
Twin and integrating the PHM of the production resources. not able to operate, in a breakdown or under maintenance)
Robustness in production scheduling is defined as the (Von Hoyningen-Huene and Kiesmüller, 2015; Wu, Sun and
capability of finding solutions that remain (close to the)
Xiao, 2018).
optimum also in presence of small perturbances or disruptions
OBJECTIVE 2: This work provides a more complex and
of the production resources (Vieira et al., 2017). Robustness is
achieved by the simheuristics approach. The reactive approach realistic health model, that additionally includes a degraded
to uncertainty instead is achieved through the real-time state, meaning that the resource is able to operate at a slower
monitoring and elaboration of field data, obtained through the pace or lower efficiency.
synchronization of the Digital Twin, and through the
possibility to react in very short time to perturbances and
disruption proposing a new scheduling solution to be
implemented in the production system (Zhang et al., 2021;
Zhang, Tao and Nee, 2021). A deep review of the previous
works on metaheuristics-based scheduling optimization and
robust production scheduling may be found in (Negri et al.,
2021). In particular, it highlighted the role of simulation to
model uncertainty to develop robust approaches to scheduling,
but it also pointed out how the paradigm encountered in
simulation-based optimization for scheduling problems has
been rarely coupled with Digital Twins. Also, the research
works found in literature investigated how the uncertainty
impacted the search of the optimum solution by the
optimization algorithm and the research progress in robust
scheduling was not based on becoming more “reactive”.
Finally, although the research works deal with production
scheduling, the vast majority investigated and validated the
scheduling techniques in not much “realistic” systems. Most
of the validations were done neither in production systems as
real setting, nor in laboratories resembling real-life conditions.
Consequently, field-synchronization to collect shop floor data
was not proposed in literature works.

3. ADDRESSED GAPS AND PROPOSED

CONTRIBUTIONS
The current work is framed within a wider investigation on the
integration of PHM with production scheduling based on
Digital Twins. The present paper aims at addressing the first
progresses with respect to the mentioned paper (Negri et al.,
2021), contributing to answer the two gaps reported below. In
particular, the objective of the paper is to demonstrate a robust
production (re)scheduling framework, which leverages on the
Digital Twin simulation-based metaheuristics optimization
empowered with a PHM model describing the health
conditions of the production resources. The gaps found in
literature and the relative objectives that this work aims to
address are the following:
Figure 1. Proposed Digital Twin-based framework
4. PROPOSED FRAMEWORK
Figure 1 presents the proposed scheduling framework. It is
composed of various modules: one collecting input data, one
performing the optimization, one with the simulation and the
resource-synchronized PHM that together compose the Digital
Twin. The framework also includes a human-machine
interface and the physical production resources of the system.
The present article presents a progress with respect to the
 Elisa Negri et al. / IFAC PapersOnLine 55-19 (2022) 31–36 33
previous work, where it is possible to find a detailed
description of each module (Negri et al., 2021). In fact it
improves the scheduling optimization by leveraging on a more
precise real time estimation of production asset/system health,
thanks to the Digital Twin capabilities.
4.1 Prognostics and Health Management module
The modelling of the states of the production resources
includes three possible health states and three possible
operating states, as reported in Figure 2. The health states are:
(i) Healthy, the production resource is able to work at nominal
efficiency (e.g. the processing times are standard); (ii)
Degraded, the production resource is able to work at lower
efficiency; (iii) Failed, the production resource is not able to
work. The three operating states are: 1) In Operation, the
production resource is currently working; 2) Idle, the
production resource is able to work but not required to at the
moment; 3) Under Maintenance, the production resource is not
able to work and currently undergoing maintenance actions.
Figure 2. Production Resource State Map
Since the health and operating states affect the productivity of
the production system, also the production capacity and the
schedule are impacted. They are carefully modelled in the
Digital Twin simulation with the synchronization of field data,
in this way the evaluation of alternative schedules results more
robust. To this end, the Digital Twin module embeds several
mathematical models to describe the transition from one
production resource state to another one, which represents one
of the contributions of this work.
In particular, the Health Assessment Model (HAM)
characterizes the transitions from a health state to another one
(from healthy, to degraded, to failed, to healthy again, as
expressed in Figure 2). The HAM outputs the time the
production resources spend in each state, through the use of
three models:
a) the Healthy Production Model (HPM) identifies when the
resource is in healthy state and if it is in idle or operating
state;
b) the Degraded Production Model (DPM) identifies when
the resource is in degraded state and if it is in idle or
operating state;
c) the Maintenance Model (MM) identifies when the
resource is back to healthy state.
As in Figure 3, the HAM assesses the Conformance Value
(CV) of a production resource through monitoring signals of
the physical resource, which come from synchronized
controllers, accelerometers and other sensors of the system
following a series of analyses that includes data preprocessing,
model training and model testing. Based on the evaluated CV,
the Digital Twin will use the processing time of the resource
computed through the HPM or the DPM. In case the CV
indicates the resource is failed, the repair time is computed
through the MM. The HAM considers several possible
degradation modes at once, as a progress with respect to the
previous mentioned work in which only one failure mode was
considered. Since the authors assumed normal processing
times, the parameters of the normal distribution should be
adapted to the health and operating state in which the
production resources are currently in and must be estimated
based on field data. Field data are segmented through pre-
processing methods into two broad categories of features:
time-based features and condition-based features. In
particular, during healthy and degraded operating states, time-
based features are collected to build the HPM and DPM. The
condition-based features are collected during all resources
states and are used to build the HAM.
Figure 3. Components of the HAM
4.2 The role of Simulation Model and Genetic Algorithm
The Simulation model is a Discrete Event Simulation model
which receives the normally distributed processing times and
repair times, with the parameters computed starting from the
synchronized field data by the HAM. The uncertainty is then
treated by repeating simulation runs various times in a Monte
Carlo approach, in order to evaluate schedules and to find a
robust solution. When scheduling is triggered, the processing
times distributions are input in real-time into the simulation
model, achieving the field-synchronization of the simulation
model with the current operating and health conditions of the
production resources. The other input into the simulation
model is an alternative production schedule generated by the
GA as one individual of a population. The simulation model
simulates the sequence of jobs entering the system and
evaluates the makespan of the alternative schedule, repeating
several runs in order to get an average and standard deviation
of makespan to be compared with the ones of the other
alternative schedules of the population. The GA will then
evaluate whether the best individual of the current population
is the desired “sufficiently good” schedule or if it is necessary
to proceed with the generation of a new population of
alternatives. This approach has been explained in (Fumagalli
et al., 2019).
5. APPLICATION CASE
5.1 Laboratory and problem description
The framework has been applied to a laboratory assembly line
installed at the Industry 4.0 Lab at the School of Management
34 Elisa Negri et al. / IFAC PapersOnLine 55-19 (2022) 31–36

of Politecnico di Milano (Luca Fumagalli et al. 2016). The line are repeated with the same degradation mode for the drill
assembles a simplified smartphone, in seven workstations. The station for robust investigation. In all four experiments, the
validation of the framework is more realistic with respect to fault is induced in between a production order after three
the previous work, as it encompasses more than one jobs have already been produced from the order. At this
degradation modes at a time and also considers more than one time, the re-scheduling is automatically triggered as a
critical workstations. In particular, the most critical result of the real-time condition-monitoring by the Digital
workstations, whose degradation modes are considered in the Twin and its HAM. The remaining sequence of five jobs is
HAM, are the drilling station (where holes are drilled on the re-optimized in real-time taking into consideration the new
front cover) and the back cover station (where the back cover conditions of the production system. A new predicted
of the phone is positioned over the other assembled makespan is compared with the actual makespan obtained
components). Three possible degradation modes in the drill after the failure.
station are considered in the experiments, as follows: (i)
Drilling tool wear; (ii) Lubrication wear on handling axis; and
(iii) Track wear on handling axis. On the back cover station
the degradation mode consisted of wear on the axis used in
locking of the cover mechanism. In order to have more
frequent conditions recognized as failures or degraded health
states, manipulations during the operations are artificially
induced by a vibration shaker on the drilling axis, by
manipulating the pressure valve of the handling axis and by
manipulating the valve on the cover mechanisms. Monitored
signals received in real-time from the field (i.e. from the PLC
of the two critical workstations through the OPC-UA
architecture) are electricity, power, air flow rate and air
pressure. The scheduling problem formulated for this
validation aims at minimizing the total makespan to produce
eight jobs in the assembly line, each of which is different from
the others in terms of variations in the tasks to be done at the
workstations of the line. The optimization module and the
parameters of the GA did not change from the previous work
(Negri et al., 2021).
5.2 Processing of data and training of models
Samples are collected in healthy and non-healthy states.
Preprocessing of data is done through a segmentation process
which extracted 66 features (frequency domain features for
signals from the accelerometer and time domain features for
the rest of signals). Features are normalized and Principal
Component Analysis (PCA) is implemented to observe
clusters in the healthy, degraded and faulty states and PCA T2
test is used to detect failures (Wise et al., 1999). As a result of
the data preprocessing and the training of the models, the
different distributions of the processing times in the drill
station and the back cover station related to the different
degraded modes are clearly identified, see Figure 4.
5.3 Description of the experiment
The validation experiments are ten, distributed in two
Figure 4. Processing times distributions
scenarios (see Table 1).
A) The first scenario corresponds to always-healthy 5.4 Experimental results
production resources. Six experiments are conducted to The results of the scenarios are represented in Table 1, in
compare the predicted with the actual makespan when the which it is clear how the production schedule follows the real-
workstations of the assembly line are always healthy (no time conditions of the production system and the predicted and
component degradation is present in these experiments). actual makespans are consistent. The predicted makespan is
B) The second scenario corresponds to the scenario where a the one predicted by the optimization algorithm of the schedule
production resource is operating in a degraded state. by predicting the processing times of the production resources
Considering the scope of this work, a fault is artificially starting from the field-data taken in real time from the HAM
induced (as explained in Section 5.1) only in the drill and its sub-models. The actual makespan is the makespan of
station for track wear on handling axis. Four experiments the actual physical assembly of the eight jobs, directly
 Elisa Negri et al. / IFAC PapersOnLine 55-19 (2022) 31–36 35

measured on the assembly line. It is worth noting that the
second scenario has lower makespans compared with the first
scenario, because the first scenario considers the whole
production schedule made of eight jobs to be assembled. The
second scenario considers the predicted and actual makespans
of the remaining five jobs, that are re-scheduled (see
description B of Section 5.3). The implemented optimal
sequences are reported in the last column of Table 1. It is
important to note that, even though a single degradation is
induced in the second scenario, the HAM is receiving data
from all the four signals from both the drill and back cover
stations and is able to identify the correct degradation mode
and correspondingly choose DPM. The accuracy of the HAM
is reported in Fig. 5. The confusion matrix of the HAM here
reported demonstrates the capability of detecting the deviation
in the monitoring signals, to diagnose the correct failure mode
and to estimate the time worsening due to degradation of the
production resource accordingly (Fig. 6), so estimating the
new total makespan in the degraded scenario in real-time. As
shown in the table, the predicted makespan lies always within
an error percentage comprised between -3,5% and +3,5% of
the actual makespan (where error percentage = [predicted
makespan – actual makespan]/actual makespan).
Table 1. Experimental results
demonstrating the capability of the scheduling framework to
detect monitoring signals deviations, diagnose the relative
degradation mode, and estimate a consistent makespan,
compared to the actual makespan of the assembly process,
both in the healthy scenarios and in the degraded ones. The
experiments have demonstrated that the framework works in a
(quasi) real-time fashion, allowing to reschedule within
seconds, providing a convincing near-optimum solution to the
scheduling problem. This work has both industrial and
research implications: representing a step forward to
understanding how the Digital Twin may be practically used
for industrial decision making. A limitation of this work is that
during the re-scheduling activity triggered by a failure or
degradation in the production, the resources keep working, so
the conditions and the monitoring signals may change,
especially for large production orders, making the new optimal
schedule obsolete.

Predicted Actual
# Makespan Makesp Error % Sequence
[sec] an [sec]
1st SCENARIO

1 208 215 -3,3% 8-2-7-1-5-3-4-6

2 206 207 -0,5% 1-8-7-2-5-3-4-6
3 209 213 -1,9% 5-2-6-8-1-7-3-4
4 207 210 -1,4% 2-8-7-1-5-3-4-6
5 204 208 -1,9% 2-5-7-8-1-3-6-4
6 210 212 -0,9% 8-3-2-7-5-1-4-6
Figure 5. Confusion matrix of the HAM. Class 0 represents the
2nd SCENARIO

7 177 183 -3,3% 1-5-3-4-5

healthy state, while classes from 1 to 3 represent the three
8 185 190 -2,6% 1-8-6-3-4 degradation modes of the drill station.
9 183 181 1,1% 1-7-5-4-6

10 186 181 2,8% 8-1-5-4-6

6. CONCLUSIONS
The paper proposes a progress with respect to a previous
published work, in the research investigation that explores the
new potentials coming from including a real-time field-
synchronized Digital Twin into the simulation-based
optimization for production scheduling for a convergence of
the benefits of reactive scheduling into a robust scheduling
approach. Doing this, the Digital Twin hosts a Health
Assessment Model that elaborates the health states and
corresponding Production Models that elaborate the
processing times of production resources by gathering real- Figure 6. DPM as estimated from the HAM results for the drill
time data from the field, based on failure mode and signal station, track wear on handling axis.
monitoring. For this reason, the prediction of the makespan Future works may go in the direction of solving this issue, by
results in being realistic and connected to the conditions of the leveraging even more on the power of synchronization of the
physical resources, that leads to three possible health states Digital Twins, by constantly monitoring and updating the
(healthy, degraded, and faulty) and to three possible operating HAM for the new schedule optimization, also during the
states (in operation, idle, under maintenance). The paper scheduling activity itself. Moreover, future work will include
reports the first experimental results of this framework, more experiments to demonstrate the behavior of this
scheduling framework under different scenarios, degraded and
36 Elisa Negri et al. / IFAC PapersOnLine 55-19 (2022) 31–36
failed conditions. Last but not least, it is worth remarking the
future role of the framework for decision-making support: the
framework will guide the implementation of a supervisory
control to dynamically adjust the asset performances and the
production schedules. It will require both to embed prognostic
capabilities on the PHM side, and a predictive workload on the
production side. In its entirety, the predictive approach should
effectively lead to optimize, as a joint practice, production, and
maintenance objectives in the presence of disturbances and
perturbations from the shop floor.
REFERENCES
Aytug, H., Lawley, M.A., McKay, K., Mohan, S., and Uszoy,
R. (2005). ‘Executing production schedules in the face of
uncertainties: A review and some future directions’,
European Journal of Operational Research, 161(1), pp.
86–110.
Baker, K. (1974) Introduction to sequencing and scheduling.
John Wiley & Sons.
Duenas, A. and Petrovic, D. (2008) ‘An approach to
predictive-reactive scheduling of parallel machines
subject to disruptions’, Annals of Operations Research,
159(1), pp. 65–82.
Fumagalli, L., Macchi, M., Pozzetti, A., Taisch, M., Tavola,
G. and Terzi, S. (2016) ‘New methodology for smart
manufacturing research and education: The lab approach’,
Proceedings of the Summer School Francesco Turco, 13-
15-Sept, pp. 42–47.
Fumagalli, L., Polenghi, A., Negri, E., Roda, I. (2019)
‘Framework for simulation software selection’, Journal of
Simulation, 13(4), pp. 286–303.
Gonzalez-Neira, E. M., Ferone, D., Hatami, S., and Juan,
A.A. (2017) ‘A biased-randomized simheuristic for the
distributed assembly permutation flowshop problem with
stochastic processing times’, Simulation Modelling
Practice and Theory, 79, pp. 23–36.
Hatami, S., Calvet, L., Fernández-Viagas, V., Framiñán,
J.M., and Juan, A.A. (2018) ‘A simheuristic algorithm to
set up starting times in the stochastic parallel flowshop
problem’, Simulation Modelling Practice and Theory,
86(April), pp. 55–71.
Hopp, W. J. and Spearman, M. L. (1996) Factory Physics:
foundations of manufacturing management. Homewood,
Illinois: Irwin.
Von Hoyningen-Huene, W. and Kiesmüller, G. P. (2015)
‘Evaluation of the expected makespan of a set of non-
resumable jobs on parallel machines with stochastic
failures’, European Journal of Operational Research,
240(2), pp. 439–446.
Juan, A. A., Faulin, J., Grasman, S.E., Rabe, M., Figueira, G.
(2015) ‘A review of simheuristics: Extending
metaheuristics to deal with stochastic combinatorial
optimization problems’, Operations Research
Perspectives, 2, pp. 62–72.
Kritzinger, W., Karner, M., Traar, G., Henjes, J., and Sihn,
W. (2018) ‘Digital Twin in manufacturing: A categorical
literature review and classification’, IFAC-PapersOnLine.
Elsevier B.V., 51(11), pp. 1016–1022.
Negri, E., Ardakani, H.D., Cattaneo, L. Singh, J., Macchi,
M., and Lee, J. (2019) ‘A Digital Twin-based scheduling
framework including Equipment Health Index and
Genetic Algorithms’, in 13th IFAC Workshop on
Intelligent Manufacturing Systems, pp. 43–48.
Negri, E., Pandhare, V., Cattaneo, L. Singh, J., Macchi, M.,
and Lee, J. (2021) ‘Field-synchronized Digital Twin
framework for production scheduling with uncertainty’,
Journal of Intelligent Manufacturing, 32, pp. 1207–1228.
Negri, E., Fumagalli, L. and Macchi, M. (2017) ‘A review of
the roles of Digital Twin in CPS-based production
systems’, Procedia Manufacturing, 11, pp. 939–948.
Neufeld, J. S., Gupta, J. N. D. and Buscher, U. (2016) ‘A
comprehensive review of flowshop group scheduling
literature’, Computers and Operations Research. Elsevier,
70, pp. 56–74.
Osman, I. H. and Laporte, G. (1996) ‘Metaheuristics: A
bibliography’, Annals of Operations Research, 63, pp.
513–623.
Ruiz, R. and Maroto, C. (2005) ‘A comprehensive review and
evaluation of permutation flowshop heuristics’, European
Journal of Operational Research. Elsevier, 165(2), pp.
479–494.
Vieira, G. E., Kück, M., Frazzon, E. and Freitag, M. (2017)
‘Evaluating the Robustness of Production Schedules
using Discrete-Event Simulation’, in IFAC-
PapersOnLine, pp. 7953–7958.
Wise, B. M., Gallagher, N.B., Watts Butler, S., White D.D.,
and Barna, G.G. (1999) ‘A comparison of principal
component analysis, multiway principal component
analysis, trilinear decomposition and parallel factor
analysis for fault detection in a semiconductor etch
process’, Journal of Chemometrics, 13(3–4), pp. 379–
396.
Wu, Z., Sun, S. and Xiao, S. (2018) ‘Risk measure of job
shop scheduling with random machine breakdowns’,
Computers and Operations Research, 99, pp. 1–12.
Zandieh, M., Mozaffari, E. and Gholami, M. (2010) ‘A
robust genetic algorithm for scheduling realistic hybrid
flexible flow line problems’, Journal of Intelligent
Manufacturing, 21(6), pp. 731–743.
Zhang, J., Deng, T., Jiang, H., Chen, H., Qin, S., and Ding,
G. (2021) ‘Bi-level dynamic scheduling architecture
based on service unit digital twin agents’, Journal of
Manufacturing Systems, 60(May), pp. 59–79.
Zhang, M., Tao, F. and Nee, A. Y. C. (2021) ‘Digital Twin
Enhanced Dynamic Job-Shop Scheduling’, Journal of
Manufacturing Systems, 58(PB), pp. 146–156.