Mariagrazia Dotoli,
Mariagrazia Dotoli,
Maria Pia Fanti,
Maria Pia Fanti,
Giorgio
Giorgio
Iacobellis
Iacobellis
, Agostino M.
, Agostino M.
Mangini
Mangini
DEE
DEE
-
-
Politecnico di Bari
Politecnico di Bari
Petri Net Models for Supply Chain
Petri Net Models for Supply Chain
Operational Management and
Operational Management and
Performance Evaluation
Performance Evaluation
Presentation Outline
Presentation
Presentation
Outline
Outline
•
• IntroductionIntroduction toto SupplySupply ChainsChains (SC)(SC) •
• AimAim ofof the the researchresearch •
• SC SC operationaloperational modelmodel byby First First OrderOrder HybridHybrid Petri Petri NetsNets (FOHPN)(FOHPN) •
• SC SC operationaloperational modelmodel byby ColouredColoured TimedTimed Petri Petri NetsNets (CTPN)(CTPN) •
Presentation Outline
Presentation
Presentation
Outline
Outline
•
• IntroductionIntroduction toto SupplySupply ChainsChains (SC)(SC) •
• AimAim ofof the the researchresearch •
• SC SC operationaloperational modelmodel byby First First OrderOrder HybridHybrid Petri Petri NetsNets (FOHPN)(FOHPN) •
• SC SC operationaloperational modelmodel byby ColouredColoured TimedTimed Petri Petri NetsNets (CTPN)(CTPN) •
Introduction to Supply Chains (SC)
Introduction to Supply Chains (SC)
Introduction to Supply Chains (SC)
Suppliers Manufacutrers Distributors Retailers Customers Coordination center
Information Flow
Material Flow
Recyclers
SC (Supply Chain):
SC (Supply Chain): collectioncollection of of independent
independent companiescompanies withwith complementary
complementary skyllsskylls integratedintegrated with
with transportationtransportation and and storagestorage systems
systems
Aims of SC:
converting raw material into
finished products
improving profits
SC definition: Operative level
SC definition: Operative level
SC definition: Operative level
•• EntitiesEntities ofof a SC:a SC: •
• 1 1 -- SuppliersSuppliers:: provideprovide rawraw material, componentsmaterial, components and semiand semi--finishedfinished products.products. •
• 2 2 –– ManufacturersManufacturers//assemblersassemblers: : transformtransform input material/componentsinput material/components intointo desired
desired productsproducts. . •
• 3 3 -- DistributorsDistributors: : are intermediate are intermediate nodesnodes ofof materialsmaterials flow.flow. •
• 4 4 –– RecyclersRecyclers//dede--manufacturersmanufacturers:: theythey feedfeed recoveredrecovered material. material. •
• 5 5 –– CustomersCustomers or or retailersretailers:: are sinkare sink nodesnodes ofof material flow.material flow. •
• 6 6 -- LogisticsLogistics and and transporterstransporters:: connectconnect the facilitiesthe facilities. . The SC
The SC dynamicsdynamics dependsdepends on the management methodologieson the management methodologies specifyingspecifying the the business
business modelmodel, the information and material flow., the information and material flow.
• Make-to-Stock (MTS): A push strategy so that customers are satisfied from stocks of inventory of finished goods.
• (R.Q) inventory control rule for MTS: fixed part quantities Q are ordered any time the stock level drops below the reorder level R.
Presentation Outline
Presentation
Presentation
Outline
Outline
•
• IntroductionIntroduction toto SupplySupply ChainsChains (SC)(SC) •
• AimAim ofof the researchthe research •
• SC SC operationaloperational modelmodel byby First First OrderOrder HybridHybrid Petri Petri NetsNets (FOHPN)(FOHPN) •
• SC SC operationaloperational modelmodel byby ColouredColoured TimedTimed Petri Petri NetsNets (CTPN)(CTPN) •
Aim of the research
Aim
Aim
of the
of the
research
research
Models
Models based on PNsbased on PNs proposed in the related literature:proposed in the related literature:
1)
1) ViswadhamViswadham 2000, Dotoli 2000, Dotoli etet al. 2005 -al. 2005 - GeneralizedGeneralized StochasticStochastic Petri Petri NetsNets, , resource
resource orientedoriented model : model : placesplaces are SC are SC facilitiesfacilities, , tokenstokens are are lotslots of of productsproducts.. Limits:
Limits: state space of the SC model is excessively largestate space of the SC model is excessively large, the , the differentdifferent typestypes of of products
products in the system are notin the system are not modeled.modeled. 2)
2) ChenChen etet al. 2005 -al. 2005 - Modeling and performance evaluation of supply chains using Modeling and performance evaluation of supply chains using batch deterministic and stochastic Petri nets
batch deterministic and stochastic Petri nets ..
Limits:
Limits: the paper does not face the basic aspect of the SC management and the paper does not face the basic aspect of the SC management and optimization problems at the operative level
optimization problems at the operative level. .
We
We propose two effective and efficient modular models for SC propose two effective and efficient modular models for SC management and control at the operative level:
management and control at the operative level:
•
• 1) First Order1) First Order HybridHybrid Petri Petri NetsNets (FOHPN). (FOHPN). •
• 2) Coloured2) Coloured TimedTimed Petri NetsPetri Nets (CTPN).(CTPN). •
• The first may be used for testing /selection of SC operational control strategies.The first may be used for testing /selection of SC operational control strategies. •
• The second may be employed for performance evaluation and verification of the The second may be employed for performance evaluation and verification of the SC operation, enlightening the SC profitability, productivity, a
Presentation Outline
Presentation
Presentation
Outline
Outline
•
• IntroductionIntroduction toto SupplySupply ChainsChains (SC)(SC) •
• AimAim ofof the the researchresearch •
• SC operationalSC operational modelmodel byby First OrderFirst Order HybridHybrid Petri Petri NetsNets (FOHPN)(FOHPN) •
• SC SC operationaloperational modelmodel byby ColouredColoured TimedTimed Petri Petri NetsNets (CTPN)(CTPN) •
A First
A First
Order
Order
Hybrid
Hybrid
Petri Net Model
Petri Net Model
for
for
Supply
Supply
Chain
Chain
Management
Management
IEEE Trans. Autom. Science and Engineering, 2009
Mariagrazia Dotoli, Maria Pia Fanti, Giorgio Iacobellis, Agostino Marcello ManginiDipartimento di Elettrotecnica ed Elettronica, Politecnico di Bari
First
First
-
-
order hybrid Petri nets.
order hybrid Petri nets.
An application to distributed manuf. systems
An application to distributed manuf. systems
Nonlinear Analysis: Hybrid Systems, 2008
Mariagrazia Dotoli, Maria Pia Fanti, Alessandro Giua, Carla SeatzuDipartimento di Elettrotecnica ed Elettronica, Politecnico di Bari
Model with FOHPN (I):
Model
Model
with
with
FOHPN (I):
FOHPN (I):
FOHPN are introduced by
• Balduzzi, Giua, Menga, (2000). Modelling and Control with First order Hybrid Petri Nets. IEEE Trans. Robotics and Automation, 4 (16), pp. 382 – 399.
Key features of FOHPNs:
• increasing in computational efficiency with respect to place/transition models;
• reducing the dimension of the state space due to fluid approximation;
• the possibility of using gradient information to speed up optimization;
• We use a hybrid model to describe the dynamics of the SC that is traced by the flow of products between facilities and transporters. • The continuous dynamics models the flow of material, the
manufacturing, the assembling of different products and the storage of products in the buffers.
• The discrete event dynamics models the occurrence of stochastic events:
the blocking of the raw material supply and transport operations due to strikes, shifts accidents, etc..
Model with FOHPN (I):
Model
•
• A FOHPN is a bipartite digraph described by the sevenA FOHPN is a bipartite digraph described by the seven--tupletuple
PN
PN
=
=
(
(
P,
P
,
T
T
,
,
Pre,
Pre
,
Post,
Post
,
Δ
Δ
,
,
F, RS
F, RS
).
).
Model with FOHPN (II):
Definition of PN
Model
Model
with
with
FOHPN (II):
FOHPN (II):
Definition
Definition
of PN
of PN
•
•
P
P
==P
P
cc∪∪P
P
dd isis the set of the set of placesplaces thatthat modeling modeling resourcesresources:: placesplaces
P
P
cc modelmodel flow of material (flow of material (buffersbuffers and and relatedrelated capacitycapacity)) places
places
P
P
dd modelmodel preferencespreferences, , constraintsconstraints and state and state (operative,(operative, failedfailed) of SC ) of SC entitiesentities
T
T
==T
T
II∪∪T
T
cc∪∪T
T
S S ∪∪T
T
D D transitionstransitions setset •• transitionstransitions
T
T
II modelsmodels preferencespreferences •• transitionstransitions
T
T
cc modelsmodels the the continouscontinous flow of materialflow of material (production,(production, assemblingassembling, , recyclingrecycling, , loadload//unloadunload operationsoperations)) •
• transitionstransitions
T
T
SS modelmodel the discrete flow of the discrete flow of materialsmaterials ((transporttransport, , distributiondistribution), ), stochasticsstochastics eventsevents ((blockingblocking, , requestsrequests) ) •
Model with FOHPN (III):
Definition of PN
Model
Model
with
with
FOHPN (III):
FOHPN (III):
Definition
Definition
of PN
of PN
•
• FunctionFunction ΔΔ specifies the timing associated to timed transitionsspecifies the timing associated to timed transitions (zero (zero for
for immediate immediate transitionstransitions, , exponentialexponential or or triangulartriangular distributiondistribution forfor stochastic
stochastic transitionstransitions, , constantconstant valuesvalues forfor deterministicdeterministic transitionstransitions).). •
• FunctionFunction F F specifiesspecifies VVmjmj and and VVMjMj the minimum and the minimum and maximummaximum firingfiring speeds
speeds associatedassociated toto jj--thth continuous transitioncontinuous transition .. •
• FunctionFunction RS RS associates a probability value called random switch to associates a probability value called random switch to conflicting discrete transitions
conflicting discrete transitions •
A
A macro eventmacro event occurs when: occurs when:
• a discrete transition fires;
• a continuous place becomes empty;
• a continuous place, whose marking is increasing (decreasing), reaches a flow level that enables (disables) one or more discrete transitions
Any admissible IFS (Instantaneous Firing
Speeds, IFS) vector v at m is a feasible solution of the following set of linear set S(PN,m): :
Model with FOHPN (IV):
The net dynamics: time driven and event driven
Model
Model
with
with
FOHPN (IV):
FOHPN (IV):
The net
The net
dynamics: time
dynamics
: time driven
driven
and event
and
event
driven
driven
( )
(
,
)
( )
j C i i j j t Tm
C p t v
∈τ =
∑
τ
The equation governing the dynamics of theThe equation governing the dynamics of the
discrete transitions is:
discrete transitions is:
The equation governing the dynamics of the
The equation governing the dynamics of the
continuous transitions: continuous transitions: ( ) ( ) ( ) ( ) ( ) ( ) k k k k k k − − τ = τ + τ τ = τ + τ c c cd d d dd m m σ m m σ C C
•
• The FOHPN is described by a discreteThe FOHPN is described by a discrete--time state equation:time state equation:
1
(τk+ ) = A(τk) (τ + τk) B( k) (τk)
x x u
Model with FOHPN (V):
The net dynamics: time driven and event driven
Model
Model
with
with
FOHPN (V):
FOHPN (V):
The net
The net
dynamics
dynamics
: time
: time
driven
driven
and
and
event
event
driven
driven
1 1 1 1 1 ( ) 0 0 ( ) ( ) ( ) 0 0 ( ) 0 ( ) ( ) ( ) 0 0 ( ) ( ) ( ) 0 k k cc k dc k k k k dd k k k k k k + + + + + ⎡ τ ⎤ ⎡ ⎤⎡ τ ⎤ ⎡ τ ⎤ τ − τ ⎡ ⎤ ⎢ τ ⎥ = ⎢ ⎥⎢ τ ⎥+⎢ τ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎣ τ ⎦ ⎢ τ ⎥ ⎢⎣ τ ⎥⎦⎢ τ ⎥ ⎢⎣ τ ⎥⎦ ⎣ ⎦ ⎣ ⎦ c c d d v σ ν m I m C C m I m C D v f σ state input •
• The behavior of the system is described in the The behavior of the system is described in the macroperiodmacroperiod [[ττkk,,ττk+1k+1) by ) by the following equations:
the following equations:
y
(
τ
k) =
Ex
(
τ
k)
•
• The net state is given by the net marking and by the vector The net state is given by the net marking and by the vector νν((ττkk) of timers ) of timers associated to discrete timed transitions at time
associated to discrete timed transitions at time ττkk.. •
• The net input includes the macro-The net input includes the macro-period duration and the firing vector period duration and the firing vector σσ((ττk+1k+1) ) associated to the firing of transition
associated to the firing of transition ttjj at at ττkk .. •
• The elements of matrix The elements of matrix DD((ττkk) and vector ) and vector ff((ττkk) are equal to 0/1 and depend on the ) are equal to 0/1 and depend on the macro
macro--event occurring at the sampling instant.event occurring at the sampling instant. •
p
p
BB: the buffer of finite capacity: the buffer of finite capacityp
p’
’
BB: the capacity : the capacitym
m
BB++m’
m
’
BB=C=CBBt
t
TTi
i : the : the trasporttrasport
t
t
DDi
i: the demand: the demand
Q
Qii, , QQ’’ii: fixed order quantity : fixed order quantity R
RBB: reorder level: reorder level p’B tTn tT1 tD1 tDm pC pB Q1 Qn t1 Q’1 Q’m Q’m Q’1 CB - RB CB 0 CB - RB -Q1 CB - RB -Qn
…
…
Input buffer
Input buffer
module
module
by
by
the (R,Q) policy
the (R,Q) policy
Model with FOHPN (VI):
Subsystems model of the SC
Model
Model
with
with
FOHPN (VI):
FOHPN (VI):
Subsystems
p
p
BB: : the buffer of finite capacitythe buffer of finite capacityp
p’
’
BB: : the capacity the capacitym
m
BB++m
m
’
’
BB=C=CBBt
t
11 : the arrival of material: the arrival of materialp
p
kk: the supplier in operative state: the supplier in operative statep
p’
’
kk:: the supplier blocked the supplier blockedt
t
kk:
:
the blocking of blocking of materilmateril supplysupplyt
t’
’
kk: the : the restoretionrestoretion of of materilmateril supplysupplyt
t
TT: : trasporttrasport operationoperation p’k pk t’k tk p B p’B t1 tT 0 CB Q Q Q QSupplier
Supplier
module
module
Model with FOHPN (VII):
Subsystems model of the SC
Model
Model
with
with
FOHPN (VII):
FOHPN (VII):
Subsystems
p
Bi: the input buffer of finite capacitythe input buffer of finite capacityp’
Bi: avaibleavaible capacitycapacitym
m
BiBi++m
m
’
’
BiBi==CCBiBip
B1: the output buffer of finite capacitythe output buffer of finite capacityp’
B1: avaibleavaible capacity capacitym
m
B1B1++m
m
’
’
B1B1=C=CB1B1Manufacturer
Manufacturer
/assembler
/assembler
module
module
Model with FOHPN (VIII):
Subsystems model of the SC
Model
Model
with
with
FOHPN (VIII):
FOHPN (VIII):
Subsystems
Subsystems
model of the SC
model of the SC
CB2 – RB2 CB3 – RB3 CBn – RBn Q2 Q3 Qn CB2 – RB2 –Q2 CB3 – RB3 –Q3 CBn – RBn –Qn tj p’B2 0 pB2 CB2 p’B1 CB1 pB1 0 Q1 p’B3 0 pB3 CB3 … CBn 0 pBn Q2 Q2 Q1 Q3 Q3 Qn Qn p’Bn
p
pBiBi: downstream facility’: downstream facility’s input s input buffer
buffer
p
p’’BiBi: downstream facility: downstream facility’’s s avaibleavaible capacity
capacity mmBiBi++mm’’BiBi=C=CBiBi
p
pkk: transporter in the operative : transporter in the operative state
state
p
p’’kk: transporter in the blocking : transporter in the blocking state
state
t
tkk: transporter bloking: transporter bloking
t
t’’kk: transporter becomes : transporter becomes operative
operative
p
pcici: control places that : control places that determines the
determines the choisechoise of of material to transport
material to transport
p
p11: it indicates if the transporter : it indicates if the transporter is
is avaibleavaible or not avaibleor not avaible
t
tTT i
i : transporters of i items, : transporters of i items,
triangular distribution triangular distribution
Transport
Transport
module
module
Model with FOHPN (IX):
Subsystems model of the SC
Model
Model
with
with
FOHPN (IX):
FOHPN (IX):
Subsystems
Subsystems
model of the SC
model of the SC
p’k pk t’k tk tT1 pC1 . . . pCn tTn t1 tn p1 p’B1 pB1 p’Bn pBn Q Q CB1-R1 CB1-R1-Q 0 0 CB1 CBn CBn-Rn CBn-Rn-Q Q Q Q Q . . . . . .
p
pBB: : buffer of finite capacitybuffer of finite capacity
p
p’’BB: : the the capacitycapacity mmBB++mm’’BB=C=CBB
t
tTT i
i : transporter of i items: transporter of i items
t
tDD i
i: request of i items: request of i items
P
Pc c : control place : control place
Distributor
Distributor
module
module
Model with FOHPN (X):
Subsystems model of the SC
Model
Model
with
with
FOHPN (X):
FOHPN (X):
Subsystems
Subsystems
model of the SC
model of the SC
p’B tTn tT1 tD1 tD m pC pB Q1 Qn t1 Q’1 Q’m Q’m Q’1 CB - RB CB 0 CB - RB -Q1 CB - RB -Qn
…
…
•
•
p
p
BB: : the buffer of finite the buffer of finite capacitycapacity ••
p
p
’
’
BB: the : the capacitycapacitym
m
BB++m’
m
’
BB=C=CBB. . ••
t
t
11 : product request, exponential : product request, exponential distributiondistribution •
•
p
p
FF: retailer: retailer ••
t
t
LL:
:
detereoretiondetereoretion of finished products of finished products ••
ps
ps
: products to be de: products to be de--manufacturedmanufactured•
•
p
p
DD: : goods to be goods to be descerdeddescerded•
•
t
t
TT:
:
transport operationtransport operationpB p’B t1 pF Q CB – RB - Q CB – RB CB 0 Q1 Q1 Q1 pS Q2 tL µQ2 tT Q3 pD (1-µ)Q2
Retailer
Retailer
module
module
Model with FOHPN (XI):
Subsystems model of the SC
Model
Model
with
with
FOHPN (XI):
FOHPN (XI):
Subsystems
De
De
-
-
manufacturer
manufacturer
module
module
Model with FOHPN (XII):
Subsystems model of the SC
Model
Model
with
with
FOHPN (XII):
FOHPN (XII):
Subsystems
Subsystems
model of the SC
model of the SC
pB2 p’B2 t1 pB1 CB2 0 pBn p’Bn 0 CBn
…
Q2 p’B1 CB1 0 Q2 Qn Qn pB p’B3 0 CB3 Q3 Q3 Q’2 Q’2 Q’3 Q’3 Q’n Q’n CB1 – RB1 Q1 CB1 – RB1 –Q1 ppBiBi: the input buffer of finite : the input buffer of finite capacitycapacity
p
p’’BiBi: the input buffer : the input buffer capacitycapacity mmB1B1++mm’’B1B1=C=CB1B1
p
pBiBi: the output buffer of finite : the output buffer of finite capacitycapacity
p
p’’BBii: the output buffer : the output buffer capacitycapacity mmBiBi++mm’’BiBi=C=CBiBi
t
Model with FOHPN (XIII):
SC control
Model
Model
with
with
FOHPN (XIII):
FOHPN (XIII):
SC control
SC control
1(τk+ ) = A(τk) (τ + τk) B( k) (τk)
x x u •• AA((ττkk) and ) and BB((
τ
τ
kk) describe the SC in [) describe the SC in [ττkk,,ττk+1k+1) ) and depend on the event occurring in t= and depend on the event occurring in t=ττkk•
• The vector IFS The vector IFS vv((ττkk))∈∈
S
S
(PN,(PN,mm((τ
τ
kk)) affects the matrix )) affects the matrix BB((ττkk))Ö
Ö Choice of Choice of v(v(ττkk)) in each macroin each macro--period on the basis of the knowledge of period on the basis of the knowledge of
the SC state in order to optimize a particular objective functio
the SC state in order to optimize a particular objective function with n with
y
y((ττkk)=)=EmEmcc((ττ k
k) and with ) and with mmrra reference vector a reference vector
Ö
Ö Input vector Input vector uu((ττkk) determined by a decision maker) determined by a decision maker
1 1 ( ) ( ) k k k k + + τ − τ ⎡ ⎤ τ = ⎢ τ ⎥ ⎣ σ ⎦ u Control Action Decision Maker B(τk) zk E A(τk) x(τk) y(τk) x(τk+1) + + - +v(τk) u(τk) mr
Model with FOHPN (XIV):
SC control
Model
Model
with
with
FOHPN (XIV):
FOHPN (XIV):
SC control
SC control
•• Controller 1 (C1): flow maximization.Controller 1 (C1): flow maximization.
m
mrr=0=0, y, y=m=mcc, max, max
v
vJJ11=max=maxvv((11TT··vv) s.t) s.t. . vv∈∈S(PN,S(PN,m)m)
•
• Controller 2 (C2): stored materials minimization.Controller 2 (C2): stored materials minimization.
m
mrr=0=0, y, y, min, minvvJJ22 s.t. s.t. vv∈∈S(PN,S(PN,m) withm) with 2 ( ) ( , )v ( ))
i j c c i k i j j k p Y t T J m p t ∈ ∈ ⎡ ⎤ = ⎢ τ + τ ⎥ ⎢ ⎥ ⎣ ⎦
∑
∑
C •• Controller 3 (C3): buffer inventory control.Controller 3 (C3): buffer inventory control.
m
mriri reference value, reference value, yy,, min
minvvJJ33 s.t. s.t. vv∈∈S(PN,S(PN,m) withm) with
2 3 m ( ( ) ( , )v ( )) i j c ri i k i j j k p Y t T J m p t ∈ ∈ ⎡ ⎤ = ⎢ − τ + τ ⎥ ⎢ ⎥ ⎣ ⎦
∑
∑
C Control Action Decision Maker B(τk) zk E A(τk) x(τk) y(τk) x(τk+1) + + - +v(τk) u(τk) mrs7 s4 s5 s3 s2 l1 l3 l4 l5 l2 l6 l9 l7 l8 l10 l11 s1 l12 s6 u1 M, K M, K C, H C, H, M PC PC PC PC PC PC C si lk Sito Trasportatore Cliente u2 uj H 3 suppliers 2 manufacturers 2 retailers 1 de-manufacturer 1 distributor
Case Study A (I):
Desktop Computer production
Case
Case
Study
Study
A (I):
A (I):
Desktop Computer production
Desktop Computer production
1 product (PC)
3 suppliers
2 manufacturers 2 retailers
1 de-manufacturer 1 distributor
Case Study A (II):
FOHPN model
Case
Case
Study
Study
A (II):
A (II):
FOHPN model
FOHPN model
The model selects the production rates by solving the controller optimization problem
• T= Throughput average number of products obtained in a time unit
•
SI= System Inventory
average number of products storedstored in in allall the the systemsystem buffersbuffers duringduring the the runrun time TPtime TP
• LT=SI/T =Lead Time a measure of the time spent by the SC to
convert raw material into final products
Case Study A (III):
Performance indices and Controller
Case
Case
Study
Study
A (III):
A (III):
Performance
Performance
indices
indices
and Controller
and Controller
• Controller C1 (flow maximization)
• Controller C2 (stored materials minimization)
P
Y={p
1,p
3,…,p
13,p
31,p
33}.• Controller C3-1 (buffer inventory control)
P
Y={p
31,p
33},m
r31=m
r33=80• Controller C3-2 (buffer inventory control)
Case Study A (IV):
Data
Case
Case
Study
Study
A (IV):
A (IV):
Data
Data
Buffers and transporters capacity, reorder level and
fixed order quantities
Transitions continuous (discrete) firing speed
Case Study A (V):
Simulation
Case
Case
Study
Study
A (V):
A (V):
Simulation
Simulation
•• The model The model isis implementedimplemented and and simulatedsimulated in MATLAB in MATLAB environmentenvironment..
•
• 50 50 indipendentindipendent replicationsreplications of TP=600 h, of TP=600 h, withwith a a transienttransient periodperiod of TR=100 h.
of TR=100 h. WeWe obtainobtain a a halfhalf widthwidth thatthat in the in the worstworst case case equalsequals 4.7%
4.7% forfor a 95% a 95% confidenceconfidence intervalinterval.. •
• At the beginning of each macroAt the beginning of each macro--period (t=period (t=ττkk) the program describes ) the program describes and solves the optimization problem on the basis of the knowledg and solves the optimization problem on the basis of the knowledge e
of
of
x
x
((ττkk) and fixes ) and fixes vv((ττkk), ), AA((ττkk), ), BB((ττkk), ), ττk+1k+1 (next macro(next macro--event). After this event). After this it determines the next stateit determines the next state
x
x
((ττk+1k+1) by the state equation and the ) by the state equation and the procedure is reCase Study A (VI):
Simulation results
Case
Case
Study
Study
A (VI):
A (VI):
Simulation
Simulation
results
results
•
• ThroughputThroughput maxmax withwith C1, min withC1, min with C2,
C2, hitghthitght withwith C3-C3-1 e C31 e C3--22
•
• System System InventoryInventory min withmin with C1, C1, max
max withwith C1 and intermediate C1 and intermediate with
with C3-C3-1 e C31 e C3--2 (2 (lessless thanthan C3-C3-2) 2)
1.75 1.41 1.72 1.74 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 C1 C2 C3-1 C3-2 Controller Th roug hput [ P ar ts per hour ] 1278.20 925.93 1194.60 1026.30 200.00 400.00 600.00 800.00 1000.00 1200.00 1400.00 S y s tem In ven to ry [P a rts]
Case Study A (VII):
Simulation results
Case
Case
Study
Study
A (VII):
A (VII):
Simulation
Simulation
results
results
731.29 657.28 694.22 588.61 0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 C1 C2 C3-1 C3-2 Controller Lead Ti m e [Hour s ] 1000 200 300 400 500 600 50 100 150 Time [hours] inve ntory le vel [p arts] C1 C2 C3-1 C3-2 •
• LeadLead timetime min withmin with C3-C3-2.2. •
• LeadLead timetime higherhigher withwith C3-C3-1 1 that
that withwith C3-C3-22
•
• buffer buffer markingmarking forfor final final productsproducts pp3131 into
into manufacturermanufacturer M1 .M1 . •
• C1: buffer alwaysC1: buffer always fullfull •
• C2: buffer alwaysC2: buffer always emptyempty •
• C3-C3-1 and C3-1 and C3-2 2 tendtend toto keepkeep the stock the stock levels
levels aroundaround the imposedthe imposed set-set-pointpoint (80
Presentation Outline
Presentation
Presentation
Outline
Outline
•
• IntroductionIntroduction toto SupplySupply ChainsChains (SC)(SC) •
• AimAim ofof the the researchresearch •
• SC SC operationaloperational modelmodel byby First First OrderOrder HybridHybrid Petri Petri NetsNets (FOHPN)(FOHPN) •
• SC operationalSC operational modelmodel byby ColouredColoured TimedTimed Petri NetsPetri Nets (CTPN)(CTPN) •
A Coloured Timed Petri Net Model
A Coloured Timed Petri Net Model
for Supply Chain Management
for Supply Chain Management
and Performance Evaluation
and Performance Evaluation
“Modeling, Control, Simulation and Diagnosis of
Complex Industrial and Energy Systems”,
ISA/O3NEDIA, Luca Ferrarini and Carlo Veber,
Editors, 2009, ISBN 978-1-934394-90-8 2009
Mariagrazia Dotoli, Maria Pia Fanti, Giorgio Iacobellis Dipartimento di Elettrotecnica ed Elettronica, Politecnico di Bari
•
• Compact and Compact and resourceresource orientedoriented modelmodel toto describedescribe the SC the SC dynamicsdynamics byby the flow the flow of
of batchesbatches ofof productsproducts amongamong SC SC facilitiesfacilities byby transporterstransporters.. •
• The discrete The discrete eventevent dynamicsdynamics modelsmodels stochasticstochastic events: client events: client demansdemans, , block/
block/restartrestart ofof supplies, block/supplies, block/restartrestart ofof transportationstransportations due todue to strikes, strikes, accidents
accidents.. •
• The The tokentoken color color definesdefines the the productproduct characteristicscharacteristics (e.g., type(e.g., type, , quantityquantity))
Model with CTPN (I):
Characteristics
Model
Model
with
with
CTPN (I):
CTPN (I):
Characteristics
Characteristics
• Jensen, Colored Petri Nets: Basic Concepts, Analysis Methods and Practical Use,
EATS Monography and Theoretical Computer Science, Springer Verlag, 1992.
• Formalism using timed PN in which a color is associated to any token
Advantages:
• Modularity
• Model suitable for simulation • Applicability of control policies
• Possibility to distinguish the SC products: different dimensions and quantities, different manufacturing/transportation paths
•
•
A bipartite digraph described by the 9
A bipartite digraph described by the 9
-
-
tuple
tuple
CTPN
CTPN=
=
(
(
P,
P
,
T
T
, Co,
, Co,
Pre,
Pre
,
Post,
Post
,
Ω
Ω
,
,
F, RS,M
F, RS,M
00).
).
•
•
P
P
set of places modelling activities: transportation,
set of places modelling activities: transportation,
storage in manufacturers, buffers availability,
storage in manufacturers, buffers availability,
presence of orders
presence of orders
•
•
T
T
=
=
T
T
FF∪
∪
T
T
IIset of transitions where
set of transitions where
Transitions in
Transitions in
T
T
FFmodel the flow of jobs
model the flow of jobs
Transitions in
Transitions in
T
T
IImodel the input of jobs in the
model the input of jobs in the
system
system
Model with CTPN (III):
PN Definition
Model
Model
with
with
CTPN (III):
CTPN (III):
PN
<
<
λ
λ
j,
j,
θ
θ
j
j
>
>
•
•
CTPN=
CTPN
=
(
(
P,
P
,
T
T
, Co,
, Co,
Pre,
Pre
,
Post,
Post
,
Ω
Ω
,
,
F, RS,M0
F, RS,M0
)
)
Coloured tokens represent jobs and orders.
Coloured tokens represent jobs and orders.
•• Each token represents a batch of parts or of orders Each token represents a batch of parts or of orders ((rawraw material, material, semi
semi--finishedfinished partsparts).).
•
• The marking M denotes the state of the CTPN:The marking M denotes the state of the CTPN: M(p
M(p) gives the colours associated with each token in p.) gives the colours associated with each token in p. M0 is the initial marking
M0 is the initial marking •
• The colour of each token j is a couple:The colour of each token j is a couple: <<λλjj,,θθjj> , > , where where λλjj is the batch is the batch type and
type and θθjj the corresponding assigned quantity the corresponding assigned quantity θθjj.. •
•
Co
Co
mapsmaps eacheach placeplacep
p
∈∈P
P
intointo the set of the set of admissibleadmissible coulorcoulorCo(p)
Co(p)
and and eacheach transitiontransition tt∈∈
T
T
intointo the set of the set of admissibleadmissible fireablefireable coulorcoulorCo
Co
(t). (t). LetLet ΩΩdefined as
defined as: : ΩΩ ==∪∪xx∈∈PP∪∪T{Co(x)}. T{Co(x)}.
Model with CTPN (IV):
PN Definition
Model
Model
with
with
CTPN (IV):
CTPN (IV):
PN
•
•
CTPN=(
CTPN
=(
P,
P
,
T, Co,
T
, Co,
Pre,
Pre
,
Post
Post
,
,
Ω
Ω
,
,
F, RS,M
F, RS,M
00).
).
+ ⎧ × → ⎪ ⎨ × → ⎪⎩ \ ` c d P T Pre,Post : P T
•
•
F
F
: specifies the firing time associated to each
: specifies the firing time associated to each
transition (exponential distribution or null value).
transition (exponential distribution or null value).
•
•
Matrix
Matrix
Pre
Pre
gives the enabling condition:
gives the enabling condition:
t
t
∈
∈
T
T
is enabled at marking M with respect to a colour
is enabled at marking M with respect to a colour
c
c
∈
∈
Co(t
Co
(t
)
)
iff
iff
for each p in input of t, it holds
for each p in input of t, it holds
M(p)
M(p)
≥
≥
Pre
Pre
(p,t)(c
(p,t)(c
),
),
•
•
Matrices
Matrices
Pre
Pre
and
and
Post
Post
update the marking
update the marking
: upon firing,
: upon firing,
t leads to a new marking M
t leads to a new marking M
’
’
:
:
M
M
’
’
(p
(p
)=
)=
M(p)+
M(p)+
Post
Post
(p,t)(c)
(p,t)(c)
-
-
Pre
Pre
(p,t)(c
(p,t)(c
).
).
Model with CTPN (V):
PN Definition
Model
Model
with
with
CTPN (V):
CTPN (V):
PN
The model of a
The model of a
general
general
purpose
purpose
site
site
(manufacturer(manufacturer/assembler, /assembler,distributor
distributor, de, de--manufacturermanufacturer, , suppliersupplier))
The CTPN modelling the SC sites
The CTPN
The CTPN
modelling
modelling
the SC
the SC
sites
sites
Inventory control rule: (R,Q)
•
•
p
p
ii: the product is in the : the product is in the input bufferinput buffer ••
p
p
’
’
icic: the : the available capacityavailable capacity. . ••
t
t
tt :the transport of material.:the transport of material. ••
t
t
II :the uploading in the production buffer.:the uploading in the production buffer. ••
p
p
PP: the : the product is in productionproduct is in production ••
p’
p
’
PcPc: the capacity of the production facility: . . ••
t
t
pp:
:
the production operation the production operation ••
p
p
oo: : the product is in the output bufferthe product is in the output buffer •The input buffer can accomodate
The input buffer can accomodate
n1
n1
lots
lots
of
of
type
type
λ
λ
1 and of
1 and of
θ
θ
1
1
pieces
pieces
Capacities
Capacities
,
,
reorder
reorder
functions
functions
and
and
reorder
reorder
quantities
quantities
can
can
be
be
different
different
for
for
each
each
type
type
of
of
product
product
:
:
Cap
Cap
(
(
p
p
IcIc)= (n
)= (n
11<
<
λ
λ
1,
1,
θ
θ
1>,n
1>,n
22<
<
λ
λ
2,
2,
θ
θ
2>
2>
…
…
)
)
The model of a
The model of a
general
general
purpose
purpose
site
site
(manufacturer(manufacturer/assembler, /assembler,distributor
distributor, de, de--manufacturermanufacturer, , suppliersupplier))
The CTPN modelling
the SC sites (II)
The CTPN
The CTPN
modelling
modelling
the SC
UP
UP isis a a functionfunction thatthat updatesupdates the the tokentoken colors
colors withwith the new color the new color obtainedobtained afterafter the
the operationoperation.. No
No labellabel standsstands forfor ““the the functionfunction makes
makes no no transformationtransformation in the in the elementselements””..
The model of a
The model of a
general
general
purpose
purpose
site
site
(manufacturer(manufacturer/assembler, /assembler,distributor
distributor, de, de--manufacturermanufacturer, , suppliersupplier))
The CTPN modelling
the SC sites (III)
The CTPN
The CTPN
modelling
modelling
the SC
The model of a
The model of a
general
general
purpose
purpose
site
site
(manufacturer(manufacturer/assembler, /assembler,distributor
distributor, de, de--manufacturermanufacturer, , suppliersupplier))
The output
The output inventoryinventory of of productproduct λλ1 in 1 in ppOO is
is higherhigher thanthan the output the output reorderreorder pointpoint,, i.e. M(
i.e. M(ppOO)>R()>R(ppOcOc)(<)(<λλ1,1,θθ1>):1>): Transition
Transition ttII isis inhibitedinhibited butbut ttPP continue continue to
to produce.produce. The output
The output inventoryinventory of of productproduct λλ1 in 1 in ppOO drops
drops belowbelow the output the output reorderreorder pointpoint,, i.e. M(
i.e. M(ppOO)< R()< R(ppOcOc)(<)(<λλ1,1,θθ1>):1>): Transition
Transition ttII isis enabledenabled withwith respectrespect toto the color <
the color <λλ1,1,θθ1>: 1>: Q(
Q(ppOcOc)(<)(<λλ1,1,θθ1>) 1>) productsproducts of of typetype λλ11 will be produced.
The CTPN modelling
the SC sites (IV)
The CTPN
The CTPN
modelling
modelling
the SC
•
•
p
p
TT: batch under transport
: batch under transport
•
•
p
p
TcTc:
:
transport capacity
transport capacity
.
.
•
•
t
t
TT: transport operation
: transport operation
•
•
t
t
OO:
:
the transport uploading
the transport uploading
•
•
p
p
UU: available products
: available products
•
•
p
p
DD:
:
outstanding orders
outstanding orders
.
.
•
•
t
t
oDoD: arrival of customer demand
: arrival of customer demand
•
•
t
t
oo:
:
marketing of product of a
marketing of product of a
•
•
recycler
recycler
The model of a
The model of a
transporter
transporter
The model of a
The model of a
customer
customer
The CTPN modelling
the SC sites (V)
The CTPN
The CTPN
modelling
modelling
the SC
3 suppliers
2 manufacturers
2 retailers
1 de-manufacturer 1 distributor
Case Study B (I):
Case
Case
Study
Study
B (I):
B (I):
• Two connections from suppliers
to retailers are added
3 suppliers
2 manufacturers
2 retailers
1 de-manufacturer 1 distributor
Case Study B (II):
CTPN model
Case
Case
Study
Study
B (II):
B (II):
CTPN model
CTPN model
The model is compact and is able to distinguish among
• T= Throughput The average number of products per time unit
•
Inventory=INin+INout
inventory level of buffers• INin Average nr of products in the input buffers • INout Average nr. of product in the output buffers
• CTij=
Order fulfilment cyle time of the product j-th for the
i-th customer,
i.e. the average time elapsing from the i-th customer request and the j-th product delivery.• CT=
Order fulfilment cyle time,
i.e. the average time elapsing from a customer request and the product delivery.Case Study B (III):
Performance indices
Case
Case
Study
Study
B (III):
B (III):
Performance
Case Study B (IV):
Data
Case
Case
Study
Study
B (IV):
B (IV):
Data
Data
Data buffers and transporters Average firing time (hours) of exponential transitions of CTPNCase Study B (V):
Simulation results
Case
Case
Study
Study
B (V):
B (V):
Simulation
Simulation
results
results
•
• The CTPN The CTPN isis simulatedsimulated in the ARENA in the ARENA environmentenvironment •
• PlacesPlaces model model resourceresource withwith limitadedlimitaded capacitycapacity , timed, timed transitionstransitions model model delay
delay, , tokenstokens modelmodel entitiesentities and coulorsand coulors model model attributsattributs of entitiesof entities..
20
20 independentindependent replicationsreplications of of TP=11856 h,
TP=11856 h, withwith a transienta transient period
period of TR=3025 h.theof TR=3025 h.the halfhalf width
width of the of the confidanceconfidance interval
interval, , beingbeing 2.7% in the 2.7% in the worst
worst casecase Since
Since customerscustomers havehave the the same
same orderorder frequencyfrequency, , throughputs
Case Study B (VI):
Simulation results
Case
Case
Study
Study
B (VI):
B (VI):
Simulation
Simulation
results
results
•
• The CT are The CT are aboutabout 7h 7h consistentconsistent withwith the
the firingfiring timestimes •
• The CT42 and CT52 The CT are lowerare lower forfor the
the lowerlower valuesvalues of the firingof the firing times. times.
•
• The input buffersThe input buffers tend
tend toto accommodateaccommodate more
more partsparts thanthan the the output
Presentation Outline
Presentation
Presentation
Outline
Outline
•
• IntroductionIntroduction toto SupplySupply ChainsChains (SC)(SC) •
• AimAim ofof the the researchresearch •
• SC SC operationaloperational modelmodel byby First First OrderOrder HybridHybrid Petri Petri NetsNets (FOHPN)(FOHPN) •
• SC SC operationaloperational modelmodel byby ColouredColoured TimedTimed Petri Petri NetsNets (CTPN)(CTPN) •
Conclusions and Future Research
Conclusions
Conclusions
and Future
and Future
Research
Research
We represent SC using the PN discrete event systems formalism
We propose two alternative SC modular models, devoted to the efficient description, management, and performance evaluation of SC at the
operational level. The FOHPN model
¾ describes material as continuous flows and the stochastic events occurring
in the system as discrete events;
¾ is resource oriented, modular and efficient;
¾ enables the designer to impose an optimal SC dynamics according to
objective functions and the knowledge of the system state.
The CTPN model
¾ describes in detail the SC products flow; ¾ is resource oriented, modular and efficient; ¾ is suitable for implementing control strategies.
Mariagrazia Dotoli,
Mariagrazia Dotoli,
Maria Pia Fanti,
Maria Pia Fanti,
Giorgio
Giorgio
Iacobellis
Iacobellis
, Agostino M.
, Agostino M.
Mangini
Mangini
DEE
DEE
-
-
Politecnico di Bari
Politecnico di Bari
Petri Net Models for Supply Chain
Petri Net Models for Supply Chain
Operational Management and
Operational Management and
Performance Evaluation
Performance Evaluation