D i i R b t V l C ti Designing Robust Value-Creating Supply Chain Networks Supply Chain Networks Alain Martel Interuniversity Research Center on Enterprise Networks, Logistics and Transportation (CIRRELT) and Faculté des sciences de l’administration Université Laval, Québec, Canada Université Laval, Québec, Canada May 2010 CIRRELT
53
Embed
Designing Robust Value-Creating Supply Chain Networksneumann.hec.ca/chairedistributique/common/Martel.pdf · Robust and efficient design under uncertainty n sion ... Designing Robust
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
D i i R b t V l C ti Designing Robust Value-Creating Supply Chain NetworksSupply Chain Networks
Alain MartelInteruniversity Research Center on Enterprise Networks, Logistics and Transportation (CIRRELT)g p ( )andFaculté des sciences de l’administrationUniversité Laval, Québec, CanadaUniversité Laval, Québec, Canada
May 2010
CIRRELT
SCN-StudioSCN StudioTool developped in collaboration withModellium and DRDCModellium and DRDCduring the DRESNET project(Design of Robust and Effective Supply(Design of Robust and Effective SupplyNetwork Engineering Tools)
A. Martel - Designing Robust Value-Creating Supply Chain Networks 2
1. Problem Context 2. Modeling SCN Processes and 2. Modeling SCN Processes and
Structures3. Taking Uncertainty into Account4 SCN D i M h d l4. SCN Design Methodology5 Research Challenges5. Research Challenges
A. Martel - Designing Robust Value-Creating Supply Chain Networks 3
1. Generic Design ProblemgRaw material sources
Deployed Supply...
ManufacturingProcess
Chain Network
SC Risk ExposureFinished Products
DistributionChannels
V l C tiValue Destruction
DisruptionsMarkets
...
Value CreationRobustness ?
A. Martel - Designing Robust Value-Creating Supply Chain Networks 4
Re-Valorization Network
SUPPLY
VALORIZED PRODUCT MARKETS
N
N
SUPPLY
SOURCESOURCE
RE
RERERE--DISTRIBUTEDISTRIBUTE
CH
AIN
CH
AIN
WO
RK
WO
RK
MANUFACMANUFAC--TURETURE
EE--V
AL
OV
AL
ON
ET
NE
T
VALORIZEVALORIZERe-manufacture
DisassemblyRepairVal e Creation
PP
LY
CP
PL
Y C
NE
TW
ON
ET
WO
OR
IZA
TO
RIZ
AT
TW
OR
KT
WO
RK
RepairValue CreationNetwork
SUP
SUP NN
T
DISTRIBUTEDISTRIBUTE
TIO
NT
ION
KK
RECOVERRECOVER
RETURNRETURNDEMAND-RETURNDISPOSAL
--RETURNRETURN
A. Martel - Designing Robust Value-Creating Supply Chain Networks 5
SCN Design Modeling
Robust and efficient design under uncertainty
nsio
n
Robust and efficient design under uncertainty • Hazardous/catastrophic disruptive events (risk modeling)• Adequate user decisions anticipation• Resilience strategies for sustainable stakeholder value
RobustDesign
cal E
xpan
Stochastic problems over a long planning horizon • Business as usual context (random events modeling)• Real options and hedging
g
Large
Location- • Generic production-distribution networksM lti l f ilit / t t ti tiod
olog
ic • Real options and hedging• Positioning by anticipation to capture promising contractsSPs
Allocation Problems
• Multiple facility / transportation options• Externalization decisions with partner selection• Multiple value offers to product-markets• Value maximization …
Met
hoLargeMIPs
Functional Expansion
A. Martel - Designing Robust Value-Creating Supply Chain Networks 6
2. Modeling SCN Process and Structures Structures
A. Martel - Designing Robust Value-Creating Supply Chain Networks 7
Modeling SCN Process and Structures Structures
SupplySupply1
RMStorage
2
3 ComponentsVS
Raw materials3
1 2 3
2 1 2p ∈ RM
fabrication
SAsmanufacture
414Manufactured prod cts
22
11
4 5
Final assembly
manufacture
Kitting5
6products
2 3 4
16 7p ∈ MP
SupplyDemand
FPStorage
7
8Finished products
8 9
p ∈ FPp ∈ P
Process graph with recipes
pp y
Bill-of-material (BOM) forassembly /disassembly process
pp
A. Martel - Designing Robust Value-Creating Supply Chain Networks 8
recipesassembly /disassembly process
Modeling SCN Process and StructuresStructures
Raw MaterialVendors
IntermediateProduct Plants
Fournisseurs res (v∈ V)To the network facilities
(s∈ S)
Sources (v ∈ V)Fournisseurs resFournisseurs res (v∈ V)
(s S)
SupplySources (v ∈ V)
FinalProduct Plants
(s ∈ S)(s S)
DCs /Warehouses
Distribution CentersProduction-Distribution Centers
To the demand zones
DemandZones
E h l t tZones dedemande(d∈ D)Zones dedemande(d∈ D)Zones dedemande(d∈ D)DemandZones (d∈ D)
G l t tEchelon structure vs General structure
A. Martel - Designing Robust Value-Creating Supply Chain Networks 9
Generic Process Modeling Formalism
ActivityStems SupplySupply
RM Suppl
Activity types:
Activity Graphfor the
Bucking
RMStorage Logs
Supply:
Transformation:
Consolidation/ transfer:
for theForest
dSawing
Rejects
Logs
Storage:
Demand:
Transformation:
ProductSawing
Drying
Chipping
Rejects
Rough Lumber
Green Lumber Demand:
Movement types:Inter location:
Industry Planing/Grading
Chips
g Inter-location:(transportation)
Intra-location:(material handling)
SupplyDemand
FPStorage
p
d
Both: (inter or intra location)
A. Martel - Designing Robust Value-Creating Supply Chain Networks 10
Boards
Activity Graph for CF Prepositioning SCN Case
SupplyDomestic/LocalSupply
(all) (4)(2,3)(6,7) (8,9,10,11)
RepairStaging-Transfer
PalletStorage
HazmatStorage
(11)Refrigerated
Storage(8,9,10,11)
Lane Meter Storage
(6,7) (2,3) (4)(all) (8,9,10,11)(all)
SupplyTheatre
Demand(13)
(1 2 3 4 5 8 9 10 11 14)(12)
Initial provisioning transportation (I)Deployment transportation (D)Sustainment transportation (S )
A. Martel - Designing Robust Value-Creating Supply Chain Networks 19
BoardsSpot market
zonesPotentialcontracts VMI agrem. …d
a A∈ l L V S D∈ = ∪ ∪
Potential Supply Chain Network Vendor 1 Vendor l
Node (l,1). . .s
V( ,1),l l L∈
pp y
inl L∈. . .( , )l a N∈ S V
(1, ) ( ,1)( , ) ( ) ( )a l p l np s P P S l L∀ ∈ ∩ × ∈
Sup
ply
arcs
T( , )n l a N= ∈
( , ')a ap P∀ ∈
alar
cs . . . ex'l L∈( , ') '( , ) a a pnnp s P S∀ ∈ ×
lc C∈
Inte
rna
C' ( ', ')n l a N= ∈ W' ( , ')n l a N= ∈
D( )l a l L∈ eman
dar
cs 'lc C∈( , ) '( , )( , ) ( ' , )k p l pn l a jplj s J S n s NS←∀ ∈ × ∈
. . .( , ), tl a l L∈
SupplyZone 1 SupplyZone lNode ( , )l a
. . .. . .. . . lkp P∈
K( )lk K∈
. . .
De
A. Martel - Designing Robust Value-Creating Supply Chain Networks 20
Modeling Value
Consider a given SCN Design
g
Consider a given SCN Design
A. Martel - Designing Robust Value-Creating Supply Chain Networks 21DOMTAR CASE
Design ObjectiveEfficient-Frontier for a given P&C System
TotalDiscounted
Qualifying requirementsCosts requirements
Dominated Design
Failed Design
Efficient Design
Response Time (or other value attribute)
A. Martel - Designing Robust Value-Creating Supply Chain Networks 22
Design ObjectiveDesign for Value
DesignDiscounted Total Revenue Maximize
Costs(Revenues)
MarginMargin
EconomicValueMarginMargin Value
AddedExpenditures
Design Response Time (or other value attribute)
A. Martel - Designing Robust Value-Creating Supply Chain Networks 23
Type of Model ObtainedMax ∑Countries{(1-Tax) ∑Sites[Revenues
- (Platform costs(Platform costs+ System & flexible capacity costs+ Facilities operations costs+ Procurement and Inventory costs+ Procurement and Inventory costs+ Transportation costs + Duties...)]}
subjet toNetwork configuration constraintsNetwork configuration constraintsPlatform/system selection constraintsSupply / Capacity / Demand constraintsM t i l i t t i tMaterial requirement constraintsInventory accounting constraintsFlow conservation constraintsLocal content and transfer price constraints…
• Coping with randomness in a business as usual context• Stochastic demands, prices, exchange rates…• Static stochastic program with recourse for a typical planning cycle• Static stochastic program with recourse for a typical planning cycle• Multi-stage (cycles) stochastic program with recourse
A. Martel - Designing Robust Value-Creating Supply Chain Networks 2626
• Demand surge for first aid and construction material
• Significant decrease of demand for luxury productsA i t l 58 000 t • Approximately 58,000 troops coming from all 50 US states assigned to the theaterassigned to the theater
A. Martel - Designing Robust Value-Creating Supply Chain Networks 27
Supply Chain Network Risk AnalysisThree Fundamental Questionspp y y
• What can go wrong?• Vulnerability sources identification and filteringVulnerability sources identification and filteringWhat is the likelihood of that happening?
Multihazard zones risk exposure levelsMultihazard zones risk exposure levelsStochastic multihazard arrival processesA i b bili i Attenuation probabilities
What are the consequences?• Incidents damage on supply/resources /demand
A. Martel - Designing Robust Value-Creating Supply Chain Networks 28
DSN Risk Analysis: What can go wrong?y g g
M ltih d
Natural disasters
Geopolitical failures× Multihazardsfailures
Market failures
Industrial accidents
×Vulnerability SourcesVulnerability Sources
Considered
A. Martel - Designing Robust Value-Creating Supply Chain Networks 29
DSN Risk Analysis:What is the likelihood of that happening ?Exposure Levels for Natural Disasters
f pp g
Derived from data provided by the
A. Martel - Designing Robust Value-Creating Supply Chain Networks 30
Center for Research on the Epidemiologyof Disasters (CRED)
Vulnerability-Exposure Relationship
A. Martel - Designing Robust Value-Creating Supply Chain Networks 31
Inter-arrival Time DistributionFor Natural Catastrophes’ Exposure Level 5
Exponential distribution with mean λ = 350 daysExponential distribution with mean λ = 350 days
DaysDays
A. Martel - Designing Robust Value-Creating Supply Chain Networks 32
Intensity Distributionte s ty st but oFor Natural Catastrophes’ Exposure Level 5L N l di t ib tiLog-Normal distribution
Loss level in $
A. Martel - Designing Robust Value-Creating Supply Chain Networks 33
A. Martel - Designing Robust Value-Creating Supply Chain Networks 34
gp
Attenuation Probabilities
• The occurrence of a multihazard in a zone does il l i k hinot necessarily translate into a network hit
• Canadian Forces Case:P b bili h i i i i i i d i h Probability that a mission is initiated in response to the occurrence of a multihazard in a given country
Hazard/Mission type Location Probability… …Natural disaster/Humanitarian assistance Belgium 0,023Natural disaster/Humanitarian assistance Botswana 0,023Natural disaster/Humanitarian assistance Chile 0,027… …Quarrel/Peacekeeping Greece 0,700Quarrel/Peacekeeping Herzegovina 0,700Quarrel/Peacekeeping Algeria 0,600… …War/Peace making Haiti 0,450War/Peace making Poland 0,400War/Peace making Cyprus 0,350… …
A. Martel - Designing Robust Value-Creating Supply Chain Networks 35
What are the consequences?SCN Risk Analysis: consequences?
Multihazard Incidents Severity Profile
y
y
1) 2) 3) 4) 5) 6)
Suppliers Plants DCs First-aid Sustainment Luxury
Capacity-based Vulnerability Sources S c = {1, 2, 3} Demand-based Vulnerability Sources S d = {4, 5, 6}
Suppliers Plants DCs Product-markets Product-markets y
Product-markets a) Natural disasters
Unfilled supply rate
Capacity lossrate
Capacity lossrate
Demand inflation rate
Demand deflation rate
b) Market Unfilled supply Demand deflation Demand deflation
ihaz
ards
a, b
, c}
Impact i i failures rate rate rate
c) Industrial accidents
Capacity lossrate
a) Natural disasters
Time to restoring supplies
Time to restarting production
Time to restarting distribution
Surge duration Drop duration
Mul
tiH
= {
rds
c}
intensity
disasters supplies production distribution
b) Market failures
Time to restoring supplies
Drop duration Drop duration
c) Industrial accidents
Time to restarting production M
ultih
azar
H =
{a,
b, c
Time to recovery
p
A. Martel - Designing Robust Value-Creating Supply Chain Networks 36
What are the consequences?SCN Risk Analysis: qy
Recovery Function ExampleCapacity loss recovery function
A lifi ti A priori percentagesAmplification percentage
100%
A priori percentages
ρ
Recovery function
Amplitude based on β
Working periods1τ ξ+ −'τ ' τ
τρ
Time to recovery
c( )' , ',..., ' 1; ( , , ); , , h h h h
lp lp lp l lp sp g l l lp s sc c r s S p P l Lτ τρ τ τ τ ξ β ξ= = + − = ∈ ∈ ∈ρ ρ
A. Martel - Designing Robust Value-Creating Supply Chain Networks 37
Plausible Future Scenarios• The superposition, over the planning horizon, of
• An instance of these multihazard processes • An instance of the business-as-usual random variables
Yields a probabilistic scenario
= Set of probabilistic scenarios
Pω∈ΩPΩ p
= Probability of occurrence of scenario• Deeply uncertain scenarios can also be considered
( )p ωΩ
Pω∈Ω• Deeply uncertain scenarios can also be considered
• Incorporate isolated, non repetitive, extreme events for which a likelihood of occurrence cannot be evaluatedwhich a likelihood of occurrence cannot be evaluated
= Set of deeply uncertain scenariosUΩ
A. Martel - Designing Robust Value-Creating Supply Chain Networks 38
l bl• Depots vulnerable to extreme events• How many warehouses and where ?
Stochastic H d
PlantPlant
gn
el
Hazard Process
⇓Potential DC
locations lx lx
l L∈
Potential DClocations
lx lx lx lxl L∈ DCD
esig
Lev
Respond from
Demand zones
(d ? D)Ship to l ti
ax
ryry rDemand zones
(d ? D)Ship-topoints
ax
ryry rDemand zones
(d ? D)Ship to l ti
axax
ryryryry rrDemand zones
(d ? D)Ship-topoints
axax
ryryryry rrer
vel
back-up depot
Demand zones locations y ryDemand zones points y ry
p P∈
Demand zones locations yy ryryDemand zones points yy ryry
p P∈Use
Lev
A. Martel - Designing Robust Value-Creating Supply Chain Networks 44Compound Poisson Demand Process
Strategic Decision Framework g
Design ( ) ( ){ } ( ) ( ) ( )1 1max x x xx ˆ, , , ,, . , d duC C CC I ω ω ω ωδΩ = + ∈ΩRModel
( ) ( ){ } ( ) ( ) ( )1 1 11 1
1 1max x X
x x xx , , , ,, . , C C CC I ω ω ω ωδ∈
+ ∈ΩR
( )1I δΩ
1x( )1 , duC ω ω∈Ωxˆ , 1( )
( )( ) ( )( ) ( )
( )( ) ( )( ) ( )( )2 1
11
opt N n
du u d ut n t
nt T t T
C C C Cω ω ω ω>∈ ∈
⎡ ⎤= + +⎢ ⎥
⎢ ⎥⎣ ⎦∑ ∑ ∑
x xx y x y
ˆ ˆ. ,..., . ˆ ˆ
ˆ ˆ ˆ ˆˆ ˆ ˆ,Anticipated ( ) ( )
11
n
T
⎣ ⎦y y ˆˆ ˆ. ,..., .
( ) ( ) ( ) ( )1s.t 1 n tnn n t tn tω ω ω ω−∈ ∀ > ∈ ∀xxx X y Y ( )ˆ ˆˆ ˆ,
ω∈Ω
Anticipated Adaptation-Response Model
& non-anticipativity of ( )xn ω
U R( )uI τ( ) ( ){ }opt yu uC I τR
*1x
User Response Decisions
( )( )
( ) ( ){ }y Y
opt yx*
n
C Iτ
τ τ
τ τ∈
RuTτ ∈
A. Martel - Designing Robust Value-Creating Supply Chain Networks 45
Generic SCN Design ModelUsing Stochastic Programming (Shapiro, 2007), Robust Optimization(Kouvelis et al., 1997) and Risk Analysis (Haimes, 2004) concepts, the design problem can be formulated as follows:
( ){ } ( ){ } ( ){ }{ }1 1
1 1 1maxx X
x x x, . , , . , , .A S UA S UC C C
Ω Ω Ω∈R R R
Conditional dispersion measure
Conditional return measure
Conditional expected value measure
{ } { } { }( ){ } ( ){ } ( ){ } [ ]1 1 1 0 1x x x, . , . , . , ,A A AA AA A AC C Cϕ ϕ
Ω Ω Ω= + ∈ER D
( ){ } ( ){ } ( ){ } [ ]1 1 1 0 1x x x, . , . , . , ,S S SS SS S SC C Cϕ ϕ
Ω Ω Ω= + ∈ER D
Multiparametric program
Robustness criterion( ){ } ( ){ }1 1, . ,U U UUC Min C
ωω
Ω ∈Ω=x xR D
Multiparametric program( ){ } ( ){ } ( ){ }
1 11 1 1 1max ( ) (1 ) ,. ,. ,.A S UA SA S U
R w C w C Cψ ψΩ Ω Ω∈
⎡ ⎤= − + +⎣ ⎦x Xx x x xR R R
A. Martel - Designing Robust Value-Creating Supply Chain Networks 46
Complexity Reduction Approachp y pp
• Use approximate anticipations of adaptation-response decisions to simplify the combinatorial structure of the design model (based on spatio-temporal aggregations)U l i di l t d l d i k i• Use only primordial expected value and risk aversion criteria associated to probabilistic scenarios
• Assuming that the SCN design problem will be solved on a• Assuming that the SCN design problem will be solved on a rolling horizon basis, reduce the design model to a multi-cycle two-stage stochastic program with recourse y g p g
• Solve the design model for several small samples of scenarios generated using Monte Carlo methodsg g
• Evaluate the designs obtained using a user response model
A. Martel - Designing Robust Value-Creating Supply Chain Networks 47
( ) ( ) ( )1 1 1 , , , ,d du Mj j jC C C ωω ω ω ∈+= Ωx x x Worst-case scenarios
Multi-criteria evaluation
Effective and Robust D i
level (κ) &Risk-aversionovershoot (Δ)
( )1 , ,du MjC ωω ∈Ωx
UMΩ
• Filteringand selection
( ){ }1 1,. , , , ; ( )MP
j jP C P A S U RΩ =x xR
Wo st case sce a os
Historical scenario
Design Ω
0( )ω
A. Martel - Designing Robust Value-Creating Supply Chain Networks 48
Design Generation for the Multi-depot location-transportation problemMulti depot location transportation problem
Several types of models and solution methods can b d l i d ibe used to generate alternative designs
• Exact solution to static deterministic location-transportation modelE t l ti t t ti d t i i ti l ti ll ti d l• Exact solution to static deterministic location-allocation model
• Multi-period versions of the previous models• Previous models with customers aggregated into demand zonesPrevious models with customers aggregated into demand zones• Stochastic versions of the previous models with different scenario
samples• Stochastic models with resilience structures (back-up depots…)• Heuristic solutions to the previous models• Solution with different expected value / dispersion weights• Solution with different expected value / dispersion weights• …
A. Martel - Designing Robust Value-Creating Supply Chain Networks 49
Design Evaluation/Selection for the Multi-depot location-transportation problemMulti depot location transportation problem• Based on a large sample of Monte-Carlo, worst-case and
historical scenarioshistorical scenarios• For each design x1 and each day τ of scenario ω :
• Assign customer orders to depots based on response policy • Assign customer orders to depots based on response policy • Request transportation at depots for truckloads• Solve depots routing problemsp g pTo get the design value
• Compute adequate performance measures1( , )R ω x
• Expected values• Dispersion measures (mean-semideviation, conditional value at risk… )
R ili• Resilience measures …
• Select the best design using multi-criteria decision methods
A. Martel - Designing Robust Value-Creating Supply Chain Networks 50
SCN St diSCN-Studio
1) Plausible Future Scenario Generation
2) Design Generation3) Design Evaluation and ) g
Selection
A. Martel - Designing Robust Value-Creating Supply Chain Networks 51
4- Research Challengesg
• SCN risk analysis• SCN multihazard modeling
• Scenario development and importance sampling• Scenario development and importance sampling
• Value based SCN design models• Dependence on value attributes & Financing
• Modeling for robustness• Modeling for robustness• Modeling resilience and responsiveness
• Solution methods
A. Martel - Designing Robust Value-Creating Supply Chain Networks 52
ReferencesReferences• Klibi Walid, Martel Alain, Guitouni Adel, The Design of Robust Value-Creating Supply Chain Networks: A
Critical Review, European Journal of Operational Research, 203(2), 283-293, 2010Klibi W lid L ll F i M l Al i I h S i Th h i l i i d l i• Klibi Walid, Lasalle Francis, Martel Alain, Ichoua Soumia, The stochastic multi-period location-transportation problem, Transportation Science, 2010 (v1-trsc.1090.0307 )
• Klibi Walid Martel Alain The design of effective and robust supply chain networks Document CIRRELT• Klibi Walid, Martel Alain, The design of effective and robust supply chain networks, Document CIRRELT-2009-28, 2009
• Martel Alain, The desing of production-distribution networks : a mathematical programming approach, Springer, in J. Geunes and P.M. Pardalos (eds.), Supply Chain Optimization, pp. 265-306, 2005
• Martel Alain, M'Barek W., D'Amours Sophie, L'influence des facteurs internationaux sur la compétitivité , , p , pdes réseaux de création de valeur multinationaux : le cas des compagnies canadiennes de pâtes et papiers, Revue Gestion, vol 31, no 3, pp. 85-96, 2006
• Martel Alain, Benmoussa A., Ezzedine I., Klibi Walid, Berger Jean, Boukhtouta A., Chouinard M., Girard S., Kettani Ossama, Military Missions Scenario Generation for the Design of Logistics Support Networks, I t ti l C f I f ti S t L i ti d S l Ch i (ILS) C bl M 2010International Conference on Information Systems, Logistics and Supply Chain (ILS), Casablanca, Maroc, 2010
• Vila D., Martel Alain, Beauregard Robert, Designing logistics networks in Divergent Process Industries: A Methodology and its Application to the Lumber Industry, International Journal of Production Economics, Vol. 102, pp. 358-378, 2006
• Vila D Martel Alain Beauregard Robert Taking market forces into account in the design of production-• Vila D., Martel Alain, Beauregard Robert, Taking market forces into account in the design of production-distribution networks: A positioning by anticipation approach, The Journal of Industrial and Management Optimization, Special edition on Supply Chain Optimization, Volume 3, No. 1, pp. 29-51, Février, 2007
A. Martel - Designing Robust Value-Creating Supply Chain Networks 53