Helia Sohrabi
Laval University, CIRRELT, Canada
Walid Klibi
BEM- Bordeaux Management School, CIRRELT, France
Benoit Montreuil
4th International conference on Information Systems, Logistics and Supply Chain
Quebec City August 26-29, 2012
Modeling a Scenario-based
Distribution Network Design Problem
Helia Sohrabi, Walid Klibi and Benoit Montreuil
Presentation plan
Current distribution context
Physical Internet initiative: Introduction
Research methodology
Future business environment modeling
Distribution network design generation
Mathematical model
o Generic model
o Context-driven model
Current distribution context
Closed distribution system
(Private Distribution Network; PDN)
Cost-service trade offs
Addressing changes in the business environment
(e.g. customer expectations, economic instability and energy crisis)
Collaborative distribution system (Shared Distribution web; SDW)
Helia Sohrabi, Walid Klibi and Benoit Montreuil
Physical Internet Initiative:
An Introduction
What is Physical Internet?
An open global logistics system founded on physical and digital interconnectivity through encapsulation, protocols and interfaces
http://physicalinternetinitiative.org
What is Physical Internet goal?
To reduce by an order of magnitude the environmental, economic and social efficiency and sustainability of the way physical objects are moved, stored, realized, supplied and used
Mobility Web Moving goods & people
Interconnected open unimodal & multimodal infrastructures,
vehicles, hubs and transits
Distribution Web Deploying, storing
products
Interconnected open warehouses & distribution centers
Realization Web Realizing products
Interconnected open production, personalising & retrofit centers
Supply Web Supplying goods
Interconnected open suppliers and subcontractors
Service Web Enabling and sharing
access and usage of services rendered
by goods & people
Interconnected open users and service providers
Logistics Web
Set of openly interconnected physical, digital, human, organizational and social agents and networks aiming to serve efficiently and sustainably
the logistics needs of people, organizations, territories and society
The Physical Internet aims to enable an efficient and sustainable Logistics Web
Research question:
How to design distribution networks exploiting the open distribution web? Research goal:
To investigate the performance improvement (both from economic and
environmental perspectives) achievable by switching from current distribution context to Physical Internet enabled context.
Helia Sohrabi, Walid Klibi and Benoit Montreuil
Research methodology
Distribution Network performance investigation (simulation) Distribution Network design generationat each distribution context Future business
scenario generation
(Monte Carlo approach)
Distribution Network
design generation
at each distribution context Distribution Network
design generation
at each distribution context
ODW SDW PDN Distribution Network performance investigation (simulation) Distribution Network performance investigation (simulation) For company 1 company m
A single final design in PDN context SDW ODW Kli bi , W . an d Ma rt el, A ( 2 0 1 2 )
Future business environment modeling
• Applying a novel scenario planning approach based on evolutionary paths
(Klibi and Martel, 2011).
• An evolutionary path is a possible view of the future characterized by the future-shaping variables.
• Future-shaping variables defined:
Demand-market,
Available capacity, (including production , storage capacity)
Helia Sohrabi, Walid Klibi and Benoit Montreuil
Future business environment modeling
Three
defined evolutionary paths
I.
Sustainability-unconscious
II. Sustainability-prone
III. Sustainability-engaged
.
• A pessimistic view of the future world
• Based on the unstable situation of several economies in the world and the scramble energy scenario described by Shell (2009).
• A world that is conscious about the need to become more sustainable, yet is slow in moving towards sustainability.
• anticipated to emerge in the parts of the world, mostly in North American and Western European developed countries.
• Moderate improvement in economy and local actions addressing energy security, in line with the blueprint scenario elaborated by Shell (2009)
• A more optimistic view
Sustainability-unconscious Sustainability-prone Sustainability-engaged
Future business environment modeling
future shaping variables evolution through each evolutionary path
Cost of investment Accessible capabilities
Planning
horizon Planning horizon
Using a
Monte Carlo
procedure
For each evolutionary path
For each future-shaping variable
A set of scenarios will be generated
This procedure will be
repeated for each company
case with
different characteristics
independently
from each other.
Helia Sohrabi, Walid Klibi and Benoit Montreuil
Distribution network design generation
Strategic reengineering process
Yearly/quarterly planning period 𝐭 ∈ 𝑻
Planning horizon
Weekly/Daily working period τ ∈ 𝒕(τ)
Transportation cost in a scenario
Demand cost in a scenario
Available capacity cost in a scenario
A reengineering cycle
Schematic representation of three distribution
contexts
Planning horizon PDN ODW Planning horizon SDWHelia Sohrabi, Walid Klibi and Benoit Montreuil
Facilities fixed and usage costs in different contexts
Initial state Do not use the site Use the site
Decision Fixed cost Decision Fixed cost Usage cost
Current facility
Own (o)
Close Transfer and closing cost
Use during all
the horizon T Capital recovery
Operating cost Rent/Use service (r) Stop for a period Closing cost plus lease penalty Renew rental contract for a period
Lease cost Operating cost
Potential site
Purchase/
Build Do not use Zero
Open and use during all the
horizon T
Setup cost plus capital recovery
Operating cost
Rent Do not use Zero
Agree a rental contract for a
period
Setup cost plus lease cost Operating cost c lL o c l L L l a l a l a r c l L L tT , lt a t T tT , a t T a t, T p lL o p l L L l a l a r p l L L tT , lt a t T a tlt, T
The generic mathematical model
Customer-DC assignment constraints:
, r o o t lt lt lt lt lt lt lt lt t T l L l L l l l l l L E NOP Max R x a x a x a x a x a x a x
Subject to: lst ltx
x
1 lt lt x x DC state constraints (for rented facilities):
(1) , l L s S and t T (2) (3) , r l L t T
Helia Sohrabi, Walid Klibi and Benoit Montreuil
The generic mathematical model
continued
DC opening/closing constraints:
Non negativity constraints:
( 1)
0
lt lt lt l tx
x
x
x
1 l l x x DC state constraints : 0 0 lt t l l l x x x x T
DC state constraints : , , , , , {0,1} lt lst lt lt l l x x x x x x (4) , r l L t T ol
L
(5) ol
L
(6) (7) , , l L sS t TThe generic mathematical model
continued
Transportation cost
Inventory cost
Recourse capacity acquirement cost
Where for a given planning period
, , , , [( ( ) ( ) ( ( ) ( )) ( )) t l l t T p l s S ps pls e e ps ps ps p s S R Max u F u f F
x
' ' ' , , , , ( ) ,)]
pvl pvl p v l pll pll p l l L S pl pl p l l l lf
F
f
F
g
I
k
Z
Total income – external sourcing cost
t T
Helia Sohrabi, Walid Klibi and Benoit Montreuil
The generic mathematical model
continued
Subject to:
Capacity constraints:
Customer assignment constraints:
Inventory/ product flow constraints:
( ) ( ( )) ( ) ( ) pl p pls l l pI pc s S F b Z
, t pls ( ) , t ps ( ) lst pT F pT d x
( 1) ( ) ( ) ( ) ( ) ( ) ( ) pl pl l L V pl l l L S pll I I
F
F , , lL sS
(9) , t, lL T (10) , , t, pP lL T (11)The generic mathematical model
continued
Initial inventory constraints:
Demand satisfaction flow constraints:
Recourse capacity constraints:
0
( )
0
plI
( ) e ( ) ( ) ( ) ps ps l L pls ps d
F
F
d
Production capacity constraints:
Integrality and non negativity constraints:
( ) ( ) pvl pv l L F b
( ), ( ), e ( ), ( ), ( ) 0 pvl pll ps l pl F
F
F
Z
I
1 , , , pP lL T (12) , , t, pP sS T (13) , t, lL T (14) , , t, pP vV T (15) ( ) ( ) l l l lt Z
b
xHelia Sohrabi, Walid Klibi and Benoit Montreuil
Context-driven design models
Current context(closed)
•
Private distribution network
The design model in PDN is similar to the generic model (1)-(16)
The optimal design solution is denoted by and the value of this solution is denoted by .
*
x
* Group of M partnering companies
Each company designs a distribution web maximizing its own performance subject to joint capacity restrictions for each shared facility
The sets of locations dedicated to each company
The set of shared facilities
The portfolio of available resources
The optimal solution and its value
Context-driven design models
Current context(collaborative)
•
Shared distribution web
* m x
* m E NOP x m mL
L
V L Sm, m, m
' ' \ m m M m m mL
L
Helia Sohrabi, Walid Klibi and Benoit Montreuil
Context-driven design models
Open distribution web context
• Companies mostly deploy their products independently from each other even though collaborative behavior such as in SDW is possible.
• Dynamic available capacity
• Two additional set of resources :
All private facilities operated by other companies
Public facilities whose storage capacity is completely open to the market,
denoted by
Ultimately, the open distribution web includes both companies, and third-party logistics providers openly offering distribution storage capacity. ( ) l b m L w m L w m m m m L L L L
Final remarks and future work
Final remarks
:
•
Sample Average Approximation (SAA)
(period sampling, scenario
sampling and/or aggregations)
•
Exact approach
(OPL-CPLEX) / Tabu based
metaheuristics
Future work:
To validate the model on a representative case
Comparison between the three alternative designs in different distribution contexts studied using a large set of scenarios and a various set of criteria such as: delivery speed, eco-efficiency, resiliency, etc.
Helia Sohrabi, Walid Klibi and Benoit Montreuil