Transcript Introduction to logistics & supply chain - univ

Agent-based supply chain
planning in the forest products
industry
Sophie D’Amours Ph.D.
Professor, Université Laval
General Director, Research Consortium FOR@C
Canada Research Chair on
planning value creation network
Basys’06, Niagara Falls, Ontario, Canada, 2006
Agenda
 FOR@C Research Consortium
 Forest products industry
 Supply chain planning challenges in the
forest products industry
 Supply chain scheduling: application to
the lumber industry
 FOR@C V-Lumber Experimental
Platform
 Agent-based simulation in supply chain
Basys’06, Niagara Falls, Ontario, Canada, 2006
Mission of the Consortium
T o become a Canadian and
International centre of expertise in the
development of the knowledge and
skills required to integrate and optimize
value creation networks in the forest
products industry by taking advantage
of the potential of new technologies
and electronic business models.
Basys’06, Niagara Falls, Ontario, Canada, 2006
Partners
Basys’06, Niagara Falls, Ontario, Canada, 2006
Supply chain
Basys’06, Niagara Falls, Ontario, Canada, 2006
Forest products supply chain
Basys’06, Niagara Falls, Ontario, Canada, 2006
Canadian Industry Snapshot
 3% GDP
 Exports for 45 billion $ of lumber, pulp
and paper every year
 Contributing 60% to the net export of
Canada
 900 000 direct and indirect jobs
 More than 350 localities depend
economically on the industry
Source: FPAC, March 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
Québec
80% is public land
 The forests of the province
of Quebec cover
750 000 km², that is the
equivalent of Sweden and
Norway combined.
 It counts for 20 % of
forested land in Canada
and 2 % of all the world’s
forests.
 This is why the vast
majority of foreigners see
Quebec as a huge green
carpet.
Basys’06, Niagara Falls, Ontario, Canada, 2006
Fiber flow
Basys’06, Niagara Falls, Ontario, Canada, 2006
Fiber transformation
Basys’06, Niagara Falls, Ontario, Canada, 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
customers customers
customers
Forest supply chain
Pulp and paper supply chain
Basys’06, Niagara Falls, Ontario, Canada, 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
Transportation in the supply chain
Basys’06, Niagara Falls, Ontario, Canada, 2006
Supply chain planning
challenges in the forest
products industry
Basys’06, Niagara Falls, Ontario, Canada, 2006
 In the United States at December 31, 2005,
the Company operated 23 pulp, paper and
packaging mills, 93 converting and
packaging plants, 25 wood products
facilities, six speciality chemicals plants and
270 distribution branches.
Top 5
International paper (~$26 B)
Weyerhaeuser (~ $20 B)
Georgia Pacific (~ $20 B)
Stora Enso (~ $15 B)
Kimberly Clark (~ $15 B)
PWC – Global Forest and Paper Industry Survey 2005
Basys’06, Niagara Falls, Ontario, Canada, 2006
Domtar supply chain
Merchants
Converters
Mills
Satellite Warehouses
Distribution Centers
Ship to points
Basys’06, Niagara Falls, Ontario, Canada, 2006
Harvesting/procurement plan
2006
2007
2008
Sustainable development
Road construction
Mixed of products, uneven aged
Plantation
Basys’06, Niagara Falls, Ontario, Canada, 2006
Alternative divergent processes
 Trees are cut to produce a set of logs
 Logs are cut to produce a set of lumbers
 Chips are mixed to produce different grades of
pulp and paper
 Rolls are cut to produce a set of rolls or sheets
Recipe/cutting pattern
Recipe/cutting pattern
Recipe/cutting pattern
Productivity not always linear
Sequence dependent set-ups
Basys’06, Niagara Falls, Ontario, Canada, 2006
Attribute based
products
Commodity Price Trends
N. American Consumption/Real GDP
Global Consumption/Real GDP
140.00
140.00
130.00
130.00
120.00
120.00
110.00
110.00
100.00
100.00
90.00
90.00
.
80.00
.
80.00
70.00
70.00
60.00
60.00
50.00
50.00
Source: RISI, CIBC World Markets
Containerboard
New sprint
2002
1998
1994
1990
1986
1982
1978
1974
UFS
2004
2003
2002
2001
2000
1999
1998
1997
1996
New sprint
40.00
1970
Containerboard
1995
1994
1993
1992
1991
1990
1989
1988
1987
40.00
P&W
Source: RISI, CIBC World Markets
• In North America, the link between consumption and real GDP is falling for all the
major grades of paper, but worst for newsprint.
• Even globally, the link between consumption and real GDP plateaued in the mid1990s.
Source; Roberts, 2005, Vision 2015 FOR@C
Basys’06, Niagara Falls, Ontario, Canada, 2006
Demand/supply propagation
Mix of spot market and contracts
Facilities
Basys’06, Niagara Falls, Ontario, Canada, 2006
Markets
Advanced Planning System for the
Pulp and Paper Industry (APS-PPI)
Basys’06, Niagara Falls, Ontario, Canada, 2006
Distributed planning systems
Top level
planning problem
Anticipation
functions
Reaction
RE*
Instructions
Final
Set of
decisions
IN*
Anticipation model of
the base level
planning problems
Instruction
IN*
Base level planning model
Schneeweiss (2003)
Basys’06, Niagara Falls, Ontario, Canada, 2006
Supply chain scheduling:
application to the lumber
industry
Basys’06, Niagara Falls, Ontario, Canada, 2006
Scheduling
 Decide what to do, when to do it
and how to do it
 Support mixed mode: Pull & Push
– Satisfy demand (committed orders & contracts)
– Maximize throughput value
 Constraints:
– Planned available inventory
– Machine capacity (potential bottlenecks)
Basys’06, Niagara Falls, Ontario, Canada, 2006
The lumber supply chain
Customers
Customers
Customers
Basys’06, Niagara Falls, Ontario, Canada, 2006
Log Requirement
Basys’06, Niagara Falls, Ontario, Canada, 2006
Sawing Line Plan
Solved using
mathematical
programming
(MIP or LP)
Basys’06, Niagara Falls, Ontario, Canada, 2006
Sawing
Cutting Pattern #9
Type 1
2x3
Cutting Pattern #25
2x4
2x6
Type 2
1x6
Cutting Pattern #12
Type 3
Cutting Pattern #57
Basys’06, Niagara Falls, Ontario, Canada, 2006
Drying Plan
Solved using
a constraint
programming
model
Basys’06, Niagara Falls, Ontario, Canada, 2006
Drying
Different Loading Patterns
(products distribution)
Green
Kiln
Kiln
Drying
Dried
Different Drying Process
Kiln
Kiln
Air Drying
Kiln
Kiln
Kiln Drying
Kiln
Kiln
Yard
Equalizing
Basys’06, Niagara Falls, Ontario, Canada, 2006
Finishing Line Plan
Solved using
heuristics
Basys’06, Niagara Falls, Ontario, Canada, 2006
Finishing
 Co-Products Management:
– Finishing 1 product type can results in 11 different product types
simultaneously
– All of them can have demand: they are not by-products
 Campaign Optimization (Setup management)
96 “
925/8 “
88 “
96 “
925/8 “
88 “
11,71 %
4,93 %
5/8
96 “
92 “
Premium
Premium
Stud
Premium
Stud
No 3
Stud
No
3
Economy
72 “
72 “
88 “
72 “
11,71
4,93 %10,68 -%
35,70
% % 14,15
6,81 -%
%%
4,93
% %
- 6,81 % 14,15
35,7011,71
%
10,68
%
6,81 %
35,70 %
14,15 %
10,68 %
6,81 %
6,81
4,49
% % - 2,68 -%
1,61 -%
4,49
0,50
%
No(TH
3 > 19 %)
KilnEconomy
wet
0,50
KilnEconomy
wet (TH > 19 %)
Kiln wet (TH > 19 %)
% - %6,81
--2,68 %
%
4,49 %
-
0,50 %
-
-
-- 1,61 %
2,68 %
-
1,61 %
-
Basys’06, Niagara Falls, Ontario, Canada, 2006
Shipment Orders
Solved using
a linear
programming
model
Basys’06, Niagara Falls, Ontario, Canada, 2006
Integration and system
 Simple integration
dynamics order
 Limited information exchanged
 Impact of the bullwhip effect
 Minimum return – local optimisation
Decentralised
Supplier
Production
site
Warehouse
Sales
material
Centralised
Planning centre
 Multi-site integration
 Standardisation of exchanges and management objectives
 Global optimisation
 Large quantity of information (collect and maintain)
Production
 Transactional
technologies available
Supplier
Warehouse
Sales
Sitereturn – but little success
 Great potential
material
Basys’06, Niagara Falls, Ontario, Canada, 2006
Planning challenges
Global Performance of the
entire supply chain network
(avoid local optimum et information
distortion)
Synchronization of decisions
Operation plans feasibility
(avoid plans that are not feasible)
Specialization of decisions
models and algorithms
Manufacturing and logistic Agility
(ability to re-plan quickly)
Decisions distribution and localization
where events must be managed
Basys’06, Niagara Falls, Ontario, Canada, 2006
Raise the needs for tools
designed
 To evolve in a decentralized, dynamic and
specialized environment
 To support demand and supply propagation with
optimization (e.g. revenue management)
 To integrate real-time execution information
(e.g. event management systems, contingency
planning)
 To support collaboration (e.g. collaborative
workflows)
Basys’06, Niagara Falls, Ontario, Canada, 2006
FOR@C V-Lumber
Experimental Platform
Basys’06, Niagara Falls, Ontario, Canada, 2006
Distributed & Specialized Tools
Basys’06, Niagara Falls, Ontario, Canada, 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
Supply Chain Planning
Agents
Data
Analysis
Tools
Tactical
planning unit
Planning Unit
Planning Unit
Planning Unit
Source
Agent
Deliver
Agent
Source
Agent
Make
Agent
Deliver
Agent
Make
Agent
Source
Agent
Deliver
Agent
Make
Agent
Basys’06, Niagara Falls, Ontario, Canada, 2006
Agent Architecture
Basys’06, Niagara Falls, Ontario, Canada, 2006
Conversation
Conversation
Conversation
Need
Need
Offer
Need
Offer
Offer Accepted
Offer
Offer Refused
Offer Accepted
Offer Refused
Offer Accepted
Offer Refused
Event
Event
Supplier Agent
Event
New Customer
Demand
New Supplier
Demand
New Supplier
Supply
Workflow
Workflow
Workflow
Event
Customer Agent
New Customer
Supply
Engins en approvisionnement fini
Engins en approvisionnement fini
Allocations
Engins en approvisionnement fini
Allocations
Engins en approvisionnement infini
Allocations
Engins en approvisionnement infini
Allocations
Engins en approvisionnement infini
Allocations
Allocations
Planning
© FOR@C – experimental platform
Basys’06, Niagara Falls, Ontario, Canada, 2006
Definition of collaboration
 An intended cooperative action
between two or more entities that
exchange or share resources in order
to take decisions or pursue an activity
that will generate benefits or loss that
are to be shared.
From an intra-organizational perspective
all resources can be view as shareable resources
D’Amours et Frayret (2003)
Basys’06, Niagara Falls, Ontario, Canada, 2006
Concepts of collaboration
 Main characteristics of inter-organizational
collaboration (from literature):
– Common goals and objectives, shared or jointly decided
Jacobs (2002)
– Implication of decision makers
–
–
–
Three important dimensions
of collaboration :
Mutual trust
Humain
Jacobs (2002)
Organisationnal
Through organisational structures
(strategy & process)
Pollard (2002)
Technology
Shared operation planning and execution
Pollard (2002)
Simatupang and Sridharan (2002), Jacobs (2002), Schrage (1990)
– Sharing of risk, rewards and responsibilities
Lambert and al. (1999)
– Be more efficient, get a competitive advantage
Simatupang and Sridharan (2002), Lambert and al. (1999), Pollard (2002)
Basys’06, Niagara Falls, Ontario, Canada, 2006
Concepts of collaboration
Nature of
exchanges
complex
Co-evolution
Collaborative operation
planning and execution
•Contracts & mechanisms
Joint planning
•Collaborative rules
•Allocation
Information exchange
•Pricing
relationship
•Incentives…
•Local & collective goals
Transactionnal relationship
•Information & decision
technologies
•Protocols & workflows
simple
weak
Intensity of the collaboration
strong
Frayret, D’Amours and D’Amours 2003
Basys’06, Niagara Falls, Ontario, Canada, 2006
Value of collaboration
 What to share? Information sharing
– Information
– Product
– Antitrust law
 How to share? Collaboration mechanism
– Minimum cost solution
– Equal % of benefit (e.g. Shapley value, Nucleus, externalities, etc.)
– Equilibrium in between?
 How to motivate? Contract and incentive designs
– Premium
– Volume guarantee
Basys’06, Niagara Falls, Ontario, Canada, 2006
Strategic game
 Precisely, a strategic game consists
of
– a set of players
– for each player, a set of actions (sometimes
called strategies)
– for each player, a payoff function that gives the
player's payoff to each list of the players'
actions.
http://www.chass.utoronto.ca/~osborne/2x3/tutorial/SGAME.HTM
Basys’06, Niagara Falls, Ontario, Canada, 2006
Retailer
Wood
Complex
Sawing
Wholesaler
Wood
FOREST
The wood
supply game
Retailer
Paper
Wholesaler
Paper
Saw Mill
Satisfy demand
Minimize inventory
Basys’06, Niagara Falls, Ontario, Canada, 2006
There is always an
equilibrium where players
demonstrate collaborative
behavior.
This equilibrium is almost
always as good as the
minimum cost solution.
Moyaux
et al. 2004 order transmission
1.
Traditional
2. Decoupled demand/order transmission
3. Real-time end customer demand transmission
Basys’06, Niagara Falls, Ontario, Canada, 2006
Moving toward collaboration
 Order based relationship
 Continous replenishment
– Transportation based
– Capacity based
 Vendor managed Inventory
 Collaborative planning, forecasting
and replenishment
Basys’06, Niagara Falls, Ontario, Canada, 2006
Agent-based simulation in
supply chain
Basys’06, Niagara Falls, Ontario, Canada, 2006
Knowledge-based supply chain planning
systems
Forget et al. 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
Multi-behavior agent
Forget et al. 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
Basys’06, Niagara Falls, Ontario, Canada, 2006
Concluding remarks
 Building the agent-based
simulation ability will permit to
model and test emerging supply
chain planningTechnical
approaches
in
a
challenges
dynamic, distributed,
specialized
Event management
delay
and stochasticDecision
environment.
Execution up-date
Players behaviours
Debugging
Basys’06, Niagara Falls, Ontario, Canada, 2006
Thank you
www.forac.ulaval.ca
Basys’06, Niagara Falls, Ontario, Canada, 2006