Description
The scope of this thesis is modelling and solving large-scale planning problems in the supply chain within the forest industry. Five research papers are included, the first three of which focus on the modelling, and the last two on the solution methods.
Link¨ oping Studies in Science and Technology. Dissertations
No. 1105
Supply chain optimization
in the forest industry
Helene Gunnarsson
Division of Optimization
Department of Mathematics
Link¨ opings universitet, SE-581 83 Link¨ oping, Sweden
ISBN 978-91-85831-85-2 ISSN 0345-7524
Copyright c 2007 Helene Gunnarsson unless otherwise noted
ISBN 978-91-85831-85-2 ISSN 0345-7524
Printed by LiU-Tryck, Link¨ oping, Sweden 2007
Acknowledgements
First of all, I would like to give special thanks to my supervisor Mikael R¨onnqvist
for giving support, inspiration and guidance. Your positive attitude and enthu-
siasm have meant much to me.
I would also like to thank my co-supervisor Jan Lundgren for your support,
encouragement and guidance in writing.
Thanks to Dick Carlsson at S¨odra Cell AB for very good cooperation and friend-
ship.
I am grateful to Torbj¨ orn Larsson, for always taken time for discussions con-
cerning both research and teaching.
I appreciate my friendship with Maud G¨othe-Lundgren.
Thanks to Mathias Henningsson for good cooperation in teaching and J¨ orgen
Blomvall for helping me with computer problems. Thanks to Kaj Holmberg for
interesting discussions about decomposition methods. Thanks to Oleg Burdakov
and Sven Erlander for encouragement.
The group of PhD students has been important for me. Thanks to Maria
Daneva, Elina R¨onnberg, Per-
?
Ake Andersson and Kristian Lundberg for sup-
porting me and for all fun discussions. Thanks also to my former colleagues, I
miss you all.
Thanks to David Bredstr¨ om for generating the routes in the second paper.
Thanks also to Anders Folkesson, Kent Hjelm, and Bertil Thunstr¨ om at Sydved
Energileveranser AB for good cooperation.
Thanks to Pamela Vang and Bj¨ orn Lidestam for helping me improving the
English.
Thanks to Monika Gustafsson and the other administrative sta? who always
are willing to provide.
Many thanks to my parents Ingemar and Britta and my sister Jennie and her
family for their support and concern about my well being.
Last, but not least, I would like to express my deepest gratitude to Bj¨ orn (my
husband-to-be and best friend) and our lovely children Adina (the actress),
Gunnar (the mathematician) and Martin (the humorist).
III
IV
Abstract
The scope of this thesis is modelling and solving large-scale planning problems
in the supply chain within the forest industry. Five research papers are included,
the ?rst three of which focus on the modelling, and the last two on the solu-
tion methods. All problems included are tactical multi-commodity problems
expressed as mixed integer programming (MIP) models. The work has been
done in collaboration with two Swedish companies within the forest industry.
In Paper I, a problem concerning the supply chain of forest fuel for Sydved
Energileveranser AB is modelled and solved. We study the problem of deciding
when and where forest residues are to be converted into wood chips, and how
the residues and chips are to be transported and stored in order to satisfy energy
demand at heating plants. The company has long-term contracts with forest
owners and saw mills. Decisions in the model include whether or not additional
harvest areas and saw mills are to be contracted and which terminals to use.
The planning horizon is one year and monthly time periods are used.
Papers II–V are based on planning problems at S¨ odra Cell AB. The planning
horizon is normally one year. Papers II–III consider only one time period.
In Paper II the supply chain from pulp mills to customers is modelled and
the combined problem of deciding terminal locations and which ship routes to
use is studied. Shipping vessels chartered on short or long term are used to
transport products to terminals in Europe. From each terminal, the products
are transported to customers by truck, train, or a combination of both. In
addition, trains and trucks can be used for transports directly to customers from
mills. In Paper III the entire supply chain, from harvest areas to customers, is
considered. Decisions included are transportation of raw materials, production
mix, the distribution of pulp products, and the selection of potential orders and
their quantities at customers. The ship routes are considered as ?ow links.
In Papers IV–V the problems in Papers II–III are combined into one model
and several time periods are used. Lagrangian heuristics based on Lagrangian
decomposition are used as solution methods in both papers. In Paper IV, the
approach leads to subproblems for each time period, whereas in Paper V, an-
other approach that results in subproblems for di?erent parts of the supply
chain is developed.
All models are based on real data from the companies. The models are detailed
and describe the problems accurately. The solution methods are developed such
that the solution time is kept within practical limits. Results from Papers II–
III have been used by S¨odra Cell AB to support the change of the terminal
structure as well as in budget planning.
V
VI
Optimering av logistikproblem inom
skogsindustrin
Sammanfattning
Denna avhandling presenterar matematiska modeller och l¨ osningsmetoder f¨or
optimering av olika logistikproblem inom skogsindustrin. Vi studerar f¨ ors¨orj-
ningskedjor f¨ or skogsbr¨ansle och massaproduker, och beaktar den ?arliga planer-
ingen i syfte att optimera ?¨ odet.
Det f¨ orsta problemet behandlar skogsbr¨ ansle och ¨ar ett samarbete med Sydved
Energileveranser AB. R?amaterial i form av grenar och toppar fr?an avverk-
ningsplatser ska ?isas och transporteras till v¨ armeverk, eventuellt via termi-
naler. Det ?nns m¨ojlighet att ?isa b?ade i skogen och p?a terminaler. Biproduk-
ter fr?an s?agverk kan ocks?a anv¨andas som r?amaterial. Vid behov kan utbudet av
r?amaterial ut¨ okas genom att ?er avverkningsplatser och s?agverk kontrakteras.
V¨armeverken har en efterfr?agan, angiven i kWh per m?anad, som ska uppfyllas.
Exempel p?a beslut som ska tas ¨ ar var ?isning ska ske, om nya avverkningsplatser
ska kontrakteras, var lagring ska ske, samt hur och n¨ar skogsbr¨anslet ska trans-
porteras.
N¨astf¨oljande problem behandlar massaprodukter och ¨ ar ett samarbete med
S¨ odra Cell AB. Olika sorters massaved fr?an skogen och biprodukter fr?an s?agverk
utg¨ or r?amaterial f¨ or produktion av massaprodukter. R?amaterialet transporteras
till massabruk f¨ or tillverkning enligt speci?cerade recept. De f¨ardiga produk-
terna transporteras sedan med fartyg till terminaler i Europa. Fr?an terminalerna
transporteras produkterna vidare till pappersbruk, vilka ¨ ar f¨ oretagets slutkun-
der. Massaprodukterna transporteras i vissa fall med lastbil eller t?ag direkt fr?an
massabruken till kunderna. Efterfr?agan ¨ ar angiven inom vissa gr¨anser i olika
order. Vissa order ¨ar fasta, vilket inneb¨ ar att dess efterfr?agan m?aste uppfyllas,
medan andra order ¨ ar fria. Exempel p?a beslut som ska tas ¨ar vilka bruk olika
produkter ska produceras p?a, hur m?anga och vilka terminaler som ska anv¨andas,
samt hur transporterna ska utf¨ oras f¨ or att ge b¨ asta resultat.
Utifr?an ovanst?aende beskrivningar har matematiska modeller formulerats. Ge-
nom att l¨ osa dessa kan vi f?a svar p?a logistik- och transportfr?agorna och ett
optimalt ?¨ ode kan hittas. F¨ or att l¨ osa modellerna har kommersiell program-
vara anv¨ants. Heuristiker och mer avancerade optimeringsmetoder har ocks?a
utvecklats i syfte att producera bra l¨ osningar snabbare.
VII
VIII
Papers included
The following ?ve papers, which will be referred to in the text by Roman nu-
merals, constitute the basis for the thesis.
Paper I: Supply chain modelling of forest fuel
Helene Gunnarsson, Jan Lundgren, and Mikael R¨ onnqvist
European Journal of Operational Research, 158 (2004) 103–123.
Paper II: A combined terminal location and ship routing problem
Helene Gunnarsson, Mikael R¨ onnqvist and Dick Carlsson
Journal of the Operational Research Society, 57 (2006) 928–938.
Paper III: Integrated production and distribution planning for
S¨ odra Cell AB
Helene Gunnarsson, Mikael R¨ onnqvist and Dick Carlsson
Journal of Mathematical Modelling and Algorithms, 6 (2007)
25–45.
Paper IV: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on time
periods
Helene Gunnarsson and Mikael R¨ onnqvist
Manuscript in preparation.
Paper V: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on physical
stages
Helene Gunnarsson and Mikael R¨ onnqvist
Manuscript in preparation.
IX
X
Contents
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V
Sammanfattning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII
Papers included . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX
Introduction and overview
Supply chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Forest supply chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Planning levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Optimization of supply chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
The Papers included
Overview of the Papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Paper I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Paper II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Paper III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Paper IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Paper V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Suggestions for future research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Paper I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Paper II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Paper III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Paper IV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Paper V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Introduction and overview
Introduction and overview
We start by giving a short introduction to the ?eld of supply chains. Thereafter
the forest supply chain is in focus and di?erent planning levels are introduced.
Then some approaches of optimizing the supply chain are presented. Last the
papers included in this thesis and their contributions are described.
Supply chains
A supply chain starts with the supply of raw materials for the products, passes
through one or several steps for manufacturing, storing and distribution, and
ends at the customer for the ?nal products. Examples of supply chains are often
presented in a network form. A simple example of a supply chain is given in
Figure 1.
Figure 1: Illustration of a general supply chain.
Since the 1990’s supply chain management (SCM) has been very popular and
useful concept in industrial planning. The SCM can be viewed from di?erent
perspectives, see O’Brien et al. [54]. One perspective is the industrial organi-
zation economics considering market structure and its importance. The other
perspective is the analytic modelling of supply chains focusing on ?nding good
planning measured for example in costs. An attempt to combine these perspec-
tives in order to bene?t the overall planning of the SCM is presented in O’Brien
et al. [54].
The area of SCM can also be divided into two parts, the integration of organi-
1
Introduction and overview
zational units and the coordination of ?ows, see Stadtler [73]. The aim of both
parts is to provide customer service and to be competitive in order to increase
pro?ts. In order to make the coordination of ?ows more e?cient, advanced
planning systems (APS) are used. Ingredients for operating the supply chain
regarding the coordination of ?ows, successfully, can be, for example, operations
research (OR) and logistics, see Stadtler [73].
A basic description of supply chain modelling is found in Shapiro [69]. A more
detailed description of industrial cases can be found in Stadtler and Kilger [74].
Examples of application in the textbook are a case of computer assembly by
Kilger [42] and a case about food and beverages by Wagner and Meyr [77].
The e?ciency of the logistics operations e?ciency is crucial for the successful
management of a supply chain. It is important to view and understand the
overall supply chain to be able to make decisions that can imply, for example,
shorter lead times and less need of safety stocks. A survey of SCM with regard
to Swedish manufacturing ?rms, can be found in Olhager and Selldin [55]. The
paper investigates the supply chain management strategies and practices in a
number of companies.
We will from now only consider the supply chain including ?ow of materials.
The focus will be on analytic supply chain modelling in general and forest supply
chains in particular.
Forest supply chains
An illustration of the supply chain valid for forest products in Sweden can be
studied in Figure 2, below.
The activities that have to exist in order to get the correct amount of raw mate-
rials from the forest without reducing future harvest possibilities, are planting,
cleaning, thinning, and harvesting. The assortments obtained from forest raw
materials can be classi?ed according to their use. Saw logs, pulp wood, and
forest residues are the major parts of the assortments. Each part can be further
divided into several subgroups according to their qualities and dimensions.
After harvesting, timber is transported to saw mills. The timber can also be
transported to terminals for storage before transportation to the saw mills. Sim-
ilar conditions hold for pulp wood, it is either transported directly to pulp mills,
or is taken via terminals for intermediate storing. The harvesting leaves some
forest residues in the form of tops and branches. The residues are left for around
a year in the forest and are then chipped. Then, after storage at terminals (if
necessary), the residues can be transported to heating plants. Byproducts from
saw mills, such as chips, are transported to pulp mills and heating plants for
further use. The making of pulp also produces byproducts, such as bark. These
byproducts are used as fuel at the pulp mill or at heating plants. The ?nal
pulp products are transported to paper mills to make paper products. The
last transportation (which is not included in the illustration), is transportation
2
Supply chain optimization in the forest industry
Figure 2: Parts of the forest supply chain.
from the saw mills, paper mills, and heating plants to the ?nal customer or to
distribution nodes. Excluded from the illustration above is also the transporta-
tion of wood ashes from heating plants back to the forest. The wood ashes are
dispersed over the ground to improve the quality of the soil.
Planning levels
There are di?erent planning levels, which depend mainly on the planning hori-
zon, to consider. Di?erent planning levels and horizons have di?erent needs for
supporting tools. Even if the number of computer based supporting tools is
increasing, most of the planning work is still done manually. These di?erent
planning levels are described, for example, in R¨onnqvist [64] and Weintraub
and Romero [80]. The strategic level often describes decisions that concern
several years or decades. Strategical decisions include investment planning and
infrastructure planning, for example, road building. Tactical planning is often
connected to annual decisions, but it could be for shorter or longer periods,
depending on the problems considered. Tactical decisions include, for example,
road upgrading, annual production planning related to the budget processing,
route structure and equipment utilization. The last level to be mentioned is
operative planning, which could span over a couple of months or involve daily
planning. Examples of operative decisions are detailed production planning and
truck dispatching. The di?erences in time for each level depending on the type
of problem are not clearly de?ned. We will now describe the planning levels
related to the supply chain in the forest industry. We start by harvest problems
3
Introduction and overview
and continue with routing problems. Finally, planning levels in production are
discussed.
Harvesting problems have other horizons regarding planning levels. Here, op-
erative planning includes decisions for months, tactical planning could be from
half a year to a few years, and strategical planning involves decisions for sev-
eral decades. The long horizons within harvesting of course are due to the long
rotation times of trees. MEDFOR, a planning tool for strategical decisions in a
horizon of 50 years, is described in Epstein et al. [20]. Another model for strate-
gical harvest planning can be found in Gunn and Rai [25]. A system based on
simulating conditions in forestry is Indelningspaketet, which is used in Sweden,
primarily for the strategical planning of harvesting over large harvest areas [39].
This system is intended to be an aid in decision operations, such as how much
to harvest each year, proportions of thinning and ?nal harvesting, and in which
order to work the areas. For making tactical decisions, the OPTIMED tool,
which is a tool for support harvest planning for a period of two to ?ve years can
be used, see Epstein [20]. Harvest scheduling, road access and adjacency con-
straints are considered simultaneously in Richard and Gunn [62]. They address
the tactical planning problem with a tabu search method. The design of the
tabu search method is further discussed in Richards and Gunn [63]. A harvest
planning model for the short term can be found in Karlsson et al. [40], where
a planning period of four to six weeks was considered and harvest schedules for
the di?erent harvest crews were decided. Other models for short term harvest
planning can be found in Epstein et al. [19]. A model integrating long term
harvest planning with short term harvest planning can be found in Nelson et
al. [52].
With regard to routing problems, the operative level often involves daily de-
cisions about the routes. The tactical level in routing problems can be from
two weeks up to a month, and the strategical level considers decisions of one
year or longer periods. In Weintraub et al. [78], an operative and computerized
system to support daily truck scheduling decisions is developed, and the system
has been implemented in some of the largest forest ?rms in Chile. Another
approach to the problem of optimizing daily transports within forestry is used
in Palmgren et al. [56], where the aim was to ?nd e?cient routes for all the
trucks involved. Several studies have been made within this area, for example,
R¨onnqvist and Ryan [65].
Within the area of production planning problems, the levels often depend on the
products. Usually, operative planning in production problems depends on the
use of working shifts, times, or daily planning. Tactical planning is often annual,
and the strategical level concerns several years. An integrated optimization
system of the bucking, sawing and planing processes, CustOpt, is presented in
Lid´en and R¨ onnqvist [46], where the aim was to get chain forestry - saw mill -
planing mill to work in a more customer oriented manner. CustOpt is a support
system on a tactical level with a three months planning horizon [46]. Problems
regarding the sawing of the logs at saw mills can be found in Todoroki and
4
Supply chain optimization in the forest industry
R¨onnqvist [75, 76]. The logs should be sawed in such a way that the amount
of residue is minimized. Trying to adjust the sawing with regard to customer
demand would to a great extent both save time and raw materials. Another
similar problem can be found in Reinders [61], where a decision support system
which focuses on tactical planning was developed.
Optimization of supply chains
In the past, ?rms tried to optimize parts of their supply chain separately. Nowa-
days approaches integrating di?erent parts of the supply chain have become
popular. The basic ideas behind most of the solution methods in the papers
described below, can be found in Wolsey [81].
One way of describing the supply chain is to focus on the di?erent activities
involved, such as production and distribution. Reviews of the integration of
production and distribution can be found in Sarmiento and Nagi [66] and Sel¸cuk
Ereng¨ u¸ c [68]. In Chang and Lee [10], a two-stage problem is de?ned, one prob-
lem for the manufacturer and one for the distributor. Two practical approaches
were applied to solve the logistic scheduling problem, the forward approach and
the backward approach. The problems are solved sequentially using a heuristic
method. Another heuristic method used for solving similar problems can be
found in Park [57]. A heuristic consisting of two phases is introduced. In the
?rst phase, a tentative production and distribution plan is established, while in
the second, an improvement of the plan is attempted. Integrated design method-
ology for production and distribution based on the Benders decomposition can
be found in Dogan and Goetschalckx [18] and Cordeau et al. [14]. Some types
of integrated production and distribution problems of reasonable sizes can be
solved directly using CPLEX, see Dhaenens-Flipo and Finke [17]. In Jayara-
man and Pirkul [38], a heuristic which utilizes linear programming techniques
as well as results from the Lagrangian relaxation procedure in order to solve
an integrated production and distribution problem, is presented. The supply
chain of a nursery company is optimized in Rantala [59], where a capacited
MIP problem for solving an integrated production-distribution problem for a
Finnish company is presented. The main decision to consider is if the company
should expand or closure its facilities. The original model is strategical, but by
applying applicable constraints, the model can also be used for solving tactical
and operational level problems.
Another way of describing the supply chain is to focus on di?erent kinds of
decisions at di?erent levels. In Shen [70], the strategical level is connected
to the supply chain design phase and after these kinds of questions have been
answered, the focus shifts to decisions at tactical about operational levels such as
inventory management, and decisions on supply and demand distribution. There
exist several integrated models of any two of the three important supply chain
decisions; location, routing and inventory [70]. A survey of models integrating
the location and routing decisions can be found in Salhi and Nagy [51]. They
5
Introduction and overview
propose a classi?cation scheme and look at both exact and heuristic algorithms.
Other reviews of the location-routing problem can be found in Balakrishnan et
al. [2] and in Min et al. [49]. Most of the earlier studies focus on heuristic
methods. A location-routing model which considers several kinds of products,
a multi-commodity model, is presented in Bookbinder and Reece [4].
The integration of inventory and routing (IR) involves inventory management
and vehicle routing decisions. A survey of IR models is provided by Kleywegt
et al. [43]. They present a classi?cation of the IR models based on the type
of demand, ?eet size, length of the planning horizon and the number of visited
demand points on a vehicle trip. The research contributions of the di?erent
models are also presented. Other more recent reviews can be found in Kley-
wegt et al. [44]. A logistical overview of inventory routing problems is given
in Moin and Salhi [50]. They classify the research according to the planning
horizon employed in the models; single period, multi period and in?nite hori-
zon. Christiansen and Nygreen [11, 12] used a column generation approach for
solving a combined ship routing and inventory problem. A real ship planning
problem, including both inventory management and routing with time windows
is presented in Christiansen [13] and solved by approaching a Dantzig-Wolfe
decomposition combined with branch-and-bound. Campbell and Savelsbergh
[6] presented a decomposition approach for an inventory-routing problem. A
two-stage approach based on the decomposition of the set of decisions and a
combination of branch-and-bound and heuristics were used as solution meth-
ods.
Models integrating location and inventory decisions can be found in Shen et al.
[71] and Daskin et al. [15]. The model presented in Daskin [15] is a non-linear
mixed integer programming model and Lagrangian relaxation is proposed as a
solution method together with heuristics in order to ?nd feasible solutions. The
model presented in Shen et al. [71] is solved using a column generation method.
A review of integrated location and inventory models of a smaller extent can be
found in Daskin et al. [16]. They also present a basic model for the integration
of location and inventory decisions.
The most common ways to express the objective function in supply chain mod-
els, are in cost and pro?t units. There are other options to express the objective
function. Relevant criteria to consider can be lead times and storage safety
levels. Li and O’Brien [45] presented a model which measured the quality of a
supply chain considering four criteria; pro?t, lead time performance, delivery
promptness and waste elimination. The model analyzes the performance at two
levels, the chain level and the operations level
A description of the supply chain in the forest from the perspective of Chilean
forestry ?rms can be found in Weintraub [79]. The use of models and computer
systems related to each physical stage in the chain is discussed as well as the
coordination between stages. In Epstein et al. [20] a more detailed description
of the OR system in the Chilean forest is presented. A general overview of OR
work within forestry can be found in Martell et al. [48] and R¨onnqvist [64]. In
6
Supply chain optimization in the forest industry
Weintraub and Romero [80] the OR models in forestry are compared to those
in agricultural. They concluded that more models managing the entire supply
chain in the forest are to be expected in the future. With regard to Swedish
conditions, several wood ?ow problems in forestry are described in Carlsson and
R¨onnqvist [9].
The supply chain of wood ?bre ?ow is studied in Carlsson et al. [7], where an
overview of the planning problems is presented in a matrix. A description of
OR tools used in practise is also provided as well as trends in the ?eld. Supply
chain management in the pulp and paper industry can be found in Carlsson et
al. [8]. The overall supply chain and its participants are described, and the
need for more information and decision support system is discussed.
Daily supply chain decisions regarding production and distribution of pulp prod-
ucts for the Swedish pulp company S¨odra Cell AB, are optimized in Bredstr¨om et
al. [5]. The planning horizon is three months and optimal production length for
each pulp product is decided as well as storage levels and distribution schemes.
A MIP model is developed, and then solved using branch and price, where the
subproblem is a shortest path problem.
In Philpott and Everett [58] a model for the optimization of the supply chain in
the paper industry is presented. The model is a large MIP model known as the
Paper Industry Value Optimization Tool (PIVOT). The model ?nds an optimal
allocation of supplier to mill, product to machines, and machines to customer,
and also the details in the supply chain are optimized.
Although many models have been developed in this area, there is still more
research to be done. The tactical models for the entire supply chain of forest
fuel and pulp products presented in this thesis contribute to the research ?eld
of optimization of supply chains.
The Papers included
In this section we start by giving an overview of the included papers. We will
continue by summarizing the papers in chronological order. Then the contri-
butions will be presented and ?nally some suggestions for future research are
presented.
Overview of the Papers
The main purpose of this thesis is to model large problems within the forest
industry, and to solve them in reasonable time using di?erent solution methods.
The creation and development of the mathematical models have in all papers
followed the operations research (OR) process. The OR area is wide and there
exists many books covering the subject, e.g. Bertsimas and Tsitsiklis [3], Wolsey
[81], Ahuja et al. [1] and Rardin [60]. The OR process is illustrated in Figure
7
Introduction and overview
3.
Figure 3: The OR process.
To de?ne a relevant problem has been an iterative process in close connection
with the planners at the companies. All the papers consider the supply chain
problem in the Swedish forest industry. Paper I describes the supply chain of
forest fuel and Paper II - V consider the supply chain of pulp products. The
problems discussed in the papers are multi-commodity problems, which means
that several products are included. The problems considered are on a tactical
level and the planning horizon is normally one year. After the problem is clearly
de?ned, a model is to be formulated.
In the modelling process several issues have to be taken into consideration.
Questions of how to express limitations and include decisions are addressed.
The models often have to be reformulated due to changed or updated conditions
regarding the problems. The models in all papers are large and detailed. The
modelling language AMPL is used for all model formulation in the papers. All
included models are mixed integer programming (MIP) models and they include
continuous and binary variables.
After the models are stated, all data is to be collected. This required time and
e?ort in all models due to their sizes and the di?culties to get the adequate
data. It is very important to make sure that the data is correct. The models
become useless if wrong data is used.
All of the included problems are solved using the commercial program CPLEX
directly or as a module in developed methods. In Papers I and II heuristics based
on the linear programming solution are developed. In Paper IV two variants of
a heuristic based on decomposition in time are used and ?nally in Paper V a
8
Supply chain optimization in the forest industry
heuristic based on decomposition of parts in the supply chain is used.
The large sizes of the models in Papers IV–V, motivated developing of more
advanced solution methods. For that reason, di?erent variants of Lagrangian
heuristics based on Lagrangian decomposition are developed. The idea of a
Lagrange relaxation is to relax constraints and move them together with a
multiplier to the objective function. General presentations and early papers
describing Lagrangian heuristics can be found in Geo?rion [23] and Fisher [22].
The Lagrangian decomposition method is presented in, for example, Guignard
and Kim [24] and N¨ asberg et al. [53]. The idea is to duplicate variables and
relax the equality constraints which say that they should have the same value.
This will lead to the division of the problem into subproblems; for each time
period in Paper IV and for each physical stage in the supply chain in Paper V.
The last step of the OR process is the analyzing of results. The reasons to
strange results can be several. The data can be incorrect or the model can have
been formulated in an inappropriate way. Conditions that no one involved in
the model process thought of can appear. It is important to discover the source
of unexpected results in order to be able to use the model in a correct way. The
results of Papers I–III have been analyzed to varying extent. The models in
Papers II and III have been used and developed at the company S¨ odra Cell AB.
An overview of the circumstances regarding the papers is presented in Table 1.
Di?erent properties that represent the problems in the papers are given. Ex-
amples of properties are conditions of the supply and the demand, time periods
included and the sizes of the problems. The aim with Table 1 is to make the
comparison of papers easier to the reader.
9
Introduction and overview
Table 1: Overview of the properties of the problems and models in the papers
included in this thesis.
Properties Paper I Paper II Paper III Paper IV Paper V
Company Sydved Energi- S¨odra S¨odra S¨odra S¨odra
leveranser AB Cell AB Cell AB Cell AB Cell AB
Product Forest fuel Pulp Pulp Pulp Pulp
Number of
products 6 30 18 5-19 5-19
Number of
time periods 12 1 1 4, 12 4, 12
Decomposition No No No Yes Yes
Heuristics Yes Yes No Yes Yes
Production Yes No Yes Yes Yes
Terminal
location Yes Yes No Yes Yes
Focus Model Model Model Algorithm Algorithm
Demand Fixed Fixed Within Within Within
limits limits limits
Eligible demand
contracts No No Yes Yes Yes
Eligible supply
contracts Yes No No No No
Objective Min. costs Min. costs Max. pro?t Max. pro?t Max. pro?t
Modes of Trucks Trucks Trucks Trucks Trucks
transportation Trains Trains Trains Trains
Vessels Vessels Vessels Vessels
Barges Barges Barges Barges
Routes No Yes No Yes Yes
Number of 63,122 11,309 37,165 71,289 – 71,289 –
constraints 274,217 274,217
Number of
binary
variables 10,201 2,449 174 675 – 825 675 – 825
Number of
continuous 494,911 422,796 2,019,384 370,811 – 370,811 –
variables 3,489,036 3,489,036
10
Supply chain optimization in the forest industry
Below, the papers included in this thesis are summarized. We start by describing
the problems and modelling issues. Then the sizes of the problems are given,
and ?nally the developed solution methods are presented.
Paper I: Supply chain modelling of forest fuel
The aim of this paper is to formulate and solve an optimization model describing
the planning problem for a company supplying forest fuel. The test data is from
the Swedish company Sydved Energileveranser AB. The planning horizon is one
year and the period is divided into 12 time periods (months).
The supply sources consist of a number of harvest areas and saw mills. They can
either be owned by the supplying company or be possible to contract. The forest
residues in contracted harvest areas can be made available if the company enters
into a contract with a supplier of residues. To contract a saw mill implies taking
care of byproducts in the form of sawdust, bark, and dry chips, continuously
over the year. After the harvesting, a lot of forest residues, consisting of forest
residues from hardwood and softwood, as well as decay damaged wood, are left
in the harvest area. The residues have to be forwarded and removed from the
area. Before using the residues as fuel, they have to be chipped. The chipping
can take place either in the forest or at terminals. Figure 4 gives an overview of
the problem. The division of the terminals into two sections is meant to show
the storage of non-chipped and chipped forest residues, respectively.
Figure 4: An illustration of the supply chain problem in Paper I.
The demand at the heating plant is speci?ed in kWh, in contrast to the supply
11
Introduction and overview
which is given in m
3
. Therefore, the forest fuel has to be converted into energy
values. The energy value depends, for example, on the moisture contents in the
forest residues.
The choices of chipping possibilities together with the conversion into energy
are new challenges, that are to be considered in the modelling processes. A
mixed integer programming (MIP) model is formulated, and then solved using
CPLEX. In addition, a heuristic method based on the LP solution is devel-
oped. The basic case includes 494,911 variables, whereof 10,201 are binary and
63,122 constraints. Several other cases, besides the basic case, are studied and
evaluated.
Paper II: A combined terminal location and ship routing problem
The aim of this paper is to formulate and solve a model, that includes both
location decisions and routings. This is a case study from S¨ odra Cell AB, one of
the world’s leading manufacturers of market pulp [72]. The problem described
is how to transport and distribute pulp products from pulp mills to customers
as cheaply as possible. Figure 5 gives an overview of the supply chain.
Figure 5: Illustration of the supply chain in Paper II.
Most of the pulp products are transported from a pulp mill to the nearest har-
bour and then further by ship to a harbour terminal outside Scandinavia. All of
the customers located in Scandinavia are supplied by train or truck transporta-
tion. Trains and trucks are also used to supply some of the customers outside
Scandinavia.
The pulp can be transported from the harbour terminals to inland terminals.
Barges are often used for this transportation, but trains and trucks can also
12
Supply chain optimization in the forest industry
be used. The transportation from the harbour and inland terminals to the
?nal customer is made by train and truck. Three vessels chartered long term,
so-called TC-vessels, are used. The routes used by the TC-vessels are divided
into direct routes, called A-routes, and composite routes, called B-routes. The
A-routes go from one pulp mill to one terminal and the B-routes visit several
places. In addition, vessels chartered short term, spot vessels, are used from a
pulp mill to one destination.
Due to the large size of this problem, only one time period is used. One of the
di?culties and challenges in modeling this problem is that we need to integrate
long term decisions on terminal usage with short term decisions on which routes
to choose. Some terminals are connected to few customers with small demands.
These terminals need to get deliveries throughout the year in order to satisfy
their orders. Only one direct delivery using a A-route is not practical. Since
we only have one time period we can model a continuous delivery through the
use to B-routes. The smaller the terminal is, the larger proportion of B-routes
is needed. On the other hand, if the terminal is large we can use A-routes with
direct deliveries from the mills. Since the B-routes are more expensive, we need
to link the proportion of B-routes against the size of the terminal. Otherwise
all ?ow would be using A-routes. Therefore di?erent levels on the terminals are
introduced. The levels are modeled through binary variables and only one level
can be chosen. The levels are coupled to the received ?ow at the terminal. High
levels mean large ?ow and more deliveries on A-routes are accepted. Otherwise,
B-routes, must be used in a particular proportion relating to the level.
The combined terminal location and ship routing problem is formulated in an
MIP optimization model. Five di?erent cases, which are modi?cations of the
basic case, have been tested and evaluated. The basic case includes 425,245
variables, whereof 2,449 are binary. The number of constraints included is
11,309. The solution approach is to use CPLEX directly and to use heuristics
based on the LP solution to ?nd a good solution in an acceptable time.
Paper III: Integrated production and distribution planning for S¨ odra
Cell AB
Paper III is related to Paper II. The problem this paper considers is the whole
supply chain from the forest districts to the customer. Di?erent kinds of raw ma-
terials originating from hardwood or from softwood, are transported by trucks
from the forest districts to the pulp mills. Pulp products are produced according
to di?erent recipes at the pulp mills. The composition and usage of raw mate-
rials are restricted within certain limits. The choice of using di?erent recipes at
di?erent pulp mills creates alternatives. A change over cost is connected to each
alternative and only one of the alternatives can be chosen. Another di?erence
from the Paper II is that all of the terminals are ?xed, meaning that we do not
include any choice of terminals. The reason why this problem includes fewer
terminals is that S¨odra Cell AB has changed their policy regarding the number
13
Introduction and overview
of terminals to use. The model does not include routes in the same sense as
before. In order to simplify the problem, the direct routes are in this paper
considered as direct ?ow without restrictions.
The demand, speci?ed in orders, is given within certain limits. The orders can
be ?xed, meaning that related demand have to be ful?lled, or the orders can
be free. Contracts are speci?ed by orders connected to them. The choice of
accepting a contract or not have to be made. Accepting a contract means that
the demand in all included orders have to be ful?lled within the given limits.
An overview of the problem is illustrated in Figure 6.
Figure 6: Illustration of the supply chain in Paper III.
An MIP model for the entire supply chain problem is presented and ?ve di?erent
alternatives are tested and evaluated. The complete problem, including all
alternatives, includes over 2,000,000 variables, whereof 174 are binary. The
number of constraints is 37,165. The model was used to support the annual
budgeting for 2005-2007.
Paper IV: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on time periods
The problem described in this paper is a combination of the problems considered
in Paper II and Paper III. The mathematical model developed includes all kinds
of routes, (A-routes, B-routes and spot trips), as well as terminal location and
production of pulp products. The components which have been added compared
to the previous models of the supply chain are the time periods and the storage
possibilities. The raw materials can be stored in the forest. To a limited extent,
The pulp products can be stored at the pulp mills and at the terminals. All of
the pulp products have to be distributed during the planning horizon. Di?erent
14
Supply chain optimization in the forest industry
demand structures are taken into consideration. The supply chain for S¨ odra
Cell AB in is illustrated in Figure 7.
Figure 7: Illustration of the supply chain in Paper IV.
The focus in this paper is on the solution methods. Due to the large size of this
problem, more advanced methods are needed to be able to solve the problem
within reasonable time limits. The aim of this paper is therefore to develop
a Lagrangian heuristic method based on Lagrangian decomposition for solving
the mathematical model. The problem is decomposed into one subproblem per
time period. In addition, two variants of a heuristic based on LP solutions, are
developed. Cases including di?erent numbers of variables and constraints are
tested and evaluated. The number of binary variables included is 675–825 and
the number of continuous variables is 370,811–3,489,036. The included number
of constraints ranges from 71,289 to 274,217.
Paper V: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on physical stages
The problem in Paper V is the same as in Paper IV. A Lagrangian heuristic
based on Lagrangian decomposition is used as well. The aim of this paper is to
decompose the supply chain in another way compared to Paper IV, in order to
improve solutions and solution times. The problem in this paper is decomposed
into one subproblem per physical stage in the supply chain. In addition a
heuristic based on division into physical stages are developed. Cases included
di?erent numbers of variables and constraints are tested and evaluated. The
included number of binary variables is 675–825 and the number of continuous
variables is 370,811–3,489,036. The included number of constraints is 71,289–
274,217.
15
Introduction and overview
Contributions
My contribution to the research presented in this thesis includes major involve-
ment in the modelling process and in writing the papers. I have developed and
implemented the Lagrangian relaxation algorithms in the Papers IV and V as
well as the heuristics in all of the papers. I have also taken part in the data
collection. Mikael R¨onnqvist and Jan Lundgren have participated in the work
as my supervisor and co-supervisor, respectively. Dick Carlsson, industrial lo-
gistics manager at S¨ odra Cell AB, complemented the work with the important
industrial approach, mainly in Paper II and Paper III. Results from these papers
are already used in practice by S¨ odra Cell AB.
The papers have previously been presented by me at conferences in Budapest
[28], Aspen [29],
?
Are [30], V¨axj¨ o [37], Copenhagen [36], [34], Link¨ oping [31],
Molde [32], Reykjavik [33] and Pittsburgh [35]. Some of the papers have also
been presented at internal seminars in the Optimization group in Link¨ oping.
Paper I and an earlier version of Paper II were presented at my licentiate sem-
inar in February 2004 [26]. A condensed version of Papers IV and V will be
presented at the EURO XXII conference on operational research in Prague in
July 2007, [21]. Other related research not included in this thesis can be found
in Gunnarsson et al. [27] and Karlsson et al. [41].
The main contributions of each paper are summarized below:
Paper I
• A new model describing the processes of handling forest fuel.
• Solution of the model using both a commercial solver and a new
developed heuristic.
• Analyzing the model using real data.
Paper II
• A new model combining terminal location and ship routing.
• Solution of the model using a commercial solver and two new developed
heuristics.
• Analyzing the model using real data.
• The result of the paper has been used in practice by S¨ odra Cell AB
Paper III
• A new model describing the entire supply chain for a pulp company.
• Solution of the model using a commercial solver.
• Analyzing the model using real data.
• The result of the paper has been used in practice by S¨ odra Cell AB.
Paper IV
• Development of a Lagrangian heuristic method based on time periods.
• Two variants of a rolling horizon heuristic are developed.
• Analyzing the solution method using real data.
16
Supply chain optimization in the forest industry
Paper V
• Development of a Lagrangian heuristic method based on physical stages.
• A heuristic based on physical stages is developed.
• Analyzing the solution method using real data.
Suggestions for future research
New business models have been developed recently. One example is vendor
managed inventory, VMI, which is used to a growing extent. The meaning of
VMI, is that the distributor takes full responsibility for maintaining an agreed
inventory of the products, usually at the buyer´s location. Including this into the
models, makes them even more large, and motivates developing of new solution
methods to be able to solve them. That is a challenge and it is new change of
getting better control and making larger pro?ts for the involved companies.
The model for the supply chain concerning forest residues can be modi?ed in
order to be used for other kinds of biomass fuels. Since this study was completed
in 2001, the prices of forest fuel used at heating plants have increased by almost
23% [67]. It is, therefore, more than ever motivated for suppliers of forest fuel to
use the proposed model in order to be able to make the planning. An overall goal
to use more bioenergy fuel together with the lack of forest fuel at some areas,
also increase the interest of the presented model. Due to the higher prices on
forest fuel, the pulp wood can in some ways work as a substitute for forest fuel.
That can lead to strange situations and does of course impact the pulp market.
The eventual con?icts between the energy sector and the forest industries are
discussed in Lundmark [47]. He also tries to ?nd out the break-point where it
becomes more pro?table to use round wood instead of forest residues as fuel. It
could be interesting to include round wood in the presented model together with
an increased demand. Then several scenarios could be tested in order to get
more solution alternatives if the prices for forest fuel change dramatically. To
create a mathematical model including the supply chains, or parts of the supply
chains, for both forest fuel and pulp products would be a new and exciting
challenge.
The models for the supply chain problems concerning pulp products presented
in this thesis are complex and detailed, which motivates developing of more
solutions methods in order to get good enough solutions faster. In addition to
the Lagrangian heuristics method, other kinds of decomposition methods can
be used, for example the Bender decomposition method. The convergence of
the methods can probably be improved by adding and relaxing some aggregated
constraints, especially in Paper IV. The constraints can also be scaled in order
to make the values of the multipliers more alike. Another way to improve the
solution times is to change the modelling language from AMPL for example to
C.
Developing of decision support system, DSS, for the problems in the included
17
Introduction and overview
papers in this thesis, would enhance the possibilities for practical use of the
presented models.
18
Supply chain optimization in the forest industry
References
[1] R. Ahuja, T. Magnanti, and J. Orlin, Network Flows, (1993) Prentice Hall,
New Jersey.
[2] A. Balakrishnan, J.E. Ward, and R.T. Wong, Integrated facility location
and vehicle routing models: Recent work and future prospects, American
Journal of Mathematical and Management Sciences 7 (1987) 35–61.
[3] D. Bertsimas and J. Tsitsiklis, Introduction to Linear Optimization, (1997)
Athena Scienti?c, Belmont.
[4] J.H. Bookbinder and K.E. Reece, Vehicle routing considerations in distri-
bution system design, European Journal of Operational Research 37 (1988)
204–213.
[5] D. Bredstr¨om, J.T. Lundgren, R. R¨ onnqvist, D. Carlsson, and A. Mason,
Supply chain optimization in the pulp mill industry – IP models, column
generation and novel constraint branches, European Journal of Operational
Research 156(1) (2004) 2–22.
[6] A.M. Campbell and M.W.P. Savelsbergh, A decomposition approach for
the inventory-routing problem, Transportation Science 38(4) (2004) 488–
502.
[7] D. Carlsson, S. D’Amours, A. Martel and M. R¨ onnqvist, Decisions and
methodology for planning the wood ?bre ?ow in the forest supply chain,
submitted to a CEMS book.
[8] D. Carlsson, S. D’Amours, A. Martel and M. R¨ onnqvist, Supply chain
management in the pulp and paper industry, Working paper DT-2006-AM-
3, Interuniversity Research Center on Enterprise Networks, Logistics and
Transportation (CIRRELT), Universit´e Laval, Qu´ebec, G1K 7P4, Canada,
2006.
[9] D. Carlsson and M. R¨onnqvist, Wood ?ow problems in the Swedish forestry,
The Forestry Research Institute of Sweden Report no. 1. (1999).
[10] Y.-C. Chang and C.-Y. Lee, Logistics scheduling: Analysis of two-stage
problems, Journal of Systems Science and Systems Engineering 12(4)
(2003) 385–407.
[11] M. Christiansen and B. Nygreen, A method for solving ship routing prob-
lems with inventory constraints, Annals of Operations Research 81 (1998)
357–378.
[12] M. Christiansen and B. Nygreen, Modelling path ?ows for a combined
ship routing and inventory management problem, Annals of Operations
Research 82 (1998) 391–412.
19
References
[13] M. Christiansen, Decomposition of a combined inventory and time con-
strained ship routing problem, Transportation Science 33(1) (1999).
[14] J.F. Cordeau, F. Pasin, and M.M. Solomon, An integrated model for logis-
tics network design, Annals of Operations Research 144(1) (2006) 59–82.
[15] M.S. Daskin, C.R. Coullard, and Z.-J.M. Shen, An inventory-location
mode: Formulation, solution algorithm and computational results, Annals
of Operational Research 110 (2002) 83–106.
[16] M.S. Daskin, L.V. Snyder, and R.T. Berger, Facility location in supply
chain design. In: A. Langevin and D. Riopel (Eds.), Logistics systems:
Design and Optimization, (2005) Springer, 39–65.
[17] C. Dhaenens-Flipo and G. Finke, An integrated model for an industrial
production-distribution problem, IIE Transactions 33 (2001) 705–715.
[18] K. Dogan and M. Goetschalckx, A primal decomposition method for the in-
tegrated design of multi-period production-distribution systems, IIE Trans-
actions 31 (1999) 1027–1036.
[19] R. Epstein, E. Nieto, A. Weintraub, P. Chevalier, and J. Gabarr´ o, A sys-
tem for the design of short term harvesting strategy, European Journal of
Operational Research 119 (1999) 427–439.
[20] R. Epstein, R. Morales, J. Ser´ on, and A. Weintraub, Use of OR systems in
the Chilean forest industries, Interfaces 29(1) (1999) 7–29.
[21]http://euro2007.vse.cz/, retrieved March 22 (2007).
[22] M.L. Fisher, The Lagrangian relaxation method for solving integer pro-
gramming problems, Management Science 27(1) (1981) 1–18.
[23] A.M. Geo?rion, Lagrangean relaxation for integer programming, Mathe-
matical Programming Study 2 (1974) 82–114.
[24] M. Guignard and S. Kim, Lagrangean decomposition: a model yielding
stronger Lagrangean bounds, Mathematical Programming 39 (1987) 215–
228.
[25] E.A. Gunn and A.K. Rai, Modelling and decomposition for planning long-
term forest harvesting in an integrated industry structure, Canadian Jour-
nal of Forest Research 17 (1987) 1507–1518.
[26] H. Gunnarsson, Optimization approaches to tactical planning problems
in the forest industry, Licentiate thesis LiU-TEK-LIC-2003:70, Link¨oping
Institute of Technology, Sweden (2004).
[27] H. Gunnarsson, J. Lundgren, and M. R¨ onnqvist, Optimering kan s¨ anka
produktionskostnaderna f¨ or skogsbr¨ansle, The Forestry Research Institute
of Sweden, (2001) Report no. 20 (in Swedish).
20
Supply chain optimization in the forest industry
[28] H. Gunnarsson, J. Lundgren, and M. R¨ onnqvist, Transportation of fuel
wood, Abstracts of the EURO XVII: OR application I, Budapest, Hungary
(2000) 118.
[29] H. Gunnarsson, J.T. Lundgren and M. R¨ onnqvist, Supply chain modeling
of forest fuel. In: G.J. Arthaud and T.M. Barrett (Eds.), Systems Analysis
in Forest Resources: Proceedings of the Eighth Symposium, held Septem-
ber 27–30, 2000, Snowmass Village, Colorado, U.S.A., Kluwer Academic
Publishers, Norwell, U.S.A. (2003).
[30] H. Gunnarsson, J.T. Lundgren, and M. R¨ onnqvist, Supply chain modeling
of forest fuel. In: M. Palmgren and M. R¨ onnqvist (Eds.), Proceedings of
the workshop: Logistik och optimering inom skogsindustrin: Workshop –
?
Are 11–14 mars 2001, Department of Mathematics, Link¨ oping Institute of
Technology, LiTH-MAT-R-2001-16, Sweden (2001) 49–58.
[31] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Annual planning of the
supply chain and ship routing problem in the pulp industry. In: H. Aronsson
(Ed.), Nofoma 2004 Book of abstracts, Link¨ oping, Sweden (2004).
[32] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Annual Planning of the
supply chain in the pulp industry, TraLog –Transportation and Logistics,
Abstracts Book, Molde University College, Molde, Norway, 25-27 August
2004.
[33] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Integrated production and
distribution planning for S¨ odra Cell AB, Abstract in Proceedings of the
second meeting of the EURO working group on operational research (OR)
in agriculture and forest management, Journal of Agricultural Science, 144
(2006) 467–472.
[34] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Integrated production and
distribution planning of the supply chain for S¨ odra Cell AB. In: J. Clausen,
R.M. Jørgensen, N. Kohl, J. Larsen and O.B.G. Madsen, Proceedings of the
1st Nordic Optimization Symposium, Copenhagen, Denmark (2006).
[35] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Optimizing the supply
chain for a pulp industry using decomposition schemes, Abstracts of the
Informs annual meeting 2006, Pittsburgh, USA (2006).
[36] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Terminal location inte-
grated with ship routing. In: Abstracts of the 18th International Symposium
on Mathematical Programming: Facility location problems, Copenhagen,
Denmark (2003) 54.
[37] H. Gunnarsson, M. R¨ onnqvist, and J. Lundgren, Optimization of trans-
portation, storage and chipping of forest fuel, Proceedings of the 2nd For-
est Engineering Conference: Decision support system/tools, V¨axj¨ o, Sweden
(2003) 74–82.
21
References
[38] V. Jayaraman and H. Pirkul, Planning and coordination of production
and distribution facilities for multiple commodities, European Journal of
Operational Research 133 (2001) 394–408.
[39] B. Jonsson, J. Jacobsson and H. Kallur, The forest management planning
package: Theory and application, Studia Forestalia Suecica 189 (1993) 1–
56.
[40] J. Karlsson, M. R¨ onnqvist, and J. Bergstr¨om, Short-term harvest plan-
ning including scheduling of harvest crews, International Transactions in
Operational Research 10 (2003) 413–431.
[41] J. Karlsson, P. Flisberg, H. Gunnarsson, and M. R¨ onnqvist, Modelling
of road investments, Department of Mathematics, Link¨oping Institute of
Technology, LiTH-MAT-R-2002-13, Sweden (2002).
[42] C. Kilger and L. Schneeweiss, Computer Assembly, Supply Chain Manage-
ment and Advanced Planning, H. Stadtler and C. Kilger (Eds.), Springer-
Verlag, Berlin (2005) 389–406.
[43] A. Kleywegt, V.S. Nori, and M.W.P. Savelsbergh, The stochastic inventory
routing problem with direct deliveries, Transportation Science 36(1) (2002)
94–118.
[44] A. Kleywegt, V.S. Nori, and M.W.P. Savelsbergh, Dynamic programming
approximations for a stochastic inventory routing problem, Transportation
Science 38(1) (2004) 42–70.
[45] D. Li and C. O’Brien, Integrated decision modelling of supply chain e?-
ciency, International Journal of Production Economics, 59 (1999) 147–157.
[46] B. Lid´en and M. R¨onnqvist, CustOpt: A model for customer optimized
timber in the wood chain, Department of Mathematics, Link¨ oping Institute
of Technology, LiTH-MAT-R-2000-15, Sweden (2000).
[47] R. Lundmark, Cost structure of and competition for forest-based biomass,
Scandinavian Journal of Forest Research 21 (2006) 272–280.
[48] D.L. Martell, E.A. Gunn, and A. Weintraub, Forest management challenges
for operational researchers, European Journal of Operational Research 104
(1998) 1–17.
[49] H. Min, V. Jayaraman, and R. Srivastava, Combined location-routing prob-
lems: a synthesis and future research directions, European Journal of Op-
erational Research 108 (1998) 1–15.
[50] N.H. Moin and S. Salhi, Inventory routing problems: a logistical overview,
Journal of the Operational Research Society (2006) 1–10.
[51] G. Nagy and S. Salhi, Location-routing: Issues, models and methods, Eu-
ropean Journal of Operational Research 177 (2007) 649–672.
22
Supply chain optimization in the forest industry
[52] J. Nelson, J.D. Brodie, and J. Sessions, Integrating short-term, area-based
logging plans with long-term harvest schedules, Forest Science 37 (1) (1991)
101–122.
[53] M. N¨ asberg, K.O. J¨ ornsten, and P.A. Smeds, Variable Splitting–A new La-
grangean relaxation approach to some mathematical programming prob-
lems, Department of Mathematics, Link¨ oping Institute of Technology,
LiTH-MAT-R-85-04, Sweden (1985).
[54] W.J. O’Brien, K. London, and R. Vrijhoel, Construction supply chain mod-
eling: A research review and interdisciplinary research agenda, Proceedings
IGLC, (2002) Gramado, Brazil.
[55] J. Olhager and E. Selldin, Supply chain management survey of Swedish
manufacturing ?rms, International Journal of Production Economics 89
(2004) 353–361.
[56] M. Palmgren, M. R¨onnqvist, and P. V¨ arbrand, A solution approach for
log truck scheduling based on composite pricing and branch and bound,
International Transactions in Operational Research 10 (2003) 433–447.
[57] Y.B. Park, An integrated approach for production and distribution plan-
ning in supply chain management, International Journal of Production Re-
search 43(6) (2005) 1205–1224.
[58] A. Philpott and G. Everett, Supply chain optimisation in the paper indus-
try, Annals of Operations Research 108 (2001) 225–237.
[59] J. Rantala, Optimizing the supply chain strategy of a multi-unit Finnish
nursery company, Silva Fennica 38(2) (2004) 203–215.
[60] R.L. Rardin, Optimization in operations research, Prentice-Hall, New Jer-
sey (1998).
[61] M.P. Reinders, Tactical planning for a cutting stock system, Journal of
Operational Research Society 44(7) (1993) 645–657.
[62] E.W. Richards and E. Gunn, A model and tabu search method to optimize
stand harvest and road construction schedules, Forest Science 46(2) (2000)
188–203.
[63] E.W. Richards and E. Gunn, Tabu search design for di?cult forest man-
agement optimization problems, Canadian Journal of Forest Research 33
(2003) 1126–1133.
[64] M. R¨ onnqvist, Optimization in forestry, Mathematical Programming, Series
B 97(1-2) (2003) 267–284.
23
References
[65] M. R¨ onnqvist and D. Ryan, Solving truck despatch problems in real time,
in Proceedings of the 31th Annual Conference of the Operational Research
Society of New Zealand, August 31–September 1, Wellington, New Zealand
(1995) 165–172.
[66] A.M. Sarmiento and R. Nagi, A review of integrated analysis of production–
distribution systems, IIE Transactions 31 (1999) 1061–1074.
[67] SCB: Statistical Yearbook of Forestry 2006, O?cial Statistics of Sweden,
National Board of Forestry, J¨ onk¨ oping (2006).
[68] S¸. Sel¸ cuk Ereng¨ u¸ c, N.C. Simpson, and A.J. Vakharia, Integrated produc-
tion/distribution planning in supply chains: An invited review, European
Journal of Operational Research 115 (1999) 219–236.
[69] J.F. Shapiro, Modeling the Supply Chain, Duxbury (2001).
[70] Z.-J.M. Shen, Integrated supply chain design models: A survey and future
research directions, Journal of Industrial and Management Optimization 3
(1) (2007) 1–27.
[71] Z.-J.M. Shen, C. Coullard, and M.S. Daskin, A joint location-inventory
model, Transportation Science 37 (1) (2003) 40–55.
[72]http://www.sodra.com, retrieved April 20 (2007).
[73] H. Stadtler, Supply chain management and advanced planning–basics,
overview and challenges, European Journal of Operational Research 163
(2005) 575–588.
[74] H. Stadtler and C. Kilger (Eds.), Supply Chain Management and Advanced
Planning, Springer-Verlag, Berlin (2005).
[75] C. Todoroki and M. R¨onnqvist, Dynamic control of timber production at
a saw mill with log sawing optimization, Scandinavian Journal of Forest
Research 17 (1) (2002) 79–89.
[76] C. Todoroki and M. R¨ onnqvist, Log sawing optimization directed by market
demands, The New Zealand Journal of Forestry February (2001) 29–33.
[77] M. Wagner and H. Meyr, Food and beverages, in H. Stadtler and C. Kil-
ger (Eds.), Supply Chain Management and Advanced Planning, Springer-
Verlag, Berlin (2005) 371–388.
[78] A. Weintraub, R. Epstein, R. Morales, J. Ser´on, and P. Traverso, A truck
scheduling system improves e?ciency in the forest industries, Interfaces 26
(4) (1996) 1–12.
[79] A. Weintraub and R. Epstein, The supply chain in the forest industry:
Models and linkages. In J. Geunes, P.M. Pardalos, and H.E. Romeijn (Eds.),
Supply chain management: Models, applications, and research directions,
(2002), Kluwer, Dordrecht, 343–362.
24
Supply chain optimization in the forest industry
[80] A. Weintraub and C. Romero, Operations research models and the man-
agement of agricultural and forestry resources: A review and comparison,
Interfaces 36(5) (2006) 446–457.
[81] L. Wolsey, Integer programming, (1998), John Wiley & Sons, New York.
25
References
26
Supply chain optimization in the forest industry
doc_979350640.pdf
The scope of this thesis is modelling and solving large-scale planning problems in the supply chain within the forest industry. Five research papers are included, the first three of which focus on the modelling, and the last two on the solution methods.
Link¨ oping Studies in Science and Technology. Dissertations
No. 1105
Supply chain optimization
in the forest industry
Helene Gunnarsson
Division of Optimization
Department of Mathematics
Link¨ opings universitet, SE-581 83 Link¨ oping, Sweden
ISBN 978-91-85831-85-2 ISSN 0345-7524
Copyright c 2007 Helene Gunnarsson unless otherwise noted
ISBN 978-91-85831-85-2 ISSN 0345-7524
Printed by LiU-Tryck, Link¨ oping, Sweden 2007
Acknowledgements
First of all, I would like to give special thanks to my supervisor Mikael R¨onnqvist
for giving support, inspiration and guidance. Your positive attitude and enthu-
siasm have meant much to me.
I would also like to thank my co-supervisor Jan Lundgren for your support,
encouragement and guidance in writing.
Thanks to Dick Carlsson at S¨odra Cell AB for very good cooperation and friend-
ship.
I am grateful to Torbj¨ orn Larsson, for always taken time for discussions con-
cerning both research and teaching.
I appreciate my friendship with Maud G¨othe-Lundgren.
Thanks to Mathias Henningsson for good cooperation in teaching and J¨ orgen
Blomvall for helping me with computer problems. Thanks to Kaj Holmberg for
interesting discussions about decomposition methods. Thanks to Oleg Burdakov
and Sven Erlander for encouragement.
The group of PhD students has been important for me. Thanks to Maria
Daneva, Elina R¨onnberg, Per-
?
Ake Andersson and Kristian Lundberg for sup-
porting me and for all fun discussions. Thanks also to my former colleagues, I
miss you all.
Thanks to David Bredstr¨ om for generating the routes in the second paper.
Thanks also to Anders Folkesson, Kent Hjelm, and Bertil Thunstr¨ om at Sydved
Energileveranser AB for good cooperation.
Thanks to Pamela Vang and Bj¨ orn Lidestam for helping me improving the
English.
Thanks to Monika Gustafsson and the other administrative sta? who always
are willing to provide.
Many thanks to my parents Ingemar and Britta and my sister Jennie and her
family for their support and concern about my well being.
Last, but not least, I would like to express my deepest gratitude to Bj¨ orn (my
husband-to-be and best friend) and our lovely children Adina (the actress),
Gunnar (the mathematician) and Martin (the humorist).
III
IV
Abstract
The scope of this thesis is modelling and solving large-scale planning problems
in the supply chain within the forest industry. Five research papers are included,
the ?rst three of which focus on the modelling, and the last two on the solu-
tion methods. All problems included are tactical multi-commodity problems
expressed as mixed integer programming (MIP) models. The work has been
done in collaboration with two Swedish companies within the forest industry.
In Paper I, a problem concerning the supply chain of forest fuel for Sydved
Energileveranser AB is modelled and solved. We study the problem of deciding
when and where forest residues are to be converted into wood chips, and how
the residues and chips are to be transported and stored in order to satisfy energy
demand at heating plants. The company has long-term contracts with forest
owners and saw mills. Decisions in the model include whether or not additional
harvest areas and saw mills are to be contracted and which terminals to use.
The planning horizon is one year and monthly time periods are used.
Papers II–V are based on planning problems at S¨ odra Cell AB. The planning
horizon is normally one year. Papers II–III consider only one time period.
In Paper II the supply chain from pulp mills to customers is modelled and
the combined problem of deciding terminal locations and which ship routes to
use is studied. Shipping vessels chartered on short or long term are used to
transport products to terminals in Europe. From each terminal, the products
are transported to customers by truck, train, or a combination of both. In
addition, trains and trucks can be used for transports directly to customers from
mills. In Paper III the entire supply chain, from harvest areas to customers, is
considered. Decisions included are transportation of raw materials, production
mix, the distribution of pulp products, and the selection of potential orders and
their quantities at customers. The ship routes are considered as ?ow links.
In Papers IV–V the problems in Papers II–III are combined into one model
and several time periods are used. Lagrangian heuristics based on Lagrangian
decomposition are used as solution methods in both papers. In Paper IV, the
approach leads to subproblems for each time period, whereas in Paper V, an-
other approach that results in subproblems for di?erent parts of the supply
chain is developed.
All models are based on real data from the companies. The models are detailed
and describe the problems accurately. The solution methods are developed such
that the solution time is kept within practical limits. Results from Papers II–
III have been used by S¨odra Cell AB to support the change of the terminal
structure as well as in budget planning.
V
VI
Optimering av logistikproblem inom
skogsindustrin
Sammanfattning
Denna avhandling presenterar matematiska modeller och l¨ osningsmetoder f¨or
optimering av olika logistikproblem inom skogsindustrin. Vi studerar f¨ ors¨orj-
ningskedjor f¨ or skogsbr¨ansle och massaproduker, och beaktar den ?arliga planer-
ingen i syfte att optimera ?¨ odet.
Det f¨ orsta problemet behandlar skogsbr¨ ansle och ¨ar ett samarbete med Sydved
Energileveranser AB. R?amaterial i form av grenar och toppar fr?an avverk-
ningsplatser ska ?isas och transporteras till v¨ armeverk, eventuellt via termi-
naler. Det ?nns m¨ojlighet att ?isa b?ade i skogen och p?a terminaler. Biproduk-
ter fr?an s?agverk kan ocks?a anv¨andas som r?amaterial. Vid behov kan utbudet av
r?amaterial ut¨ okas genom att ?er avverkningsplatser och s?agverk kontrakteras.
V¨armeverken har en efterfr?agan, angiven i kWh per m?anad, som ska uppfyllas.
Exempel p?a beslut som ska tas ¨ ar var ?isning ska ske, om nya avverkningsplatser
ska kontrakteras, var lagring ska ske, samt hur och n¨ar skogsbr¨anslet ska trans-
porteras.
N¨astf¨oljande problem behandlar massaprodukter och ¨ ar ett samarbete med
S¨ odra Cell AB. Olika sorters massaved fr?an skogen och biprodukter fr?an s?agverk
utg¨ or r?amaterial f¨ or produktion av massaprodukter. R?amaterialet transporteras
till massabruk f¨ or tillverkning enligt speci?cerade recept. De f¨ardiga produk-
terna transporteras sedan med fartyg till terminaler i Europa. Fr?an terminalerna
transporteras produkterna vidare till pappersbruk, vilka ¨ ar f¨ oretagets slutkun-
der. Massaprodukterna transporteras i vissa fall med lastbil eller t?ag direkt fr?an
massabruken till kunderna. Efterfr?agan ¨ ar angiven inom vissa gr¨anser i olika
order. Vissa order ¨ar fasta, vilket inneb¨ ar att dess efterfr?agan m?aste uppfyllas,
medan andra order ¨ ar fria. Exempel p?a beslut som ska tas ¨ar vilka bruk olika
produkter ska produceras p?a, hur m?anga och vilka terminaler som ska anv¨andas,
samt hur transporterna ska utf¨ oras f¨ or att ge b¨ asta resultat.
Utifr?an ovanst?aende beskrivningar har matematiska modeller formulerats. Ge-
nom att l¨ osa dessa kan vi f?a svar p?a logistik- och transportfr?agorna och ett
optimalt ?¨ ode kan hittas. F¨ or att l¨ osa modellerna har kommersiell program-
vara anv¨ants. Heuristiker och mer avancerade optimeringsmetoder har ocks?a
utvecklats i syfte att producera bra l¨ osningar snabbare.
VII
VIII
Papers included
The following ?ve papers, which will be referred to in the text by Roman nu-
merals, constitute the basis for the thesis.
Paper I: Supply chain modelling of forest fuel
Helene Gunnarsson, Jan Lundgren, and Mikael R¨ onnqvist
European Journal of Operational Research, 158 (2004) 103–123.
Paper II: A combined terminal location and ship routing problem
Helene Gunnarsson, Mikael R¨ onnqvist and Dick Carlsson
Journal of the Operational Research Society, 57 (2006) 928–938.
Paper III: Integrated production and distribution planning for
S¨ odra Cell AB
Helene Gunnarsson, Mikael R¨ onnqvist and Dick Carlsson
Journal of Mathematical Modelling and Algorithms, 6 (2007)
25–45.
Paper IV: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on time
periods
Helene Gunnarsson and Mikael R¨ onnqvist
Manuscript in preparation.
Paper V: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on physical
stages
Helene Gunnarsson and Mikael R¨ onnqvist
Manuscript in preparation.
IX
X
Contents
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V
Sammanfattning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII
Papers included . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX
Introduction and overview
Supply chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Forest supply chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Planning levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Optimization of supply chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
The Papers included
Overview of the Papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Paper I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Paper II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Paper III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Paper IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Paper V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Suggestions for future research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Paper I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Paper II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Paper III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Paper IV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Paper V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Introduction and overview
Introduction and overview
We start by giving a short introduction to the ?eld of supply chains. Thereafter
the forest supply chain is in focus and di?erent planning levels are introduced.
Then some approaches of optimizing the supply chain are presented. Last the
papers included in this thesis and their contributions are described.
Supply chains
A supply chain starts with the supply of raw materials for the products, passes
through one or several steps for manufacturing, storing and distribution, and
ends at the customer for the ?nal products. Examples of supply chains are often
presented in a network form. A simple example of a supply chain is given in
Figure 1.
Figure 1: Illustration of a general supply chain.
Since the 1990’s supply chain management (SCM) has been very popular and
useful concept in industrial planning. The SCM can be viewed from di?erent
perspectives, see O’Brien et al. [54]. One perspective is the industrial organi-
zation economics considering market structure and its importance. The other
perspective is the analytic modelling of supply chains focusing on ?nding good
planning measured for example in costs. An attempt to combine these perspec-
tives in order to bene?t the overall planning of the SCM is presented in O’Brien
et al. [54].
The area of SCM can also be divided into two parts, the integration of organi-
1
Introduction and overview
zational units and the coordination of ?ows, see Stadtler [73]. The aim of both
parts is to provide customer service and to be competitive in order to increase
pro?ts. In order to make the coordination of ?ows more e?cient, advanced
planning systems (APS) are used. Ingredients for operating the supply chain
regarding the coordination of ?ows, successfully, can be, for example, operations
research (OR) and logistics, see Stadtler [73].
A basic description of supply chain modelling is found in Shapiro [69]. A more
detailed description of industrial cases can be found in Stadtler and Kilger [74].
Examples of application in the textbook are a case of computer assembly by
Kilger [42] and a case about food and beverages by Wagner and Meyr [77].
The e?ciency of the logistics operations e?ciency is crucial for the successful
management of a supply chain. It is important to view and understand the
overall supply chain to be able to make decisions that can imply, for example,
shorter lead times and less need of safety stocks. A survey of SCM with regard
to Swedish manufacturing ?rms, can be found in Olhager and Selldin [55]. The
paper investigates the supply chain management strategies and practices in a
number of companies.
We will from now only consider the supply chain including ?ow of materials.
The focus will be on analytic supply chain modelling in general and forest supply
chains in particular.
Forest supply chains
An illustration of the supply chain valid for forest products in Sweden can be
studied in Figure 2, below.
The activities that have to exist in order to get the correct amount of raw mate-
rials from the forest without reducing future harvest possibilities, are planting,
cleaning, thinning, and harvesting. The assortments obtained from forest raw
materials can be classi?ed according to their use. Saw logs, pulp wood, and
forest residues are the major parts of the assortments. Each part can be further
divided into several subgroups according to their qualities and dimensions.
After harvesting, timber is transported to saw mills. The timber can also be
transported to terminals for storage before transportation to the saw mills. Sim-
ilar conditions hold for pulp wood, it is either transported directly to pulp mills,
or is taken via terminals for intermediate storing. The harvesting leaves some
forest residues in the form of tops and branches. The residues are left for around
a year in the forest and are then chipped. Then, after storage at terminals (if
necessary), the residues can be transported to heating plants. Byproducts from
saw mills, such as chips, are transported to pulp mills and heating plants for
further use. The making of pulp also produces byproducts, such as bark. These
byproducts are used as fuel at the pulp mill or at heating plants. The ?nal
pulp products are transported to paper mills to make paper products. The
last transportation (which is not included in the illustration), is transportation
2
Supply chain optimization in the forest industry
Figure 2: Parts of the forest supply chain.
from the saw mills, paper mills, and heating plants to the ?nal customer or to
distribution nodes. Excluded from the illustration above is also the transporta-
tion of wood ashes from heating plants back to the forest. The wood ashes are
dispersed over the ground to improve the quality of the soil.
Planning levels
There are di?erent planning levels, which depend mainly on the planning hori-
zon, to consider. Di?erent planning levels and horizons have di?erent needs for
supporting tools. Even if the number of computer based supporting tools is
increasing, most of the planning work is still done manually. These di?erent
planning levels are described, for example, in R¨onnqvist [64] and Weintraub
and Romero [80]. The strategic level often describes decisions that concern
several years or decades. Strategical decisions include investment planning and
infrastructure planning, for example, road building. Tactical planning is often
connected to annual decisions, but it could be for shorter or longer periods,
depending on the problems considered. Tactical decisions include, for example,
road upgrading, annual production planning related to the budget processing,
route structure and equipment utilization. The last level to be mentioned is
operative planning, which could span over a couple of months or involve daily
planning. Examples of operative decisions are detailed production planning and
truck dispatching. The di?erences in time for each level depending on the type
of problem are not clearly de?ned. We will now describe the planning levels
related to the supply chain in the forest industry. We start by harvest problems
3
Introduction and overview
and continue with routing problems. Finally, planning levels in production are
discussed.
Harvesting problems have other horizons regarding planning levels. Here, op-
erative planning includes decisions for months, tactical planning could be from
half a year to a few years, and strategical planning involves decisions for sev-
eral decades. The long horizons within harvesting of course are due to the long
rotation times of trees. MEDFOR, a planning tool for strategical decisions in a
horizon of 50 years, is described in Epstein et al. [20]. Another model for strate-
gical harvest planning can be found in Gunn and Rai [25]. A system based on
simulating conditions in forestry is Indelningspaketet, which is used in Sweden,
primarily for the strategical planning of harvesting over large harvest areas [39].
This system is intended to be an aid in decision operations, such as how much
to harvest each year, proportions of thinning and ?nal harvesting, and in which
order to work the areas. For making tactical decisions, the OPTIMED tool,
which is a tool for support harvest planning for a period of two to ?ve years can
be used, see Epstein [20]. Harvest scheduling, road access and adjacency con-
straints are considered simultaneously in Richard and Gunn [62]. They address
the tactical planning problem with a tabu search method. The design of the
tabu search method is further discussed in Richards and Gunn [63]. A harvest
planning model for the short term can be found in Karlsson et al. [40], where
a planning period of four to six weeks was considered and harvest schedules for
the di?erent harvest crews were decided. Other models for short term harvest
planning can be found in Epstein et al. [19]. A model integrating long term
harvest planning with short term harvest planning can be found in Nelson et
al. [52].
With regard to routing problems, the operative level often involves daily de-
cisions about the routes. The tactical level in routing problems can be from
two weeks up to a month, and the strategical level considers decisions of one
year or longer periods. In Weintraub et al. [78], an operative and computerized
system to support daily truck scheduling decisions is developed, and the system
has been implemented in some of the largest forest ?rms in Chile. Another
approach to the problem of optimizing daily transports within forestry is used
in Palmgren et al. [56], where the aim was to ?nd e?cient routes for all the
trucks involved. Several studies have been made within this area, for example,
R¨onnqvist and Ryan [65].
Within the area of production planning problems, the levels often depend on the
products. Usually, operative planning in production problems depends on the
use of working shifts, times, or daily planning. Tactical planning is often annual,
and the strategical level concerns several years. An integrated optimization
system of the bucking, sawing and planing processes, CustOpt, is presented in
Lid´en and R¨ onnqvist [46], where the aim was to get chain forestry - saw mill -
planing mill to work in a more customer oriented manner. CustOpt is a support
system on a tactical level with a three months planning horizon [46]. Problems
regarding the sawing of the logs at saw mills can be found in Todoroki and
4
Supply chain optimization in the forest industry
R¨onnqvist [75, 76]. The logs should be sawed in such a way that the amount
of residue is minimized. Trying to adjust the sawing with regard to customer
demand would to a great extent both save time and raw materials. Another
similar problem can be found in Reinders [61], where a decision support system
which focuses on tactical planning was developed.
Optimization of supply chains
In the past, ?rms tried to optimize parts of their supply chain separately. Nowa-
days approaches integrating di?erent parts of the supply chain have become
popular. The basic ideas behind most of the solution methods in the papers
described below, can be found in Wolsey [81].
One way of describing the supply chain is to focus on the di?erent activities
involved, such as production and distribution. Reviews of the integration of
production and distribution can be found in Sarmiento and Nagi [66] and Sel¸cuk
Ereng¨ u¸ c [68]. In Chang and Lee [10], a two-stage problem is de?ned, one prob-
lem for the manufacturer and one for the distributor. Two practical approaches
were applied to solve the logistic scheduling problem, the forward approach and
the backward approach. The problems are solved sequentially using a heuristic
method. Another heuristic method used for solving similar problems can be
found in Park [57]. A heuristic consisting of two phases is introduced. In the
?rst phase, a tentative production and distribution plan is established, while in
the second, an improvement of the plan is attempted. Integrated design method-
ology for production and distribution based on the Benders decomposition can
be found in Dogan and Goetschalckx [18] and Cordeau et al. [14]. Some types
of integrated production and distribution problems of reasonable sizes can be
solved directly using CPLEX, see Dhaenens-Flipo and Finke [17]. In Jayara-
man and Pirkul [38], a heuristic which utilizes linear programming techniques
as well as results from the Lagrangian relaxation procedure in order to solve
an integrated production and distribution problem, is presented. The supply
chain of a nursery company is optimized in Rantala [59], where a capacited
MIP problem for solving an integrated production-distribution problem for a
Finnish company is presented. The main decision to consider is if the company
should expand or closure its facilities. The original model is strategical, but by
applying applicable constraints, the model can also be used for solving tactical
and operational level problems.
Another way of describing the supply chain is to focus on di?erent kinds of
decisions at di?erent levels. In Shen [70], the strategical level is connected
to the supply chain design phase and after these kinds of questions have been
answered, the focus shifts to decisions at tactical about operational levels such as
inventory management, and decisions on supply and demand distribution. There
exist several integrated models of any two of the three important supply chain
decisions; location, routing and inventory [70]. A survey of models integrating
the location and routing decisions can be found in Salhi and Nagy [51]. They
5
Introduction and overview
propose a classi?cation scheme and look at both exact and heuristic algorithms.
Other reviews of the location-routing problem can be found in Balakrishnan et
al. [2] and in Min et al. [49]. Most of the earlier studies focus on heuristic
methods. A location-routing model which considers several kinds of products,
a multi-commodity model, is presented in Bookbinder and Reece [4].
The integration of inventory and routing (IR) involves inventory management
and vehicle routing decisions. A survey of IR models is provided by Kleywegt
et al. [43]. They present a classi?cation of the IR models based on the type
of demand, ?eet size, length of the planning horizon and the number of visited
demand points on a vehicle trip. The research contributions of the di?erent
models are also presented. Other more recent reviews can be found in Kley-
wegt et al. [44]. A logistical overview of inventory routing problems is given
in Moin and Salhi [50]. They classify the research according to the planning
horizon employed in the models; single period, multi period and in?nite hori-
zon. Christiansen and Nygreen [11, 12] used a column generation approach for
solving a combined ship routing and inventory problem. A real ship planning
problem, including both inventory management and routing with time windows
is presented in Christiansen [13] and solved by approaching a Dantzig-Wolfe
decomposition combined with branch-and-bound. Campbell and Savelsbergh
[6] presented a decomposition approach for an inventory-routing problem. A
two-stage approach based on the decomposition of the set of decisions and a
combination of branch-and-bound and heuristics were used as solution meth-
ods.
Models integrating location and inventory decisions can be found in Shen et al.
[71] and Daskin et al. [15]. The model presented in Daskin [15] is a non-linear
mixed integer programming model and Lagrangian relaxation is proposed as a
solution method together with heuristics in order to ?nd feasible solutions. The
model presented in Shen et al. [71] is solved using a column generation method.
A review of integrated location and inventory models of a smaller extent can be
found in Daskin et al. [16]. They also present a basic model for the integration
of location and inventory decisions.
The most common ways to express the objective function in supply chain mod-
els, are in cost and pro?t units. There are other options to express the objective
function. Relevant criteria to consider can be lead times and storage safety
levels. Li and O’Brien [45] presented a model which measured the quality of a
supply chain considering four criteria; pro?t, lead time performance, delivery
promptness and waste elimination. The model analyzes the performance at two
levels, the chain level and the operations level
A description of the supply chain in the forest from the perspective of Chilean
forestry ?rms can be found in Weintraub [79]. The use of models and computer
systems related to each physical stage in the chain is discussed as well as the
coordination between stages. In Epstein et al. [20] a more detailed description
of the OR system in the Chilean forest is presented. A general overview of OR
work within forestry can be found in Martell et al. [48] and R¨onnqvist [64]. In
6
Supply chain optimization in the forest industry
Weintraub and Romero [80] the OR models in forestry are compared to those
in agricultural. They concluded that more models managing the entire supply
chain in the forest are to be expected in the future. With regard to Swedish
conditions, several wood ?ow problems in forestry are described in Carlsson and
R¨onnqvist [9].
The supply chain of wood ?bre ?ow is studied in Carlsson et al. [7], where an
overview of the planning problems is presented in a matrix. A description of
OR tools used in practise is also provided as well as trends in the ?eld. Supply
chain management in the pulp and paper industry can be found in Carlsson et
al. [8]. The overall supply chain and its participants are described, and the
need for more information and decision support system is discussed.
Daily supply chain decisions regarding production and distribution of pulp prod-
ucts for the Swedish pulp company S¨odra Cell AB, are optimized in Bredstr¨om et
al. [5]. The planning horizon is three months and optimal production length for
each pulp product is decided as well as storage levels and distribution schemes.
A MIP model is developed, and then solved using branch and price, where the
subproblem is a shortest path problem.
In Philpott and Everett [58] a model for the optimization of the supply chain in
the paper industry is presented. The model is a large MIP model known as the
Paper Industry Value Optimization Tool (PIVOT). The model ?nds an optimal
allocation of supplier to mill, product to machines, and machines to customer,
and also the details in the supply chain are optimized.
Although many models have been developed in this area, there is still more
research to be done. The tactical models for the entire supply chain of forest
fuel and pulp products presented in this thesis contribute to the research ?eld
of optimization of supply chains.
The Papers included
In this section we start by giving an overview of the included papers. We will
continue by summarizing the papers in chronological order. Then the contri-
butions will be presented and ?nally some suggestions for future research are
presented.
Overview of the Papers
The main purpose of this thesis is to model large problems within the forest
industry, and to solve them in reasonable time using di?erent solution methods.
The creation and development of the mathematical models have in all papers
followed the operations research (OR) process. The OR area is wide and there
exists many books covering the subject, e.g. Bertsimas and Tsitsiklis [3], Wolsey
[81], Ahuja et al. [1] and Rardin [60]. The OR process is illustrated in Figure
7
Introduction and overview
3.
Figure 3: The OR process.
To de?ne a relevant problem has been an iterative process in close connection
with the planners at the companies. All the papers consider the supply chain
problem in the Swedish forest industry. Paper I describes the supply chain of
forest fuel and Paper II - V consider the supply chain of pulp products. The
problems discussed in the papers are multi-commodity problems, which means
that several products are included. The problems considered are on a tactical
level and the planning horizon is normally one year. After the problem is clearly
de?ned, a model is to be formulated.
In the modelling process several issues have to be taken into consideration.
Questions of how to express limitations and include decisions are addressed.
The models often have to be reformulated due to changed or updated conditions
regarding the problems. The models in all papers are large and detailed. The
modelling language AMPL is used for all model formulation in the papers. All
included models are mixed integer programming (MIP) models and they include
continuous and binary variables.
After the models are stated, all data is to be collected. This required time and
e?ort in all models due to their sizes and the di?culties to get the adequate
data. It is very important to make sure that the data is correct. The models
become useless if wrong data is used.
All of the included problems are solved using the commercial program CPLEX
directly or as a module in developed methods. In Papers I and II heuristics based
on the linear programming solution are developed. In Paper IV two variants of
a heuristic based on decomposition in time are used and ?nally in Paper V a
8
Supply chain optimization in the forest industry
heuristic based on decomposition of parts in the supply chain is used.
The large sizes of the models in Papers IV–V, motivated developing of more
advanced solution methods. For that reason, di?erent variants of Lagrangian
heuristics based on Lagrangian decomposition are developed. The idea of a
Lagrange relaxation is to relax constraints and move them together with a
multiplier to the objective function. General presentations and early papers
describing Lagrangian heuristics can be found in Geo?rion [23] and Fisher [22].
The Lagrangian decomposition method is presented in, for example, Guignard
and Kim [24] and N¨ asberg et al. [53]. The idea is to duplicate variables and
relax the equality constraints which say that they should have the same value.
This will lead to the division of the problem into subproblems; for each time
period in Paper IV and for each physical stage in the supply chain in Paper V.
The last step of the OR process is the analyzing of results. The reasons to
strange results can be several. The data can be incorrect or the model can have
been formulated in an inappropriate way. Conditions that no one involved in
the model process thought of can appear. It is important to discover the source
of unexpected results in order to be able to use the model in a correct way. The
results of Papers I–III have been analyzed to varying extent. The models in
Papers II and III have been used and developed at the company S¨ odra Cell AB.
An overview of the circumstances regarding the papers is presented in Table 1.
Di?erent properties that represent the problems in the papers are given. Ex-
amples of properties are conditions of the supply and the demand, time periods
included and the sizes of the problems. The aim with Table 1 is to make the
comparison of papers easier to the reader.
9
Introduction and overview
Table 1: Overview of the properties of the problems and models in the papers
included in this thesis.
Properties Paper I Paper II Paper III Paper IV Paper V
Company Sydved Energi- S¨odra S¨odra S¨odra S¨odra
leveranser AB Cell AB Cell AB Cell AB Cell AB
Product Forest fuel Pulp Pulp Pulp Pulp
Number of
products 6 30 18 5-19 5-19
Number of
time periods 12 1 1 4, 12 4, 12
Decomposition No No No Yes Yes
Heuristics Yes Yes No Yes Yes
Production Yes No Yes Yes Yes
Terminal
location Yes Yes No Yes Yes
Focus Model Model Model Algorithm Algorithm
Demand Fixed Fixed Within Within Within
limits limits limits
Eligible demand
contracts No No Yes Yes Yes
Eligible supply
contracts Yes No No No No
Objective Min. costs Min. costs Max. pro?t Max. pro?t Max. pro?t
Modes of Trucks Trucks Trucks Trucks Trucks
transportation Trains Trains Trains Trains
Vessels Vessels Vessels Vessels
Barges Barges Barges Barges
Routes No Yes No Yes Yes
Number of 63,122 11,309 37,165 71,289 – 71,289 –
constraints 274,217 274,217
Number of
binary
variables 10,201 2,449 174 675 – 825 675 – 825
Number of
continuous 494,911 422,796 2,019,384 370,811 – 370,811 –
variables 3,489,036 3,489,036
10
Supply chain optimization in the forest industry
Below, the papers included in this thesis are summarized. We start by describing
the problems and modelling issues. Then the sizes of the problems are given,
and ?nally the developed solution methods are presented.
Paper I: Supply chain modelling of forest fuel
The aim of this paper is to formulate and solve an optimization model describing
the planning problem for a company supplying forest fuel. The test data is from
the Swedish company Sydved Energileveranser AB. The planning horizon is one
year and the period is divided into 12 time periods (months).
The supply sources consist of a number of harvest areas and saw mills. They can
either be owned by the supplying company or be possible to contract. The forest
residues in contracted harvest areas can be made available if the company enters
into a contract with a supplier of residues. To contract a saw mill implies taking
care of byproducts in the form of sawdust, bark, and dry chips, continuously
over the year. After the harvesting, a lot of forest residues, consisting of forest
residues from hardwood and softwood, as well as decay damaged wood, are left
in the harvest area. The residues have to be forwarded and removed from the
area. Before using the residues as fuel, they have to be chipped. The chipping
can take place either in the forest or at terminals. Figure 4 gives an overview of
the problem. The division of the terminals into two sections is meant to show
the storage of non-chipped and chipped forest residues, respectively.
Figure 4: An illustration of the supply chain problem in Paper I.
The demand at the heating plant is speci?ed in kWh, in contrast to the supply
11
Introduction and overview
which is given in m
3
. Therefore, the forest fuel has to be converted into energy
values. The energy value depends, for example, on the moisture contents in the
forest residues.
The choices of chipping possibilities together with the conversion into energy
are new challenges, that are to be considered in the modelling processes. A
mixed integer programming (MIP) model is formulated, and then solved using
CPLEX. In addition, a heuristic method based on the LP solution is devel-
oped. The basic case includes 494,911 variables, whereof 10,201 are binary and
63,122 constraints. Several other cases, besides the basic case, are studied and
evaluated.
Paper II: A combined terminal location and ship routing problem
The aim of this paper is to formulate and solve a model, that includes both
location decisions and routings. This is a case study from S¨ odra Cell AB, one of
the world’s leading manufacturers of market pulp [72]. The problem described
is how to transport and distribute pulp products from pulp mills to customers
as cheaply as possible. Figure 5 gives an overview of the supply chain.
Figure 5: Illustration of the supply chain in Paper II.
Most of the pulp products are transported from a pulp mill to the nearest har-
bour and then further by ship to a harbour terminal outside Scandinavia. All of
the customers located in Scandinavia are supplied by train or truck transporta-
tion. Trains and trucks are also used to supply some of the customers outside
Scandinavia.
The pulp can be transported from the harbour terminals to inland terminals.
Barges are often used for this transportation, but trains and trucks can also
12
Supply chain optimization in the forest industry
be used. The transportation from the harbour and inland terminals to the
?nal customer is made by train and truck. Three vessels chartered long term,
so-called TC-vessels, are used. The routes used by the TC-vessels are divided
into direct routes, called A-routes, and composite routes, called B-routes. The
A-routes go from one pulp mill to one terminal and the B-routes visit several
places. In addition, vessels chartered short term, spot vessels, are used from a
pulp mill to one destination.
Due to the large size of this problem, only one time period is used. One of the
di?culties and challenges in modeling this problem is that we need to integrate
long term decisions on terminal usage with short term decisions on which routes
to choose. Some terminals are connected to few customers with small demands.
These terminals need to get deliveries throughout the year in order to satisfy
their orders. Only one direct delivery using a A-route is not practical. Since
we only have one time period we can model a continuous delivery through the
use to B-routes. The smaller the terminal is, the larger proportion of B-routes
is needed. On the other hand, if the terminal is large we can use A-routes with
direct deliveries from the mills. Since the B-routes are more expensive, we need
to link the proportion of B-routes against the size of the terminal. Otherwise
all ?ow would be using A-routes. Therefore di?erent levels on the terminals are
introduced. The levels are modeled through binary variables and only one level
can be chosen. The levels are coupled to the received ?ow at the terminal. High
levels mean large ?ow and more deliveries on A-routes are accepted. Otherwise,
B-routes, must be used in a particular proportion relating to the level.
The combined terminal location and ship routing problem is formulated in an
MIP optimization model. Five di?erent cases, which are modi?cations of the
basic case, have been tested and evaluated. The basic case includes 425,245
variables, whereof 2,449 are binary. The number of constraints included is
11,309. The solution approach is to use CPLEX directly and to use heuristics
based on the LP solution to ?nd a good solution in an acceptable time.
Paper III: Integrated production and distribution planning for S¨ odra
Cell AB
Paper III is related to Paper II. The problem this paper considers is the whole
supply chain from the forest districts to the customer. Di?erent kinds of raw ma-
terials originating from hardwood or from softwood, are transported by trucks
from the forest districts to the pulp mills. Pulp products are produced according
to di?erent recipes at the pulp mills. The composition and usage of raw mate-
rials are restricted within certain limits. The choice of using di?erent recipes at
di?erent pulp mills creates alternatives. A change over cost is connected to each
alternative and only one of the alternatives can be chosen. Another di?erence
from the Paper II is that all of the terminals are ?xed, meaning that we do not
include any choice of terminals. The reason why this problem includes fewer
terminals is that S¨odra Cell AB has changed their policy regarding the number
13
Introduction and overview
of terminals to use. The model does not include routes in the same sense as
before. In order to simplify the problem, the direct routes are in this paper
considered as direct ?ow without restrictions.
The demand, speci?ed in orders, is given within certain limits. The orders can
be ?xed, meaning that related demand have to be ful?lled, or the orders can
be free. Contracts are speci?ed by orders connected to them. The choice of
accepting a contract or not have to be made. Accepting a contract means that
the demand in all included orders have to be ful?lled within the given limits.
An overview of the problem is illustrated in Figure 6.
Figure 6: Illustration of the supply chain in Paper III.
An MIP model for the entire supply chain problem is presented and ?ve di?erent
alternatives are tested and evaluated. The complete problem, including all
alternatives, includes over 2,000,000 variables, whereof 174 are binary. The
number of constraints is 37,165. The model was used to support the annual
budgeting for 2005-2007.
Paper IV: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on time periods
The problem described in this paper is a combination of the problems considered
in Paper II and Paper III. The mathematical model developed includes all kinds
of routes, (A-routes, B-routes and spot trips), as well as terminal location and
production of pulp products. The components which have been added compared
to the previous models of the supply chain are the time periods and the storage
possibilities. The raw materials can be stored in the forest. To a limited extent,
The pulp products can be stored at the pulp mills and at the terminals. All of
the pulp products have to be distributed during the planning horizon. Di?erent
14
Supply chain optimization in the forest industry
demand structures are taken into consideration. The supply chain for S¨ odra
Cell AB in is illustrated in Figure 7.
Figure 7: Illustration of the supply chain in Paper IV.
The focus in this paper is on the solution methods. Due to the large size of this
problem, more advanced methods are needed to be able to solve the problem
within reasonable time limits. The aim of this paper is therefore to develop
a Lagrangian heuristic method based on Lagrangian decomposition for solving
the mathematical model. The problem is decomposed into one subproblem per
time period. In addition, two variants of a heuristic based on LP solutions, are
developed. Cases including di?erent numbers of variables and constraints are
tested and evaluated. The number of binary variables included is 675–825 and
the number of continuous variables is 370,811–3,489,036. The included number
of constraints ranges from 71,289 to 274,217.
Paper V: Solving a multi-period supply chain problem for a pulp
industry using Lagrangian heuristics based on physical stages
The problem in Paper V is the same as in Paper IV. A Lagrangian heuristic
based on Lagrangian decomposition is used as well. The aim of this paper is to
decompose the supply chain in another way compared to Paper IV, in order to
improve solutions and solution times. The problem in this paper is decomposed
into one subproblem per physical stage in the supply chain. In addition a
heuristic based on division into physical stages are developed. Cases included
di?erent numbers of variables and constraints are tested and evaluated. The
included number of binary variables is 675–825 and the number of continuous
variables is 370,811–3,489,036. The included number of constraints is 71,289–
274,217.
15
Introduction and overview
Contributions
My contribution to the research presented in this thesis includes major involve-
ment in the modelling process and in writing the papers. I have developed and
implemented the Lagrangian relaxation algorithms in the Papers IV and V as
well as the heuristics in all of the papers. I have also taken part in the data
collection. Mikael R¨onnqvist and Jan Lundgren have participated in the work
as my supervisor and co-supervisor, respectively. Dick Carlsson, industrial lo-
gistics manager at S¨ odra Cell AB, complemented the work with the important
industrial approach, mainly in Paper II and Paper III. Results from these papers
are already used in practice by S¨ odra Cell AB.
The papers have previously been presented by me at conferences in Budapest
[28], Aspen [29],
?
Are [30], V¨axj¨ o [37], Copenhagen [36], [34], Link¨ oping [31],
Molde [32], Reykjavik [33] and Pittsburgh [35]. Some of the papers have also
been presented at internal seminars in the Optimization group in Link¨ oping.
Paper I and an earlier version of Paper II were presented at my licentiate sem-
inar in February 2004 [26]. A condensed version of Papers IV and V will be
presented at the EURO XXII conference on operational research in Prague in
July 2007, [21]. Other related research not included in this thesis can be found
in Gunnarsson et al. [27] and Karlsson et al. [41].
The main contributions of each paper are summarized below:
Paper I
• A new model describing the processes of handling forest fuel.
• Solution of the model using both a commercial solver and a new
developed heuristic.
• Analyzing the model using real data.
Paper II
• A new model combining terminal location and ship routing.
• Solution of the model using a commercial solver and two new developed
heuristics.
• Analyzing the model using real data.
• The result of the paper has been used in practice by S¨ odra Cell AB
Paper III
• A new model describing the entire supply chain for a pulp company.
• Solution of the model using a commercial solver.
• Analyzing the model using real data.
• The result of the paper has been used in practice by S¨ odra Cell AB.
Paper IV
• Development of a Lagrangian heuristic method based on time periods.
• Two variants of a rolling horizon heuristic are developed.
• Analyzing the solution method using real data.
16
Supply chain optimization in the forest industry
Paper V
• Development of a Lagrangian heuristic method based on physical stages.
• A heuristic based on physical stages is developed.
• Analyzing the solution method using real data.
Suggestions for future research
New business models have been developed recently. One example is vendor
managed inventory, VMI, which is used to a growing extent. The meaning of
VMI, is that the distributor takes full responsibility for maintaining an agreed
inventory of the products, usually at the buyer´s location. Including this into the
models, makes them even more large, and motivates developing of new solution
methods to be able to solve them. That is a challenge and it is new change of
getting better control and making larger pro?ts for the involved companies.
The model for the supply chain concerning forest residues can be modi?ed in
order to be used for other kinds of biomass fuels. Since this study was completed
in 2001, the prices of forest fuel used at heating plants have increased by almost
23% [67]. It is, therefore, more than ever motivated for suppliers of forest fuel to
use the proposed model in order to be able to make the planning. An overall goal
to use more bioenergy fuel together with the lack of forest fuel at some areas,
also increase the interest of the presented model. Due to the higher prices on
forest fuel, the pulp wood can in some ways work as a substitute for forest fuel.
That can lead to strange situations and does of course impact the pulp market.
The eventual con?icts between the energy sector and the forest industries are
discussed in Lundmark [47]. He also tries to ?nd out the break-point where it
becomes more pro?table to use round wood instead of forest residues as fuel. It
could be interesting to include round wood in the presented model together with
an increased demand. Then several scenarios could be tested in order to get
more solution alternatives if the prices for forest fuel change dramatically. To
create a mathematical model including the supply chains, or parts of the supply
chains, for both forest fuel and pulp products would be a new and exciting
challenge.
The models for the supply chain problems concerning pulp products presented
in this thesis are complex and detailed, which motivates developing of more
solutions methods in order to get good enough solutions faster. In addition to
the Lagrangian heuristics method, other kinds of decomposition methods can
be used, for example the Bender decomposition method. The convergence of
the methods can probably be improved by adding and relaxing some aggregated
constraints, especially in Paper IV. The constraints can also be scaled in order
to make the values of the multipliers more alike. Another way to improve the
solution times is to change the modelling language from AMPL for example to
C.
Developing of decision support system, DSS, for the problems in the included
17
Introduction and overview
papers in this thesis, would enhance the possibilities for practical use of the
presented models.
18
Supply chain optimization in the forest industry
References
[1] R. Ahuja, T. Magnanti, and J. Orlin, Network Flows, (1993) Prentice Hall,
New Jersey.
[2] A. Balakrishnan, J.E. Ward, and R.T. Wong, Integrated facility location
and vehicle routing models: Recent work and future prospects, American
Journal of Mathematical and Management Sciences 7 (1987) 35–61.
[3] D. Bertsimas and J. Tsitsiklis, Introduction to Linear Optimization, (1997)
Athena Scienti?c, Belmont.
[4] J.H. Bookbinder and K.E. Reece, Vehicle routing considerations in distri-
bution system design, European Journal of Operational Research 37 (1988)
204–213.
[5] D. Bredstr¨om, J.T. Lundgren, R. R¨ onnqvist, D. Carlsson, and A. Mason,
Supply chain optimization in the pulp mill industry – IP models, column
generation and novel constraint branches, European Journal of Operational
Research 156(1) (2004) 2–22.
[6] A.M. Campbell and M.W.P. Savelsbergh, A decomposition approach for
the inventory-routing problem, Transportation Science 38(4) (2004) 488–
502.
[7] D. Carlsson, S. D’Amours, A. Martel and M. R¨ onnqvist, Decisions and
methodology for planning the wood ?bre ?ow in the forest supply chain,
submitted to a CEMS book.
[8] D. Carlsson, S. D’Amours, A. Martel and M. R¨ onnqvist, Supply chain
management in the pulp and paper industry, Working paper DT-2006-AM-
3, Interuniversity Research Center on Enterprise Networks, Logistics and
Transportation (CIRRELT), Universit´e Laval, Qu´ebec, G1K 7P4, Canada,
2006.
[9] D. Carlsson and M. R¨onnqvist, Wood ?ow problems in the Swedish forestry,
The Forestry Research Institute of Sweden Report no. 1. (1999).
[10] Y.-C. Chang and C.-Y. Lee, Logistics scheduling: Analysis of two-stage
problems, Journal of Systems Science and Systems Engineering 12(4)
(2003) 385–407.
[11] M. Christiansen and B. Nygreen, A method for solving ship routing prob-
lems with inventory constraints, Annals of Operations Research 81 (1998)
357–378.
[12] M. Christiansen and B. Nygreen, Modelling path ?ows for a combined
ship routing and inventory management problem, Annals of Operations
Research 82 (1998) 391–412.
19
References
[13] M. Christiansen, Decomposition of a combined inventory and time con-
strained ship routing problem, Transportation Science 33(1) (1999).
[14] J.F. Cordeau, F. Pasin, and M.M. Solomon, An integrated model for logis-
tics network design, Annals of Operations Research 144(1) (2006) 59–82.
[15] M.S. Daskin, C.R. Coullard, and Z.-J.M. Shen, An inventory-location
mode: Formulation, solution algorithm and computational results, Annals
of Operational Research 110 (2002) 83–106.
[16] M.S. Daskin, L.V. Snyder, and R.T. Berger, Facility location in supply
chain design. In: A. Langevin and D. Riopel (Eds.), Logistics systems:
Design and Optimization, (2005) Springer, 39–65.
[17] C. Dhaenens-Flipo and G. Finke, An integrated model for an industrial
production-distribution problem, IIE Transactions 33 (2001) 705–715.
[18] K. Dogan and M. Goetschalckx, A primal decomposition method for the in-
tegrated design of multi-period production-distribution systems, IIE Trans-
actions 31 (1999) 1027–1036.
[19] R. Epstein, E. Nieto, A. Weintraub, P. Chevalier, and J. Gabarr´ o, A sys-
tem for the design of short term harvesting strategy, European Journal of
Operational Research 119 (1999) 427–439.
[20] R. Epstein, R. Morales, J. Ser´ on, and A. Weintraub, Use of OR systems in
the Chilean forest industries, Interfaces 29(1) (1999) 7–29.
[21]http://euro2007.vse.cz/, retrieved March 22 (2007).
[22] M.L. Fisher, The Lagrangian relaxation method for solving integer pro-
gramming problems, Management Science 27(1) (1981) 1–18.
[23] A.M. Geo?rion, Lagrangean relaxation for integer programming, Mathe-
matical Programming Study 2 (1974) 82–114.
[24] M. Guignard and S. Kim, Lagrangean decomposition: a model yielding
stronger Lagrangean bounds, Mathematical Programming 39 (1987) 215–
228.
[25] E.A. Gunn and A.K. Rai, Modelling and decomposition for planning long-
term forest harvesting in an integrated industry structure, Canadian Jour-
nal of Forest Research 17 (1987) 1507–1518.
[26] H. Gunnarsson, Optimization approaches to tactical planning problems
in the forest industry, Licentiate thesis LiU-TEK-LIC-2003:70, Link¨oping
Institute of Technology, Sweden (2004).
[27] H. Gunnarsson, J. Lundgren, and M. R¨ onnqvist, Optimering kan s¨ anka
produktionskostnaderna f¨ or skogsbr¨ansle, The Forestry Research Institute
of Sweden, (2001) Report no. 20 (in Swedish).
20
Supply chain optimization in the forest industry
[28] H. Gunnarsson, J. Lundgren, and M. R¨ onnqvist, Transportation of fuel
wood, Abstracts of the EURO XVII: OR application I, Budapest, Hungary
(2000) 118.
[29] H. Gunnarsson, J.T. Lundgren and M. R¨ onnqvist, Supply chain modeling
of forest fuel. In: G.J. Arthaud and T.M. Barrett (Eds.), Systems Analysis
in Forest Resources: Proceedings of the Eighth Symposium, held Septem-
ber 27–30, 2000, Snowmass Village, Colorado, U.S.A., Kluwer Academic
Publishers, Norwell, U.S.A. (2003).
[30] H. Gunnarsson, J.T. Lundgren, and M. R¨ onnqvist, Supply chain modeling
of forest fuel. In: M. Palmgren and M. R¨ onnqvist (Eds.), Proceedings of
the workshop: Logistik och optimering inom skogsindustrin: Workshop –
?
Are 11–14 mars 2001, Department of Mathematics, Link¨ oping Institute of
Technology, LiTH-MAT-R-2001-16, Sweden (2001) 49–58.
[31] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Annual planning of the
supply chain and ship routing problem in the pulp industry. In: H. Aronsson
(Ed.), Nofoma 2004 Book of abstracts, Link¨ oping, Sweden (2004).
[32] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Annual Planning of the
supply chain in the pulp industry, TraLog –Transportation and Logistics,
Abstracts Book, Molde University College, Molde, Norway, 25-27 August
2004.
[33] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Integrated production and
distribution planning for S¨ odra Cell AB, Abstract in Proceedings of the
second meeting of the EURO working group on operational research (OR)
in agriculture and forest management, Journal of Agricultural Science, 144
(2006) 467–472.
[34] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Integrated production and
distribution planning of the supply chain for S¨ odra Cell AB. In: J. Clausen,
R.M. Jørgensen, N. Kohl, J. Larsen and O.B.G. Madsen, Proceedings of the
1st Nordic Optimization Symposium, Copenhagen, Denmark (2006).
[35] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Optimizing the supply
chain for a pulp industry using decomposition schemes, Abstracts of the
Informs annual meeting 2006, Pittsburgh, USA (2006).
[36] H. Gunnarsson, M. R¨ onnqvist, and D. Carlsson, Terminal location inte-
grated with ship routing. In: Abstracts of the 18th International Symposium
on Mathematical Programming: Facility location problems, Copenhagen,
Denmark (2003) 54.
[37] H. Gunnarsson, M. R¨ onnqvist, and J. Lundgren, Optimization of trans-
portation, storage and chipping of forest fuel, Proceedings of the 2nd For-
est Engineering Conference: Decision support system/tools, V¨axj¨ o, Sweden
(2003) 74–82.
21
References
[38] V. Jayaraman and H. Pirkul, Planning and coordination of production
and distribution facilities for multiple commodities, European Journal of
Operational Research 133 (2001) 394–408.
[39] B. Jonsson, J. Jacobsson and H. Kallur, The forest management planning
package: Theory and application, Studia Forestalia Suecica 189 (1993) 1–
56.
[40] J. Karlsson, M. R¨ onnqvist, and J. Bergstr¨om, Short-term harvest plan-
ning including scheduling of harvest crews, International Transactions in
Operational Research 10 (2003) 413–431.
[41] J. Karlsson, P. Flisberg, H. Gunnarsson, and M. R¨ onnqvist, Modelling
of road investments, Department of Mathematics, Link¨oping Institute of
Technology, LiTH-MAT-R-2002-13, Sweden (2002).
[42] C. Kilger and L. Schneeweiss, Computer Assembly, Supply Chain Manage-
ment and Advanced Planning, H. Stadtler and C. Kilger (Eds.), Springer-
Verlag, Berlin (2005) 389–406.
[43] A. Kleywegt, V.S. Nori, and M.W.P. Savelsbergh, The stochastic inventory
routing problem with direct deliveries, Transportation Science 36(1) (2002)
94–118.
[44] A. Kleywegt, V.S. Nori, and M.W.P. Savelsbergh, Dynamic programming
approximations for a stochastic inventory routing problem, Transportation
Science 38(1) (2004) 42–70.
[45] D. Li and C. O’Brien, Integrated decision modelling of supply chain e?-
ciency, International Journal of Production Economics, 59 (1999) 147–157.
[46] B. Lid´en and M. R¨onnqvist, CustOpt: A model for customer optimized
timber in the wood chain, Department of Mathematics, Link¨ oping Institute
of Technology, LiTH-MAT-R-2000-15, Sweden (2000).
[47] R. Lundmark, Cost structure of and competition for forest-based biomass,
Scandinavian Journal of Forest Research 21 (2006) 272–280.
[48] D.L. Martell, E.A. Gunn, and A. Weintraub, Forest management challenges
for operational researchers, European Journal of Operational Research 104
(1998) 1–17.
[49] H. Min, V. Jayaraman, and R. Srivastava, Combined location-routing prob-
lems: a synthesis and future research directions, European Journal of Op-
erational Research 108 (1998) 1–15.
[50] N.H. Moin and S. Salhi, Inventory routing problems: a logistical overview,
Journal of the Operational Research Society (2006) 1–10.
[51] G. Nagy and S. Salhi, Location-routing: Issues, models and methods, Eu-
ropean Journal of Operational Research 177 (2007) 649–672.
22
Supply chain optimization in the forest industry
[52] J. Nelson, J.D. Brodie, and J. Sessions, Integrating short-term, area-based
logging plans with long-term harvest schedules, Forest Science 37 (1) (1991)
101–122.
[53] M. N¨ asberg, K.O. J¨ ornsten, and P.A. Smeds, Variable Splitting–A new La-
grangean relaxation approach to some mathematical programming prob-
lems, Department of Mathematics, Link¨ oping Institute of Technology,
LiTH-MAT-R-85-04, Sweden (1985).
[54] W.J. O’Brien, K. London, and R. Vrijhoel, Construction supply chain mod-
eling: A research review and interdisciplinary research agenda, Proceedings
IGLC, (2002) Gramado, Brazil.
[55] J. Olhager and E. Selldin, Supply chain management survey of Swedish
manufacturing ?rms, International Journal of Production Economics 89
(2004) 353–361.
[56] M. Palmgren, M. R¨onnqvist, and P. V¨ arbrand, A solution approach for
log truck scheduling based on composite pricing and branch and bound,
International Transactions in Operational Research 10 (2003) 433–447.
[57] Y.B. Park, An integrated approach for production and distribution plan-
ning in supply chain management, International Journal of Production Re-
search 43(6) (2005) 1205–1224.
[58] A. Philpott and G. Everett, Supply chain optimisation in the paper indus-
try, Annals of Operations Research 108 (2001) 225–237.
[59] J. Rantala, Optimizing the supply chain strategy of a multi-unit Finnish
nursery company, Silva Fennica 38(2) (2004) 203–215.
[60] R.L. Rardin, Optimization in operations research, Prentice-Hall, New Jer-
sey (1998).
[61] M.P. Reinders, Tactical planning for a cutting stock system, Journal of
Operational Research Society 44(7) (1993) 645–657.
[62] E.W. Richards and E. Gunn, A model and tabu search method to optimize
stand harvest and road construction schedules, Forest Science 46(2) (2000)
188–203.
[63] E.W. Richards and E. Gunn, Tabu search design for di?cult forest man-
agement optimization problems, Canadian Journal of Forest Research 33
(2003) 1126–1133.
[64] M. R¨ onnqvist, Optimization in forestry, Mathematical Programming, Series
B 97(1-2) (2003) 267–284.
23
References
[65] M. R¨ onnqvist and D. Ryan, Solving truck despatch problems in real time,
in Proceedings of the 31th Annual Conference of the Operational Research
Society of New Zealand, August 31–September 1, Wellington, New Zealand
(1995) 165–172.
[66] A.M. Sarmiento and R. Nagi, A review of integrated analysis of production–
distribution systems, IIE Transactions 31 (1999) 1061–1074.
[67] SCB: Statistical Yearbook of Forestry 2006, O?cial Statistics of Sweden,
National Board of Forestry, J¨ onk¨ oping (2006).
[68] S¸. Sel¸ cuk Ereng¨ u¸ c, N.C. Simpson, and A.J. Vakharia, Integrated produc-
tion/distribution planning in supply chains: An invited review, European
Journal of Operational Research 115 (1999) 219–236.
[69] J.F. Shapiro, Modeling the Supply Chain, Duxbury (2001).
[70] Z.-J.M. Shen, Integrated supply chain design models: A survey and future
research directions, Journal of Industrial and Management Optimization 3
(1) (2007) 1–27.
[71] Z.-J.M. Shen, C. Coullard, and M.S. Daskin, A joint location-inventory
model, Transportation Science 37 (1) (2003) 40–55.
[72]http://www.sodra.com, retrieved April 20 (2007).
[73] H. Stadtler, Supply chain management and advanced planning–basics,
overview and challenges, European Journal of Operational Research 163
(2005) 575–588.
[74] H. Stadtler and C. Kilger (Eds.), Supply Chain Management and Advanced
Planning, Springer-Verlag, Berlin (2005).
[75] C. Todoroki and M. R¨onnqvist, Dynamic control of timber production at
a saw mill with log sawing optimization, Scandinavian Journal of Forest
Research 17 (1) (2002) 79–89.
[76] C. Todoroki and M. R¨ onnqvist, Log sawing optimization directed by market
demands, The New Zealand Journal of Forestry February (2001) 29–33.
[77] M. Wagner and H. Meyr, Food and beverages, in H. Stadtler and C. Kil-
ger (Eds.), Supply Chain Management and Advanced Planning, Springer-
Verlag, Berlin (2005) 371–388.
[78] A. Weintraub, R. Epstein, R. Morales, J. Ser´on, and P. Traverso, A truck
scheduling system improves e?ciency in the forest industries, Interfaces 26
(4) (1996) 1–12.
[79] A. Weintraub and R. Epstein, The supply chain in the forest industry:
Models and linkages. In J. Geunes, P.M. Pardalos, and H.E. Romeijn (Eds.),
Supply chain management: Models, applications, and research directions,
(2002), Kluwer, Dordrecht, 343–362.
24
Supply chain optimization in the forest industry
[80] A. Weintraub and C. Romero, Operations research models and the man-
agement of agricultural and forestry resources: A review and comparison,
Interfaces 36(5) (2006) 446–457.
[81] L. Wolsey, Integer programming, (1998), John Wiley & Sons, New York.
25
References
26
Supply chain optimization in the forest industry
doc_979350640.pdf