HO CHI MINH CITY INTERNATIONAL UNIVERSITY

SCHOOL OF INDUSTRIAL ENGINEERING & MANAGEMENT

SEMESTER PROJECT REPORT:

A PAPER REVIEW ON:
A STRATEGIC MODEL FOR EXACT
SUPPLY CHAIN NETWORK DESIGN AND ITS
APPLICATION TO A GLOBAL MANUFACTURER

Professor: PhD. Nguyen Van Hop

Member of Group 04 Student ID Contribution

Trần Nhật Quang IEIEIU21170 36%

Mai Quốc Lộc IEIEIU21028 36%

Lê Nguyễn Quốc An IEIEIU21084 28%

Ho Chi Minh city, Vietnam

05/2025
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

List of Content
List of Content .................................................................................................................................................. 1
List of Figures................................................................................................................................................... 3
List of Tables .................................................................................................................................................... 3
Chapter 1. Introduction................................................................................................................................... 4
1. Background ............................................................................................................................................ 4
2. Problem Statement ................................................................................................................................. 4
3. Objective ................................................................................................................................................ 4
4. Scope And Limitation ............................................................................................................................ 5
Chapter 2. Related Work............................................................................................................................ 5
1. Overview ................................................................................................................................................ 5
2. Literature Review................................................................................................................................... 6
3. Key References ...................................................................................................................................... 6
Chapter 3. Methodology ............................................................................................................................. 7
1. Approach Comparison and Selection..................................................................................................... 7
2. Proposed Concept Design .................................................................................................................... 10
Chapter 4. Solution Development ............................................................................................................ 11
1. Mathematical Model Formulation ....................................................................................................... 11
1.1. Sets ............................................................................................................................................... 11
1.2. Parameters .................................................................................................................................... 12
1.3. Decision Variables ....................................................................................................................... 14
2. Objective Function and Constraints..................................................................................................... 16
2.1. Objective Function ....................................................................................................................... 16
2.2. Constraints ................................................................................................................................... 17
3. Model Validation ................................................................................................................................. 20
3.1. Validation Approach .................................................................................................................... 20
3.2. Validation Groups ........................................................................................................................ 20
3.3. Comprehensive Model Validation ............................................................................................... 21
3.4. Performance Comparison............................................................................................................. 21
Chapter 5. Results ..................................................................................................................................... 22
1. Result Illustration and Explanation...................................................................................................... 22
1.1. Case study .................................................................................................................................... 22
1.2. Redesigning the current supply chain network ............................................................................ 22
2. Scenario Analysis................................................................................................................................. 25
2.1. Key Findings ................................................................................................................................ 26

1
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
3. Study on the Model’s Scalability ......................................................................................................... 28
3.1. Introduction.................................................................................................................................. 28
Chapter 6. Conclusion .............................................................................................................................. 31
1. Discussion on Assumptions ................................................................................................................. 31
1.1. Model Assumptions ..................................................................................................................... 31
1.2. Input Data Assumptions............................................................................................................... 31
1.3. Transportation Assumptions ........................................................................................................ 31
2. Solution Method and Analysis of Advantages and Disadvantages ..................................................... 31
2.1. Solution Method........................................................................................................................... 31
2.2. Advantages of the Solution Method ............................................................................................ 32
2.3. Disadvantages of the Solution Method ........................................................................................ 32
3. Conclusion ........................................................................................................................................... 32
4. Recommendation For Future Research................................................................................................ 33
References ....................................................................................................................................................... 34

2
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

List of Figures
Figure 1 Illustration of the as-is SCND .......................................................................................................... 23
Figure 2 Illustration of the to-be SCND at the end of the planning horizon of five years. ............................ 24
Figure 3 Impact Asia demand reduction on (a) Asia manufacturing and procurement activities, (b) global
transport modal share, and (c).......................................................................................................................... 26
Figure 4 Impact of EU demand reduction on (a) EU manufacturing, procurement activities, (b) global
transport modal share, and (c) total.................................................................................................................. 27
Figure 5 Impact on increasing costs of outsourcing inventory on outsourcing inventory and total
warehousing cost.............................................................................................................................................. 28

List of Tables
Table 1 Approaches Comparisons ..................................................................................................................... 8
Table 2 Approaches Comparisons (Cont) .......................................................................................................... 9
Table 3 Sets of nodes ....................................................................................................................................... 11
Table 4 Other sets ............................................................................................................................................ 11
Table 5 Sets of Arcs ......................................................................................................................................... 12
Table 6 Costs parameter................................................................................................................................... 12
Table 7 Capacities parameter ........................................................................................................................... 13
Table 8 Time parameter ................................................................................................................................... 14
Table 9 Products parameter.............................................................................................................................. 14
Table 10 Manufacturing facilities (plants) decision variables ......................................................................... 14
Table 11 Distribution facilities (DCs) decision variables ................................................................................ 15
Table 12 Transportation types decision variables............................................................................................ 15
Table 13 Objective function's breakdown and abbreviations .......................................................................... 16
Table 14 Constraint's breakdown and abbreviations ....................................................................................... 19
Table 15 Validation Group .............................................................................................................................. 20
Table 16 Models' performance......................................................................................................................... 21
Table 18 Comparison of major SCND costs (as-is design): Actual (from the company’s official financial
statements) versus those obtained those obtained from the proposed MILP model (a single year has been
modelled). ........................................................................................................................................................ 24
Table 19 . Comparison of major SCND costs resulting from the to-be design of the proposed MILP model
versus the as-is design...................................................................................................................................... 25
Table 20 Major activities and total cost of SCN .............................................................................................. 25
Table 24 Solving Cases in Which the Number of Suppliers and Customers is Gradually Increased Up to the
Actual Numbers ............................................................................................................................................... 28
Table 25 Review of SCND Industrial Applications ........................................................................................ 29

3
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Chapter 1. Introduction
1. Background
Supply chains are essential for ensuring efficient operations in businesses, with key drivers such as cost,
reliability, responsiveness, and responsibility playing a major role (Lee & Billington, 1995). Globalization has
pushed businesses to build more integrated and flexible supply chains, while advancements in communication
and information technology present opportunities to improve supply chain operations (Thomas & Griffin,
1996).

The design of a supply chain network (SCND) is a critical factor in determining how well a supply chain
performs. SCND involves making strategic decisions about the structure of the supply chain, such as selecting
suppliers, deciding on production locations, establishing distribution centers, and choosing transportation
methods. This process is complex and has received considerable attention in both industry and academic
research (Chopra & Meindl, 2007; Klibi, Martel, & Guitouni, 2010). While traditionally seen as a strategic
decision, SCND also includes tactical and operational planning elements that help businesses optimize their
global reach (Ernst & Young, 2013; Farahani et al., 2014).

One of the challenges in SCND is effectively integrating the various elements of the supply chain—from
suppliers to end customers. Integration involves coordinating production, inventory control, transportation,
and warehousing. Research in this area has focused on improving these processes to enhance profitability and
efficiency (Beamon, 1998; Yan, Yu, & Edwin Cheng, 2003). Furthermore, supplier selection, production
planning, and logistics have been key topics explored in past studies (Melo, Nickel, & Saldanha-da-Gama,
2009; Farahani et al., 2014).

This paper builds on previous work in SCND, offering a comprehensive model to address the complexities of
designing an efficient supply chain. By considering factors like sourcing, manufacturing, and transportation,
this model helps businesses optimize their operations and make better decisions in a competitive global
market.

2. Problem Statement
This paper investigates the strategic design and long-term transformation of a manufacturing firm’s supply
chain network (SCN), which comprises suppliers, manufacturing facilities (both owned and subcontracted),
distribution centers (DCs), transportation links, and end customers. The central objective is to meet forecasted,
deterministic demand for multiple product families while minimizing the total cost incurred across the
network. These costs include production, labor, transportation, facility investment and expansion, as well as
idle resource costs. The problem formulation requires a broad set of interrelated decisions, including the
selection of suppliers and subcontractors, the location and capacity planning of new manufacturing plants and
DCs, the expansion of existing facilities, workforce sizing, and the optimization of transportation routes and
modes—including multimodal options. The proposed model accounts for operational constraints such as
capacity limitations, production routings, storage capacities, and service-level requirements. By integrating
these aspects, the study aims to support the development of an efficient, scalable, and cost -effective SCN
configuration over an extended planning horizon.

3. Objective
In this paper, the author developed a comprehensive and exact optimization model for strategic supply chain
network design (SCND) that minimizes investment and operational costs. By formulating the problem as a
Mixed Integer Linear Programming (MILP) model, the study supports integrated decision-making of multiple
strategic dimensions, which comprises of selection and sizing of production facilities and distribution centers,
4
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
sourcing production, subcontracting options, warehousing and transportation planning. Moreover, by solving
a large scale SCND case for a global refrigerator manufacturer shows its practical value over handling real
industry level problems effectively and efficiently

4. Scope And Limitation

This study introduces a mathematical model designed to support the strategic planning and redesign of a
manufacturing firm’s supply chain network. The model aims to minimize both investment and operational
costs over a multi-period planning horizon, while ensuring all forecasted customer demand is met. The model
covers key elements of the supply chain including suppliers, manufacturing plants (owned and subcontracted),
distribution centers (DCs), and end customers. It supports decision-making in multiple areas such as facility
selection and capacity planning, establishment of new facilities, supplier and subcontractor selection,
workforce sizing in owned operations, and transportation route and mode selection—including multimodal
transport options. The model takes into account product Bills of Materials (BoMs), production sequences, and
various cost elements such as procurement, production, warehousing, and transportation. It is structured to
handle multiple products, time periods, and supply chain entities, and has been tested through a practical case
involving a global commercial refrigerator manufacturer.

Despite its broad coverage, the model has a number of limitations. It assumes that customer demand is known
and constant, and does not account for uncertainty or fluctuations over time. While it is capable of addressing
large-scale problems, its complexity grows quickly with the number of variables, which can make it difficult
to solve efficiently in some cases. The production process is simplified in that it assumes each product is fully
manufactured at a single plant; it does not support distributed manufacturing or partial outsourcing of
production. Workforce-related planning is only applied to facilities owned by the firm, with third -party
subcontractors treated as external units without internal modeling. The planning is done on a period -by-period
basis without dynamic adjustments during a period. Lastly, the model focuses purely on cost and does not
incorporate any environmental or sustainability metrics, such as emissions or resource efficiency.

Chapter 2. Related Work

1. Overview
This paper tackles the complex challenge of Supply Chain Network Design (SCND) by presenting a robust
and detailed Mixed Integer Linear Programming (MILP) model. It begins with a comprehensive review of
existing SCND literature, identifying critical gaps—like limited integration across supply chain functions,
underexplored transportation options, and oversimplified planning horizons.

Building on these gaps, the paper introduces a new model that integrates all key supply chain elements:
sourcing, production, inventory, warehousing, and transportation. The model supports dynamic decision-
making across multiple time periods, products, facilities, and transportation modes. It also allows for
strategic actions like opening or closing plants and distribution centers, outsourcing, workforce adjustments,
and choosing transport routes.

To prove the model's real-world applicability, the authors apply it to a complex, large-scale case involving a
multinational commercial refrigerator manufacturer. Despite the high complexity, the model delivers
optimal solutions and provides insights that outperform the company’s existing network setup. The paper
closes with analysis on network performance, cost breakdowns, and transport mode usage, showing how the
model can support strategic decision-making in industry.

5
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

2. Literature Review
The literature on Supply Chain Network Design (SCND) highlights the strategic importance of network
structure in achieving key performance goals such as responsiveness, cost efficiency, reliability, and
adaptability. Earlier studies, including Lee and Billington (1995) and Chopra and Meindl (2007), emphasized
the increasing complexity of supply chains due to globalization and technological advancements, which has
driven a strong research focus on integrated, flexible SCN models. A wide range of studies has recognized
SCND as a complex decision-making problem spanning strategic, tactical, and operational levels (e.g.,
Erengüç et al., 1999; Melo et al., 2009; Farahani et al., 2014). Key challenges identified in the literature include
supplier integration, bill of materials (BoM) modeling, production outsourcing, and the role of transportation
decisions—including multimodal transport and economies of scale—in shaping network performance.
Furthermore, recent research has stressed the importance of multi-period planning to better capture demand
variability and long-term decision impacts (Zhang et al., 2016; Badri et al., 2013). However, despite significant
progress, many existing models remain limited in scope—failing to fully integrate critical decisions such as
facility expansion, workforce planning, third-party logistics, and idle cost considerations. In response, this
paper proposes a comprehensive mixed-integer linear programming (MILP) model that incorporates all major
SCN components and decision variables. The model supports realistic, large-scale case applications, thus
addressing both academic gaps and industrial relevance, and directly responds to the challenge of bridging
theoretical models with practical SCND implementation.

3. Key References
This study builds on several important contributions from the existing literature on supply chain network
design (SCND). Lee and Billington (1995) highlighted responsiveness, cost, reliability, and responsibility as
key drivers of effective supply chains. Chopra and Meindl (2007) framed SCND as a strategic decision-making
process that plays a critical role in overall supply chain performance. Melo, Nickel, and Saldanha-da-Gama
(2009) focused on the integration of sourcing, production, and distribution, while also pointing out the
modeling challenges that arise in multi-period planning scenarios. Vidal and Goetschalckx (1997) emphasized
the importance of incorporating Bills of Materials (BoMs) to better align supply and production. Fahimnia et
al. (2013b, 2014) addressed gaps in how models treat production costs, outsourcing decisions, and downstream
logistics. Martel (2005) developed a comprehensive mixed -integer programming model that has influenced
much of the work that followed, including this study. Later works by Badri, Bashiri, and Hejazi (2013), as
well as Zhang, Lee, and Zhang (2016), reinforced the need for multi-period models to capture shifting demand
and operational changes. Finally, Eskandarpour et al. (2015) pointed out the limitations of standard
commercial solvers in handling large, real-world SCND problems — underscoring the need for more scalable
and integrated modeling approaches.

6
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Chapter 3. Methodology
1. Approach Comparison and Selection
A key concern in Supply Chain Network Design (SCND) is achieving integration across the entire chain—
from suppliers to end customers—as it has a significant impact on profitability. Beamon (1998) emphasized
two critical, interrelated components:

• Production planning and inventory control

• Transportation, warehousing, and distribution

Efforts to integrate these areas (known as P-D integration) have been widely explored in research (e.g., Yan
et al., 2003; Yilmaz & Çatay, 2006; Fahimnia et al., 2013b). However, **full integration remains
challenging** due to the complexity of synchronizing all supply chain elements, a challenge highlighted in
numerous studies (e.g., Erengüç et al., 1999; Sarmiento & Nagi, 1999; Meixell & Gargeya, 2005).

In the context of supply chain network design (SCND), integration from sourcing to final delivery is a critical
factor for profitability and efficiency. Several key areas of integration have been explored in the literature,
each contributing to the development of more robust and realistic SCND models:

• Sourcing and Supplier Decisions: Research by Melo et al. (2009), Mula et al. (2010), and Farahani
et al. (2014) has addressed supplier selection under various simplifying assumptions, such as a single
time horizon, a single product, or a uniform manufacturing facility type.

• Material Supply and Production Alignment: The importance of incorporating Bill of Materials
(BoM) in SCND models has been emphasized by Vidal and Goetschalckx (1997) and supported by
Yan et al. (2003), as BoMs are crucial for accurate material planning.

• Production Considerations and Outsourcing: Fahimnia et al. (2013b) have identified key gaps in
SCND literature, particularly in modeling production costs at the work center level and the inclusion
of production outsourcing. Additionally, downstream logistics options such as third -party logistics
(3PL) remain underexplored (Meixell & Gargeya, 2005; Farahani et al., 2014).

• Transportation Modeling: While transportation mode selection and cost efficiencies (e.g., full
truckload or block train transport) are significant, few models incorporate multimodal options or
explore these decisions in depth (Goetschalckx et al., 2002; Mula et al., 2010; Zakeri et al., 2015).

• Temporal Planning Considerations: Multiple time periods are vital for capturing real-world
dynamics in demand and decision-making. Studies by Badri et al. (2013), Govindan et al. (2014), and
Zhang et al. (2016) highlight that considering multiple periods allows models to reflect customer
demand variability and strategic adjustments over time.

• Model Complexity and Solver Limitations: As more elements are integrated into SCND models,
computational complexity increases. Eskandarpour et al. (2015) note that exact solutions for practical,
large-scale instances are often difficult to obtain using standard solvers.

A review of the referenced models in the table below shows that while most focus on minimizing total costs,
few address multiple transportation types or paths. Additionally, elements such as plant capacity expansion,
idle production time, and subcontracting are generally underrepresented.

7
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
Martel (2005) developed a multi-echelon, multi-period mixed integer programming model that includes profit
maximization, aggregate BoMs, and distance-based transportation costs. It makes strategic decisions at the
beginning of the planning horizon and continues operational decisions throughout, serving as a strong example
of an integrated SCND approach.
Table 1 Approaches Comparisons

Obj. Function
Min Cost Multiple Multiple
Multiple Multiple Multiple Multiple Multiple Multiple
Model No. Max Profit Transportation transportation
periods products suppliers plants DCs customers
Multiple types/modes paths
Objective

1 MC • • • • • •

2 MC • • • • • •

3 MP • • • • • • •

4 MC • • • • • •

5 MC • • • • • •

6 MC • • • • • •

7 MC • • • • • •

8 MC • • • • • • •

9 MC • • • • • •

10 MC • • • • • •

11 MO • • • • • • •

12 MP • • • • • • • •

13 MC • • • • • • •

14 MP • • • • • • •

15 MC • • • • • •

16 MO • • • • • • •

17 MO • • • • • • • •

18 MO • • • • • • •

19 MC • • • • • • •

20 MC • • • • • • • •

The proposed
MC • • • • • • •
model

8
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
Table 2 Approaches Comparisons (Cont)

Obj. Function Capacity Plant DC DC

Plant Plant idle Plant DC DC idle DC
Min Cost decisions workfor capacit workfor
Model establish capacity outsourci establishm capacity outsourci
Max Profit BoM per ce y ce
No. ment considerati ng ent considerati ng
Multiple departm decision decisio decision
decisions ons decisions decisions ons decisions
Objective ent s ns s

1 MC • •

2 MC • •

3 MP •

4 MC •

5 MC • •

6 MC

7 MC • • • •

8 MC

9 MC

10 MC

11 MO

12 MP •

13 MC

14 MP • •

15 MC •

16 MO •

17 MO • •

18 MO •

19 MC

20 MC

The
propose MC • • • • • • • • • • •
d model

Notes 1:Yan, Yu, and Edwin Cheng (2003); 2: Kanyalkar and Adil (2005); 3: Martel (2005); 4: Melo, Nickel,
and Saldanha-da-Gama (2006); 5: Yilmaz and Çatay (2006); 6: Park, Choi, and Kang (2007); 7: Thanh, Bostel,
and Péton (2008); 8: Bard and Nananukul (2009); 9: Hamedi et al. (2009); 10: Manzini and Bindi (2009); 11:
Cintron, Ravindran, and Ventura (2010); 12: Bashiri, Badri, and Talebi (2012); 13: Fahimnia, Luong, and
Marian (2012); 14: Badri, Bashiri, and Hejazi (2013); 15: Fahimnia et al. (2013b); 16: Sadrnia et al. (2013);
17: Govindan et al. (2014); 18: Mota et al. (2015); 19: Zakeri et al. (2015); 20: Zhang, Lee, and Zhang (2016)

9
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

2. Proposed Concept Design

This paper introduces a comprehensive Mixed Integer Linear Programming (MILP) model for deterministic
Supply Chain Network Design (SCND), aiming to minimize both investment and operational costs across the
supply chain. Originally developed as part of a PhD thesis (Arampantzi, 2017), the model incorporates all
critical SCN characteristics identified in the literature with 3 key contributions:

The first contribution is model completeness, the model captures a wide range of real-world supply chain
features from the manufacturer's perspective. These include:

• Multiple time periods and product types

• A variety of facilities (plants and distribution centers)

• Multiple stakeholders (suppliers and customers)

• Comprehensive transportation link modeling, accounting for cost alternatives and loading efficiencies

• Inclusion of complex production factors such as Bills of Materials (BoMs), macro-routings, setup, and
processing times

The second contribution refers to integrated decision-making, the model supports a broad spectrum of strategic
and tactical decisions for each period in the planning horizon. These include:

• Facility establishment, expansion, or closure

• Production allocation and inventory distribution

• Subcontracting of manufacturing or logistics

• Workforce sizing

• Transportation mode and routing choices

• All decisions are addressed in an integrated framework, reflecting their interdependencies.

The third one contribute to the scalability and practical relevance of the model, demonstrating its real-world
applicability, the model was tested on a large-scale case study involving a global commercial refrigerator
manufacturer. Despite the complexity, the model achieved exact solutions using commercial optimization
software, outperforming the manufacturer’s existing network. The analysis also provided insights into supplier
distribution, cost structures, and transportation modal splits—useful for strategic decision-making.

10
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Chapter 4. Solution Development

1. Mathematical Model Formulation
The supply chain network (SCN) is modeled by the researchers as a directed multigraph, G=(N,A), where N
represents the set of nodes and A - the set of arcs connecting these nodes. This structure encompasses all major
supply chain entities, including suppliers, manufacturing plants, distribution centers (DCs), and customers.

To manage the complexity of the model, the researchers define all relevant sets, parameters, and decision
variables in detail. Indices i and j are consistently used to denote the main entities in the network. For each
index, the corresponding set is clearly specified. This approach was chosen to maintain clarity and conciseness,
especially given the extensive number of elements involved in the proposed model.

1.1. Sets
1.1.1. Sets of nodes
Table 3 Sets of nodes

Category Notation Description

Suppliers S Set of possible suppliers

M̃ Set of existing owned plants

M̄ Set of possible new owned plants

Plants
M̌ Set of possible subcontractor plants

M = M̃ ∪ M̄ ∪ M̌ Total set of plants

Plant department types W Set of possible types of departments in an owned plant

D̃ Set of existing owned DCs

D̄ Set of possible new owned DCs

DCs
Ď Set of possible subcontractor DCs

D = D̃ ∪ D̄ ∪ Ď Total set of DCs

Customers C Set of customers

Set of nodes N=S∪M∪D∪C Complete set of nodes in the network

1.1.2. Other sets

Table 4 Other sets

Category Notation Description

11
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Set of product families; each family comprises products with similar

F
manufacturing characteristics
Products / Raw
Set of raw material families; each family comprises raw materials with similar
materials Θ
characteristics

Θ^f Set of raw material families related to product family f ∈ F; Θ^f ⊆ Θ

T = {1, 2, ...,
Time periods Set of time periods, with t̂ the total number of time periods
t̂}

R Set of transportation types

Transportation
types
Rᵢ Set of transportation types available at node i ∈ N, where Rᵢ ⊆ R

1.1.3. Sets of arcs

Table 5 Sets of Arcs

Category Notation Description

Set of arcs (i, j) representing connections from origin to destination under different node
A types
Arc set
Extended arc set (i, j, r) where r is a transportation type available at both i and j (r ∈ Rᵢ ∩
Ā Rⱼ)

1.2. Parameters
1.2.1. Costs
Table 6 Costs parameter

Category Notation Description

Supplier costs kᵢ^θ Procurement cost per unit of raw material family θ ∈ Θ related to supplier i ∈ S

Depreciation cost per unit of capacity per period (in working time units) for
γᵢʷ
department w ∈ W of plant i ∈ M̃ ∪ M̄

σᵢ^f Set-up cost for product family f ∈ F manufactured in owned plant i ∈ M̃ ∪ M̄

βᵢ^f Production cost per unit (run time) for product family f ∈ F in plant i ∈ M̃ ∪ M̄
Manufacturing
costs
Idle equipment cost per unit of capacity per period in department w ∈ W of plant
λᵢʷ
i ∈ M̃ ∪ M̄

uᵢ Idle labour cost per unit of capacity per period in plant i ∈ M̃ ∪ M̄

fᵢ Severance cost per dismissed employee in owned plant i ∈ M̃ ∪ M̄

12
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

hᵢ Recruitment cost per new hire in owned plant i ∈ M̃ ∪ M̄

Subcontracting cost per unit of product family f ∈ F in subcontractor plant i ∈

vᵢ^f
M̌

γᵢ Depreciation cost per unit of storage (e.g., m²) in owned DC i ∈ D̃ ∪ D̄

νᵢ^f Inventory cost per unit of product family f ∈ F in owned DC i ∈ D̃ ∪ D̄

λᵢ Idle unit cost per unit of capacity in owned DC i ∈ D̃ ∪ D̄

uᵢ Idle labour cost per unit of capacity in owned DC i ∈ D̃ ∪ D̄

Distribution costs
f̂ᵢ Severance cost per dismissed employee in owned DC i ∈ D̃ ∪ D̄

ĥᵢ Recruitment cost per new hire in owned DC i ∈ D̃ ∪ D̄

Subcontracted warehousing cost per unit of product family f ∈ F in subcontracted

v̂ ᵢ^f
DC i ∈ Ď

Transportation cost per unit of product family f ∈ F from node i to node j by

τᵢⱼʳ^f
transportation type r
Transportation
costs
Transportation cost per unit of raw material family θ ∈ Θ from node i to node j
μᵢⱼʳ^θ
by transportation type r

1.2.2. Capacities
Table 7 Capacities parameter

Category Notation Description

Max number of raw material units of family θ ∈ Θ from supplier i ∈ S in period

Supplier capacity aᵢ^θt
t∈T

Initial capacity per period of department w ∈ W in owned plant i ∈ M̃ ∪ M̄ (zero

δᵢʷ
if i ∈ M̄)
Manufacturing
capacity ĉᵢ^ft Max number of products of family f ∈ F by subcontractor i ∈ M̌ in period t ∈ T

H Number of working time units per employee per time period

ρ̂ᵢ Initial capacity of owned DC i ∈ D̃ ∪ D̄ at start of horizon (zero if i ∈ D̄)

Distribution capacity q̂ ᵢ^t Available capacity of subcontractor DC i ∈ Ď in period t ∈ T (in area units)

E Employees required per unit of capacity (e.g., area in m²) in a DC

13
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
1.2.3. Time
Table 8 Time parameter

Category Notation Description

Production time (in working units) required to process one unit of product
U_f^w
family f ∈ F in department w ∈ W
Manufacturing time
Total set-up time per period (in working units) required to process product
Y_f^w family f ∈ F in department w ∈ W, based on a fixed number of setups per
strategic period

1.2.4. Products
Table 9 Products parameter

Category Notation Description

Customer demand dᵢ^ft Number of products of family f ∈ F required by customer i ∈ C at time t ∈ T

Quantity of raw material θ ∈ Θ^f needed to produce one unit of product family
BoM Q_f^θ
f ∈F

Safety stock level (as % of demand) for product family f ∈ F, used for inventory
Level of safety stock S_f
planning

K^f Number of products of family f ∈ F per unit load (e.g., m²)

Unit load
Λ^θ Number of raw materials of family θ ∈ Θ per unit load

Transport lower Minimum unit load per transport type r ∈ R; below this, transport is not
l_r
bound financially viable

1.3. Decision Variables

Table 10 Manufacturing facilities (plants) decision variables

Category Notation Description

φᵢᵗ Binary variable = 1 if new owned plant i ∈ M̄ is established at time t ∈ T; otherwise 0

Manufacturing facilities (plants)

δᵢʷᵗ Expansion of department w ∈ W capacity in owned plant i ∈ M̃ ∪ M̄ at time t ∈ T

εᵢʷᵗ Total production capacity of department w ∈ W in plant i ∈ M̃ ∪ M̄ at time t ∈ T

ψᵢᶠᵗ Binary variable = 1 if plant i ∈ M̃ ∪ M̄ manufactures product f ∈ F at time t ∈ T

ζᵢᶠᵗ Quantity of product family f ∈ F manufactured by plant i ∈ M at time t ∈ T

θᵢʷᵗ Labour staff in department w ∈ W of plant i ∈ M̃ ∪ M̄ at time t ∈ T (non-negative integers)

14
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

υᵢʷᵗ Number of dismissed employees in department w ∈ W of plant i ∈ M̃ ∪ M̄ at time t ∈ T

ηᵢʷᵗ Number of hired employees in department w ∈ W of plant i ∈ M̃ ∪ M̄ at time t ∈ T

ξᵢʷᵗ Labour idle capacity of department w ∈ W in plant i ∈ M̃ ∪ M̄ at time t ∈ T

ξᵢʷᵗ Equipment idle capacity of department w ∈ W in plant i ∈ M̃ ∪ M̄ at time t ∈ T

Table 11 Distribution facilities (DCs) decision variables

Category Notation Description

φ̂ᵢᵗ Binary variable = 1 if new owned DC i ∈ D̄ is established at time t ∈ T; otherwise 0

δ̂ᵢᵗ Expansion of storage capacity in owned DC i ∈ D̄ ∪ D̃ at time t ∈ T (non-negative)

ε̂ᵢᵗ Total storage capacity in owned DC i ∈ D̄ ∪ D̃ at time t ∈ T (non-negative)

Binary variable = 1 if DC i ∈ D̄ ∪ D̃ is operational at time t ∈ T for product f ∈ F; otherwise

ψ̂ᵢᶠᵗ
0

Inventory of product family f ∈ F held in DC i ∈ D at end of time period t ∈ T (non-

πᵢᶠᵗ
negative)

πᵢᶠ⁰ Initial inventory in DC i ∈ D for product f ∈ F

θ̂ᵢᵗ Number of employees in owned DC i ∈ D̄ ∪ D̃ at time t ∈ T (non-negative integers)

Distribution facilities (DCs)

θ̂ᵢ⁰ Initial workforce at DC i ∈ D̄ ∪ D̃ , defined as ρ̂ᵢ ÷ E

υ̂ᵢᵗ Number of dismissed employees at DC i ∈ D̄ ∪ D̃ at time t ∈ T (non-negative integers)

η̂ᵢᵗ Number of hired employees at DC i ∈ D̄ ∪ D̃ at time t ∈ T (non-negative integers)

ξ̂ᵢᵗ Labour idle capacity in owned DC i ∈ D̄ ∪ D̃ at time t ∈ T (non-negative)

ζ̂ᵢᵗ Facility idle capacity in owned DC i ∈ D̄ ∪ D̃ at time t ∈ T (non-negative)

Table 12 Transportation types decision variables

Category Notation Description

Transportati

yᵢⱼʳᵗ Binary variable = 1 if transportation type r is operated along arc (i,j) ∈ A at time t ∈ T
on types

Quantity of products f ∈ F transferred via transport type r along arc (i,j) ∈ A at time t ∈ T
λᵢⱼʳᶠᵗ
(non-negative)

15
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Quantity of raw material θ ∈ Θ transferred via transport type r along arc (i,j) ∈ A at time t ∈
ωᵢⱼʳᶿᵗ
T (non-negative)

2. Objective Function and Constraints

2.1. Objective Function
The objective function minimizes the total investment and operational costs over the planning horizon. It
includes cost components for manufacturing, distribution, transportation, procurement, labor, and inventory
handling.

min ∑_{i ∈ M̃ ∪ M̄} ∑_{w ∈ W} ∑_{t ∈ T} γᵢʷ · εᵢʷᵗ + ∑_{i ∈ D̄ ∪ D̃ } ∑_{t ∈ T} γᵢ · εᵢᵗ +

∑_{(i,j) ∈ A | i ∈ S} ∑_θ ∈ Θ ∑_{t ∈ T} kᵢ^θ · ωᵢⱼʳᶿᵗ + ∑_{i ∈ M̃ ∪ M̄} ∑_{f ∈ F} ∑_{t ∈ T} σᵢ^f ·
ψᵢᶠᵗ + ∑_{i ∈ M̃ ∪ M̄} ∑_{f ∈ F} ∑_{t ∈ T} βᵢ^f · ζᵢᶠᵗ + ∑_{i ∈ M̃ ∪ M̄} ∑_{w ∈ W} ∑_{t ∈ T}
fᵢ · υᵢʷᵗ + ∑_{i ∈ M̃ ∪ M̄} ∑_{w ∈ W} ∑_{t ∈ T} hᵢ · ηᵢʷᵗ + ∑_{i ∈ M̃ ∪ M̄} ∑_{w ∈ W} ∑_{t
∈ T} υᵢ · ξᵢʷᵗ + ∑_{i ∈ M̃ ∪ M̄} ∑_{w ∈ W} ∑_{t ∈ T} λᵢʷ · ξᵢʷᵗ + ∑_{i ∈ M̌} ∑_{f ∈ F} ∑_{t
∈ T} vᵢ^f · ζᵢᶠᵗ + ∑_{i ∈ D̄ ∪ D̃ } ∑_{f ∈ F} ∑_{t ∈ T} νᵢ^f · πᵢᶠᵗ + ∑_{i ∈ D̄ ∪ D̃ } ∑_{t ∈ T} f̂ᵢ ·
υ̂ᵢᵗ + ∑_{i ∈ D̄ ∪ D̃ } ∑_{t ∈ T} ĥᵢ · η̂ᵢᵗ + ∑_{i ∈ D̄ ∪ D̃ } ∑_{t ∈ T} υ̂ᵢ · ξ̂ᵢᵗ + ∑_{i ∈ D̄ ∪ D̃ }
∑_{t ∈ T} λᵢ · ζ̂ᵢᵗ + ∑_{i ∈ Ď} ∑_{f ∈ F} ∑_{t ∈ T} v̂ᵢ^f · πᵢᶠᵗ + ∑_{(i,j) ∈ A | i ∈ M ∪ D} ∑_{f
∈ F} ∑_ {t ∈ T} τᵢⱼʳ^f · χᵢⱼʳᶠᵗ + ∑_{(i,j) ∈ A | i ∈ S} ∑_θ ∈ Θ ∑_{t ∈ T} μᵢⱼʳ^θ · ωᵢⱼʳᶿᵗ

Table 13 Objective function's breakdown and abbreviations

Term Expression Description

1 ∑ᵢ∈M̃∪M̄ ∑ʷ∈W ∑ₜ∈T γᵢʷ · εᵢʷᵗ Depreciation cost of owned plants

2 ∑ᵢ∈D̄∪D̃ ∑ₜ∈T γᵢ · εᵢᵗ Depreciation cost of owned DCs

3 ∑_((i,j)∈A | i∈S) ∑_θ∈Θ ∑ₜ∈T kᵢ^θ · ωᵢⱼʳᶿᵗ Procurement cost of raw materials

4 ∑ᵢ∈M̃∪M̄ ∑ₜ∈T ∑f∈F σᵢ^f · ψᵢᶠᵗ Setup costs in owned plants

5 ∑ᵢ∈M̃∪M̄ ∑ₜ∈T ∑f∈F βᵢ^f · ζᵢᶠᵗ Production (run time) costs in owned plants

6 ∑ᵢ∈M̃∪M̄ ∑w∈W ∑ₜ∈T fᵢ · υᵢʷᵗ Severance costs of owned plant employees

7 ∑ᵢ∈M̃∪M̄ ∑w∈W ∑ₜ∈T hᵢ · ηᵢʷᵗ Recruitment costs of owned plant employees

16
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

8 ∑ᵢ∈M̃∪M̄ ∑w∈W ∑ₜ∈T υᵢ · ξᵢʷᵗ Labour idle capacity costs of owned plants

9 ∑ᵢ∈M̃∪M̄ ∑w∈W ∑ₜ∈T λᵢʷ · εᵢʷᵗ Equipment idle capacity costs of owned plants

10 ∑ᵢ∈M̌ ∑ₜ∈T ∑f∈F vᵢ^f · ζᵢᶠᵗ Subcontracting manufacturing cost

11 ∑ᵢ∈D̄∪D̃ ∑f∈F ∑ₜ∈T νᵢ^f · πᵢᶠᵗ Inventory cost in owned DCs

12 ∑ᵢ∈D̄∪D̃ ∑ₜ∈T f̂ᵢ · υ̂ᵢᵗ Severance cost of owned DC employees

13 ∑ᵢ∈D̄∪D̃ ∑ₜ∈T ĥᵢ · η̂ᵢᵗ Recruitment cost of owned DC employees

14 ∑ᵢ∈D̄∪D̃ ∑ₜ∈T υ̂ᵢ · ξ̂ᵢᵗ Labour idle capacity costs of owned DCs

15 ∑ᵢ∈D̄∪D̃ ∑ₜ∈T λᵢ · ζ̂ᵢᵗ Facility idle capacity costs of owned DCs

16 ∑ᵢ∈Ď ∑f∈F ∑ₜ∈T v̂ ᵢ^f · πᵢᶠᵗ Subcontracted warehousing cost

17 ∑ₜ∈T ∑_((i,j)∈A | i∈M∪D) ∑f∈F τᵢⱼʳ^f · xᵢⱼʳᶠᵗ Transportation cost of final products

18 ∑ₜ∈T ∑_((i,j)∈A | i∈S) ∑_θ∈Θ μᵢⱼʳ^θ · ωᵢⱼʳᶿᵗ Transportation cost of raw materials

2.2. Constraints
• Supplier Capacity

o ∑_{(j,r)|(i,j,r) ∈ Â} ω_{ijr}^{θt} ≤ a_i^{θt}, ∀i ∈ S, θ ∈ Θ, t ∈ T (2)

• Material Requirements

o ∑_{(i,r)|(i,j,r) ∈ Â} ω_{ijr}^{θt} = ∑_{f ∈ F|(θ ∈ Θ^f)} Q_f^θ · z_j^{ft}, ∀j ∈ M̄ ∪ M̃, θ ∈

Θ, t ∈ T (3)

• Demand

o ∑_{i ∈ M} ∑_{b=1}^{t} z_i^{fb} ≥ ∑_{j ∈ C} ∑_{b=1}^{t} d_j^{fb}, ∀f ∈ F, t ∈ T (4)

• Demand and Continuity

o ∑_{(i,r)|(i,j,r) ∈ Â} x_{ijr}^{ft} = d_j^{ft}, ∀j ∈ C, f ∈ F, t ∈ T (5)

o z_i^{ft} = ∑_{(j,r)|(i,j,r) ∈ Â} x_{ijr}^{ft}, ∀i ∈ M̄ ∪ M̃, f ∈ F, t ∈ T (6)

• Plants and Plant Capacities

o ∑_{t ∈ T} φ_i^t ≤ 1, ∀i ∈ M̄ (7)

o ∑_{b=1}^{t} φ_i^b ≥ ψ_i^{ft}, ∀i ∈ M̄, f ∈ F, t ∈ T (8)

o ε_i^{wt} = ∑_{b=1}^{t} δ_i^{wt} + δ̂_i^w, ∀i ∈ M̄ ∪ M̃, w ∈ W, t ∈ T (9)

17
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
• Production and Production Time

o z_i^{ft} ≤ B · ψ_i^{ft}, ∀i ∈ M̄ ∪ M̃, f ∈ F, t ∈ T (10)

o ε_i^{wt} − ∑_{f ∈ F} (U_f^w · z_i^{ft} + Y_f^w · ψ_i^{ft}) = ξ_i^{wt}, ∀i ∈ M̄ ∪ M̃, w ∈

W, t ∈ T (11)

• Workforce

o θ_i^{wt} = θ_i^{w(t−1)} − ι_i^{wt} + η_i^{wt}, ∀i ∈ M̄ ∪ M̃, w ∈ W, t ∈ T (12)

o H · θ_i^{wt} − ∑_{f ∈ F} (U_f^w · z_i^{ft} + Y_f^w · ψ_i^{ft}) = ξ_{iwt}, ∀i ∈ M̄ ∪ M̃,

w ∈ W, t ∈ T (13)

• Subcontracting Production

o z_i^{ft} ≤ ĉ_i^{ft}, ∀i ∈ M̆, f ∈ F, t ∈ T (14)

o ∑_{b=1}^{t} z_i^{fb} ≥ ∑_{(j,r)|(i,j,r) ∈ Â} ∑_{b=1}^{t} x_{ijr}^{fb}, ∀i ∈ M̆, f ∈ F, t ∈

T (15)

• Inventory Levels and Continuity

o ∑_{i ∈ D} p_i^{ft} ≥ S_f · ∑_{j ∈ C} d_j^{ft}, ∀f ∈ F, t ∈ T (16)

o p_i^{f(t−1)} + ∑_{(k,r)|(k,i,r) ∈ Â} x_{kir}^{ft} − ∑_{(j,r)|(i,j,r) ∈ Â} x_{ijr}^{ft} = p_i^{ft},

∀i ∈ D, f ∈ F, t ∈ T (17)

• Warehousing

o ∑_{t ∈ T} φ̂_i^t ≤ 1, ∀i ∈ D̄ (18)

o ∑_{b=1}^{t} φ̂_i^b ≥ ψ̂_i^{ft}, ∀i ∈ D̄, f ∈ F, t ∈ T (19)

o ∑_{(j,r)|(j,i,r) ∈ Â} (x_{jir}^{ft} + p_i^{ft}) ≤ B · ψ̂_i^{ft}, ∀i ∈ D̄ ∪ D̃ , f ∈ F, t ∈ T (20)

o ε̂_i^t = ∑_{b=1}^{t} δ̂_i^t + ρ̂_i, ∀i ∈ D̄ ∪ D̃ , t ∈ T (21)

o ∑_{f ∈ F} p_i^{ft}/K^f ≤ ε̂_i^t, ∀i ∈ D̄ ∪ D̃ , t ∈ T (22)

o ε̂_i^t − ∑_{f ∈ F} p_i^{ft}/K^f = ξ̂_i^t, ∀i ∈ D̄ ∪ D̃ , t ∈ T (23)

o θ̂_i^t = θ̂_i^{t−1} − ι̂_i^t + η̂_i^t, ∀i ∈ D̄ ∪ D̃ , t ∈ T (24)

o 1/E · ∑_{f ∈ F} p_i^{ft}/K^f ≤ θ̂_i^t, ∀i ∈ D̄ ∪ D̃ , t ∈ T (25)

o ∑_{f ∈ F} p_i^{ft}/K^f ≤ ψ̂_i^{ft}, ∀i ∈ D̄, t ∈ T (26)

• Transportation

o ∑_{f ∈ F} 1/K^f · x_{ijr}^{ft} ≥ l_r · y_{ijr}^t, ∀(i,j)|(i,j,r) ∈ Â, r ∈ R, t ∈ T (27)

o ∑_{θ ∈ Θ} 1/Λ^θ · ω_{ijr}^{θt} ≥ l_r · y_{ijr}^t, ∀i ∈ S, j|(i,j,r) ∈ Â, r ∈ R, t ∈ T (28)

18
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
o ∑_{f ∈ F} 1/K^f · x_{ijr}^{ft} ≤ B · y_{ijr}^t, ∀(i,j)|(i,j,r) ∈ Â, r ∈ R, t ∈ T (29)

o ∑_{θ ∈ Θ} 1/Λ^θ · ω_{ijr}^{θt} ≤ B · y_{ijr}^t, ∀i ∈ S, j|(i,j,r) ∈ Â, r ∈ R, t ∈ T (30)

Table 14 Constraint's breakdown and abbreviations

Constraint No. Category Explanation

2 Supplier capacity Limits the quantity of raw materials procured from suppliers to not exceed
supplier capacity.

3 Material requirements Ensures raw materials shipped to a plant match what is needed for final product
production (JIT procurement).

4 Demand Guarantees that cumulative production meets or exceeds cumulative customer

demand.

5 Demand continuity Prevents early fulfillment; matches shipped quantity to customer demand at
each time.

6 Shipping only after production Ensures shipments are based on completed production.

7 Plant establishment Limits each candidate plant to be established at most once.

8 Plant operation timing Ensures a plant can only operate after it is established.

9 Plant capacity Defines production department capacity over time.

10 Production setup Requires setup for production to begin.

11 Idle production capacity Calculates unused capacity in a plant department.

12 Workforce variation Models change in workforce over time in plant departments.

13 Idle labor in plant Captures unused labor capacity in plant departments.

14 Subcontracting limit Limits outsourced production to subcontractor capacity.

15 Subcontracting consistency Ensures shipped outsourced products do not exceed produced amount.

16 Inventory safety stock Maintains safety stock levels for each product family at each time.

17 Inventory flow Tracks inventory changes over time based on incoming and outgoing flows.

18 DC establishment Limits DC establishment to at most once per candidate location.

19 DC operation timing Ensures DCs operate only after being established.

20 DC product flow constraint Restricts incoming and stored product amounts to DC capacity.

21 DC capacity expansion Updates DC capacity with expansions over time.

22 DC capacity use Limits inventory stored to DC capacity.

19
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

23 DC idle space Calculates unused space in owned DCs.

24 DC workforce variation Models changes in DC staffing over time.

25 DC idle labor Computes unused labor capacity in DCs.

26 DC inventory load Ensures subcontracted DC inventory does not exceed available capacity.

27 Transport min load (final Ensures minimum shipment load is met when using a transport type.
product)

28 Transport min load (raw Same as (27) but for raw material flows.
material)

29 Transport max load (final Ensures shipped quantity does not exceed allowed transport load.
product)

30 Transport max load (raw Same as (29) but for raw material flows.
material)

3. Model Validation
The proposed SCND model was implemented in Matlab R2014b and solved using the Gurobi
optimizer 6.6.0. Tests were performed on a system with an Intel Core i7 PC (3.4 GHz, 8GB RAM).

3.1. Validation Approach

Model validation was performed by:

• Testing on simple cases to validate sub-models.

• Comparing results with published solutions from Park, Choi, and Kang (2007).

3.2. Validation Groups

Table 15 Validation Group

Group Purpose

Group A Validates selection of suppliers, own production facilities, and production

subcontractors.

Group B Validates selection of own distribution facilities and warehousing

subcontractors.

Group C Validates selection of transportation types.

20
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
3.3. Comprehensive Model Validation
The model was validated using the smallest and largest test cases from Park, Choi, and Kang (2007).
The authors reproduced the experimental environment and tuned their model to match the Park et al.
formulation.

The smallest instance involved 2 suppliers, 2 plants, 2 DCs, 2 products, and 5 time periods. The results
of the proposed model matched Park et al.’s results exactly.

For the largest test, 100 instances were generated using Park et al.'s generator. Each instance involved
10 suppliers, 10 plants, 10 DCs, 10 products, and 10 time periods.

3.4. Performance Comparison

Table 16 Models' performance

Model Average Cost Time (s)

Proposed MILP model 6,021,500 0.26

Genetic algorithm (Park et al. 2007) 6,209,590 465

Difference 3% ×1788 faster

The proposed MILP model achieved the same optimal results as Park et al.'s model but did so with
drastically better performance. It solved the problem 1788 times faster with slightly lower average cost. This
confirms both the correctness and efficiency of the proposed SCND model.

21
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Chapter 5. Results
1. Result Illustration and Explanation
1.1. Case study
To test the proposed model, the authors applied it to a global commercial refrigerator manufacturer
with eight plants across Europe, Asia, and the US. The company’s supply chain includes 850 suppliers and
350 customers worldwide. Data for the case study was sourced from publicly available stock exchange reports,
with additional manufacturing details drawn from literature (e.g., Braglia, Carmignani, and Zammori, 2006),
including information on raw materials, plant departments, production times, and staffing ratios.

The goal of the case study was to redesign the company’s supply chain over a five-year period, aiming
to minimize costs while meeting projected demand. The demand forecast for the first year was based on the
company’s actual sales, distributed across countries in proportion to their GDP. Sales in countries with
multiple customers were equally divided. For the following years, sales were assumed to grow by 1.4% in
Europe, 2.9% in the US, and 5.5% in Asia (Global Economic Outlook 2015).

1.2. Redesigning the current supply chain network

The approach examines the present supply chain to identify inefficiencies and cost factors, ultimately
revising the MILP model for accuracy. Redesigning the network is critical to reducing costs and meeting
market demands. The case study demonstrates how poor location selection and resource allocation increases
costs. The solution is to optimize facility location, production, and inventories.

1.2.1. As-is Design and Model Calibration:

To model the firm's existing supply chain network (SCN), additional assumptions were made beyond publicly
available data, as detailed in Arampantzi and Minis (2016b).

[Link]. Network:

The authors modeled a simplified version of the real SCN, including 35 suppliers, 8 production plants, and
one main distribution center (DC). They selected 14 key customers responsible for 64% of sales, with demand
and parameters scaled accordingly.

[Link]. Products and Materials Supply:

The company categorized its products into four refrigerator families, each with a bill of materials (BoM) using
six raw material categories, which make up about 70% of the total product cost. Sourcing materials from Asian
suppliers reduces procurement costs by 15% (BCG 2004).

[Link]. Manufacturing:

Manufacturing involves five steps: metal, glass, plastic, assembly, and painting, with each department’s
capacity based on employee numbers. These were derived from company reports and department -specific
ratios. No plant expansions were considered, and labor and production cost data came from public sources,

22
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
including Institut der Deutschen Wirtschaft (2010) and Global-production Inc. (2014). Severance and hiring
costs were estimated at half and three months' salary, respectively.

[Link]. Warehousing and Inventory:

The model includes the primary DC with its known storage capacity. A minimum safety stock of 15% of
annual demand was set. Warehousing costs covered inventory, labor, HR turnover, idle capacity, and
depreciation. Outsourced warehousing costs were assumed to be 30% higher than in-house costs, excluding
labor mobility.

Figure 1 Illustration of the as-is SCND

[Link]. Transportation assumptions:

The model includes typical transport modes like road, rail, and sea, used alone or combined (multimodal).
Each transport type has minimum shipment sizes to stay cost-effective. Figure 1 shows the current supply
chain layout and where procurement, production, inventory, and sales happen.

To model the current system, some decision variables (Error! Reference source not found.) were fixed based o
n known data, while others like transport modes, routes, and workforce levels were left open for the model to
optimize.

The model’s cost estimates for the first year were compared with the company’s actual financial data.
Differences ranged from about –4% to –12%, mainly because the model found better solutions in areas where
it could choose freely and due to lower assumed procurement costs in Asia.

1.2.2. To-be design

For the proposed redesign, the authors used the same parameters as in the "as-is" design but allowed flexibility
in certain areas. Notably, the proposed design focused on optimizing the network to minimize costs over a
five-year planning horizon, while meeting projected demand.
23
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
Table 17 Comparison of major SCND costs (as-is design): Actual (from the company’s official financial statements) versus those obtained those
obtained from the proposed MILP model (a single year has been modelled).

Cost Category Actual (from Financial Model Results (from MILP Difference
Statements) model) (%)

Production €32,000,000 €30,722,072 -4.0%

Warehousing €22,000,000 €19,786,251 -10.1%

Procurement €110,000,000 €97,027,413 -11.8%

Transportation €11,000,000 €10,311,143 -6.3%

Total €175,000,000 €158,187,067 -9.6%

The design introduced the possibility of outsourcing warehousing and adjusting distribution center (DC)
capacities. However, no new production plants were included, as the company aimed to consolidate its
manufacturing operations. Instead, existing plants could expand their capacities to meet growing demand.

Figure 2 Illustration of the to-be SCND at the end of the planning horizon of five years.

The model identified three potential outsourcing DCs in addition to the company’s existing DC. These were
strategically located in Amsterdam (near the European plants), Hong Kong (for Asian operations), and
Pennsylvania (serving US customers). The locations were chosen based on global logistics hubs and the
company’s practice of placing DCs near manufacturing sites.

The new design significantly altered the supply chain operations:

• Europe: Production in Europe was reduced, with one plant in Greece no longer in use.

24
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
• Asia: A substantial shift occurred, with 65% of total production and 63% of inventory being handled
by Asian facilities. A logistics partner in Hong Kong managed the majority of this inventory.

• United States: All procurement and production activities were eliminated, with only minimal
warehousing remaining on the East Coast.
Table 18 . Comparison of major SCND costs resulting from the to-be design of the proposed MILP model versus the as-is design.

Cost Category As-Is Design To-Be Design Difference (%)

Production €31,535,932 €28,761,223 -8.8%

Warehousing €22,092,702 €21,051,450 -4.7%

Procurement €102,908,329 €97,883,715 -4.9%

Transportation €11,443,317 €10,197,060 -10.9%

Total €167,980,280 €157,893,448 -6.0%

The final proposed design, after optimization, demonstrated a clear shift towards lower-cost regions, mainly
Asia, where labor and procurement costs were more competitive. This led to lower transportation costs due to
increased use of sea freight, which is cost-effective for long-distance shipping. The redesign resulted in
significant cost savings across multiple supply chain functions.

2. Scenario Analysis
The scenario analysis investigates how changes in three key parameters affect the redesigned supply chain
network. These parameters include:

• Demand Reduction in Asia: assess the impact of reducing demand in Asia by varying percentages
(10%, 20%, 50%).

• Demand Reduction in Europe: To understand the effect of reduced demand in Europe by 10%, 20%,
and 50%.

• Increase in Outsourcing Inventory Costs: : To explore the impact of increasing inventory

outsourcing costs by 40%, 50%, and 60%.
Table 19 Major activities and total cost of SCN

SCND major activity KPIs Unit

Procurement Distribution of procurement activity per continent %

Production Distribution of production activity per continent and per plant %

Total warehousing cost * €

Warehousing
Total outsourced warehousing activity %

Transportation Distribution of total transportation per type (in product units per mode) %

25
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Normalised total cost Total cost (as per Equation (1)) per product unit €

2.1. Key Findings

2.1.1. Demand Reduction in Asia:

Figure 3 Impact Asia demand reduction on (a) Asia manufacturing and procurement activities, (b) global transport modal share, and (c)

Minimal Impact on Manufacturing and Procurement: Even with a 50% reduction in demand,
manufacturing activities decreased by only 10%, and procurement activities fell by 5%. This demonstrates the
strong position of Asian manufacturing and procurement in the supply chain, primarily due to lower labor and
material costs.

Shift in Transportation Modes: The global share of combined truck, rail, and sea transport decreased,
especially due to less production in Indonesia and increased production in Turkey, where transportation was
less reliant on multimodal solutions.

26
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
2.1.2. Demand Reduction in Europe:

Figure 4 Impact of EU demand reduction on (a) EU manufacturing, procurement activities, (b) global transport modal share, and (c) total

Significant Impact on Manufacturing and Procurement: A 50% reduction in demand led to a 62% decrease
in European manufacturing activities and a 66% reduction in procurement activities.

Shift in Transportation Modal Split: A 23% increase in the global share of sea transportation was observed.
This is due to the reduced transportation need for European manufacturing, with fewer products available to
meet the cost-effective multimodal transport thresholds (particularly land -based transport modes like truck
and rail).

Increase in Transportation Costs: The total cost per product unit increased by 4%, as fewer products from
Europe met the minimum thresholds for cost-efficient transport modes, particularly multimodal transport.

27
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
2.1.3. Increase in Outsourcing Inventory Costs:

Figure 5 Impact on increasing costs of outsourcing inventory on outsourcing inventory and total warehousing cost.

Preference for Outsourcing Remained: Despite a significant rise in outsourcing inventory costs, the model
still preferred outsourcing due to the higher costs associated with internal warehousing at the company’s
existing DC in the Czech Republic. Outsourcing was more cost-effective because of the lower storage costs
in countries like China.

Cost Differential: The existing internal warehousing had much higher costs, including labor and depreciation,
compared to outsourcing, making outsourcing the more attractive option even with increased costs.

3. Study on the Model’s Scalability

3.1. Introduction
To assess the scalability of the proposed Supply Chain Network Design (SCND) model, the authors conducted
experiments with progressively larger datasets. The objective was to determine the model’s capacity to handle
large-scale supply chain problems and deliver optimal solutions efficiently.

3.1.1. Experiment Setup

Five experiments were conducted, progressively increasing the number of suppliers, customers, and decision
variables. The final experiment, representing the actual company scale, involved:

• 850 suppliers

• 350 customers

This test case mirrored a global supply chain network.

Table 20 Solving Cases in Which the Number of Suppliers and Customers is Gradually Increased Up to the Actual Numbers

Suppliers Customers Decision Variables Constraints CPU (s) MIP Gap Tolerance (%)

35 14 25915 29442 256 0

70 28 56025 56112 4554 0

28
 IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

140 56 88765 109452 17859 0

280 112 172525 216132 10400 0.01

560 224 340045 429492 11613 0.01

850 350 653195 674967 6808 0.01

3.1.2. Results and Analysis

[Link]. Solution Quality

The model produced optimal or near-optimal solutions in all cases, with deviations of only 0.01% from the
optimal solution. This shows the model’s ability to handle complex, large-scale problems accurately.

[Link]. Computational Time

The largest problem, involving 850 suppliers and 350 customers, was solved in about 1.9 hours (6806
seconds). This time is considered reasonable for the complexity of the problem. The time spent building the
model’s constraints was notably higher (385,200 seconds), highlighting the challenge in model formulation.

[Link]. Efficiency Improvements

Several techniques were applied to enhance efficiency:

• Network Link Reduction: Redundant links were removed to simplify the network and reduce
complexity.

• Entity Clustering: Grouping similar entities together helped reduce the dimensionality of the problem.

• Pareto Principle: Focusing on the most impactful factors streamlined the problem-solving process.
Table 21 Review of SCND Industrial Applications

Model Size # Products # Suppliers # Plants # DCs # Customers # Types # Paths # Variables
Parameters

1 15 9 0 5 0 19 1 18500

2 5 1 5 5 32 75 1 32268

3 7 1 0 2 1 464 2 78291

4 1 2 7 7 0 4 3 N/A

5 1 3 6 8 0 12 1 464

6 4 2 2 2 2 2 1 1136

7 12 4 0 3 6 5 1 2916

29
 IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

8 1 1 0 6 7 14 1 N/A

9 12 3 2 2 5 9 1 2916

10 1 1 56 9 23 47 3 1823

11 1 7 0 1 12 278 2 153128

12 1 3 3 5 5 5 3 N/A

13 1 3 0 2 3 5 3 N/A

3.1.3. Conclusion
The SCND model successfully handled large-scale supply chain network problems, providing optimal
solutions within a reasonable timeframe. Techniques such as network link reduction, entity clustering, and the
Pareto principle improved efficiency, making the model suitable for real-world applications in global supply
chain optimization.

30
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

Chapter 6. Conclusion
1. Discussion on Assumptions
1.1. Model Assumptions
The model relies on several assumptions to simplify the SCND problem and make it applicable to real-world
scenarios:

• Product Families and Bill of Materials (BoM): The model uses product families, but can easily be
adapted to handle individual products if required. However, increasing product detail would raise the
model's complexity.

• Manufacturing Plants and Production Routings: While the model uses macro-routings for
production, these can be refined to greater detail, though this increases the model's computational
complexity.

• Distribution Channels and Warehousing: The model assumes three distribution channels, but
additional channels can be incorporated with minor adjustments. It also assumes a fixed quantity per
unit load (e.g., pallets), which can be modified without significant model changes.

1.2. Input Data Assumptions

The model primarily relies on publicly available data, such as stock exchange reports, for cost estimation and
demand forecasting. Where specific data is lacking, assumptions and estimates are used.

1.3. Transportation Assumptions

The model assumes various transportation modes (road, rail, sea, multimodal) with the ability to add capacity
constraints if needed. Adjustments for capacity are relatively simple to implement.

2. Solution Method and Analysis of Advantages and

Disadvantages
2.1. Solution Method
The proposed solution method employs an integrated Mixed Integer Linear Programming (MILP) model
to address the Supply Chain Network Design (SCND) problem. The model incorporates key strategic
decisions over a long-term horizon, such as:

• Supplier Selection

• Establishment or expansion of production facilities

• Subcontracting

• Warehousing (DCs) management

• Transportation mode and route selection

31
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management
The objective is to minimize overall costs related to procurement, production, inventory, warehousing, and
transportation while meeting customer demand. The model handles multiple time periods, products, and
complex supply chain relationships through appropriate constraints.

2.2. Advantages of the Solution Method

• Comprehensive and Integrated: The model captures all key elements of the supply chain from the
manufacturer’s perspective, providing a holistic approach to decision-making across various activities
such as sourcing, production, logistics, and distribution.

• Flexibility: The model can easily accommodate different supply chain configurations, including
varying transportation types and multimodal transport options, as well as outsourcing warehousing
activities.

• Scalability: The MILP model is capable of solving large-scale supply chain problems, as demonstrated
by its application to a global manufacturer with 850 suppliers and 350 customers. It efficiently
handles significant complexities while delivering optimal solutions within reasonable timeframes.

• Real-World Applicability: The model has been validated using real data from a global manufacturer,
proving its relevance and reliability for practical applications in industry.

2.3. Disadvantages of the Solution Method

• Computational Complexity: As the number of entities and decision variables increases, the model’s
complexity grows, making it computationally expensive. For large-scale supply chain problems,
solving the model can require significant computational resources, especially when accounting for
numerous constraints.

• Data Dependency: The model’s effectiveness relies heavily on the quality and accuracy of the input
data. Inaccurate or incomplete data, such as demand forecasts or cost estimates, can result in
suboptimal solutions, impacting the model’s reliability.

• Limited Outsourcing Flexibility: The current model assumes that manufacturing is either in-house
or fully outsourced for entire product families. However, this may not reflect real-world practices,
where components or subassemblies are often outsourced independently. Additionally, the model does
not account for more granular operational decisions, such as adjustments in production at the part or
subassembly level.

• Lack of Uncertainty Handling: The model assumes deterministic demand and supply parameters. As
such, it does not account for uncertainty in demand, supply disruptions, or external market changes
— factors that frequently affect real-world supply chains. This limitation can reduce the model's
applicability in environments with high uncertainty.

3. Conclusion
The MILP-based solution method provides a robust, flexible, and scalable approach for solving complex
supply chain network design problems. However, its reliance on accurate data, computational demands, and
assumptions on supply chain operations should be considered when applying it to real-world problems. Future
improvements could involve incorporating uncertainty, fine-tuning the outsourcing options, and reducing
computational complexity.

32
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

4. Recommendation For Future Research

Future research in this area may address some limiting assumptions made in this work, as well as new
exciting opportunities in sustainable SCND. As an example, the proposed model addresses either in-house or
outsourced manufacturing of a product family. However, in practice, subassemblies, parts, or even operations
in a part’s routing may be outsourced. The proposed strategic model does not address these finer cases, and
appropriate extensions are necessary. Future research may also address sustainability concerns in the SCND
process, as well as considerations on risk, quality, and lead times. Such enhancements may affect the network
architecture and have significant implications in a firm’s competitive advantage.

33
IS087IU VNU-International University
Logist ics En gineering & Supply Chain Design School of Industrial Engineering & Management

References
Arampantzi, C., Minis, I., & Dikas, G. (2018). A strategic model for exact supply chain network design and
its application to a global manufacturer. International Journal of Production Research, 56(17), 5575–5597.
[Link]
Kanyalkar, A. P., and G. K. Adil. 2005. “An Integrated Aggregate and Detailed Planning in a Multi-Site
Production Environment Using Linear Programming.” International Journal of Production Research 43 (20):
4431–4454.
Melo, M. T., S. Nickel, and F. Saldanha-da-Gama. 2006. “Dynamic Multi-Commodity Capacitated Facility
Location: A Mathematical Modeling Framework for Strategic Supply Chain Planning.” Computers &
Operations Research 33 (1): 181–208.
Manzini, R., and F. Bindi. 2009. “Strategic Design and Operational Management Optimization of a Multi
Stage Physical Distribution System.” Transportation Research Part E: Logistics and Transportation Review
45 (6): 915–936.
Ramudhin, A., A. Chaabane, and M. Paquet. 2010. “Carbon Market Sensitive Sustainable Supply Chain
Network Design.” International Journal of Management Science and Engineering Management 5 (1): 30–38.
Wang, F., X. Lai, and N. Shi. 2011. “AMulti-Objective Optimization for Green Supply Chain Network
Design.” Decision Support Systems 51 (2): 262–269.
Chaabane, A., A. Ramudhin, and M. Paquet. 2012. “Design of Sustainable Supply Chains Under the Emission
Trading Scheme.” International Journal of Production Economics 135 (1): 37–49.
Fahimnia, B., L. Luong, and R. Marian. 2012. “Genetic Algorithm Optimisation of an Integrated Aggregate
Production–Distribution Plan in Supply Chains.” International Journal of Production Research 50 (1): 81–96.
Pishvaee, M. S., J. Razmi, and S. A. Torabi. 2012. “Robust Possibilistic Programming for Socially
Responsible Supply Chain Network Design: A New Approach.” Fuzzy Sets and Systems 206: 1–20.
Fahimnia, B., R. Z. Farahani, R. Marian, and L. Luong. 2013a. “A Review and Critique on Integrated
Production–Distribution Planning Models and Techniques.” Journal of Manufacturing Systems 32 (1): 1–19.
Fahimnia, B., E. Parkinson, N. P. Rachaniotis, Z. Mohamed, and A. Goh. 2013b. “Supply Chain Planning for
a Multinational Enterprise: A Performance Analysis Case Study.” International Journal of Logistics Research
and Applications 16 (5): 349–366.
Sadrnia, A., N. Ismail, N. Zulkifli, M. K. A. Ariffin, H. N. Pour, and H. Mirabi. 2013. “A Multiobjective
Optimization Model in Automotive Supply Chain Networks.” Mathematical Problems in Engineering.
Mota, B., M. I. Gomes, A. Carvalho, and A. P. Barbosa-Povoa. 2015. “Towards Supply Chain Sustainability:
Economic, Environmental and Social Design and Planning.” Journal of Cleaner Production 105: 14–27.
Rezaee, A., F. Dehghanian, B. Fahimnia, and B. Beamon. 2015. “Green Supply Chain Network Design with
Stochastic Demand and Carbon Price.” Annals of Operations Research 250 (2): 1–23.
Zakeri, A., F. Dehghanian, B. Fahimnia, and J. Sarkis. 2015. “Carbon Pricing Versus Emissions Trading: A
Supply Chain Planning Perspective.” International Journal of Production Economics 164: 197–205.
Goetschalckx, M. 2011. Supply Chain Engineering. New York: Springer.
Goetschalckx, M., C. J. Vidal, and K. Dogan. 2002. “Modeling and Design of Global Logistics Systems: A
Review of Integrated Strategic and Tactical Models and Design Algorithms.” European Journal of Operational
Research 143 (1): 1–18.
Melo, M. T., S. Nickel, and F. Saldanha-da-Gama. 2009. “Facility Location and Supply Chain Management
– A Review.” European Journal of Operational Research 196 (2): 401–412.
Zhang, L. L., C. Lee, and S. Zhang. 2016. “An Integrated Model for Strategic Supply Chain Design:
Formulation and ABC-Based Solution Approach.” Expert Systems with Applications 52: 39–49.
Mota, B., M. I. Gomes, A. Carvalho, and A. P. Barbosa-Povoa. 2015. “Towards Supply Chain Sustainability:
Economic, Environmental, and Social Design and Planning.” Journal of Cleaner Production 105: 14–27.

34